Contribution of Adipose-Derived Factor D/Adipsin to Complement Alternative Pathway Activation: Lessons from Lipodystrophy Free

Complement is an important arm of innate immunity against bacteria and viruses and its activated products also serve as a bridge between innate and adaptive immunity. Three distinct but related pathways are responsible for complement activation: the classical pathway (CP), the alternative pathway (AP), and the lectin pathway. All three enzymatic pathways converge on the cleavage of C3. This results in the generation of proinflammatory complement fragment C3a and a major opsonin C3b. The latter also serves as gateway to the membrane attack complex and a substrate to engage the AP’s amplification loop. In the CP, C1, via its C1q subcomponent, binds Ab–Ag complexes, thus attacking foreign agents. In the lectin pathway, a plasma lectin, such as Abs ficolins, collectins, or mannose-binding lectin, binds to the microbial surface. In the AP, upon C3b’s covalent binding to a microbial surface, it engages the protease precursor (zymogen) factor B (FB) to form C3bB. This bimolecular complex is cleaved by factor D (FD), a serine protease unique to AP. The cleavage results in C3bBb, known as the AP C3 convertase, which is stabilized by properdin (P), an AP-specific positive regulator. FD itself is cleaved from a pro-FD zymogen to a mature FD. This cleavage is mediated by members of the mannose-binding protein-associated serine protease (MASP) family. Mice deficient in both MASP1 and MASP3 only have pro-FD in the serum and no AP activity, suggesting that MASP1/3 serve as proteases to cleave pro-FD into FD (1). MASP3 is thought to be the major pro-FD convertase whereas MASP1 is the minor convertase (2–4).

Owing to its low concentration in human serum [2–4 μg/ml (5, 6) and our data in this study] and highly selective function, FD is a strong candidate for pharmaceutical approaches to block the powerful AP and its feedback loop for the treatment of numerous inflammatory diseases such as atypical hemolytic uremic syndrome (7, 8), paroxysmal nocturnal hemoglobinuria (8), and age-related macular degeneration (AMD) (9–13). Abs to FD can be used to neutralize AP activity in mice (14) and more recently have been developed for the treatment of patients with AMD (15, 16). An mAb to FD was shown to block AP in primates without affecting the CP. AP components FB and C3 in blood are largely produced by the liver but also locally by bone marrow–derived and many other cell types (17). FD is also known as adipsin because it was independently identified as an abundant transcript in 3T3-L1 adipocytes (18), but it was soon identified as identical to FD (19, 20). In addition to the high level of mRNA expression of adipsin/FD in adipose tissue, adipsin/FD expression has been described in other tissues, including muscle and lung in humans (19) and the sciatic nerve (21) as well as by the adrenal gland and ovary in mice (22). Spiegelman and colleagues (23–26) went on to demonstrate that paradoxically adipsin mRNA is decreased in genetic and chemical models of obesity. More recently, they demonstrated that adipsin plays a role in pancreatic β cell function (27).

We have assessed FD levels and AP activity in a variety of lipodystrophic mice. These data, along with those derived from a series of dilution and serum-mixing experiments, have enabled us to determine that little FD is required for AP activity. In other words, the “safety factor” for FD appears to be high. These results have clinical implications for the treatment of AP-mediated disease with FD-blocking Abs or small FD inhibitory molecules (28).

Lipodystrophic mice were created by crossing adiponectin-Cre mice (29) with homozygous lox-stop-lox–ROSA–diphtheria toxin A mice (30). Of note, we observed that these lipodystrophic mice die when born at room temperature (RT) and to survive must be maintained at thermoneutrality (30°C) until weaning age. Following weaning, animals were maintained in a temperature-controlled room (22°C) on a 12-h light/dark cycle. FD^−/−, FB^−/−, P^−/−, and C3^−/− mice have been described (31–34). Animal work was performed according to the policies of the Animal Studies Committee at Washington University School of Medicine in St. Louis. Mice were analyzed under approved protocols and were provided appropriate care while undergoing research, which complies with the standards in the Guide for the Use and Care of Laboratory Animals and the Animal Welfare Act.

Primary mouse embryonic fibroblasts (MEFs) were prepared from wild-type (WT) C57/BL6 embryonic day (E)14 embryos as described (35). Cells were allowed to become confluent and then differentiated 3 d later. White adipocytes were generated by using a mixture of 1 μM dexamethasone, 5 μg/ml insulin, and 500 μM 3-isobutyl-1-methylxanthine (DIX) mixture for 3 d followed by insulin alone for 3 d and then base media (high-glucose DMEM plus 10% FBS) for 3 d such that adipocytes are harvested 9 d after differentiation. Brown adipocytes were generated by adding the thiazolidinedione troglitazone (15 μM) to the white adipocyte mixture (DIXγ mixture for 3 d and insulin-γ for 3 d) (35). All adipocytes were harvested on day 9 of differentiation with radioimmunoprecipitation assay buffer containing protease and phosphatase inhibitors. We have previously demonstrated that cells differentiated with DIX express white adipose tissue (WAT) markers and not brown adipose tissue (BAT) markers, whereas cells differentiated with DIXγ express BAT markers without suppressing WAT markers (S. Mascharak and C.A. Harris, unpublished observations).

Primary MEFs were prepared from WT histocompatible C57/BL6 E14 embryos as described (35). MEFs were injected s.c. at the sternum as reported (36). Mice were sacrificed 2 mo after transplantation and sera were used to measure FD levels and AP activity.

Recombinant mouse FD from Sino Biological (Beijing, China) and freshly isolated mouse serum were used in assays to remove N-linked glycosylated sugars by peptide-N-glycosidase (PNGase) F (QA-Bio, Palm Desert, CA). Following the manufacturer’s protocol, 1 μl of serum and various amounts of recombinant mouse FD were mixed with 30 μl of H₂O, 10 μl of 5× buffer, and 2.5 μl of denaturant prior to being heated at 100°C for 5 min. After cooling for 10 min and then adding 2.5 μl Triton X-100 and 2.0 μl of PNGase F to the reaction, the sample was incubated for 3 h at 37°C. Reaction solution (50 μl) was then mixed with 50 μl of SDS sample buffer containing reducing reagent (Sigma-Aldrich, St. Louis, MO). After heating for 60°C for 5 min, 30 μl of solution, equivalent to 0.3 μl of serum, was added and analyzed by SDS-PAGE.

Plasma samples were harvested from mice using sterile technique. ELISA plates (catalog no. 3855; Thermo Fisher Scientific, Waltham, MA) were coated with LPS (2 μg per well, catalog no. L2762; Sigma-Aldrich) at 4°C overnight. After washing three times in buffer (PBS/0.05% Tween 20) and blocking with BSA (250 μl per well, 1% in PBS buffer), diluted mouse plasma (1:5 or 1:10 in Mg²⁺-EGTA buffer) was incubated on an ELISA plate at 37°C for 1 h (50 μl per well). The plates were washed three times in washing buffer and then goat anti-mouse C3 Ab (100 μl per well, 1:4000 dilution in 1% BSA in PBS buffer; MP Cappel; MP Biomedicals, Santa Ana, CA) was added for 1 h. Wells were washed three times as above and incubated with HRP-conjugated anti-goat IgG (100 μl per well, 1:2000 dilution in 1% BSA in PBS buffer; Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h. After three washes, substrate reagent (R&D Systems, Minneapolis, MN) was added for 10 min (100 μl per well). The reaction was stopped with 50 μl of 1 M H₂SO₄, and the OD of samples was measured at 450 nm. Similar experiments were performed with normal human serum and FD-depleted human serum supplemented with purified human FD (CompTech, Tyler, TX). The primary and secondary Abs used in the LPS assay with human serum are: mouse anti-human C3d (no. HYB 005-01-02; Thermo Scientific) and HRP donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories), respectively.

A standard rabbit RBC hemolysis assay was employed to measure AP activity of mouse serum (34). Rabbit RBCs (1 μl) (Colorado Serum Company, Denver, CO) were placed in AP buffer (GVB with 20 mM MgCl₂ and 8 mM EGTA). Following centrifugation, the pellet was resuspended in the AP buffer and aliquots of 20% serum from lipodystrophy (LD) and various complement-deficient mice as well as human serum samples were added as described. After a 2 h incubation at 37°C, released hemoglobin was measured at an OD of 405 nm. Lysis of rabbit RBCs in water served as the positive control whereas rabbit RBCs in AP buffer served as the negative control. Hemolysis was determined by an OD ratio, that is, (20% serum with RBC in AP buffer − 20% serum without RBC in AP buffer)/(RBC in water − RBC in AP buffer). To account for minor differences in hemolysis between experiments, in each experiment the values were normalized to WT sera being 100% hemolysis.

As sera makes up 20% of the rabbit RBC hemolysis assay (above), we kept the total serum concentration the same, but we varied the contribution of WT serum and serum from knockout mice (C3, FB, FD, or P). For human serum mixing experiments, normal human serum was mixed with serum depleted of C3 or FD (CompTech).

Fresh serum from mice was employed to measure C3, FB, P, or FD by Western blotting. Goat anti-M C3 (MP Biomedicals), goat anti-Hu FB (CompTech), rabbit anti-M P (37), sheep anti-M FD (R&D Systems), and rabbit anti-Hu FD (Abcam, Cambridge, MA) were incubated with the membranes for 2 h at RT or overnight at 4°C. Secondary HRP-conjugated rabbit anti-goat IgG (Sigma-Aldrich), goat anti-rabbit IgG (GE Healthcare UK), donkey anti-sheep IgG (R&D Systems), or rabbit anti-human IgG (Jackson ImmunoResearch Laboratories) was added for 1 h at 37°C. Membranes were developed with a SuperSignal West kit (Pierce).

For in vitro experiments, 10 μl of serum, 10 μl of 30 μg/ml cobra venom factor (CVF; Quidel, San Diego, CA) and 10 μl of AP buffer (GVB with 20 mM MgCl₂ and 8 mM EGTA) were incubated at 37°C for 30 min. For in vivo experiments, 30 μg of CVF was injected i.p. into the mice and serum was harvested 2 h after the CVF injection. All samples were subjected to Western blots under reducing conditions.

Patients were enrolled in a study to evaluate the effects of recombinant human methionyl leptin (metreleptin) therapy on LD (https://clinicaltrials.gov/ct2/show/NCT00025883). We only used baseline sera samples (before leptin treatment) for this study, which was conducted at the Clinical Center of the National Institutes of Health and was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. Written informed consent was obtained from all participants. Samples were also collected from healthy volunteers with informed consent under a protocol approved by the Washington University Human Studies Committee. Human FD was determined by an ELISA kit from R&D Systems according to the manufacturer’s protocol as well as by Western blots with anti-human FD Ab from Abcam.

Both endogenous FD of mouse serum and recombinant mouse FD were a mixture of glycosylated isoforms with two prominent bands at an M_r between 37 and 50 kDa (Fig. 1A). A similar sized full-length isoform was observed in whole-cell lysates from inguinal and epididymal WAT as well as BAT (Fig. 1B). We also saw similar sized full-length isoforms in whole-cell lysates of MEFs that had been differentiated into white as well as brown adipocytes, whereas no FD was observed in undifferentiated MEFs (Fig. 1C). Treatment of sera or recombinant mouse FD with PNGase F resulted in a single, 25-kDa deglycosylated isoform (Fig. 1A). This also occurred when whole-tissue lysates from visceral epididymal WAT, superficial inguinal WAT, and BAT (which have different physiological roles) were treated with PNGase F (Fig. 1D).

We set out to create lipodystrophic mice by crossing two publicly available lines (29, 30). When adiponectin-cre mice were crossed to homozygous lox-stop-lox–diphtheria toxin A mice and pups were delivered at RT, adiponectin-cre⁺ mice were not found at Mendelian frequency (3 of 46 mice were Cre⁺ and lipodystrophic, p < 0.05 by χ² test). To assess whether this was due to nonspecific leakage of adiponectin-cre in another tissue or whether it was a specific functional consequence due to the loss of adipose tissue, we postulated that the latter might be rescued by allowing mice to be born in a thermoneutral setting; that is, the reduced viability was due to the loss of BAT-mediated thermoregulation and resultant hypothermia. The cross was repeated with the dam moved from RT to thermoneutrality once they appeared pregnant (∼E14). Mutant double-transgenic mice were now born at the expected Mendelian frequency. Subsequently, we found that pregnant mice carrying mutant progeny could be moved from RT to thermoneutrality up until the night before parturition without loss of viability in lipodystrophic pups (27 of 62 progeny Cre⁺, not statistically different from the Mendelian expected 50%). Double-transgenic lipodystrophic mice can be recognized soon after birth due to an interscapular defect normally containing BAT (Fig. 2A). LD mice had undetectable adiponectin (Fig. 2B), grossly absent WAT (Fig. 2C–F) and BAT (Fig. 2G, 2H), and massive hepatic steatosis (Fig. 2I–L)

We next asked whether FD could be detected in the plasma of LD mice. Western blotting revealed FD in the serum of WT mice, but FD was undetectable in the plasma of LD mice (Fig. 3). We also examined the relative abundance of the other AP complement proteins in serum. C3 and P levels were similar in WT and LD mice, but, interestingly, the FB concentration was approximately twice the level in LD compared with WT mice (Fig. 3).

We subsequently used two functional assays to determine the activity of the AP activity in the LD mouse. There was no detectable AP in LD plasma in an LPS-coated microtiter plate assay whereas WT plasma demonstrated robust activity (Fig. 4A). In this assay, plasma is added to LPS-coated wells and then the quantity of deposited C3 fragments is determined. We performed a second AP assay testing the ability of serum to lyse rabbit RBCs. In this experiment LD sera AP activity was 15% of the WT activity (Fig. 4B).

We followed these experiments by investigating whether purified human FD could rescue the AP defect in sera from LD mice and FD^−/− mice. FD (20 ng/ml) rescued ∼50% of the AP defect in both LD and FD-null mice and 100 ng/ml FD completely rescued the AP defect in sera from both LD and FD^−/− mice (Fig. 5A). These data further indicate that LD mice have virtually no FD and that relatively small amounts of FD are sufficient to rescue an FD deficit.

We next examined sera from mice heterozygous for C3, FB, or FD. There was a modest decrease in AP activity of C3 (63.3 ± 22.1%, p < 0.05 versus WT) and FB heterozygous mice (76.7 ± 18.8%, p < 0.05 versus WT), but no difference between FD heterozygous mice (109.3 ± 4.3%) and WT mice (100 ± 18.9%, Fig. 5B).

To more directly investigate the relationship between mouse FD levels and AP activity, we conducted mouse serum mixing experiments. In hemolysis assays, the total amount of serum was held constant at 20% (i.e., 20 μl of sera in a 100-μl RBC lysis reaction) but the contribution of WT serum and mutant serum was altered. This was accomplished by mixing WT sera with that from FD^−/− mice, P^−/−, FB^−/−, or C3^−/− mice. The absence of P only had a small effect on the AP activity as measured by RBC hemolysis assay (Fig. 6). Reduction of FB or C3 in the reaction by mixing WT serum with either FB^−/− or C3^−/− sera resulted in a linear relationship between the amount of FB or C3 and AP activity. However, FD showed a strikingly different relationship. When the FD concentration was lowered to 50, 25, or 12.5% of the WT levels (by mixing with FD^−/− sera), the AP activity remained normal (Fig. 6A). Experiments in which mixing serum at a lower range of WT sera showed that 2.5% of normal FD levels maintained full AP activity (Fig. 6B). A further mixing experiment showed that a FD level of 0.5% (10 ng/ml using our measured value of FD of 10 μg/ml in mouse serum) supported 43% of AP activity (Fig. 6C).

To gain additional insight into the relationship among adipose mass, FD levels, and AP activity, we examined an additional mouse strain with partial LD, fatty liver dystrophy (fld). Fld mice have ∼80% loss of adipose tissue (38). FD levels were reduced by ∼70% in sera of fld mice (compared with WT mice, Fig. 7A, 7B), whereas levels of other complement components, C3, FB, and P, were similar to WT mice (Fig. 7A, 7B). Titration of plasma from fld mice also showed that FD levels were ∼30% of WT (Fig. 7C). Despite only having ∼30% of normal FD levels, serum from fld mice reconstituted 77.5 ± 8.1% of WT AP activity (nonsignificant compared with WT) in the rabbit RBC hemolysis assay (Fig. 7D).

The other approach we used to probe the relationship among adipose mass, FD levels, and AP activity was to perform rescue experiments with LD mice. We harvested MEFs from histocompatible C57/BL6 E14 embryos and injected the cells s.c. under the sternum of LD mice (36). A large fat pad developed at the injection site within 4 wk of injection. Mice were sacrificed 8 wk after injection. The fat pads weighed 500–1000 mg, and therefore constituted 10–20% of the normal fat mass of age-matched WT mice. Remarkably, many of the metabolic derangements (e.g., fatty liver) were rescued by the formation of the ectopic fat pad. We were now also able to detect circulating FD in LD mice that had been rescued by MEF injection. However, the FD level in transplanted mice only represented ∼1–5% of WT circulating FD levels (Fig. 8A). Despite these reduced FD levels, sera from the rescued LD mice were able to support AP activity at a level of ∼30% of WT levels (Fig. 8B).

CVF is similar to host-derived C3b in that it can work together with FB to trigger the AP. Consequently, we examined the ability of CVF to induce C3 cleavage (to C3b) in sera of WT, LD, FD^−/−, or FB^−/− mice in vitro. CVF robustly induced cleavage of C3 in WT and LD sera, whereas very little cleavage occurred in sera from either FD^−/− or FB^−/− mice (Fig. 9A). The FB and FD blots confirmed the absence of FD in LD and FD^−/− sera and the increased FB in LD sera.

We used the ability of CVF to induce C3 cleavage in vivo when injected into WT, LD, FD^−/−, or FB^−/− mice. CVF injection into WT mice resulted in C3 cleavage. CVF injection into FB^−/− mice resulted in no cleavage of C3, indicating that FB is absolutely required for C3 cleavage. Surprisingly, injection of CVF into LD or FD^−/− mice resulted in cleavage of C3, suggesting that FD may be dispensable for CVF-induced cleavage of C3 in vivo (Fig. 9B). This observation is similar to that in experiments performed when FD^−/− mice were initially described (31). For these experiments, Western blotting again demonstrated an absence of FD in the sera of FD^−/− and LD mice and increased FB in the sera of LD mice.

We next pursued whether the absolute requirement of adipose tissue for circulating FD levels observed in the mouse system applied to humans. We tested circulating FD levels in 1) normal subjects, 2) patients with congenital generalized LD (CGL) due to mutations in acyl-glycerol acyltransferase 2 (n = 3) or Berardinelli–Seip congenital LD (BSCL)2 (n = 1), 3) patients with familial partial LD (FPL) due to mutations in lamin A (LMNA, n = 7), and 4) patients with acquired generalized LD (AGL) (n = 5) using an ELISA. We found that FD was detectable in patients with these forms of LD, but that FD levels were significantly lower in CGL than in controls. Specifically, CGL patients had circulating FD levels ∼50% of that in controls (Fig. 10A). Patients with FPL or AGL had levels of FD between those of controls and CGL. We performed a Western blot on the sera from patients with CGL and controls and observed that FD levels were reduced in CGL patients compared with controls (see Fig. 10B). Our data on human FD with Western blots showed that there was a single unglycosylated band with a molecular mass of ∼25 kDa. Because of the variability in fat mass in these patients, we created a plot of adipose mass versus FD levels and saw a linear relationship between adipose mass and FD levels (Fig. 10C, R² = 0.388). We performed the RBC hemolysis assay for AP on patient sera and observed equivalent levels of RBC hemolysis with CGL and control sera.

To determine whether the low concentrations of FD were also sufficient for activation of AP activity in the human system, we performed mixing experiments between normal human sera and human sera that had been immunodepleted of either C3 or FD. These experiments mirrored what was seen in mice; namely, a low concentration of human FD was needed to activate AP (∼1% for half-maximal activity) whereas there was more of a linear relationship between human C3 for AP activation (Fig. 10D).

The literature from the 1970s suggests that FD is the limiting factor in AP activation (39). That view was based on rabbit hemolysis assays performed using the addition of purified human FD or human serum to substitute for diisopropylfluorophosphate-inactivated FD in human plasma (39). Those experiments showed that adding more than physiological amounts of FD (up to 9-fold) resulted in increased AP activity. Those experiments, however, were carried out with incubation times of only 5 min. We have performed equivalent assays using purified human FD and FD-depleted human serum with a similar result at 5 min incubation (Fig. 11A, 11B). In contrast, at longer incubation times much less added FD was required to attain WT-level AP activity (Fig. 11C). Note that in this lytic system employing RBCs, by 10 min there was >90% lysis when using 400 ng/ml human FD (hFD; Fig. 11B). Moreover, substantial lysis was obtained at 25 ng/ml FD in the standard 30-min incubation commonly employed in this assay system (Fig. 11C). We also performed a dose response curve for FD in the LPS microplate assay using different incubation times both in human (adding hFD to FD-depleted serum) and mouse (mixing WT and FD-null sera) systems (Fig. 11D, 11E). When we added various amounts of hFD back to FD-depleted human serum we observed an EC₅₀ of 76 ng/ml for the activation of AP with a 30-min incubation and an EC₅₀ of 5 ng/ml with a 2-h incubation (Fig. 11D). Addition of hFD to WT serum did not increase AP activity (data not shown). We also conducted similar experiments by mixing WT mouse sera with FD-null mouse sera. For this experiment the total percentage of serum in the assay was held constant (20%) but the proportion of WT and FD-null serum varied. Using our measured values of 10 μg/ml for WT serum, the amount of FD required for 50% activation in this assay was 1.8% (36 ng/ml) and 0.3% (6 ng/ml) for the 30-min and 2-h incubation times, respectively (Fig. 11E). These numbers are quite similar to what we saw with the RBC hemolysis assay.

Our studies using lipodystrophic and FD^−/− mice establish that FD/adipsin is almost entirely produced from adipose tissue. FD was not detected in LD mice. Although FD was previously described to be highly expressed in adipose tissues, others had reported significant expression in nonadipose tissues such as by the sciatic nerve in mice (21) as well as by muscle and lung in humans (19). Our experiments indicate though that although other tissues might have the capacity to produce FD, it is the adipose tissue that accounts for almost all of the circulating FD in mice. Note that when comparing the mouse and human systems, circulating mouse FD is found at higher concentrations [10–30 μg/ml; our data and Ref. (1)] than human FD (2–4 μg/ml).

In contrast to the mouse, there were reduced (∼50%, p < 0.05) FD levels in the sera of patients with CGL. This is most likely because the patients with LD still have a small amount of adipose tissue as compared with the lipodystrophic mice that were genetically engineered to be completely devoid of adipose tissue. Alternatively, either FD is normally secreted from nonadipose tissues in humans, or nonadipose tissues [e.g., the liver, which highly expresses peroxisome proliferator-activated receptor (PPAR)-γ and some adipocyte markers in the lipodystrophic state] (40) compensate to secrete FD in the lipodystrophic state. FD levels were reduced to a lesser extent in the sera of patients with AGL or FPL, a state of partial loss of adipose mass. It is unclear why patients with AGL did not have as severe a reduction in FD levels compared with CGL patients. One potential explanation is the fact that AGL patients retain variable amounts of visceral adipose tissue whereas it is completely absent in CGL patients (41–43). Additionally, marrow adipose tissue is retained in AGL and absent in both CGL1 and CGL2 (41–43), the patients included in the present study. Also, it is perhaps not too surprising that patients with CGL have detectable FD and AP activity, as an increased incidence of Neisseria infections is not observed in these patients as has been described in patients with loss-of-function FD mutations (44–46).

Another interesting observation was that lipodystrophic mice had increased levels of circulating FB. This is unlikely to be due solely to a compensatory effect in the absence of FD, as it was not seen in FD^−/− mice (Fig. 8). FB is an acute phase response protein (17). We favor a hypothesis that the hepatic steatosis in LD mice induces hepatic inflammation and this in turn stimulates hepatic FB expression.

A finding of note is that CVF was able to facilitate in vitro C3 cleavage to a greater extent in LD than FD^−/− sera in vitro (Fig. 9A). This may be due to a small amount of FD in sera of LD mice that is not detected by our assay systems (LPS microplate assay, RBC hemolysis, or Western blotting). Another possibility is that the increased FB in sera of LD mice accounts for augmented CVF-mediated C3 cleavage. However, when CVF was injected into WT, FD^−/−, LD, or FB^−/− mice, C3 was cleaved in WT, FD^−/−, and LD mice, but not in FB^−/− mice (Fig. 9B). These data point to an FD-independent mechanism of CVF-induced C3 cleavage. One possibility is that a second “compensatory” protease is responsible (31). Kallikrein is a possible candidate given its similar protease activity to FD (47).

Using both mixing experiments with sera from FD^−/− mice and using LD mice rescued with preadipocyte transplant, we found that only a small percentage of the normal FD level is required to support significant in vitro AP activity in a time-dependent manner (Fig. 11 and see below). This was observed with two different assays: rabbit RBC hemolysis and the LPS microplate C3 deposition assay. Utilizing the LPS assay for FD, there was scant, if any, AP activity detected at 5 min (even with “normal” FD levels). At 30 min, 3–10% of normal FD levels were required for maximal activity. Remarkably, at 120 min nearly maximal deposition occurred and required ∼1% of the serum FD levels. These data strongly support the concept that the FD dosage for maximum AP activation is not assay specific (as for P) but is especially dependent on kinetics (duration of the incubation). Thus, time of incubation and FD concentration are key features in AP assays and will play out in a distinctive fashion at various tissue sites based in part as well on the type and magnitude of cellular/tissue damage or degeneration. We are in the process of comparing other assays for the AP.

This finding likely has clinical implications. Near complete inhibition of FD may be necessary to fully suppress AP complement activation in autoimmune and innate immune-mediated inflammatory human diseases, a conclusion in contrast to that of prior studies that FD is the rate-limiting component of AP activation (39). As we have shown, this dissimilarity is based on differences of in vitro reaction times. FD-blocking Abs have been and are being developed for the treatment of diseases featuring AP activation such as AMD and have entered clinical trials (15, 16). This approach was prompted by the observation of complement system proteins being deposited in the retina of patients with AMD (48) and then followed by the discovery that common (9–12) and rare variants (49, 50) in complement factor H are associated with risk for AMD. Future studies will determine whether FD inhibition will result in immunosuppression and opportunistic infections (which could be addressed prophylactically with meningococcal vaccination) (51, 52) and whether low levels of functional FD that may survive FD inhibition are sufficient to sustain pathogenesis in AP-mediated disease models, especially in chronic inflammatory conditions such as AMD.

We further speculate that these data are likely to affect how we envision complement activation in extravascular tissue sites. Rather amazingly, the results indicate that low levels of FD are sufficient to mediate AP activation. We propose that this process is a fundamental mechanism by which the host responds to altered self, that is, cellular and tissue debris that is continuously generated (both for intracellular and extracellular disposal).

Following this line of reasoning, we hypothesize that this disposal activity was part of the original complement system. It was then adapted by the host to facilitate recognition of a foreign agent. With the subsequent evolutionary development of circulatory systems, reaction speed becomes more critical. However, the AP, in particular, has maintained its ability to recognize and opsonize debris.

We thank Abhimanyu Garg for assistance in genotyping patients with LD, Brian Finck for providing the fld mice, and Madonna Bogacki and Lorraine Schwartz for manuscript editing.

The authors have no financial conflicts of interest.