Comparative Analysis of Novel Complement-Targeted Inhibitors, MiniFH, and the Natural Regulators Factor H and Factor H–like Protein 1 Reveal Functional Determinants of Complement Regulation

The complement cascade is increasingly recognized as an important mediator of immunological and inflammatory processes (1). Triggering of the lectin pathway (LP), the classical pathway (CP), or the alternative pathway (AP) of complement activation leads to the production of bimolecular C3 convertases. These convertases proteolytically activate the central complement component C3 by cleaving it into the anaphylatoxin C3a and the opsonin C3b. C3b can then covalently associate with both foreign and host surfaces. In the absence of regulation, C3b deposition is rapid, and progression from the early cascade (C3 activation) to the terminal pathway occurs with the formation of C5 convertases that split C5 into C5a (a potent anaphylatoxin) and C5b. C5b nucleates the assembly of the lytic membrane-attack complex.

Activation of the CP or LP requires recognition of pathogen- or danger-associated molecular patterns. Whereas the AP response may be initiated and enhanced by the positive regulator properdin (2–4), the AP also has the unique property of remaining continuously and indiscriminatingly activated, albeit at a low level (referred to as ‘tick-over’). In the AP, C3b self-propagates via a positive-feedback amplification loop (a comprehensive scheme of the cascade is given in Ref. 5). This self-amplification feature of the AP as well as the indiscriminate nature of C3b deposition during tick-over necessitates very tight regulation that is specific to host cells. Factor H (FH) and its splice product FH like-1 (FHL-1) are the key soluble AP regulators and act together with membrane-bound regulators on self-cells (i.e., CD35, CD46, and CD55). FH is composed of 20 complement control protein domains (CCPs), whereas FHL-1 consists of FH CCPs 1–7 and an additional four C-terminal residues. FH occurs in the blood at concentrations of ∼2–3 μM (6–8), whereas FHL-1 is less abundant (∼1 μM) (8). Both regulators specifically adhere, via a polyanion-binding site in CCP 7 and another in CCP 20 in the case of FH (9, 10), to glycosaminoglycans and sialic acids on host surfaces. Thus, FH and FHL-1 not only prevent complement depletion from plasma (because C3b amplification can occur in the fluid phase as well as on surfaces) but also directly protect host cell surfaces from accumulating C3b (9, 11, 12). FH and FHL-1 prevent the formation of AP C3 convertases and accelerate its dissociation (decay-accelerating activity [DAA]) and also promote factor I–mediated proteolysis of C3b (cofactor activity [CA]).

Failure to control the AP can result in disease (13). Examples include the kidney conditions, for example, atypical hemolytic uremic syndrome and C3 glomerulopathy with dense deposit disease, as well as a hematologic disorder, paroxysmal nocturnal hemoglobinuria (PNH). In these diseases, mutations in complement genes or Abs against complement components lead to an imbalance between activation of the AP and its regulation. Age-related macular degeneration (AMD) is another prominent AP-mediated disease. AMD is a major cause of visual loss in the Western world and, in regard to pathophysiology, is only poorly understood. AMD is associated with common and rare genetic variants in FH (among other complement proteins; reviewed in Ref. 14), supporting the idea that AP regulation is critical to maintaining a physiological steady-state in the eye. Most studies, which have investigated the common AMD-associated single nucleotide polymorphism in CCP 7 (Y402H), have centered on FH as a major genetic contributor to this disease. Recently, FHL-1 was identified as the predominant complement regulator within Bruch’s membrane, a structure in the eye that is implicated in the pathophysiology of AMD (11), highlighting the need for more studies of FHL-1.

A promising therapeutic strategy for diseases linked to inadequate complement regulation is to block C3b amplification by using engineered versions of FH (15–17). Examples are miniFH and the FH-complement receptor type 2 fusion protein TT30, which have both been successfully tested in PNH erythrocyte lysis assays (16, 18).

A clearer understanding of the relationships between structure and function in FH, and in FHL-1, would address fundamental questions of complement regulation and has the potential to further improve the AP regulatory potency of engineered inhibitors.

Within FH, CCP 7 and CCP 20 harbor binding sites for host-specific polyanionic surface markers (see Fig. 1A) (9). The site in CCP 20 recognizes specific sialic acid moieties (19) that appear to be critical for establishing the host-surface specificity of FH (12, 16, 20–23). There are two principal binding sites for the C3 activation product C3b within FH, lying in CCPs 1–4 and 19–20 (Fig. 1) (9). Whereas the N-terminal CCPs contain all the complement regulatory functions (10), that is, cofactor and DAA, the C-terminal domains 19–20 increase the avidity for C3b by binding to its thioester-containing domain (TED) (9, 24, 25). Thus, the absence of the 13 C-terminal CCPs in the splice product FHL-1 results in the loss of a key functional site.

Although the TED is also present in the C3b-inactivation products iC3b and C3dg, FH does not bind iC3b or C3dg efficiently in vitro (16, 26, 27). This suggests that the TED-binding site within CCPs 19–20 is at least partly unavailable, or cryptic (16, 22), within a compact conformation of full-length FH (28, 29). Interestingly, engineered small versions of FH (miniFH) consisting of the N-terminal four CCPs linked directly (17) or via an optimized peptide linker (16) to CCPs 19–20 have higher AP regulatory activity than does FH in several assays, including erythrocyte lysis assays. This regulatory enhancement of miniFH over FH was attributed to an increased availability of CCPs 19–20 (16), arising from the loss of the compact conformation of FH when its central modules are deleted.

Why the C-terminal binding site of full-length FH should be effectively unavailable for engagement with iC3b or C3dg is unknown. However, if a “released” C terminus promotes regulatory activity, this feature could be designed into engineered proteins intended to serve as therapeutic complement regulators. Good accessibility of CCPs 19–20 should facilitate binding to the late-stage opsonins iC3b and C3dg, which accumulate on vulnerable host tissues under complement attack (30). To investigate this possibility further, we produced a deletion variant of FH that lacks six central CCPs (FHΔ10–15). The central CCPs 10–15 are structurally compact (28) and may specifically orient the remainder of the molecule so as to mask the C-terminal C3b/TED-binding site (16). Thus, removal of CCPs 10–15 is hypothesized to decrypt this site, thereby yielding a gain in AP regulation. We further hypothesized that a duplicated version of miniFH (termed midiFH) that links two miniFH molecules together and has four potential binding sites for C3b and two for sialic acid should display even higher regulatory activity. To test these hypotheses, we compared our engineered constructs miniFH, midiFH, and FHΔ10–15 to the native proteins FH and FHL-1 in a wide range of assays. Collectively, by correlating complement regulatory functions with the overall inhibitory activity of selected engineered and native FH-based regulators, we have elucidated features that determine high AP regulatory activity.

Erythrocytes from a PNH patient were collected by venipuncture and used immediately in the AP lysis assay. This project was conducted under ethical approval 279/09, granted by the local Ethics Commission of Ulm University.

Plasma-purified FH was purchased from CompTech (Tyler, TX) and used for all experiments except the surface plasmon resonance (SPR)-based C3b-binding assay, which used a plasma-purified FH preparation previously prepared in-house as described (16). Soluble complement receptor type 1 (sCR1) was a gift from Celldex Therapeutics and was used as a control (see Supplemental Fig. 2). Recombinant versions of FHL-1 (11), miniFH (16), and FH18–20 (24, 31) were produced and purified as previously described. To construct the DNA for midiFH, the codon-optimized miniFH-encoding DNA was amplified in two separate PCRs. In the first reaction, the 5′ and 3′ primers contained the PstI and XmaI restriction sites, respectively. The second PCR reaction used primers with XmaI and XbaI restriction sites. Restriction enzyme digestion of the PCR products and the pPICZαB Pichia pastoris expression vector (Invitrogen) was performed according to the manufacturer’s instructions (New England Biolabs). Following agarose gel analysis, DNA was extracted using the kit from Qiagen. The purified coding DNAs were ligated into the vector, which was then used to transform competent Escherichia coli cells for plasmid propagation. After sequence verification, the vector containing the midiFH-coding sequence was used to transform P. pastoris strain KM71H (Invitrogen) according to the manufacturer’s instructions. MidiFH protein was produced according to the expression and purification protocols used for miniFH. In the case of FHΔ10–15, a synthetic, codon-optimized gene was obtained from GeneArt and expressed in P. pastoris as described (32). Cells were grown up in a fermenter, and FHΔ10–15 was purified by consecutive cation- and anion-exchange chromatography steps, including the removal of N-linked glycans with the endoglycosidase EndoH_f (New England Biolabs). The native amino acid sequences of all recombinant constructs (Fig. 1A) are preceded by the nonnative sequence EAEAAG, EAAG, or AG, with the EA sequences being the remnants of the processing of the yeast secretion signal peptide (33), and AG being a cloning artifact.

All experiments were carried out at 25°C using a Reichert SR7500DC SPR instrument (Reichert Technologies, Buffalo, NY). As a running buffer, 10 mM HEPES (pH 7.4) with 150 mM NaCl, 1 mM MgCl₂, and 0.005% (v/v) Tween 20 was used throughout. To analyze the interactions of C3b with the natural and engineered FH-based inhibitors, C3b was deposited using a combination of amine coupling and convertase-driven cleavage of C3 to C3b following published protocols (16, 34). Thus, most C3b molecules were attached to the SPR chip surface via their thioester groups. In brief, 800 response units (RU) C3b (CompTech) were immobilized onto one flow cell of a carboxymethyldextran hydrogel biosensor chip (CMD500m, purchased from XanTec Bioanalytics, Duesseldorf, Germany) using standard amine coupling according to the manufacturer’s instructions. The reference surface was blank immobilized. AP C3 convertases (i.e., C3bBb) were then formed on the chip by injecting a mix of 600 nM factor B and 100 nM factor D in running buffer, followed by injections of C3 at 0.5 μM, which resulted in 1130 RU physiologically immobilized C3b molecules (total C3b loading, 1930 RU). To probe binding, a concentration series of FH analytes was injected (at 10 μl/min for 3.5 min) followed by buffer flow for 300 s and a regeneration step consisting of an injection of 1 M NaCl for 30 s. The lowest concentration of each series was injected twice to assess reproducibility. The SPR responses were corrected for the differing molecular weights of the analytes by multiplying the primary SPR responses by the quotient of the molecular masses of miniFH and the respective test analytes. For the supplemental SPR figure, the correction for molecular mass was achieved by multiplying the primary SPR responses by the quotient of the molecular masses of FH18–20 and the respective test analytes. The following molecular masses were used: miniFH (42.9 kDa), FH (155 kDa), FHΔ10–15 (96.0 kDa), FHL-1 (49.3 kDa), midiFH (86.1 kDa), FH18–20 (21.4 kDa), and sCR1 (212.0 kDa). Where appropriate, affinity constants were extracted by plotting the response at steady-state against the molar concentration and subsequently fitting the affinity with TraceDrawer software using a 1:1 steady-state affinity model. Because steady-state conditions were not reached for the interaction between FHΔ10–15 and C3b, a kinetic analysis was performed, with responses fitted to a 1:2 (FHΔ10–15/C3b) binding model (two independent binding sites).

To test binding to C3d (CompTech), 1900 RU C3d were immobilized on a CMD500m chip via standard amine coupling according to the manufacturer’s instruction. On a reference surface, blank immobilization was carried out. Binding of FH analytes to C3d was probed by duplicate injections of 1 μM solutions. The buffer, flow, contact time, and regeneration conditions were as specified above. Reference-subtracted sensorgrams are shown throughout.

The DAA of the various FH analytes was assessed by SPR as described previously (16). All experiments were carried out at 25°C on a Reichert SR7500DC SPR instrument in running buffer consisting of 10 mM HEPES (pH 7.4) with 150 mM NaCl, 1 mM MgCl₂, and 0.005% (v/v) Tween 20. The flow rate was 10 μl/min. Reference-subtracted sensorgrams (based on blank-immobilized reference surfaces) are shown throughout. To assemble the C3 convertases of the AP, a mix of 600 nM factor B and 100 nM factor D in running buffer was injected for 180 s onto a CMD500m chip that had been immobilized either “physiologically” with C3b (see above) or with amine-coupled C3b (800 RU). After building the convertase, running buffer was applied for 60 s, followed by injection of the analytes (all at 100 nM) for 210 s, then again by buffer flow for 300 s. Finally, the surface was regenerated with an injection of CR1 CCPs 1–3 at 0.8 μM for 30 s and a 30-s injection of 1 M NaCl. To allow visualization of the C3bBb decay process, SPR-binding signals of the FH analytes, in the absence of convertase formation (which corresponds to the response obtained solely from C3b binding), were subtracted as described previously (16, 35). Processed, superimposed sensorgrams were normalized (to compensate for the small drift in signal resulting from the convertase-mediated immobilization procedure) to facilitate an overlay of sensorgrams at the time point corresponding to analyte injection; the normalization was <5% for all sensorgrams.

Cofactor activity was tested as previously described (16). In brief, reaction mixes in PBS containing C3b (0.7 μM), factor I (0.01 μM), and various complement inhibitors (0.1 μM) were incubated on a rotatory shaker (60 rpm) at 37°C for 10, 20, 40, and 80 min. Reactions were stopped by adding 5 μl reducing SDS-PAGE loading buffer. Samples were analyzed for cleavage products of the α′-chain on a 9% SDS-PAGE gel stained with Coomassie brilliant blue (Sigma-Aldrich). As a negative control, a sample lacking cofactors was used. Densitometry was performed on scanned gel images using the ImageJ software. The absolute OD was estimated for each C3α′ 43-kDa band and, in negative control reactions only, also for the uncleaved C3α′ band. Then, the relative optical densities of all the C3α′ 43-kDa bands on a gel were calculated by comparing them to the OD of the C3α′ band of the negative control present on the same gel. This normalization procedure facilitated comparison of C3α′-43 optical densities between different gels.

To assay AP activation, ELISAs were performed as described previously (16). In brief, 50 μl LPS solution (50 μg/ml) from Salmonella typhimurium (Sigma-Aldrich) was coated onto 96-well plates in PBS (pH 7.4) for 2 h at room temperature or overnight at 4°C, followed by washing (with PBST: PBS containing 0.05% [v/v] Tween 20) and blocking (1% [w/v] BSA in PBS). In a 96-well plate, analytes in PBS (30 μl) were mixed with 30 μl 50% serum containing 10 mM MgEGTA to block CP/LP activation. The only exception was FHΔ10–15, which was not sufficiently soluble in PBS to be added into the assay as a concentrated solution (when solubilized in PBS). Therefore, FHΔ10–15 was solubilized in a buffer composed of 20 mM glycine at pH 10.5 and 150 mM NaCl, and was concentrated to 21.5 μM. To achieve the final FHΔ10–15 concentration of 0.5 μM in the assay, a 1:42 dilution (containing 5 mM MgEGTA) of the FHΔ10–15 stock was performed to yield a serum content of 25%. Thus, samples contained 2.3% (v/v) of the 20 mM glycine buffer for the highest concentration point of 0.5 μM FHΔ10–15 in the assay. We tested whether the presence of this concentration of glycine buffer negatively influenced complement activation in this assay and found that it did not. Other FHΔ10–15 concentration points were achieved via a 1:1 dilution series of this 0.5 μM FHΔ10–15 solution with 25% serum (containing 5 mM MgEGTA). The mixtures of analytes in serum (all at a final serum content of 25%) were added to the LPS-coated microtiter plate and incubated for 1.5 h at 37°C prior to washing (2× with PBST) and detection by reaction with HRP conjugated goat anti-human C3 (MP Biomedicals) at 1:1000 dilution in 1% BSA/PBS. Further washing (3× in PBST) was followed by detection with a solution of ABTS (0.5 mg/ml; Roche) and 0.03% (v/v) H₂O₂ in 0.1 M sodium citrate buffer at pH 4.3. Absorbance was read at 405 nm as a measure of C3b and iC3b deposition. PBS and EDTA (final concentration, 5 mM) were used as positive and negative controls, respectively. All responses were normalized to the response achieved when only PBS was added in the absence of a regulator.

The natural and engineered FH proteins were probed for their efficiency in protecting lysis of erythrocytes from a PNH patient undergoing eculizumab therapy. Erythrocytes from PNH patients can be subdivided into three categories based on their number of GPI-anchored proteins, which includes the complement regulators, CD55, and CD59: type III cells completely lack GPI-anchored proteins, type II cells display reduced numbers, and type I cells display normal numbers of these proteins. Unlike type I erythrocytes, type II and type III cells are more susceptible to complement-mediated hemolysis than are healthy cells. After informed consent was obtained, PNH patient blood was collected into EDTA tubes to obtain PNH erythrocytes. The erythrocytes were washed three times in PBS prior to use in the assay. ABO-matched serum was obtained from healthy volunteers (ethical approval 155/12, Ulm University) or from the German Red Cross Blood Donor Service, Baden-Wuerttemberg, and shock frozen and stored at −80°C prior to use. In this assay, the pH of the serum was lowered to 6.6–6.9 to trigger brisk activation of the AP (36). An aliquot of 2 μl washed and packed PNH erythrocytes was incubated with a mixture of 8 μl test proteins in PBS and 30 μl HCl-acidified serum (pH 6.6–6.9) containing MgCl₂ to yield a final concentration of 1.5 mM MgCl₂. The only exception was FHΔ10–15, which was not sufficiently soluble in PBS to be added into the assay as a concentrated solution. Therefore, FHΔ10–15 was added to acidified serum as a concentrated solution solubilized in 150 mM NaCl and 20 mM glycine at pH 10.5 as described above. PBS and MgCl₂ were added to the FHΔ10–15/serum mixture to achieve a serum concentration of 79% (v/v) and a MgCl₂ concentration of 1.58 mM. Finally, 2 μl washed and packed PNH erythrocytes was added to 38 μl FHΔ10–15/serum/PBS mixture, yielding a final serum concentration of 75% in the test tube. The percentage of the 20 mM glycine buffer for the highest concentration point of 0.8 μM FHΔ10–15 in the assay equates to 3.75% (v/v); we ascertained that this did not negatively influence complement activation. The hematocrit was ∼8%. The samples were incubated for 24 h at 37°C, after which the reaction mixtures were centrifuged at 2000 × g for 60 s to harvest the cell pellets and supernatants. Hemolysis of PNH type II and type III erythrocytes was measured by flow cytometry with a PE-coupled anti-CD59 Ab (clone OV9A2; eBioscience) as described previously (16, 37), with the level of hemolysis being normalized to the value for the lysis of PNH type II and type III erythrocytes observed in acidified serum that had been mixed with 8 μl PBS instead of inhibitor. Classification of PNH erythrocytes into PNH type I, II, or III was achieved by employing two gates. Patient-derived erythrocytes that exhibited the same level of fluorescence as a PE isotype control (for the anti-CD59 Ab) were characterized as PNH type III cells (lacking CD59), whereas erythrocytes with same level of fluorescence as healthy control cells were characterized as PNH type I cells (normal copy number of CD59). Any cells that lay between these two gates were classified at PNH type II cells (reduced copy number of CD59). The level of hemolysis was also assessed by determining the release of hemoglobin, by measuring the OD₄₀₅ of the diluted assay supernatants (typically a 1:4 dilution in PBS).

The recombinant variant FHΔ10–15 was designed to test the structural and functional effects of deleting six central CCP modules of FH. MidiFH allowed us to investigate the consequences of having additional C3b and sialic acid–binding sites. Finally, FHL-1 allowed us to characterize a version of FH that retained a single C3b-binding site and a solitary glycosaminoglycan-binding region along with decay-accelerating and cofactor activities, but it lacks both central and C-terminal CCPs. Recombinant proteins (Fig. 1A) were produced in P. pastoris and purified using a combination of cation-exchange, anion-exchange, or size-exclusion chromatography steps (32).

Analytical SDS-PAGE (Fig. 1B) showed that all of the analytes were >95% pure. The FHΔ10–15 preparation had a minor impurity manifesting as an extra band present under both reducing and nonreducing loading conditions. The midiFH preparation had a small proportion of proteolytically cleaved material, revealed only under reducing conditions. As shown previously, miniFH contained no contaminants that could be detected by Coomassie staining on SDS-PAGE gels.

To investigate the structural consequences of deleting the central CCPs, we employed dynamic light scattering to compare FHΔ10–15 and wild-type FH (Supplemental Fig. 1). Despite differing in the number of CCPs, the two proteins had nearly identical hydrodynamic radii (11.8 nm for FHΔ10–15, 11.6 nm for FH). This increased ratio of hydrodynamic radius to number of CCPs suggests a looser or more open arrangement of the 14 CCPs remaining in FHΔ10–15 as compared with the organization of the 20 CCPs in FH. This result reinforces the notion that the central CCPs of FH favor a compact, probably bent-back, arrangement of FH (28, 38).

Binding to C3b is required for the complement-regulating activities of FH. Therefore, differential binding affinities of the analytes for C3b are likely to affect the AP regulatory activities. To investigate this, we deposited C3b onto a hydrogel biosensor surface by applying a combination approach of amine coupling followed by on-chip formation of C3 convertase in the presence of C3. Following decay of the convertase complexes and washing away of Bb, this strategy results in the deposition of C3b in a physiological orientation via its reactive thioester. We first compared the relative binding responses of all five analytes at 62 nM: FH, FHL-1, miniFH, midiFH, and FHΔ10–15 (Fig. 2A). MidiFH produced by far the highest response. Both the association and dissociation phases of the midiFH and FHΔ10–15 responses were slower than for the other proteins tested (Fig. 2B). To compare affinities for C3b, we determined K_D values by fitting the response versus concentration plots to a 1:1 steady-state affinity model (Fig. 2C–I).

It was necessary to exclude FHΔ10–15 from this analysis because the C3b-binding curve of FHΔ10–15 did not reach steady-state under the conditions assayed (Figs. 2B, 3, Supplemental Fig. 2). Therefore, we employed a kinetic analysis of the C3b interaction with FHΔ10–15 by fitting the responses to a 1:2 binding model (black lines), which predicts that one molecule of FHΔ10–15 can bind (via each of its termini) to two independent targets on the C3b surface (i.e., to either the FH19–20 or the FH1–4 binding site on different C3b molecules). The model yielded affinity constants of 1.7 and 5.6 μM. These results are in excellent agreement with the published K_D for FH19–20 of ∼5 μM (9, 16, 39) and for FHL-1 of 1.5 μM (see Table I). As also implied by its large SPR response and slow dissociation rate, the midiFH/C3b complex had the highest affinity, at 27 nM. This affinity was ∼25-fold tighter than that of FH or miniFH, which had K_Ds (consistent with Refs. 9, 16) of 650 and 710 nM, respectively. With an affinity of 1.5 μM, the binding strength of FHL-1 for C3b was ∼2-fold weaker than that of FH or miniFH, but roughly 10-fold stronger than the published K_D for FH CCPs 1–4 (9, 16, 35). In summary, there is a correlation between the number of C3b-binding sites an analyte possesses and its affinity for C3b, but there was a much bigger difference between FH/miniFH (both with two sites) and midiFH (four sites) than between FH/miniFH and FHL-1 (one site).

In vivo, C3b is degraded by factor I and its cofactors (including FH) to iC3b and then to C3dg via cleavage at three proteolytic sites in the CUB domain within the α′-chain of C3b. Finally, C3dg may be cleaved further by plasma proteases to generate the small C3g fragment and the larger C3d fragment. C3d corresponds to the TED of C3b and, similar to iC3b and C3dg, remains bound to the surface after cleavage. C3d and C3dg have previously been shown to interact equally tightly with recombinant FH 19–20 (24). We investigated the C3d binding properties of FH and its engineered derivatives by flowing 1 μM solutions of our FH-derived proteins over a SPR sensor chip that had been coated with C3d using standard amine coupling. As expected, FHL-1, consisting of CCPs 1–7, did not bind C3d sufficiently tightly to be measurable by SPR (Fig. 4). Consistent with the weak binding displayed by full-length FH for C3d in earlier studies (16, 26, 27), at the 1 μM concentration tested in this assay, full-length FH also failed to bind detectably to C3d, despite the presence within FH of the C3d/C3b TED-binding site in CCPs 19–20 (16, 26, 27). In contrast, FHΔ10–15 bound appreciably to C3d on the SPR chip, as did miniFH and midiFH, with the last of these being by far the strongest binder judging by the size of the response.

We also elucidated the relative binding responses of FHΔ10–15 and FHL-1 (as well as of the controls, sCR1, and recombinant triple CCP construct FH18–20) to physiologically deposited C3b, iC3b, and C3dg (Supplemental Fig. 2) (as described in Ref. 16). FH18–20 includes CCPs 19–20 and is known to bind TED almost identically to FH19–20 (24, 31), and it served as a positive control for the recognition of TED in C3b, iC3b, and C3dg. Indeed, FH18–20 bound to all of these with a similar affinity (judging by the response sizes), consistent with its recognition of the TED in all three fragments. sCR1 served as control for evaluating the completeness of processing of C3b, because sCR1 is known to bind only weakly to iC3b (when compared with C3b) and fails to bind C3dg (27, 40). Neither sCR1 nor FHL-1 bound well to iC3b, and negligible responses were subsequently obtained for binding to C3dg. In contrast, FHΔ10–15 bound effectively to iC3b and C3dg, in line with the data (described above) for the binding of FHΔ10–15 to amine-coupled C3d. These data are all consistent with an unavailable or cryptic C3d/TED-binding site in the C terminus of FH that is exposed upon deletion of the central CCPs from the parent protein.

Full-length FH is an efficient cofactor for the factor I–mediated proteolysis of C3b to iC3b that stalls the AP C3b amplification loop. The α′-chain of C3b is cleaved twice, but the β-chain remains unaltered. The first cleavage produces 68- and 46-kDa polypeptides; the latter is then cleaved to yield a 43-kDa product (41). The N-terminal four CCPs of FH (or FHL-1) are known to be necessary and sufficient for this process in vitro. We submitted FH and our FH-derived proteins (all at 100 nM) to a fluid-phase cofactor assay in which the disappearance of the α′-chain and the appearance of the 68-, 46-, and 43-kDa products were observed over a time course using SDS-PAGE (Fig. 5A, Supplemental Fig. 3). Of the proteins tested, FH showed the highest cofactor activity, followed by midiFH. The cofactor activities of FHL-1, miniFH, and FHΔ10–15 were all noticeably lower and, despite small differences, in a similar range.

In addition to its cofactor activity, FH accelerates the irreversible dissociation (decay) of the already unstable bimolecular AP C3 convertase, C3bBb. We used a well-established SPR-based assay to measure in vitro the DAA of FH and the FH-derived proteins (at 100 nM). In this assay, C3b was first immobilized on a CMD500m sensor chip. The immobilization was achieved either by using standard amine coupling, to a final density of 800 RU (Fig. 5B), or via a strategy of amine coupling (800 RU) followed by convertase-driven generation and immobilization of C3b via its thioester to a density of 1930 RU (Fig. 5C). Using FH18–20 as a negative control, we showed that all of our other FH-derived proteins had DAA as expected, because they all possess CCPs 1-4. Of note, while the relative ranking of DAA remained constant, the activity differences observed on the surface with amine-coupled C3b were substantially bigger when compared with the surface with physiologically immobilized C3b molecules. With this effect in mind, we tentatively propose the following ranking of our proteins by their DAA: midiFH and miniFH exhibited the highest activity, followed by FHL-1, FHΔ10–15, and FH. Thus, the ranking of our proteins by their DAA is distinctly different from the ranking by their CA.

We assayed the ability of our proteins to prevent the deposition of C3b on microtiter plates following complement activation via the AP. Proteins were added, at a range of concentrations, into human serum, which was then exposed to LPS-coated microtiter plates to initiate AP activation. In this assay, FHL-1 exhibited an inhibitory activity similar to that of FH (Fig. 6A). The observation that FHL-1 and FH have similar regulatory activities in our serum assay is surprising because a previous study reported that FH outperforms FHL-1 ∼100-fold in DAA (42). The different findings likely arise from differences in the assays. The previous report solely investigated the DAA functionality on preformed convertase-loaded sheep erythrocytes, whereas in the present study, the overall AP regulatory activity (including DAA and CA) in serum was probed on LPS-coated microtiter plates (Fig. 6A) and on PNH erythrocytes (see below and Fig. 6B, 6C). Both native proteins, FH and FHL-1, were less active as regulators than was miniFH, as expected from previous work (16). Interestingly, FHΔ10–15 and midiFH were each substantially more active AP regulators than miniFH. Thus the rank order is: FHΔ10–15 and midiFH (highest), followed by miniFH, and then FHL-1 and FH.

We also probed the capacity of our proteins to prevent AP-mediated lysis of human erythrocytes derived from a PNH patient (Fig. 6B, 6C). In vivo, AP-mediated lysis of PNH erythrocytes is usually preceded by triggers such as pregnancy, surgery, or an infection. In our in vitro assay, we used the established method of acidification of human serum (to pH 6.6–6.9) to spontaneously trigger AP-mediated hemolysis (16, 18, 36, 43). We employed two different readouts to evaluate the protection of PNH erythrocytes from lysis. On a cellular level, we employed FACS to monitor the relative proportions of PNH types I, II, and III erythrocytes prior to and after incubation in serum, with or without added FH-derived proteins (Fig. 4B). As a more direct readout of hemolysis, we measured the OD₄₀₅ of the supernatants to estimate the amount of hemoglobin released (Fig 6C). Both readouts provided highly consistent results (Table II) that were also in very good agreement with a previous report on the activities of FH and miniFH in this assay (16). FH and FHL-1 had comparable regulatory capacity, but all of the engineered inhibitors were more potent. MidiFH did not outperform miniFH as it did in the microtiter plate C3b-deposition assay, but instead showed almost identical activity, whereas FHΔ10–15 was less active than miniFH or midiFH, but still substantially more active than FH. Thus the rank order was: miniFH and midiFH (highest), followed by FHΔ10–15 and then FH and FHL-1.

We have compared the complement-regulating activities of full-length native FH to a set of smaller versions of the protein designed to incorporate various configurations of C3b-binding and self-surface recognition sites. Our study provides important novel data on the physiologically important splice variant FHL-1, probes the role of the 13 CCPs that are missing from FHL-1, and suggests routes for designing more efficacious complement-suppressing proteins with therapeutic potential.

The two novel recombinant proteins, midiFH and FHΔ10–15, formed complexes with C3b immobilized on a SPR chip surface that had slower off-rates than did those formed by miniFH, full-length FH, or FHL-1 (Fig. 2A, 2B). Similar to miniFH, but unlike FH (or FHL-1), both midiFH and FHΔ10–15 bound to immobilized C3d (equivalent to the TED of C3b) on the SPR chip. The new proteins were also better at preventing C3b deposition on LPS-coated, complement-activating microtiter plates that had been exposed to human serum (Fig. 6A). Alternatively, midiFH and FHΔ10–15 were not more active than miniFH in protecting human PNH erythrocytes (Fig. 6B, 6C) from complement-mediated lysis. Notably, all three engineered proteins (FHΔ10–15, midiFH, and miniFH) performed better in this assay than did the native proteins FH and FHL-1.

As expected, in our SPR-based assays, full-length FH bound to C3b with a K_D of 650 nM but did not bind with a measurable K_D to C3d. In line with previous work, truncated fragments bound C3b less well than did full-length FH. For FHL-1, which is effectively FH1–7, the SPR-derived K_D for C3b was 1.5 μM. This binding affinity is tighter than the K_D of ∼10–14 μM reported by several authors for FH1–4 (9, 16, 35), pointing to a possible contribution of one or more of the CCP domains 5–7 in FHL-1 to C3b binding. The crystal structure of FH1–4/C3b indicated that CCP 5 could make direct contact with C3b. This possibility is in line with a recent study (44) in which FH CCPs 1–5 were observed to have slightly higher regulatory cofactor activity than did CCPs 1–4. The higher affinity for C3b of full-length FH as compared with FHL-1 can be ascribed to a contribution from the C-terminal C3b/C3d-binding site. Although this site is envisaged to be partially occluded or inaccessible initially (hence, no measurable C3d binding), it seems that once FH engages C3b through its N-terminal site, the C-terminal site is, in effect, released. Thus, the overall affinity of FH for immobilized C3b on an SPR chip surface is enhanced (when compared with FHL-1) as a result of bivalency, but we suggest that this effect is rather small (a 2-fold gain in K_D) because of the conformational adjustment required for bivalent binding. In summary, this benchmark portion of our study extends the N-terminal C3b-binding site to include at least CCP 5, but otherwise is in good agreement with previous studies that have suggested that the C-terminal C3b-binding site of native circulating FH is not initially fully accessible for engagement with C3b.

In our SPR-based study, we confirmed a previous observation that miniFH and FH have similar K_D values for C3b binding, even though miniFH lacks CCP 5 (as well as CCPs 6–17). Given that FH1–4 had only a 10–14 μM affinity, this result suggests that the C-terminal C3b-binding site of miniFH makes a stronger contribution to the overall K_D (making it 20-fold stronger) than it does in full-length FH (because the K_D of the full-length protein was only 2-fold stronger than that of FHL-1). This finding is consistent with the observation (in the present study and in the previous study) that miniFH binds C3d whereas FH does not, implying that the C-terminal binding site of miniFH is fully accessible. Thus, we conclude that the overall affinity of miniFH for immobilized C3b is enhanced when compared with FH1–4 as a result of bivalency, and this effect is significant, because in the case of miniFH no major conformational adjustment is required for simultaneous binding to C3b by both sites.

MidiFH is effectively two copies of miniFH spliced together, with four C3b-binding sites and the theoretical ability to bind simultaneously to adjacent C3b molecules on a surface, or to a C3d and a nearby C3b molecule. Hence, an avidity effect could explain its very high affinity for C3b in the SPR assay. Its relatively high occupancy and long residency on microtiter plates displaying C3b (and C3b breakdown products) could therefore explain its enhanced ability to prevent further C3b deposition in this assay.

The central CCP modules 10–15 form a compact structural region that could favor bent-back conformers of the FH molecule (28, 29, 38). We therefore reasoned that deletion of these CCPs to create FHΔ10–15 would result in a less compact, more open structure. This change was confirmed by dynamic light scattering measurements. Deletion of these modules also resulted in a gaining in affinity (measured by SPR) for immobilized C3d. This is consistent with our model, in which opening up of the FH molecule disfavors conformers whose C-terminal C3b/C3d-binding site is inaccessible to ligand. Alternatively, the artificially extended structure of FHΔ10–15 might also disfavor the binding of both N-terminal and C-terminal C3b-binding sites to the same C3b molecule. This model is supported by the outcome of the kinetic analysis. The fit to a 1:2 interaction model returned affinities of 1.7 and 5.6 μM for the two independent C3b binding sites within FHΔ10–15 (Fig. 3), which closely match the determined values for the individual proteins FH19–20 and FHL-1 assayed in isolation. Thus, the longer contact times of FHΔ10–15 for C3b are based on the avidity for C3b-deposited surfaces that was introduced by connecting to two C3b-binding sites (CCPs 1–7 and 19–20) in a way that disfavors simultaneous C3b binding.

The binding of FH to C3b is a prerequisite for its key regulatory functions: competing with factor B for binding to C3b, accelerating the decay of C3bBb, and acting as a cofactor for the factor I–mediated degradation of C3b. FH and miniFH bound C3b with similar affinities, yet FH had a significantly higher fluid-phase cofactor activity (Fig. 5A and Ref. 16). Indeed, FH had a higher cofactor activity than that of any of the shorter proteins, including midiFH, which bound 25-fold better to a C3b-coated surface than did FH.

These findings imply that fluid-phase cofactor activity is not directly correlated with C3b-binding affinity and therefore suggests that CCPs absent from the shorter proteins contribute to the recruitment of factor I prior to its cleavage of the α′-chain of C3b.

That miniFH, midiFH, FHL-1, and FHΔ10–15 each have a slightly greater DAA than FH when measured by SPR (Fig. 5B, 5C) could arise from their open conformations when compared with the inherently closed conformation of FH. The engineered constructs miniFH and midiFH exhibited the highest DAA, probably benefitting from CCPs 1–4 being readily available for binding and causing C3bBb decay, whereas FH CCPs 19–20, connected by a poly-Gly linker of optimized length, provide an additional binding anchor for C3bBb. In FH, this second site is cryptic; in FHL-1, it is absent; and in FHΔ10–15, it is not positioned appropriately and (on the SPR chip surface) may bind to an adjacent C3b molecule. Of note, although the relative ranking remained identical, the differences in DAA levels were substantially bigger on the sensor chip surface with randomly amine-immobilized C3b molecules (Fig. 5B) than on the surface with the more physiological, convertase-driven C3b immobilization (Fig. 5C). This difference indicates that regulatory activities may be underestimated when activities on surfaces are probed with randomly immobilized and oriented C3b molecules.

Of relevance to the development of therapeutic proteins, both new variants in this study, midiFH and FHΔ10–15, showed enhanced regulatory activity (compared with FH) with respect to preventing C3b deposition on microtiter plates, effectively a foreign surface. Extrapolating from SPR experiments, this might be connected with a higher affinity of midiFH for C3b and a slower off-rate of the FHΔ10–15/C3b complex. Alternatively, relative to miniFH, these proteins failed to enhance regulatory activity on the self-surface represented by human PNH cells (compare the activities of miniFH to midiFH and FHΔ10–15 in Fig. 6A, 6B). This is despite the increased avidity and stability of their complexes with C3b on an SPR chip surface, suggesting that the mechanism of action of these proteins is subject to different limitations when they are acting on host/self versus foreign (or artificial) surfaces. On PNH cells, all three engineered FH-based molecules were more effective than the natural regulators FH and FHL-1 (Fig. 6B, 6C). Remarkably, the lack of the 13 C-terminal FH domains in FHL-1, relative to FH, did not affect its ability to protect PNH erythrocytes.

Despite the similar levels of CA and DAA when compared with FHL-1, FHΔ10–15 was five times more active than FH in protecting PNH cells from AP attack, perhaps because FHΔ10–15 can use its liberated C-terminal TED-binding site to bind C3d/g and thereby target sites of ongoing complement activation and regulation, a mode of action also ascribed to miniFH (16). Apart from targeting to such sites via CCPs 19–20, the presence of CCP 20 endows the advantage of specifically targeting sialic acid groups on host cells (12, 16, 19–21, 45).

Both miniFH and midiFH exhibited similar levels of PNH cell protection, demonstrating that the higher affinities of midiFH for C3b/C3dg do not translate into a further enhanced regulation on host cells. As a consequence, we suggest that the therapeutic value of inhibitors, that act in a “catalyst”-like fashion (without getting consumed) to perform DAA and CA, may benefit more from a higher rate of turnover than from higher affinities to C3-opsonin–deposited surfaces, where a tight-binding inhibitor could become trapped.

Finally, miniFH and FHΔ10–15 both feature a nonnatural, engineered connection of the N- and C-terminal portions of FH. Despite the presence of additional modules in FHΔ10–15 that contribute to C3b binding (based on our observation that FHL-1 binds better than FH1-4 to C3b), miniFH was significantly better at protecting PNH cells and was more active in our DAA. This result demonstrates that, apart from keeping the FH C terminus available for TED/iC3b/C3dg targeting, miniFH also benefits from its optimized peptide linker, not present in FHΔ10–15, which allows miniFH to occupy both FH-binding sites on C3b simultaneously in a configuration that seems to drive displacement of Bb. Future studies may also mechanistically address how the differential decay acceleration activities toward C3bBb seen in the FH-based analytes affect properdin-stabilized or properdin-initiated C3 convertases.

Our study has general implications for understanding the physiological regulation of the AP by the FH family of proteins. The molar ratio of FH to FHL-1 is ∼2:1 in plasma, and along with the previously reported substantially poorer DAA regulatory properties of FHL-1 (42), this ratio was interpreted to mean that FH is the major AP regulator in both the fluid phase and on host cells. Our findings, however, show that FH and FHL-1 are equally active in protecting PNH erythrocytes and thus indicate that FHL-1 plays a more important role in AP regulation than acknowledged to date. Future studies may investigate whether the equal AP regulatory activity between FH and FHL-1 observed in the present study is a more general feature that also holds true for other specialized host tissues (e.g., in the eye or kidney). This new finding is of particular relevance in view of the recent discovery of AP deregulation caused by FH-related proteins. These proteins bear no regulatory CCPs but include an almost identical copy of CCPs 19–20. They compete with FH for surface binding, an observation with potential implications for various diseases (46, 47). Presumably, FHL-1, because it lacks the C-terminal CCPs, would not be outcompeted by FH-related proteins.

Clark et al. (11) have shown that FHL-1 acts as the predominant AP regulator in Bruch’s membrane, a structure in the eye that has been implicated in AMD. Further to outnumbering FH on Bruch’s membrane, our finding of almost equal regulatory activity for FHL-1 and FH reinforces the concept of FHL-1 as the major AP-implicated regulator in AMD.

In conclusion, our study provides valuable new insights into the collaboration between the two key functional sites of FH in surface protection. Increased availability of the FH C terminus appears to generally promote the complement regulatory functions ascribed to the N-terminal FH domains. This activity can be further potentiated by optimal spacing of the two C3b-binding regions. Future efforts in generating therapeutic molecules to control complement dysregulation may benefit from the structure–function relationships presented in the present study. Any active regulatory protein with DAA or CA toward C3b is predicted to benefit from an optimally spaced linkage to FH CCPs 19–20. Our comparison also indicates that about one third of the systemic complement regulation of the AP may be performed by FHL-1, a finding that has implications for the pathophysiology of diseases associated with genetic variations within the FH protein family.

We thank Dr. Deborah McClellan for excellent editorial assistance.

C.Q.S., D.R., and J.D.L. are inventors of a patent application that describes the use of miniFH for therapeutic applications. P.N.B. and C.Q.S. are inventors of a patent that describes the recombinant production of FH. J.D.L. is the founder of Amyndas Pharmaceuticals, which develops complement therapeutics. B.H., C.Q.S., M.H.-L., and H.S. received honoraria for speaking at symposia organized by Alexion Pharmaceuticals. H.S. and B.H. served on an advisory committee for and received research funding from Alexion Pharmaceuticals. The other authors have no financial conflicts of interest.