1. Introduction
The complement system forms an essential part of the innate immune response. Consisting of more than 30 serum and cell surface proteins, some of which are highly abundant, it functions as a first line of defense against infectious pathogens [
1]. Once activated by one of three activation pathways, it stimulates phagocytosis by macrophages, releases pro-inflammatory signaling molecules and elicits direct lysis of pathogens and cells [
2]. In addition, complement plays a role in the clearance of apoptotic cells by marking them for removal [
3]. Complement activation requires tight regulation by many complement inhibitors for example, CD55 and CD59, which prevent damage to healthy tissue [
4].
Mechanistic as well as genetic evidence has linked dysregulation of complement activation to several diseases. These include rare diseases such as Paroxysmal Nocturnal Hemoglobinuria (PNH), Atypical Hemolytic Uremic syndrome (aHUS) and Complement 3 Glomerulopathy (C3G), as well as common, chronic neurodegenerative pathologies like Alzheimer’s Disease [
5,
6]. Thus, there is considerable interest and opportunity in the complement system as a drug target [
7,
8].
Despite this, only a small number of complement drugs have been approved since the 2007 approval of eculizumab (Soliris
®, Alexion Pharmaceuticals, Boston, MA, USA), indicative of challenges in targeting the complement system. The major challenge is the high abundance and fast turnover rates of some complement proteins necessitating high doses and frequent dosage regimens. For biological modalities, this is particularly challenging due to the high cost of goods. The success of the humanized anti-C5 monoclonal antibody eculizumab in treating PNH and aHUS provides validation of C5 as a drug target. However, it also highlights the challenges inherent in targeting complement proteins because these are highly abundant and have rapid turnover rates [
9,
10]. The maintenance dose for treatment of aHUS is a bi-weekly intravenous infusion of 1200 mg of Eculizumab. These high doses are one of the reasons for the high cost of eculizumab at more than £300,000 per annum in the UK [
11]. 2013). More recently, (ravulizumab/Ultomiris
®, Alexion Pharmaceuticals), an engineered variant of eculizumab that somewhat addresses the dosing challenges, has been approved.
One strategy to achieve lower dosing, which could improve patient access, is to develop therapeutics that specifically target neoepitopes exclusive to the less abundant activated forms of complement proteins. This could also address the challenges of target-mediated drug disposition reported for Eculizumab as well as “C3 bypass” cleavage of C5 which causes continued C5 cleavage in the presence of adequate anti-C5 inhibition [
12,
13]. Examples of complement neoepitope targeting therapeutics include the H17 antibody developed by Elusys, which specifically binds to activated C3 complement [
14]. The H17 antibody binds to an epitope on the CUB domain of C3 which is accessible in C3b, but occluded in C3 [
15]. Similarly, the S77 antibody developed by Genentech, also recognizes a C3b neoepitope located on the MG7 domain of C3 [
16]. The IFX-1 antibody recognizes a neoepitope on one cleavage product of Complement C5, the anaphylatoxin C5a [
17], which is further downstream in the complement cascade. This antibody has been shown to be efficacious as an anti-inflammatory in the treatment of African green monkeys infected with influenza A [
18].
The membrane attack complex (MAC) is generated downstream of the complement protein C5 in the terminal cascade. Direct tissue damage by MAC is implicated in numerous pathologies with the mechanistically best-defined examples being PNH and aHUS. There is also increasing evidence to suggest that sub-lytic (i.e., non-pore forming) levels of MAC on the cell surface have numerous deleterious consequences on cell function [
19,
20]. Eculizumab inhibits cleavage of C5 and thus prevents MAC formation and C5a generation. C5a is a pro-inflammatory effector and it could be beneficial to selectively target formation of MAC, but without direct targeting of C5a. We therefore aimed to develop a MAC-specific inhibitor and set out to generate an anti-neoepitope mAb targeting C5b6, an intermediate complex in MAC formation (
Figure 1A). Computational modelling supported the hypothesis that an antibody with an affinity of 1–2 pM would enable a favorable dosing regimen of monthly subcutaneous injections.
Given that the assembly of the terminal complement complex (TCC) is governed by high-affinity protein–protein interactions and involves large molecular surfaces, the system is likely to be more amenable to inhibition using biologicals rather than small molecules. Following cleavage of C5, the unstable C5b product binds C6 in a practically irreversible interaction, resulting in the formation of the stable C5b6 complex that forms the nucleus for TCC formation [
21,
22].
Figure 1B illustrates the comprehensive structural rearrangements undergone by C5b and C6 upon complex formation indicating the formation of neoepitopes.
The aim of this study was to generate a fully human IgG1, Fc-disabled monoclonal antibody that targets a neoepitope on the C5b6 complex of the TCC for potential use in disease indications in which terminal pathway effector function is known to drive pathology. Here, we describe the approach used to identify MAC-specific inhibitors targeting C5b6 neoepitopes.
2. Materials and Methods
2.1. Complement System Dose Modelling
All modelling and simulations were performed using MatLab SimBiology v5.2. C5b6 (The MathWorks, Inc., Natick, MA, USA) complex formation and proposed antibody binding were modelled using kinetic parameters from [
23]. In addition, the effect of the faster C7 on-rate reported by Thai and Ogata, 2005 on target engagement was also modelled. Details are given in
Supplementary methods (File S1) [
24].
2.2. Antibody Selections
Antibody clones were selected from Adimab LLC platform libraries or newly generated libraries with re-diversified CDRs, according to the protocols developed by Adimab LLC (Adimab, Lebanon, NH, USA). All antigens used were purchased from Complement Technology and were biotinylated prior to use via amine coupling. Magnetic bead selections were performed using streptavidin beads from Miltenyi (MACS
®Miltenyi, Bergisch Gladbach, Germany) and FACS selections were performed on a BD ARIA II. Yeast populations were sorted based on binding to biotinylated C5b6 antigen, IgG expression, or a lack of binding to C5, C6 or a polyspecificity reagent (PSR) [
25].
The initial selections consisted of two rounds of magnetic bead selections using 10 nM C5b6 and one round of FACS using 10 nM C5b6. The heavy chains of this round were then transformed into yeast containing a diversified light chain library from Adimab LLC. These libraries were then used to perform one round of MACS using 10 nM C5b6 and five rounds of FACS. The FACS rounds were as follows: 1: 10 nM C5b6, 2: 10 nM C5, 10 nM C6 and PSR, 3: 10 nM C5b6, 4: C7 competition, 5: 1 nM C5b6. The C7 competition was applied by binding 10 nM biotinylated C5b6 to the yeast to equilibrium, the yeast was then labelled with SA-633 (Streptavidin Alexa Fluor 633, Invitrogen, Waltham, MA, USA). To block any unbound biotin binding sites on the SA-633, biotin was used as a blocking agent post labelling. The yeast was then incubated with 100 nM biotinylated C7 for 20 min and labelled with EAPE (ExtrAvidin
®−R-Phycoerythrin, Sigma Aldrich, St. Louis, MO, USA). Subsequently selecting clones that show high C5b6 binding and a lack of binding to C7. Round 4 was duplicated as above, except for using unlabeled C5b6. The plots in
Figure 2B show the dual labelling of C5b6 and C7. The final round used the outputs from both arms of round 4. The final outputs were plated out on agar plates, and 95 colonies from each library output were picked and sequence verified. Unique sequences were expressed and purified as below.
For affinity maturation, new libraries re-diversified within the CDRH1 and CDRH2 regions were generated from eight selected clones. One round of MACS was performed followed by six rounds of FACS, using the following antigen concentrations: 1: 10 nM C5b6, 2: 10 nM C5, 10 nM C6 and PSR, 3: 10 nM C5b6, 4: C7 competition, 5: 1 nM C5b6, 6: 10 nM C5b6 with competition with 400 nM parental IgG. The C7 competition was performed by pre-incubating 10 nM unlabeled C5b6 with the cells, followed by incubation with 100 nM biotinylated C7 for 20 min, labelled with EAPE and subsequent FACS selection. The parental IgG competition was performed by the pre-incubation of 400 nM parental IgG with 10 nM C5b6 for 15 min prior to incubation with the cells. The gates for the final round were positioned to collect ~0.1% of the population that also, if possible, had better binding affinities than their corresponding parent. The final outputs were plated out on agar plates and 95 colonies from each library output were picked.
2.3. Antibody Expression and Purification
Antibody clones were expressed as fully human IgG1s from a proprietary yeast strain (Adimab) and purified using protein A affinity chromatography followed by buffer exchange into PBS. Fab fragments were generated by papain digestion and subsequent passage through a protein A affinity column to remove intact IgGs and Fc fragments. Introduction of Fc disabling mutation, such as of LAGA (L235A/G237A)/LALA (L234A/L235A), would have taken place later in the molecule discovery process [
26].
2.4. Cloning, Expression and Purification of an Anti-C5 and Anti-MAC Tool Antibodies
The variable region sequences of an anti-C5 antibody were taken from European Patent EP 2359834, cloned into a pEF vector, expressed in CHO cells. This antibody has a reported affinity for C5 of <50 pM and prevents cleavage of C5 into C5a and C5b by the C5 convertase [
27,
28]. The anti-C7 mouse mAb was derived in-house from immunization of a transgenic mouse and generation of a hybridoma line with standard techniques. Both antibodies were purified by protein A affinity and size exclusion chromatography.
2.5. Biolayer Interferometry
A BLI assay was run on an Octet RED384 instrument (ForteBio, Fremont, QC, Canada). Antibodies were captured on protein A sensors (ForteBio). Following a buffer wash in PBSF, the sensors were dipped into an analyte solution for 180 s and the binding response at the end of the contact time recorded. For the screening of the naïve selection outputs, 50 nM of C5b6 (Comptech), C5 (Comptech) or C6 (Comptech) were used as analytes. For the screening of the affinity matured clones, 600 nM of C6 was used.
2.6. Terminal Complement Assay
The TCC assay was performed on a BioRobot FX. Normal human serum was diluted in 1× CFD buffer to 4% (
v/v) and 25 C was added into polypropylene 96-well plates containing the anti-C5b6 antibodies and controls and incubated for 30 min on ice. The concentrations of test antibodies and controls are detailed in the
Supplementary data. Complement was activated with the addition of Zymosan (11 mg/mL), 5 μL per well, and incubated for 30 min at 37 °C. After incubation, complement activation was stopped using 11 μL of chilled EDTA (0.5 M) followed by 100 μL of chilled D:PBS to all the wells. The plates were centrifuged at 1000×
g for 10 min at 4 °C. The supernatants were evaluated using the BD human C5b-9 ELISA Assay. The capture antibody provided within the kit was diluted 1:1000 with 0.1 M Carbonate-Bicarbonate buffer and added into ELISA Max high bind plates at 100 μL per well and incubated overnight at 4 °C. The plates were washed using a Biotek plate washer with PBS containing 0.05% Tween-20 and blocked for 1 h at room temperature with 10% heat inactivated fetal bovine serum in D:PBS at 100 μL per well. Working detection antibodies 1000× detector antibody and 1000× Streptavidin-HRP were combined in 10% heat inactivated fetal bovine serum in D:PBS. After blocking, the plates were washed with PBS with 0.05% Tween-20 and 100 μL of the prepared working detection antibodies was added and incubated for 1 h at room temperature. After incubation, the plates were washed with PBS with 0.05% Tween-20 and 100 μL of substrate solution TMB was added into each well and incubated for up to 30 min at room temperature in the dark. The reaction was stopped with 50 μL of 0.25 M HCl. The absorbance was read at 450 nm and 570 nm using a Biotek Epoch plate reader. The data were normalized against 0% inhibition.
2.7. Surface Plasmon Resonance
All Biacore experiments were performed using HBS-EP+ (Teknova) as a running buffer. An anti-C5 antibody was diluted to 50 μg/mL in 50 mM NaAc pH4.0 and amine coupled using a Biacore amine coupling kit to a Biacore CM5 chip on a Biacore 8K instrument (all GE Healthcare) according to the manufacturer’s instructions on all flow channels. For the MAC assembly assay, C5 (Quidel) or C5b6 (Comptech) were captured on the sample channels at 4 nM and nothing was captured on the reference channels. C7, C8 and C9 (all Comptech) were then flowed over all surfaces at concentrations of 4, 8, and 50 nM, respectively. For the kinetic run to determine the C7 on-rate, C5b6 was captured at 4 nM and C7 injected at 0, 1, 2 and 4 nM. For the kinetic analysis of the eight anti-C5b6 clones, C5b6 was captured at 4.5 nM and antibodies injected at the concentrations indicated in the figures. All interactions were fitted to a 1:1 model with local Rmax fitting. For the screening of the affinity matured clones, C5b6 or C5 were captured at 5 nM on a CM5 chip with immobilized anti-C5 antibody (immobilization as described above) on a Biacore4000 instrument (GE Healthcare) and antibody injected at 500 nM for 240 s. The binding response at the end of the contact time was used to plot the relative binding of each clone.
2.8. C9 Oligomerization HTRF Assay
A total of 0.5 mg of Complement C9 (Comptech) at 1.05 mg/mL was dialyzed into 100 mM sodium bicarbonate. A total of 50 ug of amine reactive fluorescein isothiocyanate (FITC) (Pierce) were dissolved in DMSO to 10 mg/mL. Then, 250 μL of the C9 solution reacted with 2.5 μL of the FITC solution, thus at a molar ratio of dye/protein of 2:1, for 2 h at room temperature. Free FITC was separated from the C9 protein using a PD10 column (GE Healthcare) pre-equilibrated in PBS. Fractions were collected and analyzed by UV/VIS-spectroscopy.
A total of 0.5 mg of Complement C9 (Comptech) at 1.05 mg/mL were mixed with 10 μL of 10 mM TCEP. Then, 100 ug of terbium maleimide (Life Technologies) were dissolved in 20 μL of PBS to 5 mg/mL. Then, 250 μL of the C9 solution were reacted with 2.2 μL of the terbium solution, thus at a molar dye/protein ratio of 3:1, for 2 h at room temperature. Free fluorophore was separated from the C9 protein using a PD10 column (GE Healthcare) pre-equilibrated in PBS. Fractions were collected and analyzed by UV/VIS-spectroscopy. The dye/protein ratios were 1.0 for the FITC-C9 and 0.7 for the Tb-C9.
In the HTRF assay, 2.5 μL of antibody were incubated with 2.5 μL of either 2 or 20 nM C5b6 for 60 min. Subsequently, 5 μL of a mixture of 20 nM C7, 20 nM C8, 50 nM FITC-C9 and 25 nM Tb-C9 were added and the HTRF signal read immediately using the following filter settings: Mirror = TRF D400/D505 dual/Bias, Emission Filter (1) = Invitrogen 520/25, Emission Filter (2) = Invitrogen 495/10. In the one-shot assay, the antibodies were used at a range of concentrations, with the majority being between 100–250 ug/mL FAC. For the dose response curves, antibody was diluted to the concentrations indicated and only the higher C5b6 concentration was used. To normalize the data, a control reaction with no C5b6 was used as low control and a reaction with no antibody was used as a high control. Low and high control were used to normalize the HTRF signal to 100% inhibition and 0% inhibition, respectively.
2.9. Liposome Leakage Assay
A liposome leakage assay was performed using a protocol adopted from [
29,
30]. A lipid mixture of 60/30/10% DOPC, DOPE and cholesterol (all Sigma Aldrich) was dissolved in methanol, dried under nitrogen and then resuspended in PBS supplemented with 50 mM sulforhodamine B (Sigma Aldrich) at a total concentration of 20 mg/mL. The suspension was then freeze–thawed three times and liposome size was reduced by 11 passages through an 800 nm pore size polycarbonate membrane (Avestin). Free sulforhodamine B was removed using a PD10 desalting column followed by gel filtration using a 16/60 S200 Superdex column (both GE healthcare). Both columns were pre-equilibrated in PBS. Antibody clones were incubated with 4 nM C5b6, 12 nM C8, 140 nM C9 (all Comptech) and a 1/25 dilution of liposomes diluted in PBS for 1 h in a black 384-well plate (Greiner) with 30 μL per well. Then, 10 μL of 20 nM C7 (Comptech) were added to start MAC formation and fluorescence intensity measured at 540 nm excitation and 590 nm emission. Fluorescence intensity was measured kinetically, in well-mode, every 0.5 s for 20 s and a linear slope fitted to the curve between seconds 8–13. An anti-MAC antibody was used as a positive control for liposome leakage inhibition. For the assay development, the anti-MAC (anti-C7) antibody was at 250 nM and the isotype control was at 500 nM final assay concentration, or as indicated in the figure. For the screening of the affinity matured antibodies, all antibodies were at ~1 μM.
4. Discussion
The data presented here illustrate the challenges inherent in antibody discovery targeting neoepitopes and possible strategies to overcome these. Although we were able to select specific neo-epitope binders, we were ultimately unsuccessful in isolating an antibody with the required potency, and several lessons have been learned which may guide future antibody discovery campaigns targeting neoepitopes on challenging antigens. Previous examples of antibody discovery approaches for complement neoepitopes include phage display (S77) or mouse immunization (H17) [
16,
33]. More recently, a study by Zelek and Morgan (2020), provided validation of the lower dosing implications of targeting an epitope of an active intermediate terminal pathway complex (C5b7) with identification of an antibody using an in vivo approach [
34].
The in vitro selections approach we used enabled us to tailor the selection strategy towards desired epitopes and specificities. By including negative rounds of selection on the FACS using both C5 and C6, we successfully identified antibodies selective for C5b6 neoepitopes with only low level C5 or C6 binding. The C7 competition in another selection round enabled us to select antibodies binding to C7-competitive epitopes. This was confirmed by functional data from the HTRF assay which shows that a subset of clones was C7-competitive. In addition, the SPR data of the affinity matured antibodies shows increased, specific binding to C5b6 over the parental clones, suggesting that the selection method deployed selects for neoepitopes on C5b6.
One of our criteria was that the selected antibodies needed to be able to outcompete the high affinity physiological binding partner C7, which is found at high concentrations in blood. Initially, a kinetic haemolysis assay using carefully titrated purified terminal pathway components (C5b6, C7, C8, C9) was employed as a primary functional screen. However, as this did not provide robust functional data on the clones, we considered the possibility that inhibition of antibodies from the naïve selection rounds may not be measurable in a haemolysis assay due to their low affinity, despite skewing of the assay condition in favour of C5b6 inhibition. Indeed, no functional inhibitors were identified in haemolysis assays using whole serum as a source of complement. Possible reasons for this are that (1) clones from naïve selection may be too low affinity to outcompete binding of C7 to C5b6 which binds with an on rate of 2.3 × 106 Ms−1, and (2) the sensitivity of this assay type is insufficient. Therefore, we designed two additional functional assays to measure the potency of our antibodies.
A solution-based assay and a membrane-based assay were developed to measure the engagement of our antibodies to C5b6 in solution and to the slightly lipophilic C5b6 when bound to a phospholipid membrane, respectively. In the haemolysis assay using serum, the C5b6 is generated in situ, requiring the antibodies to compete with C7 in real time. The assays described here used purified terminal complement proteins, allowing for pre-incubation of C5b6 with the antibodies to increase the sensitivity of detection of antibodies that outcompete C7. We showed that both solution and membrane-based assays were able to measure the in vitro potency of antibodies targeting MAC, but the antibodies generated here only showed functional activity in the solution-based assay. This may be due to different assay sensitivities or the fact that membrane-bound C5b6 presents slightly different epitopes compared to C5b6 in solution, but it confirms that the epitopes targeted were C7 competitive.
Another challenge was the propensity of C5b6 to aggregate and bind surfaces non-specifically. Identifying and selecting the highest affinity clones by FACS was made highly challenging by this behaviour. Consequently, stored and potentially non-monomeric C5b6 may display different epitopes compared to C5b6 generated de novo in serum-based assays.
We generated a bespoke SPR assay to determine the binding kinetics of our selected clones. C5b6 was captured using an anti-C5 tool antibody and candidate antibodies were used as analytes. Fab fragments were used instead of full length IgGs and the sensorgrams produced agreed with a standard 1:1 binding model and could be used to determine absolute binding kinetics. Using a tool antibody to capture a challenging antigen on an SPR surface is a promising method when working with antigens that are aggregation prone and cannot easily be regenerated, as long as the tool antibody does not occlude the desired epitopes. The measured dissociation constants ranged between 26–1500 nM, several orders of magnitude below the desired target affinity of 1–2 pM. Similarly, the measured on-rates were lower than the desired on-rate of 1 × 107 Ms−1 and did not show the desired potency. Some affinity matured clones showed stronger binding by BLI, but this did not translate to improved activity in functional assays, most likely due to the higher sensitivity and precisions of the BLI assay.
Overall, the data demonstrate the utility of a yeast-based antibody platform in tailoring selection strategies to required specificities and selectivity. We also highlight the value of bespoke assay formats and reagents to characterize challenging drug targets in light of the inherent difficulties in an antibody discovery campaign aiming to deliver high-affinity neoepitope-specific molecules. An antibody was sought that was neoepitope-specific, binds a C7 competitive epitope and shows high-affinity for its target. This work highlights the challenges of discovering an antibody that possesses all these characteristics and provides valuable lessons for antibody discovery against targets in the complement system.