Bispecific Complement Engagers for Targeted Complement Activation

This article is featured in Top Reads, p. 309

Immunotherapy that stimulates the immune system to target tumor cells is rapidly transforming cancer treatment (1). Although T-cell therapy and engagement have shown remarkable results, comparatively little focus has been put on directing the complement system toward tumor cells, though the complement system is known to represent a powerful cell-killing system. The complement system is part of the innate immune system and is involved in maintaining homeostasis by detecting and responding to pathogen-associated molecular patterns and danger-associated molecular patterns. The classical pathway (CP) of complement is initiated by the C1 complex that recognizes the Fc moiety of Ag-bound Abs. C1 comprises the hexameric pattern recognition molecule C1q and a tetramer of the serine proteases C1r and C1s (C1r₂C1s₂). C1q contains six heterotrimeric collagen-like helices, which form a stem that associates with C1r₂C1s₂. The helices diverge to form six individual branches that end in C-terminal globular heads (gC1q) that recognize the Fc region of Ag-bound Abs. C1q interacts with a single Ab with low affinity (2), and physiological binding and activation of C1 are achieved by avidity-driven multivalent Fc binding (3). This explains why potent complement activation by mAbs is observed for only a limited number of Abs and Ags (4–6), because the geometry, size, density, and mobility of the Ag influence positioning of the Fc, which generates constraints for multivalent binding of a single C1 molecule (7). The activation mechanism of C1 is not entirely understood. Activation has been suggested to occur as a result of structural changes induced in C1q upon Ag binding, which subsequently leads to C1r and C1s activation (8). Alternatively, C1 activation may be driven primarily by intermolecular activation of C1r and C1s between neighboring C1 complexes on an activator (9). Activation of C1 initiates a proteolytic cascade, resulting in the assembly of the CP C3 convertase (C4b2a) and the alternative pathway (AP) C3 convertase (C3bBb) on activator surfaces. The convertases cleave C3 into C3b and C3a, leading to C3b deposition and assembly of more AP C3 convertases. At a high density of C3b, the C3 convertases assemble into C5 convertases that cleave C5 into C5a and C5b. C5b associates with C6–C9 to form the lytic membrane attack complex (MAC). Complement is tightly controlled by membrane-associated complement regulatory proteins (mCRPs) and soluble complement regulatory proteins (sCRPs) to maintain appropriate activation and inhibition.

The role of the complement system in cancer is complex and context dependent, where a low degree of activation promotes tumor progression, whereas potent activation has antitumor effects (4–6, 10). Mechanistically, complement activation can aid with tumor clearance in a number of ways. Tumor cells are lysed directly by complement-dependent cytotoxicity (CDC) as a result of the assembly of the cytolytic MAC. Opsonization of tumor cells by C3b stimulates complement-dependent cell-mediated phagocytosis by macrophages or complement-dependent cell-mediated cytotoxicity by PBMCs and polymorphonuclear cells (11). Furthermore, the anaphylatoxins C5a and C3a act as chemokines that recruit and activate effector cells. Complement also regulates the adaptive immune system by directly and indirectly modulating B- and T-cell responses (12–14). However, tumor cells may overexpress complement regulatory proteins (CRPs) and generate an immunosuppressive environment by limiting complement activation to sublytic levels (10, 15).

Multiple approaches have been employed to increase the complement-activating potential of Abs, including (1) generation of chimeras of human IgG1 with Fc from human IgG3 (16); (2) increasing the affinity between human IgG1 and C1q by mutating residues in the Fc region where C1q binds (i.e., CH2) (17, 18); (3) combining two Abs with distinct epitopes (19–21); (4) introduction of mutations in IgG1 Fc, including E345K and E430G, that facilitate IgG1 hexamer formation upon Ag binding (22, 23); (5) monovalent binding to target Ags (24); and (6) bispecific IgG Abs that recruit C1q to Ags (25).

Single-domain Abs (sdAbs, also called nanobodies) are composed of one Ag-recognizing Ig domain and are derived from naturally occurring heavy chain–only Abs in camelids (26). They are highly stable and one-tenth the size of an IgG (15 kDa versus 150 kDa). sdAbs fulfill the criteria for therapeutic modalities, and one bivalent sdAb (caplacizumab) is currently clinically approved.

In this study, we present, to our knowledge, a novel strategy for potent activation of the complement system using bispecific recognition molecules composed of a C1q-targeting sdAb fused to a second tumor Ag binding sdAb. By way of analogy to bispecific T-cell engagers and bispecific killer cell engagers, we termed the modality “bispecific complement engagers” (BiCEs). We show that the targeting sdAb can easily be exchanged for directing complement to completely distinct Ags. We find that the efficacy of the bispecific Abs to activate complement is highly dependent on the epitope of the C1q-binding sdAb, suggesting that the spatial orientation of C1 upon Ag engagement is critical for complement activation.

The anti-human C1q nanobodies were selected by phage display using full-length C1q or gC1q as bait and subcloned into the pET-22b(+) expression vector for bacterial expression as described by Laursen et al. (27). The gene encoding the 7D12 sdAb, specific for epidermal growth factor receptor (EGFR), was purchased from GenScript and subcloned in the same vector (28). All constructs contained an N-terminal PelB signal for periplasmic secretion and a C-terminal 6xHis tag. BiCE molecules were prepared by fusing the individual C1q-specific sdAbs to 7D12 using a 2×(GGGGS) linker. Expressions of individual sdAbs and BiCEs were performed using the Escherichia coli expression strain LOBSTR (29) grown in 2xTY media. Expression was induced with 0.5 mM isopropyl β-d-thiogalactoside and carried out at 18°C overnight. The cells were pelleted and resuspended in 50 mM Tris, pH 8, 500 mM NaCl, 20 mM imidazole, pH 8.0, before lysis by sonication. Cell debris was removed by centrifugation at 12,000 × g for 30 min, and the resulting supernatant was mixed with nickel-nitrilotriacetic acid beads (Macherey-Nagel) and incubated for 1 h at 4°C. The beads were washed with 50 mM Tris, pH 8, 500 mM NaCl, 20 mM imidazole, and bound protein was eluted with 50 mM Tris, pH 8, 500 mM NaCl, 400 mM imidazole. The eluate was subsequently dialyzed against 20 mM sodium acetate, pH 5.5, 25 mM NaCl, overnight and loaded on 1 ml Source 15S (GE Healthcare) equilibrated in 20 mM sodium acetate (pH 5.5), 25 mM NaCl. Protein was eluted with a 35-ml gradient from 25 to 500 mM NaCl. A final polishing step was performed by gel filtration using a Superdex 75 10/300 column (GE Healthcare) equilibrated in 20 mM HEPES, pH 7.5, 150 mM NaCl. Purified proteins were analyzed by SDS-PAGE under nonreducing conditions.

Soluble EGFR-His (sEGFR) was produced as described by Pedersen et al. (28). In short, A431 cells were grown in RPMI 1640 media supplemented with 10% FBS and penicillin (100 U/ml) and streptomycin (100 μg/ml). sEGFR was purified from the cell media using anti-EGFR scFv425 immobilized on chitin beads (28). The beads were washed with PBS, and sEGFR was eluted with a buffer containing 100 mM acetic acid, 50 mM NaCl, pH 2.8, and immediately neutralized with 1 M Tris, pH 7.5. Relevant fractions were pooled and concentrated before being subjected to size exclusion chromatography on a 24-ml Superdex 200 increase (GE Healthcare) equilibrated in 20 mM HEPES, pH 7.5, 150 mM NaCl.

Biotinylated sEGFR was generated by mixing purified sEGFR (2 mg/ml in 20 mM HEPES, pH 7.5, 150 mM NaCl) with a 10-fold molar excess of EZ-Link N-hydroxysuccinimide biotin reagent dissolved in 10 mM DMSO. The reaction mixture was incubated at room temperature for 30 min before dialysis against 2 L 20 mM HEPES, pH 7.5, 150 mM NaCl, overnight at 4°C.

C1q was purified from outdated human citrate plasma as described previously (27). Briefly, plasma was thawed at 4°C, 5 mM EDTA was added to the plasma, and any precipitate was collected at 6000 × g for 30 min. The supernatant was gaze filtered and incubated for 2 h at room temperature with 100 ml Bio-Rex 70 beads (Bio-Rad Laboratories) equilibrated in 50 mM Na₂HPO₄, 82 mM NaCl, 2 mM EDTA, pH 7.4 (buffer A). The beads were washed extensively with buffer A and packed on an XK16 (GE Healthcare) column. The beads were washed with buffer A until baseline and eluted with a 300-ml linear gradient from 82 to 300 mM NaCl. The fractions containing C1q were pooled, and a 1:1 v/v ratio of 66% saturated (NH₄)₂SO₄ solution was dripped into the sample while stirring at 4°C. After 15 h, the precipitate was pelleted by centrifugation at 3000 × g for 30 min and resuspended in 50 mM Tris-HCl, 500 mM NaCl, pH 7.2 (storage buffer). The sample was dialyzed three times for at least 6 h at 4°C against 2.5 L storage buffer before flash freezing at −80°C.

Single-chain gC1q was prepared as described by Laursen et al. (27). Briefly, a stably transfected HEK293 cell line maintained in FreeStyle 293 media with 200–300 μg/ml geneticin (Life Technologies) was used for expression of the single-chain form of gC1q with a C-terminal His6 tag. gC1q was purified from culture media adjusted to pH 8.0 using a 5-ml HisTrap excel column (GE Healthcare). The protein was eluted with PBS containing 500 mM NaCl and 400 mM imidazole, pH 8.0, before a final polishing step on a Superdex 200 Increase 10/300 column (GE Healthcare) in 20 mM HEPES-NaOH, 150 mM NaCl, pH 7.5.

Biolayer interferometry (BLI) experiments were performed with an Octet RED96 instrument (forteBIO) at 30°C with shaking at 1000 rpm in 20 mM HEPES, pH 7.5, 150 mM NaCl, 0.02% Tween 20, 1 mg/ml BSA (assay buffer). For affinity measurement, BiCEs were immobilized on amine-reactive sensors (AR2G, forteBIO). The sensors were equilibrated in H₂O for 5 min before being activated in a mixture of 20 mM 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide and 10 mM N-hydroxysuccinimide for 5 min. The BiCE to be coupled was diluted to 20 µg/ml in 10 mM sodium acetate, pH 5.5, 25 nM NaCl, and loaded on the sensors for 10 min. After loading, sensors were quenched with 1 M ethanolamine pH 8 for 5 min. The sensors were then equilibrated in the assay buffer for 2 min before association, and dissociation was monitored for 5–10 min with gC1q (0.3–22 nM) or C1q (0.1–5.5 nM) diluted in assay buffer. Sensors were regenerated using 10 mM glycine, pH 2.2, and assay buffer for neutralization. Kinetic constants were determined for the BiCE–gC1q interaction by fitting to a 1:1 binding model using the Octet Data Analysis software.

To analyze simultaneous binding of gC1q and sEGFR to BiCE85, biotinylated sEGFR was immobilized on streptavidin sensors (SA, forteBIO). Biotinylated sEGFR was prepared at 2 µg/ml in assay buffer and loaded on the sensors for 5 min. Sensors were then washed in assay buffer for 1 min before measuring binding to BiCE085 at 10 µg/ml and/or human gC1q at 20 µg/ml in assay buffer. Association and dissociation were monitored for 120 s.

Epitope binning was performed with C1qNb75 immobilized on streptavidin sensors through a biotinylated C-terminal Avi-tag. Site-specific biotinylation of C1qNb75 was carried out as described by Zarantonello et al. (30). Biotin-C1qNb75 was prepared at 2 µg/ml in assay buffer and loaded on streptavidin sensors (SA, forteBIO) for 5 min. Sensors were washed in assay buffer for 1 min before measuring binding to gC1q (at 1.1 µg/ml in assay buffer) or C1q preincubated with a 5-fold molar excess of C1qNb31, 57, or 75. Sensors were regenerated using 10 mM glycine, pH 2.2, and assay buffer for neutralization.

BHK-21-EGFR+ cells were cultured in RPMI supplemented with 10% FBS and penicillin (100 U/ml) + streptomycin (100 μg/ml). Selection pressure was maintained by addition of 1 mg/ml geneticin (PAA, Pasching, Austria) (31). Flow cytometry was performed as described by Pedersen et al. (28). Briefly, A431, MDA-MB-468, A549, A1207 and Raji cells were cultured in RPMI-1640 at 37°C with 8% CO₂. The media were supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml). Adherent cells were detached with Accutase (Sigma-Aldrich) for 5 min at 37°C, collected by centrifugation, and resuspended in Veronal-buffered saline (VBS) (Lonza; 3.11 mM barbital, 1.8 mM sodium barbital, 145 mM NaCl, 0.83 mM MgCl₂, 0.25 mM CaCl₂) at 5 × 10⁶ cells/ml. Cells (250,000) were mixed with BiCE/sdAb/mAb to a final concentration of 83 nM in VBS supplemented with 7 mM MgCl₂ and 10% normal human serum (NHS) final concentrations. Cells were incubated for 1 h at 37°C and gently vortexed every 15 min. Cells were subsequently collected at 300 × g and washed three times in ice-cold PBS supplemented with 1% BSA and 0.01% sodium azide (PBA). Cells were resuspended in 100 μl PBA containing mouse anti-hC3c (Quidel) (diluted 1:100) or mouse anti-C1q (Quidel) (diluted 1:100) and incubated for 1 h on ice. Subsequently, cells were collected, washed three times in PBA, resuspended in 100 μl goat F(ab′)₂ anti-mouse IgG FITC Ab (Sigma-Aldrich), diluted 1:100 in PBA, and incubated for 1 h on ice. Finally, cells were collected and washed three times in PBA before they were resuspended in 200 μl PBA containing 300 nM DAPI (Thermo Fisher). Flow cytometry was performed using a NovoCyte flow cytometer (ACEA Biosciences). All cells were gated in a scatterplot using forward scatter height and side scatter height, whereas single cells were gated using forward scatter area and forward scatter height. Live cells were gated as DAPI negative. Data analysis was performed with FCS Express 6 Plus (De Novo Software). Control experiments were performed with VBS supplemented with 10 mM EDTA or hC4bNb6 (5 µg/ml) to prevent complement activation; that is, EDTA depletes for divalent metal ions inhibiting all complement pathways, whereas hC4bNb6 inhibits the classical and lectin pathways at C4 activation (32). Cell viability was calculated using the formula: % live cells = (live cell count_treatment/live cell count_buffer) × 100. Cell viability analysis was based on two independent experiments. LAM4 was used as an irrelevant control sdAb.

EGFR expression on A431, A1207, MDA-MB-468, and A549 cells was determined by flow cytometry using FITC-conjugated anti-EGFR affibody (ab81872, Abcam). Cells were cultured and collected as described above. Cells were incubated with the FITC-labeled affibody for 1 h at 4°C, washed three times with PBA, and resuspended 200 μl PBA containing DAPI. Flow cytometry was performed as described above.

CDC assays were performed as described previously (20). Briefly, target cells (BHK-31, A431, A1207, or MDA-MB-468) were first incubated with 200 μCi ⁵¹Cr for 2 h. After washing, cells were adjusted to 10⁵ cells/ml. Cells (50 µl) were transferred to round-bottomed microtiter plates (Nunc) and mixed with 50 µl RPMI 1640 (supplemented with 10% FBS) and 50 µl 100% NHS or unfractionated hirudin-treated human whole blood as the source of complement. Next, 50 µl BiCE or control Ab was added to reach the following final concentrations. For the CDC assay with A431, A1207, and MDA-MB-468 cells treated with BiCE, 7D12, matuzumab, or a CDC-enhanced version of matuzumab (matuzumab-mut, K326A/E333A) (17, 33), we used 10 µg/ml. Dose–response experiments with A431 and BHK-31 cells were performed with 0.05–666 nM BiCE and 666 nM 7D12. For the assay with BHK-31 cells treated with a single dose of BiCE or EGFR Ab cetuximab or matuzumab, their combination (cetuximab + matuzumab) or matuzumab-mut was used at 66 nM. For the combination treatment, cetuximab and matuzumab were mixed at 33 nM. The whole-blood assay was also performed with 66 nM BiCE or EGFR Ab. The C5-reacting eculizumab (100 µg/ml) (Alexion Pharmaceuticals, Boston, MA) was used to block the terminal pathway of complement. After 3 h at 37°C, plates were centrifuged, and ⁵¹Cr release from cells into the supernatant was measured as counts per minute (cpm). Three independent experiments were performed. The percentage of ⁵¹Cr release was calculated using the formula: % lysis = (experimental cpm − basal cpm)/(maximal cpm − basal cpm) × 100, where maximal ⁵¹Cr release is determined by adding perchloric acid (3% final concentration) to the target cells, and basal release is measured in the absence of BiCE or Ab.

Raji cells were cultured and collected as described above. Cells were resuspended in RPMI 1640 supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml) to get 250,000 cells/ml. Cells (50 µl) were transferred to black microtiter plates with optical bottoms (Nunc). Daratumumab or BiCE161 (25 µl) diluted in VBS supplemented with 28 mM MgCl₂ was mixed with 25 µl 100% NHS and added to the cells. The plate was incubated at 37°C, 8% CO₂, shaking. After 1 h, 10 µl alamarBlue (Thermo Fisher) was added to each well. The plate was incubated for another 60 min at 37°C before measuring fluorescence (excitation 530 nm, emission 570 nm).

Fluid-phase activation of complement was performed as described previously (27). In brief, serum samples were incubated with BiCE85 or C1qNb75 (5–50 µg/ml) for 18 h at 37°C to initiate autoactivation of complement. Samples were subsequently diluted 1:15, and C5a levels were measured by a commercial ELISA (HK349, Hycult Biotech).

A time-resolved immunofluorimetric assay was used to measure the binding of C1q sdAbs to C1q and was performed as described by Laursen et al. (27). In brief, microtiter wells were coated using 1 µg/ml C1qNb31, C1qNb35, C1qNb57, C1qNb75, or rabbit F(ab′)₂ anti-human C1q prepared by pepsin digestion of a commercial rabbit anti-C1q Ab (A136, Dako). The wells were blocked with 250 μl TBS supplemented with 1 mg/ml human serum albumin and washed three times with TBS/Tween. Purified C1q or NHS (100 µl) diluted in TBS/Tween supplemented with 5 mM CaCl₂ was added to each well and incubated for 1 h at room temperature. The wells were then washed three times with TBS/Tween before adding 100 μl biotinylated rabbit anti-C1q Ab (Dako A136, biotinylated by standard methods) diluted to 0.25 μg/ml in TBS/Tween. Microtiter wells were incubated for 1 h at room temperature and washed three times in TBS/Tween. Europium-labeled streptavidin (100 µl) at 1 µg/ml in TBS/Tween supplemented with 25 μM EDTA was added to each well and incubated for 1 h at room temperature. Wells were then washed three times in TBS/Tween before 200 μl enhancement buffer was added to each well, and time-resolved fluorescence (read as counts per second) was measured using a VICTOR3 multilabel plate counter (PerkinElmer).

Recruitment of C1q to a surface by C1q-specific Abs has previously been shown to result in activation of the CP (25, 34–37). We reasoned that CP could also be activated by recruiting C1q to a surface-localized Ag with a bispecific sdAb (Fig. 1A). sdAbs are derived from the Ag-binding VHH domain of camelid heavy chain–only Abs (26). Due to their single-domain structure, they are easily formatted into bi- and multispecific Abs by linking the C- and N-terminals of individual sdAbs with peptide linkers. To generate C1q-recruiting bispecific sdAbs, we used the previously described C1q-reacting C1qNb75 (27) and further selected two additional sdAbs against C1q (C1qNb31 and C1qNb35) and one against a single-chain form of the gC1q domain (C1qNb57). C1qNb75 binds with high affinity to gC1q and prevents CP activation by IgG and IgM through steric hindrance (27). All selected sdAbs bound both purified C1q and C1q in NHS (Supplemental Fig. 1A, 1B). As a proof of concept, we decided to target the EGFR because it is a well-characterized cancer Ag and sdAbs specific for the receptor are available (38). Bispecific Abs were generated by fusing the C-terminal of the C1q sdAb to the N-terminal of the EGFR-specific sdAb 7D12 (38) using a 10-aa GS linker (Fig. 1B, 1H).

Epitopes of the 7D12 sdAb and the EGFR mAb cetuximab overlap, justifying a direct comparison (Fig. 1C) between our BiCEs and cetuximab. We first measured binding to gC1q and C1q by BLI. BiCE38, BiCE85, and BiCE94 bound to gC1q with a K_d of 0.1–0.2 nM (Fig. 1D, 1E, 1G). BiCE38, BiCE85, and BiCE94 also bound with high affinity to C1q (Supplemental Fig. 1C–1F). The binding of BiCE85 to gC1q was comparable to the previously measured affinity of C1qNb75 for gC1q (27). Thus, linking C1qNb75 to 7D12 in BiCE85 does not notably affect the affinity of its interaction with gC1q.

We did not observe any binding of BiCE90 to gC1q (Fig. 1F), but we observed a strong association with full-length C1q (Supplemental Fig. 1E), indicating that the epitope of BiCE90 on C1q is located outside the gC1q domains. To confirm that binding of 7D12 to EGFR was maintained in BiCE format, in both the presence and absence of gC1q, EGFR-coated sensors were immersed in either BiCE85 alone or preformed BiCE85–gC1q complex (Fig. 1I). Binding to EGFR was observed for both BiCE85 and gC1q-bound BiCE, whereas the negative control gC1q alone showed no association.

Using flow cytometry, we next evaluated the ability of our BiCE molecules to recruit C1q from NHS to the breast cancer cell line MDA-MB-468 that overexpresses EGFR. All tested BiCEs efficiently recruited C1q to MDA-MB-468 cells to a comparable level, whereas no binding was observed with 7D12 alone (Fig. 2A). This agrees with their similar affinity for C1q and likely reflects the full saturation of EGFR with BiCE and C1q. We next assessed complement activation on MDA-MB-468 cells by measuring deposition of C3 fragments following incubation of the cells with BiCE and 10% NHS at 37°C. All BiCE molecules activated complement (Fig. 2B), but this led to substantially different levels of deposited C3 as a measure of complement activation. Compared with control sdAb 7D12, BiCE85 efficiently activated complement, whereas incubation with BiCE94 led to moderate C3 fragment deposition, and the addition of BiCE90 and BiCE038 resulted in substantially less complement activation. Interestingly, the histogram plots for C3 deposition showed two peaks following treatment with BiCE85, suggesting that two distinct populations of cells exist with different amounts of C3 deposited.

Complement recruitment and activation by BiCE85 were next assessed on the epidermoid carcinoma cell line A431 and the glioma cell line A1207, both expressing high levels of EGFR (Fig. 2C–2F). BiCE85 efficiently recruited C1q on A431 and A1207 cells compared with 7D12 (Fig. 2C, 2E). BiCE85 also potently activated complement on both A431 and A1207 cells (Fig. 2D, 2F). C3 deposition was abolished by the addition of EDTA in accordance with activation of the CP being dependent on calcium ions. Incubation with buffer or 7D12 led to a small increase in C3 deposition compared with BiCE85 with EDTA, possibly due to a small amount of spontaneous complement activation under these experimental conditions. C3 deposition could be inhibited by the addition of the CP inhibitor hC4bNb6, a C4 Nb with functional properties similar to those of hC4Nb8 (30, 32). Importantly, benchmarking BiCE85 against the clinically approved full IgG1 EGFR mAb cetuximab on the MDA-MB-468 cells showed that cetuximab is much less efficient in activating complement than BiCE85 (Fig. 2H). This is consistent with the observation that cetuximab recruits less C1q than BiCE85 (Fig. 2G) and prior data demonstrating that cetuximab is a poor activator of complement (20). We also investigated if the binding of BiCE85 or C1qNb75 to C1q in serum can initiate activation of complement in the absence of target cells. C5a generation was used as a measure for complement activation in the fluid phase. As shown in Supplemental Fig. 2A and 2B, no activation was observed with BiCE85 and C1qNb75 concentrations ranging from 5 to 50 µg/ml.

In summary, the data show that all the tested BICEs recruit similar amounts of C1q on EGFR-expressing cells, but they differ considerably in their ability to activate the complement system through the CP and elicit deposition of complement C3 on the targeted cell.

Deposition of C3b may lead to the assembly of C5 convertases and initiation of the terminal pathway and CDC-mediated cell killing (Fig. 1A). We evaluated the ability of BiCE85 and BiCE38 to lyse EGFR-positive tumor cells by CDC and compared it with matuzumab and an engineered version of matuzumab (K326A/E333A) that facilitates C1q binding and results in stimulation of complement activation through the CP (matuzumab-mut) (17). Neither 7D12 nor matuzumab nor BiCE38 induced CDC of A431 cells when incubated at 10 µg/ml with 25% NHS (Fig. 3A). However, both BiCE85 and matuzumab-mut efficiently lysed A431 cells to a comparable level (Fig. 3A). Using A1207 cells, we observed a similar tendency, with BiCE85 and matuzumab-mut inducing 40–51% CDC and minimal lysis with matuzumab, BiCE38, and 7D12 (Fig. 3C). In contrast, matuzumab-mut was unable to lyse MDA-MB-468 cells, whereas BiCE85 led to 38% lysis (Fig. 3D). The absence of CDC with BiCE38 is consistent with its poor ability to deposit C3 fragments, despite being able to recruit C1q to a level similar to that of BiCE85 (Fig. 2A, 2B). BiCE38, cetuximab, and matuzumab were, however, able to lyse a small fraction (12–15%) of BHK-21 cells stably transfected with EGFR (Fig. 3E, 3F). BHK-21 cells are derived from hamster kidney cells and have mCRPs that do not inhibit human complement (22).

We next investigated the BiCE molecules in a more physiological setting using whole human blood (hirudin treated) as a source of complement in CDC assays and A431 as target cells (Fig. 3G). Again, BiCE85 triggered significant lysis, whereas 7D12, BiCE38, cetuximab, and matuzumab were not effective. A combination of cetuximab and matuzumab, which has previously been shown to stimulate CDC of A431 cells (20), also induced cell lysis, although to a ∼2-fold lower level than observed for BiCE85 (Fig. 3G). Lysis could be fully inhibited by the addition of eculizumab, a complement C5–specific Ab that inhibits the terminal pathway of complement, confirming that cell killing was entirely mediated by CDC (Fig. 3G).

Although both BiCE85 and matuzumab-mut lead to similar levels of CDC of A431 and A1207 cells, only BiCE85 was able to lyse MDA-MB-468 cells. This difference in activation on the cell lines tested is most likely due to a difference in the expression level of Ag or complement regulators. Thus, we quantitated the relative expression of EGFR, CD46, CD55, and CD59 by flow cytometry (Supplemental Fig. 3). CD55 dissociates C2a and Bb from the CP and AP C3 convertases, respectively (39). This leads to inactivation of the convertase and additionally prevents assembly of new convertases. CD46 functions as a cofactor for the protease factor I that cleaves C3b and C4b into fragments that cannot function as a convertase. CD59 regulates CDC by interfering with the formation of the MAC.

EGFR expression was highest on A431, followed by A1207 and MDA-MB-468 (Supplemental Fig. 3A), and all cells expressed similar levels of CD46 (Supplemental Fig. 3B). Expression of CD59 was ∼3-fold higher on MDA-MB-468 cells than on A1207 cells, whereas the expression was 2-fold higher on A431 cells than on A1207 cells (Supplemental Fig. 3C). For CD55, the expression was 2-fold higher on MDA-MB-468 cells and 4-fold higher on A431 than on A1207 cells (Supplemental Fig. 3D). Thus, of the three cell lines tested, MDA-MB-468 has the lowest expression of EGFR and the highest expression of CD59. This suggests that BiCE85, compared with matuzumab-mut, is less sensitive to EGFR and/or CD59 expression levels. Whether this represents a more general difference in how BiCEs and mAbs activate complement needs further investigation.

Previous findings suggest that the epitope recognized by an Ab within the same Ag can profoundly influence its ability to activate complement (6, 40), with epitopes located closer to the membrane generally favoring more efficient activation (41). To explore the effect of the epitope recognized by the EGFR-directed BiCE molecules, we exchanged 7D12 in BiCE85 with the EGFR sdAb 9G8 to generate BiCE128 (Fig. 4A). The 9G8 sdAb binds to EGFR with an affinity similar to that of 7D12 but to a nonoverlapping and more membrane-proximal epitope (Fig. 4B) (42, 43).

By flow cytometry, we observed that BiCE128 and BiCE85 at 83 nM are equally potent in recruiting C1q to the surface of A431 and A1207 (Fig. 4C), resulting in a similar amount of C3 fragments deposited (Fig. 4B). We next tested whether the orientation of the C1q binding sdAb influenced CDC potency by preparing a BiCE (BiCE150) in which C1qNb75 was placed C-terminally to 7D12 (Fig. 4A). BiCE150 efficiently initiated the killing of A431 cells compared with control sdAb but was significantly less potent than BiCE85 (Fig. 4D).

Thus, under the experimental conditions tested, there was no measurable effect on C3 deposition when targeting the Ag at a different epitope, but a significant difference in CDC occurred when the order of the sdAbs was changed.

The tested BiCE molecules bind with a comparable affinity to C1q, and recruit a similar amount of C1q, but differ strikingly in their ability to activate complement (Fig. 2). This suggests that recruitment of C1q to the cell surface per se is not sufficient to activate complement efficiently. To further explore this hypothesis, the epitope diversity of the C1q sdAbs was evaluated. We performed BLI competition assays with C1qNb75 (from BICE85) loaded on the sensor and measured binding to gC1q in the presence of the C1q sdAb from BiCE38 (C1qNb31) and BiCE094 (C1qNb57) (Fig. 4E). Binding was observed for both C1qNb31 and C1qNb57, but not for the negative control C1qNb75:gC1q, demonstrating that both sdAbs have nonoverlapping epitopes with C1qNb75.

In summary, the C1q epitope and most likely also the orientation of the C1 complex upon Ag engagement are important factors to achieve a potent complement response by our BiCE molecules.

The single-domain nature of BiCEs means that they can easily be reformatted to bind other Ags. Therefore, we next asked whether the targeting sdAb in the BiCE format could be exchanged while preserving the activity. We fused C1qNb75 to the CD38-specific sdAb MU1053 (44) to generate BiCE161 (Fig. 5A). CD38 is highly expressed on multiple myeloma cells and is the target of the mAb daratumumab. Daratumumab is a potent complement activator and was selected among a panel of CD38 mAbs for its ability to efficiently lyse CD38⁺ cells by CDC (6). MU1053 recognizes an epitope on CD38 overlapping with that of daratumumab, suggesting that any differences in the activity are not related to the epitope on the tumor Ag (Fig. 5A). BiCE161 bound to gC1q with the same affinity as BiCE85 (Fig. 5B). C1q was efficiently recruited to CD38⁺ human Burkitt lymphoma Raji cells when incubated with 10% NHS (Fig. 5C) and led to complement activation, as demonstrated by C3 fragment deposition (Fig. 5D). BiCE161 also induced CDC of Raji, and even more efficiently than daratumumab, in 25% NHS (Fig. 5E). Notably, daratumumab only lysed up to 68% of the Raji cells, whereas BiCE85 lysed up to 100% of the cells under the conditions tested (Fig. 5E). Collectively, we show that the modular nature of the BiCEs allows the exchange of the targeting moiety while still preserving potency.

Complement has the ability to efficiently clear cells on which complement C3 has been deposited (11). Nevertheless, only a subset of therapeutic Abs is capable of inducing CDC, which has fueled the search for alternative strategies to potentiate complement activation (45, 46). We describe a new bispecific Ab modality that recruits the C1 complex, resulting in effective complement activation and cell killing. The modular nature of the BiCEs allows an exchange of the targeting moiety while maintaining potency, suggesting that the format may be applicable to other Ags than EGFR and CD38 that are examined in the present study. In contrast to IgG-based therapeutics, the described BiCE modality does not contain an Fc fragment that can elicit non–complement-related effector functions, including Ab-dependent cellular cytotoxicity and Ab-dependent cellular phagocytosis. The contribution from complement- and Fcγ receptor–mediated effector functions to the in vivo efficacy of anticancer mAbs is challenging to delineate and inevitably varies between different Abs and clinical settings. For soluble tumors, it is generally accepted that complement is part of the antitumor mechanism for a number of mAbs, including daratumumab, rituximab, ofatumumab, and alemtuzumab. Data obtained with variants of rituximab that bind C1q but not Fc receptors showed that complement, through complement-dependent cell-mediated cytotoxicity and complement-dependent, cell-mediated phagocytosis, contributes to clearance of tumor cells in murine models with efficacies comparable to those of the FcγR-mediated effector functions (11). Recent data also show involvement of the complement system in the therapeutic activity of HER2 mAb combinations (trastuzumab and pertuzumab) in breast cancer treatment, suggesting that, if sufficiently activated, complement can also mediate antitumor effects in solid tumors (21). Indeed, increased C1q gene expression correlates with a better survival outcome in patients with HER2⁺ breast cancer, and increased expression of complement regulators CD55 and CD59 inversely correlates with the outcome (21).

The expression levels of both Ag and CRPs affect mAb ability to activate complement and induction of CDC (22, 47), and expression of CRPs has even been suggested as a prognostic marker in different cancers, including breast (48, 49) and ovarian cancer (50).

In the present study, we directly compared BiCE molecules with IgGs targeting very similar epitopes on EGFR and CD38 (Figs. 1 and 5), and we consistently observed more potent CDC with BiCEs. This suggests that BiCE molecules are more efficient in activating complement at the same level of Ag and/or are less prone to inhibition by CRPs. To bind C1 and activate complement efficiently, mAbs need to form an Fc oligomer (23). This generates constraints because the density of the Ag, as well as the epitope, angle of approach, and flexibility of the mAb, determines whether a C1-recruiting Fc oligomer can form. BiCE molecules are not dependent on the formation of Fc oligomers, and a single BiCE molecule is likely able to recruit C1q due to the high-affinity interaction. We propose that the flexibility of the BiCE combined with the ability to target specific epitopes on C1q allows optimal positioning of the C1 complex at the cell surface, leading to potent complement activation.

A single BiCE binds a single globular head domain of C1q with subnanomolar affinity and intact C1q with very high avidity due to the presence of six globular head domains. IgG Fc regions bind with low affinity to the C1q globular domain, and multivalent binding to Fc oligomers is required for complement activation. C1 is rapidly inhibited by C1 inhibitor that binds covalently to activated C1r₂C1s₂, displacing these from Ab-bound C1q (51). Whether BiCE bound C1q can reassociate with new molecules of C1r₂C1s₂ to form an active C1 complex needs further clarification.

C1q recruiting IgG-based bispecific Abs for activating complement has previously been described (25, 34–37). In this setting, one Fab arm recognizes C1q, and another arm binds a target Ag. The bispecific Ab and BiCE modalities differ in multiple aspects, including flexibility and size, that may influence their potency. For instance, epitope distance from the cell membrane is known to affect CDC potency with Abs recognizing epitopes located more proximal to the membrane, leading to more efficient CDC than a membrane-distal epitope (41). In this regard, we expect a small BiCE to be more efficient than an IgG-based C1q bispecific Ab (Supplemental Fig. 4).

Delineating the importance of complement contribution to the in vivo effects of therapeutic Abs has remained challenging, and conflicting results have been reported. Our BiCE molecules rely exclusively on complement, and future in vivo studies may shed light on the potential of complement to clear malignant cells.

Due to their small size and lack of binding to the neonatal Fc receptor, the BiCEs are expected to be rapidly cleared from the circulation upon i.v. administration. The half-life may extend to that of C1q if the BiCE circulates in a complex with C1q. C1q is present in serum at a concentration of ∼170 nM, where the vast majority is bound to C1r and C1s, forming the C1 complex. Any BiCE present in serum below this concentration will rapidly associate with C1q. Binding of the BiCE to C1 will likely change its extravasation, distribution in extravascular tissues, and tumor penetration, considerations that are especially important when targeting solid tumors. As such, a BiCE with a lower affinity to C1q that does not associate with C1 in circulation may be more efficient in this setting. How a change in affinity toward C1q will affect the potency of the BiCE and its extravasation will need to be assessed. In hematological tumors, in which tumor cells are located in tissues where complement components and BiCEs are readily available, the optimal affinity will likely be different.

In contrast to mAbs, the BiCEs cannot induce FcγR-mediated trogocytosis, a process that can lead to the removal of Ag and Ab from the cancer cell surface. Transfer of daratumumab–CD38 complexes from multiple myeloma cells to granulocytes and monocytes reduces CD38 expression levels on multiple myeloma cells in patients treated with daratumumab, at least partly through trogocytosis (52). Likewise, trogocytosis has been claimed to be responsible for the removal of CD20 and rituximab from B cells (53, 54). This Ag shaving may have direct negative implications in cancer immunotherapy because mAbs and their Ags may be removed from tumor cells in vivo. In different settings, however, mAb-mediated trogocytosis may also have antitumorigenic effects (55, 56).

CP activity is not decreased during aging (57), in contrast to many immune effector cells, which have impaired functions in the elderly (58–60) and in immunocompromised people. Consequently, complement-activating therapeutics may be applicable in patients in whom other effector systems are weakened.

The use of complement as an antitumor mechanism of action also has several challenges that need to be considered. First, complement components can be depleted in the presence of a high tumor burden, as observed following rituximab and ofatumumab treatment (61–63). Dosing with lower levels of BiCE may limit the consumption without substantially compromising efficacy (63, 64). BiCE treatment could also be supplemented with fresh frozen plasma to provide additional complement proteins. Second, mCRP is expressed by many tumors, and increased expression is one resistance mechanism used by cancers to impede killing by complement-activating mAbs (65–67). To counteract this inhibition, one should employ more potent activators, such as the BiCEs described in this study. Alternatively, the activities of CRPs can be inhibited by Abs targeting mCRPs (68–70). Third, reduction of Ag expression is also associated with the development of resistance to mAb-based therapy because triggering of effector functions, including activation of complement, is dependent on Ag expression levels. We have not yet evaluated if BiCEs are more effective at lower Ag levels than mAbs and consequently less prone to this resistance mechanism. A different strategy to mitigate Ag escape is to simultaneously target two or more Ags (71). BiCEs are specially adapted to this multispecific approach because of their high modularity.

The BiCEs may also be applicable in nononcology settings where one wishes to deplete deleterious cells: for instance, autoantibody-producing plasma cells or B cells, or infectious pathogens such as bacteria, viruses, and fungi.

In summary, we have developed, a modality for potent complement activation. We demonstrate that BiCEs are superior to mAbs in activating complement due to a combination of their higher affinity for C1q and their higher flexibility. The BiCEs may find utility in cancer treatment and possibly also autoimmune and infectious diseases.

S.T., A.Z., N.S.L., G.R.A., and D.V.P. are coinventors in a patent application covering the BiCE molecules. D.V.P., H.G., M.B.L.W., and N.S.L. are employees and hold equity in Commit Biologics ApS. The other authors have no financial conflicts of interest.

We thank the FACS Core Facility at Aarhus University, where all flow experiments were performed.