Complement Deficiencies Result in Surrogate Pathways of Complement Activation in Novel Polygenic Lupus-like Models of Kidney Injury

This article is featured in In This Issue, p.2615

Systemic lupus erythematosus (SLE) is a complex polygenic autoimmune disease associated with more than 90 genetic risk loci (1). Despite the large number of genes implicated, distinctive pathways of immune function, such as complement activation, immune complex (IC) and apoptotic cell (AC) processing, B cell tolerance, and IFN, are affected (2). Involvement of these pathways is commonly reflected in three hallmarks of SLE: 1) low serum complement levels, 2) accumulation of AC/AC debris in the germinal centers (GC) and kidney tissue, and 3) presence of autoantibodies (3–7). How these pathways interact and lead to lupus nephritis (LN), the main contributor to morbidity and mortality in SLE patients (8, 9), remains unclear.

Both animal models and evidence in humans indicate that defective AC clearance is an important component of SLE development (4, 6, 10, 11). Cells from SLE patients show accelerated death ex vivo (12) and impaired phagocytosis of AC (13), both of which could be the contributing factors to increased accumulation of cellular debris found in the GC of SLE patients (14). Multiple opsonins, including milk fat globule epidermal growth factor 8 (MFGE8), and cell surface receptors have been shown to bind AC and enable their clearance (4, 6, 15). MFGE8 is a soluble glycoprotein that binds to phosphatidylserine on AC, facilitating the removal of AC through its interaction with the α_vβ₃ integrin on phagocytes (10, 16). Aberrant splicing of Mfge8 has been associated with dysregulated MFGE8 function in SLE patients, as have genetic polymorphisms in Mfge8 (17, 18). MFGE8-deficient mice accumulate AC in GC and develop a lupus-like disease on the mixed C57BL/6 × 129 (10) but not on the pure C57BL/6 (B6) background (19). Therefore, defective AC clearance alone is not sufficient to induce a lupus-like pathology but can fuel disease progression in a susceptible environment.

The role of complement in kidney disease is complex: early components (C1q, C2, C4, and C3) function to clear AC and ICs, whereas late components form the membrane attack complex (MAC) that acts as an effector of kidney injury in SLE (20–22). Specifically, more than 90% of C1q-deficient individuals develop SLE, including LN in a subset of those patients (3), and anti-C1q Abs in SLE patients are associated with more severe LN (23). C3 deficiency can also lead to SLE, characterized by LN in ∼26% of C3-deficient patients (24). In addition, low circulating C3 levels are strongly associated with active and recurrent renal disease in SLE patients, in part because of consumption (25, 26). Whereas previous studies have suggested that C1q and C3 play a role in AC clearance and B cell tolerance (3, 27–29), the mechanisms that lead to kidney injury in low complement states in the presence of autoantibodies and defective clearance of dead or dying cells remain unknown.

To determine the mechanisms of kidney injury in a disease model that closely mimics human SLE pathogenesis, we generated SLE-like polygenic murine models that have defective clearance of AC (Mfge8-deficiency), produce anti-chromatin autoantibodies (Sle1 interval), and have low complement levels (C1q- or C3-deficiency). We demonstrate that spontaneous nephritis and proteinuria, associated with increased IgG deposition in the kidney and enhanced B cell autoreactivity, occur only in the presence of all three abnormalities (i.e., in Sle1.Mfge8^−/−C1q^−/− [C1q^−/− triple-mutant (TM)] or Sle1.Mfge8^−/−C3^−/− [C3^−/−TM] mice, both of which exhibit glomerular and tubular structural features akin to those found in LN patients). Increased IgG2c levels, together with C3/C3d, MASP-2, and MAC deposition in the glomeruli of C1q^−/−TM mice suggest activation of the lectin complement pathway in the absence of C1q. The unexpected discovery of MAC deposition in C3^−/−TM mice introduces a novel C3-independent mechanism of complement activation in low C3 states in the kidney. Using a model of IC-mediated nephritis, we uncover that thrombin contributes to complement activation and kidney damage in the absence of C3. Enhanced fibrin deposition in the kidneys of C3^−/−TM but not C1q^−/−TM mice reveals a role for thrombin activation in low C3 states.

Mfge8^−/− mice on B6 background were kindly provided by Dr. Nagata (10). C1q^−/− mice were a gift from Dr. M. Botto (30). C3^−/− mice (31) were originally from The Jackson Laboratory and were kindly provided by Dr. J. Heinecke (University of Washington). B6.Sle1 mice were from Dr. L. Morel (32). All mice were crossed onto the C57BL/6 background for at least seven generations. Twelve murine strains were examined across two different disease models: 1) spontaneous disease (B6.Sle1, Sle1.Mfge8^−/−, Sle1.C1q^−/−, Sle1.C3^−/−, Sle1.Mfge8^−/−C1q^−/− [C1q^−/−TM] and Sle1.Mfge8^−/−C3^−/− [C3^−/−TM]) and 2) accelerated nephrotoxic nephritis (NTN) (Mfge8^−/−, C3^−/−, and Mfge8^−/−C3^−/−, on B6 background). Mice subjected to accelerated NTN were immunized with sheep IgG (1 mg s.c.; Jackson ImmunoResearch) combined with Freund’s adjuvant (Sigma-Aldrich), and 5 d later they were challenged intravascularly with one dose of nephrotoxic serum (NTS) or normal sheep serum (75 μl–5 mg/ml). The preimmunization protocol generates anti-sheep IgG Abs that bind to sheep IgG from NTS that has fixed on glomerular Ags in the kidney. Proteinuria was quantified by Bradford assay (Bio-Rad Laboratories) and urine albumin/creatinine ratio (UACR) was evaluated using Albuwell Albumin Assay and its Creatinine Companion Kit (Exocell). Survival of different mouse strains in the spontaneous model was monitored over 12 mo, and data were presented by Kaplan–Meier curve. All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Washington, Seattle.

Serum samples from different mouse strains were collected every month after birth and stored in −80°C before analysis. To detect anti-dsDNA IgG, calf thymus DNA (10 μg/ml) was dissolved in DNA Binding Solution (Pierce) and was used to coat ELISA plates at 4°C overnight. To detect anti-chromatin Abs, 10 μg/ml purified histone from calf thymus (Sigma-Aldrich) was added to the plate to form histone/DNA complexes [artificial chromatin (33)]. Pooled serum samples from 7-mo-old NZB/W F1 female mice were used as a standard for anti-dsDNA and anti-chromatin IgG detection assays. For both assays, serum samples (1:100 in PBS/10% FBS) were incubated overnight at 4°C, and IgG binding was detected using alkaline phosphatase–conjugated goat anti-mouse IgG (Sigma-Aldrich). For confirmation of anti-dsDNA IgG, Crithidia luciliae slides (Bio-Rad Laboratories) were used. Serum samples were diluted in PBS/FBS (1:50), and IgG binding was detected by Alexa 555–conjugated anti-mIgG Ab. To analyze the autoantibody repertoire, pooled serum samples from six different strains used in the spontaneous disease model (12-mo-old females, n = 15–20) were examined for reactivity to a defined autoantigen array at the University of Texas Microarray Center (University of Texas Southwestern). Normalized data were clustered according to the multidimensional principal component analysis and visualized in heatmap form using Qlucore Omics Explorer data analysis software; statistically significant expression of strain-specific autoantigens was determined (Qlucore, Lund, Sweden).

Formalin fixed kidney tissues sections (3 μm) were stained with periodic acid–Schiff (PAS) at the University of Washington Pathology Research Service Laboratory. Kidneys were scored in a blinded fashion using a modification of previously described scoring systems (34, 35). Briefly, the following were scored: glomerular proliferation/PAS-positive material, glomerular hypercellularity and/or cellular infiltration, tubular protein, and interstitial inflammation. For glomerular mesangial expansion, the following scores were assigned: 0 = absent; 1 = minimal, with thickening of the mesangium and minimal diffuse glomerular loop staining; 2 = mild, with PAS-positive material affecting <10% of glomeruli; 3 = mild-to-moderate, with 10–25% of glomeruli; 4 = moderate-to-marked, with 25–50% of glomeruli; and 5 = marked, with >50% of glomeruli. Other parameters were scored on a scale of 0–5, where 0 = absent, 1 = minimal (<5% glomerular or cortical area affected), 2 = mild (5–20%), 3 = mild-to-moderate (21–33%), 4 = moderate (34–65%), and 5 = marked (>65%). Representative images were taken using NIS-Elements BR 3.2 64-bit. Transmission electron microscopy was performed using whole kidney that was immersion-fixed in formalin. After removing the capsule, small samples of cortex were postfixed in 2% osmium tetroxide and processed by standard procedures. Thin sections (0.1 μm) were stained with uranyl acetate and lead citrate, then examined with a Jeol JEM-1230 electron microscope. Normal mouse kidney was obtained in a separate experiment, fixed in half-strength Karnovsky Fixative (2% paraformaldehyde, 2.5% glutaraldehyde in sodium cacodylate buffer), then processed, stained, and examined in a similar fashion.

Frozen spleen sections (6 μm) were fixed in acetone and blocked in TBS with 5% BSA and 0.05% Tween 20. Tissues were stained with peanut agglutinin and Alexa Fluor 488 anti-mouse CD169 Ab (BioLegend). Frozen kidney sections (5 μm) were fixed in cold acetone for 10 min and blocked (PBS/1% BSA/5% goat serum) for 30 min at room temperature. Mouse IgG deposition in the kidney was detected with FITC-conjugated goat anti-mouse IgG Ab or Alexa 555–conjugated goat anti-mouse IgG2c Ab (Invitrogen). Sheep IgG deposition was evaluated using FITC-conjugated goat anti-sheep IgG Ab (Jackson ImmunoResearch). Deposition of complement component C3 in the kidneys was detected with FITC-conjugated goat anti-mouse C3 Ab (1:100; CEDARLANE), whereas C3d staining was performed using rabbit anti-mouse C3d IgG (1:100; R&D systems) followed by Cy5-conjugated goat anti-rabbit IgG (1:1000; Jackson ImmunoResearch). Deposition of complement component C4 in the kidneys was detected with FITC-conjugated goat anti-mouse C4 Ab (1:100; CEDARLANE). Fibrin or fibrinogen staining was performed using FITC-conjugated goat anti-mouse fibrin/fibrinogen [abbreviated fibrin(ogen)] IgG (1:100; OriGene). Fibrin was distinguished from fibrinogen using biotin-labeled 59D8 anti-fibrin IgG (15 μg/ml) provided by Dr. Esmon (Oklahoma Medical Research Foundation); kidney sections were fixed in 4% paraformaldehyde, free aldehyde groups were blocked with 0.1 M glycine/PBS, and tissues were blocked with avidin and biotin-blocking buffers (Vector Laboratories); fibrin staining was detected using Cy5-labeled streptavidin (1:200). MAC deposition was detected with rabbit anti-mouse C5b-9 IgG Ab (clone B7, 1:100), developed and provided by Dr. P. Morgan (36), followed by Alexa 647–conjugated goat anti-rabbit IgG (1:1000; Jackson ImmunoResearch). Mannan-binding lectin-associated serine protease 2 (MASP-2) deposition was detected with rabbit anti–MASP-2 IgG (1 μg/ml; Abcam), followed by Alexa Cy5-conjugated goat anti-rabbit IgG (1:1000; Jackson ImmunoResearch); kidney tissue from MASP-2 knockout mice was used as negative control (37). For indirect immunofluorescence (IF) staining, nonspecific binding to Abs deposited in the kidney was prevented by preincubating both primary and secondary Abs with 5 μg/ml purified mouse IgG as well as with 5 μg/ml sheep IgG for kidneys from mice exposed to accelerated NTN. All IF staining was imaged using the digital fluorescence microscope EVOS FL, and fluorescence intensity and colocalization were quantified by Image J. AC in the spleen and kidney tissues were detected using Roche In Situ Cell Death Detection Kit, Fluorescein (Sigma-Aldrich). The total number of TUNEL⁺ cells per 50 randomly selected glomeruli was determined using Image J.

Thrombin activity in the plasma of C3^−/−TM, C1q^−/−TM, and B6 mice was measured as previously described (38) using S-2238 (H-d-phenylalanyl-l-pipecolyl-l-argininep-nitroaniline dihydrochloride; Diapharma) as a substrate. At a 1:1 ratio, plasma (25 μl) was incubated with the thrombin substrate and the amount of p-nitroaniline dihydrochloride formed was measured by absorbance at 405 nm. A standard curve was generated using bovine thrombin (Sigma-Aldrich), and data were presented as units of thrombin. RNA was isolated from C3^−/−TM, C1q^−/−TM, and B6 kidneys (QIAGEN RNAeasy MidiPrep Kit). Tissue factor (TF) mRNA levels were detected by quantitative PCR (SYBER Green; Invivogen) using the following primers: 5′-TCAAGCACGGGAAAGAAAAC-3′ (forward) and 5′-CTGCTTCCTGGGCTATTTTG-3′ (reverse).

Statistical analyses were performed using GraphPad Prism 7. Normality of the data were determined using a Shapiro–Wilk test, and statistical difference between two sample means (specified in the figures and figure legends) was determined using a Student t test for normally distributed data sets and a Mann–Whitney U test for not normally distributed data sets. Differences between three or more data groups were analyzed by multiple comparison one-way ANOVA with a Bonferroni post hoc test. Correlation strength between two parameters was determined by Pearson correlation analysis. Statistical significance was determined according to the following p values: *p < 0.05, **p < 0.01, and ***p < 0.001. The Heatmap function in Qlucore was used for unsupervised clustering of autoantigen array data.

Deficiencies in the early complement components are among the strongest disease-susceptibility genes in SLE patients (3). Most C1q (>90%) and 26% of C3-deficient (24) patients develop severe disease. However, single-complement deficiencies in mice do not lead to spontaneous disease, and ∼30–42% of C1q-deficient SLE patients develop glomerulonephritis (21, 39), which indicates that other genes and/or defects in immunologic pathways contribute to the development of autoimmunity and nephropathic effects in SLE. Similarly, the presence of anti-chromatin Abs (B6.Sle1) or impaired AC clearance (B6. Mfge8^−/−), both important abnormalities contributing to lupus pathogenesis, are insufficient to drive spontaneous disease in B6 mice (19, 40, 41). Therefore, to determine whether and how complement deficiencies exacerbated the SLE phenotype in a truly polygenic SLE-like environment (i.e., in the presence of autoantibodies and impaired AC clearance [Sle1.Mfge8^−/−]), we first generated two double-mutant (DM) strains, Sle1.C1q^−/− mice and Sle1.C3^−/− mice. We then crossed these DM strains to the Sle1.Mfge8^−/− strain to create Sle1.Mfge8^−/−C1q^−/− and Sle1.Mfge8^−/−C3^−/− TM mice. Histologic analyses of the kidneys of both TM strains showed membranoproliferative glomerulonephritis with PAS-positive deposits, glomerular hypercellularity, and mesangial expansion, as well as multifocal-to-coalescing cellular infiltrates in the kidney interstitium, which tended to be more severe in animals with more pronounced glomerular pathology (Fig. 1A, Supplemental Fig. 1A). Significantly higher pathology scores were observed in Sle1.Mfge8^−/− mice that lacked either C1q or C3 (C1q^−/− or C3^−/− TM mice, respectively), compared with the DM controls (Fig. 1B). The higher pathology scores in TM strains were due to moderate-to-marked glomerular mesangial expansion and glomerular hypercellularity (glomerular scores), as well as marked-to-moderate tubulointerstitial inflammation (interstitial scores) (Fig. 1A, 1B, Supplemental Fig. 1A). Whereas complement deficiency exacerbated nephritis in Sle1.Mfge8^−/− mice, lack of MFGE8 was sufficient to trigger changes in the kidney pathology in Sle1 mice (Fig. 1A, 1B). However, only the TM strains (9–11 mo) spontaneously developed proteinuria (Fig. 1C). Despite severe kidney pathology in both TM strains, the C1q^−/− but not the C3^−/− TM mice demonstrated reduced survival at 12 mo of age (Supplemental Fig. 1B). The reasons for the difference in survival is not certain but may be related to the broader anti-inflammatory effects of C1q (42).

IgG autoantibodies are central to the pathogenesis of glomerulonephritis in SLE patients as well as in lupus mouse models (43–45). To what extent kidney damage is instigated by circulating Ab:Ag (immune) complexes versus accumulation of dead cell debris in the kidney (planted Ag) targeted by IgG autoantibodies (46–48) is controversial. Because, in addition to MFGE8, early complement components facilitate AC clearance (27, 49), we first asked whether there were differences in AC deposition in the test strains. Although both complement-deficient TM strains had a significantly greater number of TUNEL⁺ AC in the glomeruli relative to their Mfge8-sufficient DM controls (Sle1.C1q^−/− and Sle1.C3^−/−), these levels were comparable to Sle1.Mfge8^−/− mice (Fig. 1D), indicating that reduced clearance of cell debris in the kidney was not sufficient to account for the more severe nephritis in the TM strains. In contrast, in the TM models of spontaneous nephritis, renal IgG deposition was significantly higher in both C1q- and C3-deficient TM mice compared with Sle1.Mfge8^−/− controls (Fig. 1E). Sle1.Mfge8^−/− mice had comparable levels of IgG deposition with B6.Sle1 and the complement-deficient DM strains (Sle1.C1q^−/− and Sle1.C3^−/−) (Fig. 1E).

Electron microscopic analysis of representative kidney samples from Sle1.Mfge8^−/− demonstrated increased leukocyte infiltration in the glomeruli of this strain compared with the normal control (Fig. 2A, arrows). Subendothelial electron-dense immune deposits were detected in the glomerular basement membranes of C3^−/−TM mice, resembling “wire loop” deposits seen in human LN (50) (Fig. 2B). C1q^−/− TM kidneys demonstrated features characteristic of human membranoproliferative glomerulonephritis, including mesangial immune deposits with focal extensions into the subendothelial space of the capillary walls, duplication of basement membrane matrices, and deposits interposed between duplicated membrane matrices (Fig. 2C, first and second panels). Some deposits showed a fibrillar-organized substructure, similar to that seen in patients with LN (51) (Fig. 2C, third panel). Additionally, we observed severe damage of proximal tubules, characterized by cellular dissolution and loss of interdigitating cell borders (Fig. 2C, fourth panel). In sum, these substructural abnormalities further illustrate the similarities between these novel lupus strains and human SLE.

B6.Sle1 congenic mice spontaneously produce anti-chromatin but only produce anti-DNA Abs when additional genetic intervals containing lupus susceptibility loci (sle2 and sle3) are crossed to B6.Sle1 (32). Similarly, many mouse strains deficient in opsonins, such as MFGE8 (19) or C1q (52), fail to develop anti-dsDNA Abs and overt clinical disease on the B6 genetic background, indicating that AC accumulation is insufficient to induce high-affinity pathogenic anti-dsDNA Abs in these strains. Because we observed enhanced renal IgG deposition in both TM strains (Fig. 1E), we next asked whether complement deficiency in the presence of autoantibodies and AC accumulation alters the autoantibody response. Indeed, at 5 mo of age, both TM strains developed significantly higher titers of anti-chromatin and anti-dsDNA IgG compared with the DMs (Sle1.C1q^−/−, Sle1.C3^−/−, and Sle1.Mfge8^−/−; Fig. 3A, 3B). Only female 5-mo-old Sle1.Mfge8^−/− mice had increased anti-dsDNA Ab titers compared with B6.Sle1 controls (Fig. 3A, 3B). Elevated levels of IgG anti-dsDNA were confirmed in 60% of C1q^−/−TM and 80% of C3^−/−TM mice by C. luciliae (representative images in Fig. 3C). At the same age (9 mo), ∼30% of Sle1.Mfge8^−/− mice developed anti-dsDNA autoantibodies.

To examine the effects of complement component deficiencies on the autoantibody repertoire in complement-deficient Sle1.Mfge8^−/− mice, we analyzed the specificities of mouse sera in an autoantigen microarray panel (53). As shown in Fig. 3D, the deletion of Mfge8 had a significant effect on broadening the autoantibody specificities of either Sle1 or DM strains containing Sle1 and a complement deficiency. Strikingly, the combination of Sle1 with Mfge8^−/− and either C1q or C3 deficiencies in TM mice not only increased Ab titers and led to an earlier breaking of tolerance (5 mo) but also enabled Ab production against numerous autoantigens found in other autoimmune and/or rheumatic diseases (e.g., histones, collagens, aggrecan, α-actinin, cytochrome C, hemocyanin, vimentin, complement factors, SRP54, and SP100), many of which were previously reported in murine models and SLE patients with LN (Fig. 3D) (53–55). Although many autoantibody specificities were shared between both TM strains, C1q^−/−TM mice developed strain-selective autoreactivity to Sm, SmD, ribosomal P protein, hemocyanin, vimentin, and laminin (Fig. 3D, Supplemental Fig. 2A). Autoreactivity in C3^−/−TM mice was selective for α-actinin, collagens, aggrecan, and chondroitin sulfate (Fig. 3D, Supplemental Fig. 2B). Surprisingly, anti-C1q Abs were detected in C1q^−/−TM and anti-C3 Abs were detected in C3^−/−TM mice (Fig. 3D). These could be explained by low affinity cross-reactivities [e.g., anti-DNA Abs in SLE patients cross-react with C1q (56) and/or polyclonal B cell activation, for example, generation of anti-complement Abs following virus infection (57, 58)].

The increase in autoantibody titers and repertoire in TM mice was accompanied by significant splenomegaly in 10–12-mo-old mice compared with the DM strains (Supplemental Fig. 1C). This increase was associated with an expansion in TUNEL⁺ AC in both the red pulp and the follicles of the spleen (Supplemental Fig. 1D), the organ shown to play a prominent role in the clearance of AC from the circulation (10). Together, these findings suggest that both C1q and C3 play a role in suppressing autoantibody responses to AC, and their absence enables Ag spread.

Although IC injury in the kidney is thought to be initiated by IgG2c fixing C1q and triggering classical complement pathway activation (59), whether and how complement activation contributes to nephritis in the absence of C1q is not known. We first examined the IgG subclass deposited in the kidney. Although no difference was observed in the intensity of renal IgG3 staining between strains (data not shown), IgG2c was detected in both the glomeruli and the tubular interstitium of C1q^−/−TM mice at significantly higher levels compared with the DM controls and the C3^−/−TM strain (Fig. 4A). To determine whether a complement-mediated pathway contributed to nephritis in the absence of C1q, we tested whether the MAC was present in the kidneys of C1q^−/−TM mice. Of considerable interest, we detected abundant staining of MAC in the kidneys of C1q^−/−TM mice, thus confirming terminal complement pathway activation in the absence of C1q (Fig. 4B). Glomerular MAC expression was significantly higher in the C1q^−/−TM compared with the Sle1.Mfge8^−/− DM strain (Fig. 4B), suggesting that complement activation contributes to the kidney injury seen in C1q^−/−TM mice. Consistent with activation of complement pathway despite absence of C1q, we detected abundant C3 accumulation in the glomeruli and tubules of C1q^−/−TM mice, with tubular C3 levels significantly greater compared with Sle1.Mfge8^−/− controls (Fig. 4C). Moreover, C1q^−/−TM kidneys demonstrated enhanced tubular and glomerular deposition of C3 breakdown product C3d relative to Sle1.Mfge8^−/− controls, which indicates activation of the alternative or lectin complement pathways (Fig. 4C). Immunofluorescence staining revealed deposition of MASP-2 in the tubules and glomeruli of C1q^−/−TM kidneys (Fig. 4D). MASP-2 colocalized with C3 (Fig. 4D), indicating complement activation via the lectin pathway (LP). Because we did not detect C4 in C1q^−/−TM kidneys (data not shown), LP activation occurred independently of C4, as previously reported in the ischemia-reperfusion model of kidney injury (60). Some tubular and glomerular MASP-2 deposition was also observed in Sle1.C1q^−/− mice, however, to a significantly lower degree than in C1q^−/−TM kidneys and with minimal colocalization with C3 (Supplemental Fig. 3).

Whereas the absence of C3 in the C3^−/−TM strain would suggest that nephritis in these mice is complement independent, surprisingly, we detected MAC deposition in the glomeruli of C3^−/−TM mice and at a significantly higher fluorescence intensity than in the Sle1.Mfge8^−/−DM controls (Fig. 5A, 5B), indicating that a C5 convertase had been generated. Because a handful of studies have identified thrombin as a surrogate C5 convertase (38, 61–63), we looked for evidence of thrombin activation in the kidneys of C3^−/−TM mice. Indeed, immunofluorescence staining revealed fibrin(ogen) deposition in C3^−/−TM glomeruli (Fig. 5C). Because deposition of fibrinogen could be a relatively nonspecific finding in tissue damage, we used an mAb, 59D8, that distinguishes fibrinogen from fibrin (the cleavage product of thrombin) as it binds to the N terminus of cross-linked fibrin β-chain, a site not exposed in fibrinogen (64–66). Positive staining for fibrin was observed only in C3^−/−TM but not in Sle.Mfge8^−/− DM nor in C1q^−/−TM mice (Fig. 5D). Moreover, we detected a significantly higher number of glomeruli staining positive for fibrin in C3-deficient TM mice compared with Sle1.Mfge8^−/− and C1q^−/−TM controls (Fig. 5D), indicating that fibrin deposition was selective for inflammation in the absence of C3. Significantly higher thrombin activity in the plasma of C3^−/−TM mice compared with C1q^−/−TM controls (Fig. 5E) confirmed increased systemic activation of the coagulation pathway in the absence of C3. At the kidney level, we detected greater TF expression in C3^−/−TM relative to C1q^−/−TM mice (Fig. 5E), which could play a role in thrombin activation in the kidneys of the C3-deficient TM strain (67).

Because our spontaneous model takes ∼9 mo for disease to develop, we turned to a well-described short-term model of kidney injury instigated by Ab binding to planted glomerular Ags, accelerated NTN (68), to investigate whether thrombin-triggered complement activation contributes to kidney disease in the absence of C3 (Fig. 6A). Analogous to the findings in TM mice that developed worse nephritis in the absence of C3, generation of NTN in C3-deficient mice (C3^−/− and Mfge8^−/−C3^−/−) led to worse kidney disease, characterized by an increase in the UACR, compared with their respective controls (B6 and Mfge8^−/−) (Fig. 6B). Kidney injury was independent of the glomerular deposition of sheep or mouse IgG or circulating Ab titers, as similar levels were detected in all NTS-treated strains (Supplemental Fig. 4A–D). Neither Ab deposition nor kidney disease were observed in normal serum–treated controls (data not shown). Parallel to the findings in the C3^−/−TM mice, we detected greater glomerular deposition of both MAC and fibrin(ogen) in C3-deficient mice treated with NTS (C3^−/− and Mfge8^−/−C3^−/−) compared with B6 and Mfge8^−/− controls, respectively (Fig. 6C, 6D). The levels of renal MAC and fibrin(ogen) deposition in NTS-treated C3^−/− mice were comparable to those found in Mfge8^−/−C3^−/− kidneys (Fig. 6C, 6D), indicating that increased thrombin activity associated with MAC assembly in Ab-mediated kidney injury arises specifically in the absence of C3, independent of the disease model and consistent with observations in IC-mediated lung injury (38). However, worse kidney disease in Mfge8^−/−C3^−/− mice compared with C3^−/− controls indicates that the increased AC burden because of MFGE8 deficiency contributed to additional inflammatory processes in NTS-mediated injury and increased UACR.

To verify that thrombin contributes to NTS-driven kidney injury in C3-deficient mice, we compared kidney function in the presence or absence of the thrombin-specific inhibitor argatroban (69) (Fig. 6E). Indeed, treatment of Mfge8^−/−C3^−/− mice with argatroban led to a decrease in NTS-triggered UACR compared with vehicle-treated controls (Fig. 6F). The improvement in kidney function in argatroban-treated mice was accompanied by a decrease in glomerular deposition of fibrin(ogen) (Fig. 6G) as well as in diminished glomerular MAC deposition (Fig. 6H). Argatroban-mediated effects were independent of the levels of sheep or mouse IgG deposition in the kidney or circulating Ab levels, as similar titers were detected in drug- versus control-treated animals (data not shown). Together, these findings strongly suggest that, in the absence of C3, Ab-mediated renal damage leads to thrombin activation in the kidney and is responsible for generating a C5 convertase that subsequently triggers MAC formation and kidney damage.

Significant limitations to current mouse models of lupus exist. In the spontaneous models, the genetic factors that contribute to disease are largely unknown, rendering the ability to tease apart cause and effect a challenge. In addition, single-gene knockout models do not accurately reflect the complexity of polygenic SLE in humans. We have therefore generated mouse strains that contain combinations of abnormalities in three common pathways affected in SLE: defective clearance of AC, low complement levels, and B cell dysfunction. Significantly, we observed that, whereas anti-chromatin autoantibodies and defective clearance of AC (Sle1.Mfge8^−/− mice on the B6 background) were not sufficient to induce clinical nephritis, the addition of either C1q or C3 deficiency promoted spontaneous glomerulonephritis in Sle1.Mfge8^−/− mice with “subclinical lupus” (19). Two major differences in the TM compared with DM mice were identified (Fig. 7). First, serum autoantibodies in both C1q^−/− and C3^−/− TM mice were higher, showed significantly greater Ag spread, and were associated with increased glomerular IgG deposition compared with their matched Sle1.Mfge8^−/− counterparts. Second, although Sle1.Mfge8^−/− as well as both TM strains had increased AC debris in the kidneys, the TM mice showed significantly greater activation of the MAC of the complement pathway. Examination of the mechanisms involved in generation of the MAC revealed that activation occurred through different pathways in C1q^−/− and C3^−/− TM mice. In C1q^−/−TM mice, the alternative and/or LP were likely activated, leading to generation of abundant glomerular and tubular C3d. When C3 was absent, thrombin activation was observed locally and systemically, and functional inhibition of thrombin significantly attenuated kidney injury in the NTN model, strongly suggesting that thrombin acted as an alternative C5 convertase. Our novel disease model demonstrates that the three pathways commonly abnormal in SLE combine to promote disease by enhancing B cell function and, at the target tissue level, enabling multiple nonconventional pathways of terminal complement pathway activation, which is relevant in human disease.

The increased Ag spread in the absence of the complement components C1q or C3 in addition to Sle1.Mfge8^−/− (C1q^−/−TM and C3^−/−TM strains) was surprising because of the previously described indispensable role that complement and, particularly, C3 plays in foreign Ag processing and lowering the threshold of the B cell immune response (70–73). Instead, our findings suggest that C1q and C3 prevent or modulate production of Abs to self-antigens arising from AC (Fig. 7A). Complement proteins could protect against Ag spread by several mechanisms. First, we found significantly greater levels of AC in the spleens of C1q^−/−TM and C3^−/−TM mice compared with Sle1.Mfge8^−/− controls, which supports the important role that both C1q and C3 play in the clearance of AC (27, 42, 49). Therefore, one possibility is that absence of complement enables a greater quantity and persistence of self-antigens, which leads to necrosis and activation of phagocytes and adaptive immune cells (74). C1q has also been shown to promote a tolerogenic state by modulating both the dendritic cells and macrophages toward anti-inflammatory functions (75, 76) as well as by enhancing presentation of self-antigens to immature B cells, thus promoting negative selection by BCR editing, anergy, or death (77–79). Similarly, C3 bound to self-antigens from dying cells likely facilitates immune tolerance through swift clearance and enhancement of anti-inflammatory cytokines by phagocytes (27, 80). Baudino et al. (81) have proposed that C3 modulates intracellular trafficking and processing of opsonized and phagocytosed AC Ags, thus controlling the immune responses to self-antigens. Together with excess dying cells in splenic follicles, the absence of C1q- and C3-mediated tolerogenic mechanisms might be contributing to Ag spread in C1q^−/−TM and C3^−/−TM mice. Finally, lack of C1q and C3 could influence the nature and availability of B cell epitopes. For example, C1q appears to be required to drive C1r/C1s toward proteolytic activity of nucleolar Ags (82), and C3 deficiency could lead to aberrant proteolysis by thrombin or other proteases, yielding peptides that are more immunogenic. Future studies will address these critical questions.

Because we demonstrated an expansion in the autoantibody response in TM mice relative to their DM controls, the development of nephritis in these strains could be attributed to the quality or quantity of these autoantibodies. The higher titers of autoantibodies, particularly anti-DNA, in the two TM strains may well contribute to the disease, as these autoantibodies are known to associate with renal disease exacerbations in human SLE and to possibly cross-react with other renal targets (83–85). The quality of the IgG Abs was also different, particularly in C1q^−/−TM mice [a greater deposition of IgG2c, the IgG subtype with potent effector functions for both complement fixation and engagement of the FcgR on myeloid cells (86, 87)]. Regarding differences in Ag targeting, many of the identified Ab specificities in the TM strains, including anti–α actinin, anti-aggrecan, anti-phosphatidylinositol, anti-proteoglycans, anti-collagens, anti-vimentin, anti-laminin, anti-fibrinogen, and anti-hemocyanin autoantibodies, have been associated with severe nephritis in both murine (54) and/or human studies (55, 88–90). Similar nephritogenic properties have been attributed to anti-vimentin autoantibodies, which were shown to target vimentin in situ and to associate with tubulointerstitial inflammation in LN patients (88). In our model, anti-vimentin autoantibodies were selective for C1q^−/− TM mice, which demonstrated C3d deposition similar to some SLE patients (91) and significant tubular damage by electron microscopy. Indeed, microarray analysis of kidney tissues confirmed elevated levels of IgG transcripts in situ in C1q^−/−TM mice, similar to what was observed in human tubular interstitial inflammation LN (88). Therefore, in addition to anti-DNA Abs, autoantibodies reactive to local kidney Ags likely played a role in the nephropathic effect in the TM mice lacking either C1q or C3.

Besides the consequences of complement component deficiencies on B cell tolerance in the presence of excess AC, the significant increase in nephritis in the TM strains could result from the cumulative effects that MFGE8 and C1q/C3 deficiencies have on AC and IC clearance. However, AC accumulation was similar in Sle1.Mfge8^−/− and the TM mice, indicating that AC accumulation in the kidney alone was not sufficient to account for the nephritis observed in the TM mice. These findings suggested that the early complement components play a role in preventing nephritis. This could occur through a variety of mechanisms that include direct engagement by C1q of inhibitory receptors, such as LAIR-1 and CD33 on immune cells (92–95), as well as C3-mediated dampening of innate immunity (27, 96–98). Unexpectedly, we observed that absence of C1q or C3 allowed surrogate mechanisms to activate complement and lead to the assembly of the MAC. The MAC not only mediates tissue (kidney) injury through perforin lysis but also by activation of inflammatory signaling pathways (99–101). Besides MAC, terminal complement activation generates another potent inflammatory mediator, C5a, which could contribute to kidney injury. In fact, C5a/C5aR signaling has been found instrumental in mediating nephritis in Mrl^lpr mice by driving recruitment of inflammatory cells and production of Th1 cytokines (102). Moreover, direct engagement of C5Ra on kidney endothelial cells could trigger cellular proliferation and cytokine production, further exacerbating the kidney pathology (103, 104).

In C1q^−/−TM mice, abundant deposition of C3 and its breakdown product, C3d, in the glomeruli and tubules was associated with MASP-2 deposits, indicating that C3 was activated by the lectin complement pathway. Because C3d was detected in the tubules as well as glomeruli, the role of local tubular C3 production (105, 106) and activation in the generation of interstitial nephritis will be important to explore further. Moreover, the significantly greater MASP-2 deposition in C1q^−/−TM compared with Sle1.C1q^−/− mice, supports the activation of mannose-binding lectin in the presence of increased tissue damage (107, 108).

Whereas C1q insufficiency is strongly associated with SLE development, homozygous deficiency in C3, the convergence point of all three complement pathways, has a lower penetrance (∼26%) (24). In practice, low levels of C3 resulting from disease activity or the presence of autoantibodies to C3 (∼30% of SLE patients) (109) are encountered frequently in active SLE and are associated with more severe LN, often characterized by increased glomerular MAC deposition (110–112). Greater MAC accumulation in the kidneys of C3^−/−TM mice, compared with their matched Sle1.Mfge8^−/− counterparts, raised the question as to how MAC is assembled in the kidneys in the absence of C3. Because thrombin was previously identified as an alternative C5 convertase in IC-mediated lung injury in C3-deficient mice (38), we sought evidence of thrombin activation in C3^−/−TM kidneys. Detection of fibrin in the glomeruli of C3^−/−TM but not C1q^−/−TM mice implicated thrombin as a surrogate C5 convertase responsible for MAC generation. A direct role for thrombin in the generation of the MAC in C3-deficient kidneys was confirmed by functional attenuation of this protease using the thrombin-specific inhibitor argatroban in the NTN model. Together, these findings reveal a novel mechanism of kidney injury in lupus-prone mice with low C3. Although, to our knowledge, this is the first study to identify a role for thrombin in terminal complement activation in the kidney in C3 deficiency, our findings are consistent with previous reports of increased expression and enzymatic activity of thrombin leading to surrogate C5 convertase activity in the liver and blood of C3-deficient mice (38, 63) (Fig. 7B). What triggers the compensatory mechanism in the absence of C3 is unknown. Interestingly, we detected increased expression of TF, a known initiator of the extrinsic coagulation pathway, in C3^−/−TM but not C1q^−/− TM kidneys. The elevated expression of TF in the absence of C3 was also reported in C3^−/− mice with IC-mediated lung injury (38) as well as in a model of allograft rejection in mice lacking C3 (113). This latter study demonstrated that excessive production of C5a in the absence of C3 led to increased activation of TF in endothelial cells and subsequent thrombin generation. It is plausible that tissue damage in the kidney, possibly because of Ab deposition, coupled with lack of opsonization by C3 drives TF expression in C3^−/−TM mice. Alternatively, IgG-dependent activation of thrombin in autoantibody-mediated arthritis has been reported in C3^−/−-deficient mice (63) and could also play a role.

Complement and coagulation cascades coevolved as necessary homeostatic pathways rich in serine proteases, and numerous interactions between these pathways have been documented (5, 114). Studies published decades ago revealed that C5 could be cleaved by proteases in the coagulation pathway, including thrombin and plasmin (61), a result confirmed in vivo in models of IC-mediated lung disease and Ab-mediated arthritis (38, 63). Detailed molecular studies revealed that during TF-induced clotting of plasma, thrombin efficiently cleaved C5 at the highly conserved R947 site rather than the conventional site of cleavage by C5 convertase R751, generating the intermediates C5T and C5bT, which were more effective than C5b in generating MAC (67). However, the efficiency of thrombin-mediated C5 cleavage in these in vitro studies was lower than that of the conventional C5 convertase. Whereas this might not be the case in vivo, the modest reduction in MAC deposition and kidney injury on thrombin inhibition in our model does suggest that other proteases besides thrombin might act as surrogate C5 convertase in the absence of C3. These could be other proteases in the coagulation pathway, such as factors Xa, IXa, and Xia (5), in the fibrinolysis pathway, such as plasmin (115), or independent of the coagulation cascade, such as macrophage and neutrophil proteases (116, 117).

Of relevance to LN, studies published more than 20 y ago reported increased activation of the coagulation pathway in the kidneys of SLE patients with active disease (118, 119) and efficacy of fibrinolytics on attenuating kidney disease in SLE (120). Recently, elevated levels of coagulation cascade proteins in urine strongly predicted renal disease in patients with LN (121). Interestingly, TF was one of the proteins that correlated with clinical parameters of SLE (121), suggesting the role for extrinsic activation of the coagulation pathway in the LN kidney. Recent systematic integrated (genomic, proteomic, and metabolomic) analysis has shown that cleavage of complement components associated with activation of the coagulation cascade in SLE patients, likely because of enhanced inflammatory cytokine production (122). Although the relationship between low C3 levels and thrombin activation has not been studied at the tissue level in LN, terminal complement activation in the kidney has been reported in a single C3-deficient SLE patient (123), implicating alternative pathways of C5 cleavage in the absence of C3. These clinical associations, coupled with our findings, encourage further investigation of thrombin activation and therapeutic intervention in SLE patients with low C3 levels.

In summary, we propose the following model for LN in our novel polygenic mice constructed to resemble pathway abnormalities in human SLE. In the presence of increased self-antigens and hyperresponsive B cells (124), lack of early complement components increases autoantibody production and Ag spread. Some of these autoantibodies (e.g., anti-DNA, anti-α actinin, and anti-vimentin) (45, 85) deposit in the kidney, causing cell injury and augmenting the damage-associated molecular pattern molecules arising from defective clearance of cell debris. These damage-associated molecular pattern molecules, which may include mannose-binding lectin, collectins, and ficolins and may also induce TF expression (125, 126), activate downstream complement pathways (lectin and alternative), as well as surrogate pathways such as thrombin (refer to Fig. 7). Assembly of the MAC leads to necrosis and also increases inflammatory responses in the kidney. Whereas FcgR engagement on myeloid cells may play a contributory role in these models, in Ab-mediated arthritis it was reported either that activation of both the complement pathway and FcgR were necessary (127) or that inflammation was independent of FcgR and C3, yet dependent on the downstream complement activation (C5) and thrombin generation (63, 127).

We thank Drs. Paul Morgan and Charles Esmon for providing the MAC and the fibrin Abs, respectively, as well as Drs. Dudler and Yaseen from Omeros (Seattle, WA) for providing MASP-2 knockout mouse tissues. We also thank Dr. Christian Lood for insightful discussion.

The authors have no financial conflicts of interest.