The classical complement system represents a central effector mechanism of Abs initiated by the binding of C1q to target bound IgG. Human C1q contains six heterotrimeric globular head groups that mediate IgG interaction, resulting in an avidity-driven binding event involving multiple IgG molecules binding a single C1q. Accordingly, surface bound IgG molecules are thought to assemble into noncovalent hexameric rings for optimal binding to the six-headed C1q. To study the C1q–Fc interaction of various Abs and screen for altered C1q binding mutants, we developed, to our knowledge, a novel HPLC-based method. Employing a single-chain form of C1q representing one C1q head group, our HPLC methodology was able to detect the interaction between the single-chain monomeric form of C1q and various ligands. We show that, despite a narrow window of specific binding owing to the low affinity of the monomeric C1q–IgG interaction, this approach clearly distinguished between IgG subclasses with established C1q binding properties. IgG3 displayed the strongest binding, followed by IgG1, with IgG2 and IgG4 showing the weakest binding. Fc mutants known to have increased C1q binding through oligomerization or enhanced C1q interaction showed greatly increased column retention, and IgG glycovariants displayed a consistent trend of increasing retention upon increasing galactosylation and sialylation. Furthermore, the column retention of IgG isotypes and glycovariants matches both the cell surface recruitment of C1q and complement-mediated cytotoxicity induced by each variant on an anti-CD20 Ab backbone. This methodology therefore provides a valuable tool for testing IgG Ab (glyco)variants for C1q binding, with clear relevance for therapeutic Ab development.

Therapeutic mAbs have emerged over the last 2 decades as robust and efficacious treatment agents for many indications, including cancer and autoimmune disease (1–3). Almost all of the currently approved mAbs are of the human IgG class and have the capacity through their Fc domain to interact with both FcγRs on the surface of leukocytes and C1q, a serum protein involved in the initiation of the complement cascade (4–6).

The exact mechanism of action of target-depleting mAbs in vivo is not fully understood, with complement, FcγR expressing effector cells such as NK cells and myeloid cells, and direct signaling into the tumor cell being implicated (reviewed in Reference 7). With regard to complement, the data are mixed, with evidence both supporting and denying its importance in Ab efficacy, but components of the cascade are convincingly shown to be depleted after administration of the anti-CD20 mAb rituximab in patients with chronic lymphocytic leukemia (8, 9), and in vitro chronic lymphocytic leukemia cell killing by rituximab is reduced after anti-C5 mAb eculizumab treatment (10).

Furthermore, C1q and the complement cascade are implicated in the pathology of autoimmune diseases, where autoantibodies can result in immune complex formation and complement activation on host cells and tissues, causing injury, as in systemic lupus erythematosus (11). To this end, full-length porcine C1q has been used in dialysis-type columns as a more selective alternative to plasmapheresis to remove immune complexes and autoantibodies from the plasma of patients with systemic lupus erythematosus (12). Although this treatment showed some promise in several small trials with an acceptable safety profile, it has not yet been adopted for general use (13, 14).

C1q is a large protein complex consisting of 18 polypeptides: 6× C1qa, 6× C1qb and 6× C1qc protein chains encoded by the genes C1QA, C1QB, and C1QC located on chromosome 1 (15, 16). Each molecule of C1qa, C1qb, and C1qc contains a carboxyl terminal globular domain and a collagen-like domain that assemble into a triple-helix structure with the carboxyl terminal globular domains forming a heterotrimeric globular head region (17). A disulfide bond between C1qa and C1qb stabilizes each heterotrimer, whereas a second disulfide bond between adjacent C1qc molecules couples two heterotrimers together (18). The disulfide bonds and helical interactions cause C1q to adopt a large oligomeric structure with six heterotrimeric globular head groups and a helical coiled tail, commonly referred to as a “bouquet of tulips” conformation (15, 18). It is these six head groups that are responsible for the ligand binding properties of C1q, whereas the associated proteases C1r and C1s interact with C1q to form the full C1 complex required for activation of the classical arm of the complement pathway (15, 19).

C1q binds to the upper CH2/lower hinge region of IgG, similarly to FcγRs, but with very low affinity (20–22). Accordingly, efficient recruitment of C1q by IgG requires binding to multiple IgG Fc regions immobilized to the target cell surface, whereas the naturally pentameric IgM is far more efficient at recruiting C1q (6). In keeping with this observation, IgG Fc regions have recently been shown to form ordered noncovalent hexameric rings at the target cell surface that interact with C1q (23). Disruption of this Fc–Fc interface through either mutation or the application of a competitive binding peptide prevents hexamerization and thus reduces C1q binding and complement activation.

Several residues within both IgG Fc (24) and C1q (25) have been shown to impact the affinity of the C1q–IgG interaction, and, in particular, both the presence and characteristics of the IgG Fc hinge region are critical (26). This is clearly demonstrated by the relationship between hinge size and C1q binding, with IgG3 having both the longest hinge and the strongest C1q binding, whereas IgG2 and IgG4 have a short hinge and bind C1q with very low affinity (27). Furthermore, IgG1 with the hinge region removed was unable to bind to C1q (28), and various mutations to the IgG1 hinge region can increase or decrease binding (29). Although it is thought that C1q binds to the CH2 domain, the nature of the hinge may be important for orientating the CH2 domain or distancing the Fab arms. In support of the latter, IgG4, unable to bind to C1q as a full-length Ab due to potential Fab steric clashes (30), is able to bind as an Fc alone (31).

Unlike the binding of IgG Fc to the family of FcγRs, which is of low-micromolar and high-nanomolar affinity (depending on the individual member), IgG Fc binding to C1q is of very low affinity and high micromolar (32). As such, it is thus difficult to study accurately by surface plasmon resonance (SPR) or ELISA without crosslinking or aggregating the IgG (25). Furthermore, different subpopulations within a sample that exhibit differential C1q binding are not resolved by these techniques. As such, we sought to develop an alternative method amenable to high-throughput screening of C1q ligands using a single-chain monomeric form of C1q (scC1q) head group coupled to Sepharose beads in an HPLC column. We assessed the potential of this approach for studying scC1q binding to various ligands, distinguishing altered scC1q binding of IgG mutants and glycovariants and identification of subspecies within a sample that exhibit differential binding to scC1q.

NaCl and HEPES were purchased from MerckMillipore. scC1q was produced in-house on two separate occasions in HEK293 cells and purified by affinity chromatography using a His tag to give a purified product of 50 kDa. Streptavidin Sepharose High Performance and Tricorn columns were purchased from GE Healthcare. BSA and plasma purified C-reactive protein (CRP; catalog no. C4063) were purchased from Sigma-Aldrich. Penicillin/streptomycin, glutamine, sodium pyruvate, RPMI 1640, and DMEM were all purchased from Life Technologies. FCS was purchased from Lonza. Cell lines were purchased from American Type Culture Collection: Raji (CCL-86), Ramos (CRL-1596), SU-DHL-4 (CRL-2957), and MDA-MB-453 (HTB-131). Human C1q and anti-C1q FITC Abs were purchased from Abcam (catalog nos. ab96363 and ab4223, respectively). Serum was obtained in-house from the blood of healthy donors with informed consent and stored at −80°C in glass vials.

In brief, an N-terminal gC1qC-gggsgdyka-gC1qB-meaggnikd-gC1qA construct was produced incorporating an Avi-tag sequence for enzymatic monobiotinylation after the C-terminal C1qA and a Hexa-His tag for purification. This construct was expressed using a proprietary expression vector following routine HEK293F culture. Plasmid DNA was added to FectoPRO (PolyPlus), and transient transfection was performed according to the manufacturer’s protocol. Eighteen hours after transfection, the culture was fed with 3 g/L glucose. The cell suspension was cleared by centrifugation at 4000 × g after 7 d. The cleared supernatant was used for further purification using a cOmplete His-tag column (Roche, 20 ml) at 1.5 ml/min equilibrated in 20 mM Na2HPO4, pH 7.4, 500 mM NaCl. Elution was attempted by applying a step gradient of 4%, 20%, 60%, and 100% 20 mM Na2HPO4, pH 7.4, 500 mM NaCl, 500 mM imidazole. The fractions were analyzed by SDS-PAGE and Caliper.

scC1q was biotinylated according to the manufacturer’s instructions using a BirA biotin-protein ligase kit (Avidity, Aurora, catalog no. bulk BirA). Biotinylated scC1q was dialyzed in Slide-A-Lyzer cassettes (Thermo Fisher Scientific) against 150 mM NaCl, 20 mM NaH2PO4*H2O (monohydrate), pH 7.5, and mixed dropwise with 1 ml streptavidin Sepharose beads (GE Healthcare, catalog no. 17-5113-01) while shaking. scC1q-bound Sepharose was then packed into 4.6 × 50–mm chromatography columns (GE Healthcare, Tricorn 5/50 column, catalog no. 28-4064-09) and used for chromatography experiments. A blank column was generated in the same manner using mock-treated streptavidin Sepharose beads. The columns were stored in 20 mM HEPES, pH 7.4, at 4°C.

Sample buffer (35 µl; HT Protein Express Sample Buffer) was pipetted into the wells of a microtiter plate. Five microliters (at a concentration of 1 mg/ml or comparable 5 µg) of each scC1q sample were pipetted into the well.

When finished, the plate was covered with foil to minimize evaporation. The samples (plate) were denatured at 70°C for 10 min, with or without reducing agent. Water (70 µl) was added to each sample. The sample plate was rotated at 3000 rpm for 1 min to eliminate bubbles and move the fluid to the bottom of the well. The samples were applied to the Caliper instrument, and the result was evaluated by LabChip GX software.

The samples were diluted in a reaction vessel with a placebo solution to a protein concentration of 1 mg/ml. A quantity of 20 µl (20 µg protein) of diluted sample or reference standard was mixed in the reaction vessel with 20 µl of the reducing sample buffer and incubated for 5 min at 70°C in a preheated heating block. A quantity of 16 μl (8 μg protein) of this solution was applied to each gel pocket. A quantity of 10 µl of the m.w. marker was applied undiluted to the gel. After the gel was installed in the electrophoresis apparatus, the buffer chamber was filled with ∼800 ml (1×) Tris-glycine SDS Running Buffer. The run was finished after the sample front reached the end of the gel. The gel was stained with 60 ml SimplyBlue SafeStain staining solution while gently shaking for 1 h.

To check the integrity of the product, mass spectrometry was used. Before measurement, the samples were deglycosylated by adding Rapid PNGase F from New England Biolabs and further processed according to the manufacturer’s instructions. Final incubation of the substrate/enzyme mix was done at 50°C for 15 min. The samples were prepared by an automated workstation (Biomek i7, Beckman Coulter).

The product obtained was injected and separated without pretreatment using reverse-phase ultra-HPLC (Vanquish, Thermo Fisher Scientific). A PLRP-S column (1 × 50 mm; 5-µm particle diameter, 1000 Å pore size) from Agilent was used for separation. The ultra-HPLC eluate was infused into a MaXis II ETD Q‐TOF instrument (Bruker Daltonics) operating in positive ion mode. Data were evaluated using software developed in‐house.

Size exclusion chromatography was carried out using a TSKgel UP-SW3000 column (7.6 × 300 mm, 2-µm particle size; Tosoh Bioscience). An isocratic elution using 100% running buffer (200 mM KH2PO4/K2HPO4, 250 mM KCl, pH 6.2) at a flow rate of 0.3 ml/min was used for chromatographic separation on an UltiMate3000 HPLC system (Dionex Softron GmbH) equipped with ultraviolet detection at 280 nm. A quantity of 15 µg of rhC1q was injected for the chromatographic analysis (at room temperature). Relative quantification of the protein amount was performed by manual integration of the sample elution peaks. The areas of the peaks were compared to each other, to indicate what proportion of the sample was in each elution peak. Relative quantification was performed by manual integration of the sample elution peaks within the elution window to identify the purity of the protein.

HPLC running buffers for scC1q and blank affinity columns consisted of 20 mM HEPES, pH 7.4 (eluent A), and 20 mM HEPES, 500 mM NaCl, pH 7.4 (eluent B). The primary scC1q affinity elution gradient used is shown in Table I.

The scC1q affinity column was operated on a Shimadzu 10A HPLC system with an inline degasser. Samples were detected by measuring their absorbance at 280 nm. The autosampler was maintained at 4°C, and the column was maintained at 25°C. Prior to injection of samples, running buffer (eluent A) was injected to equilibrate the column, followed by two injections of 25 µg of Ab standard. Samples (25–100 µg) were typically injected at 1 mg/ml in eluent A from Chromacol glass sample vials (Thermo Fisher Scientific), with a buffer flow rate of 0.5 ml/min. Eluent B (100%) was used to regenerate the column during each run. Chromatograms were generated and processed using Chromeleon version 7.2 SR5 (Thermo Fisher Scientific).

Anti-HER3 IgG1 and anti-P-selectin IgG4 were generated in-house (Roche) and used as a standard control and for column establishment. Anti-HER2, anti-CD20, and anti-IL-6R Abs were generated in-house, and glycoengineering was performed in vitro as previously described (33). Anti-HER2 Ab was also used as a framework for Fc and RGY Fc mutations produced by Absolute Antibody. Purified human Fab and IgG1-Fc were purchased from Jackson ImmunoResearch (catalog no. 009-000-007) and Sino Biological (catalog no. 10702-HNAH-200), respectively. Anti-CD20 subclass variants were generated in Southampton (IgG2 and IgG4) or purchased (IgG3, InvivoGen, catalog no. hcd20-mab3). Serum purified IgM was purchased from Sigma-Aldrich (catalog no. I8260).

The degree of glycan modification was confirmed using IdeS cleavage followed by mass spectrometry analysis of the IgG Fc region, as previously described (34).

The binding properties of complexed anti-CD20 Abs to scC1q were analyzed by SPR technology using a BIAcore T200 instrument (BIAcore AB, Uppsala, Sweden). scC1q was immobilized onto a BIAcore CAP chip (GE Healthcare Bioscience, Uppsala, Sweden) via biotin-streptavidin. The overall level reached 12,000 response units. The assay was carried out at room temperature with PBS containing 0.05% Tween 20, pH 6.0 (GE Healthcare Bio-Sciences) as running and dilution buffer. Complexed or monomeric Ab samples (500 nM) were injected at a flow rate of 30 µl/min at room temperature. Association time was 120 s, and the dissociation phase took 360 s. Regeneration of the chip surface was achieved through a short injection of 6 M guanidinium hydrochloride and 1 M NaOH.

Cell lines were maintained in complete RPMI 1640 (cRPMI) or DMEM (cDMEM) supplemented with glutamine (2 mM), sodium pyruvate (1 mM), streptomycin (100 U/ml), penicillin (100 µg/ml), and 10% FCS. Cells lines were split three times per week to ∼0.5 × 106 cells/ml or to ∼20% confluency (for MDA-MB-453 cells). All cells were grown at 37°C, 5% CO2, in a humidified incubator in T25 or T75 tissue culture flasks (Sigma-Aldrich).

Target cells (5 × 104; at 0.5 × 106 cells/ml) were opsonized with Ab or media (cRPMI) at room temperature for 15 min in a flat-bottomed 96-well plate (Corning, Sigma-Aldrich). Freshly thawed human serum was added to a final concentration of 30%, and cells were incubated at 37°C for 30 min in a humidified incubator at 5% CO2. The plate was then placed on ice, and samples were transferred to FACS tubes. A quantity of 25 µl of 10 µg/ml propidium iodide (Sigma Aldrich) was added, and samples were mixed. Data from single cells were acquired on a FACSCalibur flow cytometer (BD Biosciences). Serum debris was gated out, and cells positive for propidium iodide were classified as dead. Data were analyzed using FCS Express version 3 and GraphPad Prism version 6.05.

Target cells (5 × 104) were opsonized at 1 × 106 cells/ml with Ab or media (cRPMI) at room temperature for 15 min in a flat-bottomed 96-well plate. Purified human full-length C1q (Abcam) was added to a final concentration of 8 µg/ml, and cells were incubated at 37°C for 15 min. Cells were transferred to FACS tubes on ice and washed twice in 3 ml PBS with 0.1% BSA and 10 mM sodium azide for 5 min at 450 × g. Cells were stained with 1 µl per tube undiluted anti-C1q FITC (Abcam) for 30 min at 4°C and then washed as above. Data from single cells were acquired on a FACSCalibur flow cytometer. Data were analyzed using FCS Express version 3 and GraphPad Prism version 6.05.

Target expressing cells (1 × 105) were opsonized at 1 × 106 cell/ml with Ab variants or media (cRPMI) at 4°C for 30 min in FACS tubes and washed once in 3 ml of PBS containing 0.1% BSA and 10 mM sodium azide at 450 × g for 5 min. Cells were incubated with 0.1 mg/ml scC1q for 30 min at 4°C. A quantity of 10 µl of biotin binding agents (streptavidin-FITC, streptavidin-allophycocyanin, anti-biotin-PE [clone 1D4-C5], all undiluted and from BioLegend [catalog nos. 405202, 405207, and 409004, respectively]) was added at 4°C for 30 min, and cells were washed once as above. Data from single cells were acquired on a FACSCalibur flow cytometer. Data were analyzed using FCS Express version 3 and GraphPad Prism version 6.05.

A serial 1:2 dilution of scC1q from 100 µg/ml was immobilized onto flat-bottomed 96-well Nunc MaxiSorp ELISA plates (Thermo Fisher Scientific) in coating buffer (15 mM Na2CO3, 35 mM NaHCO3, both Thermo Fisher Scientific) by incubating at 37°C for 2 h. Plates were washed three times with PBS with Tween (0.05%) using a SkanWasher 300 plate washer (Skatron). Plates were blocked for 1 h with PBS and 1% BSA at 37°C, washed as above, and 1 µg/ml Ab variants (in PBS with 1% BSA) was incubated in wells for 90 min at 37°C, and the plates were washed again three times. HRP-conjugated goat anti-human Fc Ab (in PBS with 1% BSA, Jackson ImmunoResearch, catalog no. 109-035-098-JIR) was added for 1.5 h at 37°C, and plates were washed as above before addition of o-phenylenediamine dihydrochloride (Sigma-Aldrich, catalog no. P5412-100TAB) ELISA substrate. Prior to use, 40 μl of 40% H2O2 (Merck) was added to the ELISA substrate, then 100 μl of the prepared substrate was added per well. ELISA plates were incubated in the dark at room temperature for 20 min until a color change had occurred, when 50 μl 2.5 M H2SO4 (VWR) was added. Plates were read on an Epoch microplate reader (BioTek) at 450 nm. Data were analyzed using Microsoft Excel and GraphPad Prism version 7.

Statistics were analyzed in GraphPad Prism version 6.05.

In order to study the interaction of the globular C1q head domains with various ligands, first we produced a recombinant scC1q molecule. We used an approach similar to that of Moreau et al., who linked the C1q subunits in the order found in the human genome, C1qA-C1qC-C1qB (35). Here, we linked human C1qA, C1qB, and C1qC into a single chain in the order C1qA-C1qB-C1qC using short glycine-serine-rich linkers at the base of the trimer (36). This construct was produced in HEK293 cells, purified by affinity chromatography, and biotinylated. Analysis of the purified scC1q by Caliper indicated a over 97 percentage purity (Fig. 1A), and SDS-PAGE analysis of the scC1q product showed a main band at ∼45 kDa, with some higher bands likely to be multimers (Supplemental Fig. 1A). Further analysis by size exclusion chromatography (Supplemental Fig. 1B) and mass spectrometry (Supplemental Fig. 1C) also confirmed monobiotinylation of scC1q.

The biological activity of the resulting biotinylated scC1q molecule was confirmed through SPR to detect IgG Fc binding (Fig. 1B). No binding was seen with monomeric IgG or F(ab′)2 alone. Binding was only detected using F(ab′)2 anti-Fab complexed IgG1 (Fig. 1B, blue line).

We further confirmed the activity of the scC1q molecule through detection of scC1q on IgG opsonized cells by means of fluorescently labeled streptavidin and an anti-biotin Ab (Supplemental Fig. 1). Using two different Abs for opsonization, we found that scC1q was detectable, with mutations known to increase C1q binding giving increased scC1q signal (notably the K326W, E333S double mutant) (Supplemental Fig. 1D1F, 1I). A similar trend was seen when detecting the level of anti-HER2 Ab bound to both scC1q and purified human C1q using an anti-C1q Ab, with a greater signal seen using the hexameric purified human wild-type C1q (Supplemental Fig. 1G, 1H). Finally, scC1q coated onto a plate was able to capture the K326W, E333S mutant (but not wild-type or P329G mutant) anti-HER2 in an ELISA (Supplemental Fig. 1J).

On this basis, we sought to produce a column suitable for scC1q affinity chromatography that may be more sensitive and allow the detection of monomeric IgG interactions with scC1q. Biotinylated scC1q was coupled to streptavidin-coated Sepharose beads at a density of 1 mg of biotinylated C1q per 1 g of Sepharose streptavidin and packed into 4.6 × 50–mm chromatography columns. Higher densities of C1q were also tested but did not prove to be stable, because their IgG1 retention time dropped off after the column was stored (Supplemental Fig. 2A, 2B).

C1q binding to IgG is thought to be a largely charge-based interaction, so, on this basis, we selected an NaCl-based gradient to control binding/release of ligand from the column. The gradient used is shown in Table I. In this system, sample is injected at time 0 in the absence of NaCl (0% eluent B) to allow binding of weak affinity ligands, and then NaCl is gradually added to compete off the ligand on the basis of its binding strength. The column is then regenerated by increasing the NaCl concentration to 500 mM to elute all bound material.

IgG1 (anti-HER3) eluted from the column as a retention peak during the titration of NaCl at ∼50 mM, with a peak retention time of ∼25 min (Fig. 1C). Next, the specificity of the system was tested by injection of BSA onto the scC1q column. BSA showed no binding to the column and eluted entirely in the void peak at ∼2 min (Fig. 1D). The scC1q column was next assessed for its ability to distinguish between IgG1 (known to bind C1q and induce CDC) and IgG4 (which does not bind C1q or induce CDC as strongly as IgG1). Although IgG4 (anti-P-selectin) binding to the column was detected, it was substantially less than that observed in IgG1 (∼20 min versus 25 min) (Fig. 1E).

In order to further confirm the specificity of the scC1q column, a blank column containing mock-treated streptavidin-coated Sepharose was produced, and the binding of IgG1 to this blank column was tested under the same conditions as used above. IgG1 binding to the blank column gave rise to a similar retention profile, with one major elution peak at ∼22 min. Comparison of the elution profiles of IgG1 on the C1q column and the blank column indicated a peak retention time for the scC1q column ∼3 min greater than for the blank column (Fig. 1F).

Systematically increasing the NaCl concentration within the system during the Ab binding phase resulted in a progressively earlier elution peak for IgG1 injected onto the blank column (Fig. 2A, left). Similarly, however, increasing the NaCl concentration also reduced the retention time of IgG1 on the scC1q column (Fig. 2A, right) while maintaining a slightly later peak retention time for each NaCl concentration tested (see Fig. 2B for direct overlays for 5 nM and 25 nM NaCl).

C1q binding is mediated by the IgG Fc region. As such, Fab and Fc regions were tested independently on both the scC1q and the blank columns (Fig. 2C, 2D, respectively). IgG1 Fab demonstrated apparently identical binding on both the scC1q and the blank columns, with most of the material showing no binding and a small binding peak eluting at ∼19 min (Fig. 2C). IgG1 Fc showed no binding on the blank column but displayed a later elution peak on the scC1q column, eluting ∼2 min later than on the blank column (Fig. 2D).

In order to further characterize the scC1q column, we tested the effect of Ab subclass on column retention. An anti-CD20 Ab was selected, and the four different human IgG subclasses (IgG1–IgG4) were analyzed on the scC1q (Fig. 3A, left) and blank columns (Fig. 3A, right). Anti-CD20 IgG3 showed the greatest level of binding to the scC1q column, as evidenced by the window of separation between the retention on the scC1q column and the blank column (Fig. 3B, left). Anti-CD20 IgG1 showed the next highest binding, with anti-CD20 IgG4 showing the least binding (Fig. 3B, right). These data were then compared with the performance of these subclasses in assays of C1q cell surface recruitment and CDC (Fig. 3C, 3D). In addition to demonstrating the greatest specific binding to the scC1q column, anti-CD20 IgG3 also displayed the greatest C1q recruitment to the surface of opsonized cells (Fig. 3C) and induced the strongest CDC response (Fig. 3D). The pattern of results for anti-CD20 IgG1, IgG2, and IgG4 are also consistent with that seen on the scC1q binding column, with anti-CD20 IgG2 and IgG4 being the least active.

We then tested a series of different glycoforms of three different Abs, generated by in vitro glycoengineering, on the scC1q column (33) (Fig. 4). These glycoforms had quite similar retention profiles, although trends for increased retention were seen for all Abs tested. Increasingly mature glycan patterns (galactose and sialic acid) resulted in prolonged retention times (Fig. 4A4D). Samples with more galactose eluted slightly later on the scC1q column than samples with less or no galactosylation, and sialylated samples eluted later than galactosylated samples. This order of retention matches the levels of C1q recruitment to opsonized cells (Fig. 4E) and the level of CDC induced (Fig. 4F) when these glycoforms are tested on an anti-CD20 IgG1 Ab backbone. Fully galactosylated anti-CD20 IgG1 had significantly increased C1q binding recruitment compared with mock-treated anti-CD20 IgG1, with sialylated anti-CD20 IgG1 having further increased C1q recruitment (Fig. 4E). The same trend was seen for CDC induction, with fully galactosylated anti-CD20 IgG1 showing increased CDC over mock-treated anti-CD20 IgG1, and sialylated anti-CD20 IgG1 having further increased CDC (Fig. 4F).

Next, a panel of IgG1 mutants with anticipated effects on C1q binding or CDC induction was assessed (Fig. 5). Three mutant Abs showed dramatically increased binding to the scC1q column: the two double mutants E345K, K326W and K326W, E333S as well as the RGY triple mutant (E345R, E430G, S440Y), which is known to form solution state hexamers (23) (Fig. 5A, 5B). These three mutants showed broadly similar binding to the wild-type Ab on the blank column (Fig. 5C). Interestingly, the RGY mutant gave rise to two elution peaks on the scC1q column, which match by relative size the size exclusion chromatography data, indicating that the hexameric peak (80% of the sample; Supplemental Fig. 3) also has high C1q binding (Fig. 5D). Two of these high binding mutants (RGY and K326W, E333S) also showed increased resistance to the presence of NaCl during the Ab binding phase, as shown by the increased binding when tested at these increased NaCl concentrations (Fig. 5E). The hexameric RGY peak elutes at a retention time that corresponds to 500 mM NaCl, whereas the monomeric species is more sensitive and elutes earlier while retaining greater binding than seen for the wild-type Ab (Fig. 5D). Interestingly, the P329G mutant (known to have greatly abrogated Fc effector function) had highly similar retention profiles on both the scC1q and the blank column (Fig. 5A5C).

To investigate the Fab and Fc impact on scC1q column retention time, Fc-only constructs containing the RGY triple mutation, as well as various attenuating mutations, were generated and tested on the scC1q column (Fig. 5F, left) and the blank column (Fig. 5F, right). RGY Fc showed a retention peak similar to that seen for the high binding species of full-length RGY anti-HER2 on the scC1q column (∼82 min) and a slightly earlier retention peak on the blank column (∼21 min). The effect of attenuating mutations on the RGY Fc binding to C1q is shown by the progressively earlier elution of the sample with various mutations. LALA mutations attenuate the C1q binding, whereas N297Q (aglycosylation) has a more limited effect on the C1q binding. Mutation of the P329G residue has a greater effect than the two previously described mutations. Combining the P329G mutation with the LALA mutations resulted in the earliest elution time, earlier than wild-type anti-HER2. The binding of this strongly attenuated mutant was much greater than that of the IgG1 Fc, showing the effect of the avidity endowed by the presence of multiple Fc domains seen in the RGY Fc samples.

Finally, we tested two non-IgG C1q binding ligands on the scC1q column. CRP showed greater specific binding to the scC1q column versus the blank column and to a greater level than did IgG1, as exhibited by a later retention time (Fig. 6A, 6B). Serum purified IgM also showed specific binding to the scC1q column and had a greater retention than IgG1 (Fig. 6C, 6D).

The broad recognition properties of C1q stem from its six heterotrimeric globular head groups. We generated a single-chain format of the C1q head group (scC1q) using a method based on that which has been described previously to generate functional C1q head groups (35). As described, binding of a monomeric IgG to a single C1q head is of very low affinity (10−4 M) and therefore difficult to study by ELISA or SPR (35). The authors of this report commented that binding to IgG was not detectable in their SPR setup (35). Recently, an SPR method using protein L to capture IgG Abs through their L chain, thus presenting their Fc regions, has been described that facilitated detection of C1q binding to captured IgGs (37). Although promising, this method was unable to capture the IgG2 and IgG4 subclasses, thereby preventing the assessment of their C1q binding properties.

The weak affinity and the functionality of our scC1q head group were confirmed by the lack of binding to monomeric Ab by SPR measurement using immobilized scC1q, with binding seen only after the Ab was crosslinked (Fig. 1B). This approach of crosslinking/aggregation is commonly used in C1q ELISA or SPR detection methods to detect binding.

The biological activity of the scC1q construct was further demonstrated through its ability to bind to Ab Fc regions on the surface of opsonized cells (Supplemental Fig. 1C1I). Greater recruitment of the scC1q construct was seen with Abs carrying Fc mutations known to increase C1q binding (notably the K326W, E333S double mutant). When detecting bound C1q using an anti-C1q Ab, there appeared to be greater signal when using a wild-type purified human C1q molecule than with the scC1q molecule (Supplemental Fig. 1F, 1G). This could be due to the greater binding avidity of the hexameric wild-type C1q, which contains six globular Fc binding heads, compared with the single Fc binding head of the scC1q molecule. However, the scC1q molecule is demonstrably able to bind to Ab Fc regions, as measured by ELISA, SPR, and flow cytometry.

Biotinylated scC1q was packed into columns and tested for IgG1 binding. Initial tests showed that the NaCl-based elution gradient chosen allowed binding and elution of monomeric IgG1 (Fig. 1C). We found that the optimal density of scC1q for the column was 1 mg scC1q per 1 ml Sepharose column, because, although increased densities initially showed a greater IgG1 retention time than the 1-mg column, they were not found to be stable, because this dropped after storage (Supplemental Fig. 2A, 2B). Testing of BSA on the column indicated that the binding to IgG was specific, because no BSA binding was detected (Fig. 1D). Furthermore, the scC1q column showed greater binding for IgG1 than for IgG4, which matches previous data determined by other methods and the complement-activating potential of these two IgG subclasses (Fig. 1E). However, IgG4 still showed considerable binding to the column, which was surprising.

When IgG1 was tested on the blank column, considerable binding was detected, with the retention peak eluting ∼3 min earlier than on the scC1q column (Fig. 1F). We hypothesized that the absence of NaCl from the buffer during the Ab binding phase was allowing nonspecific interactions between the IgG and the Sepharose matrix. Increasing the NaCl concentration present within the buffer during the Ab binding phase resulted in a progressive decrease in binding to the blank column but also to the scC1q column (Fig. 2A, 2B). Comparison of samples tested under the same NaCl running conditions on both columns indicated that the elution of samples from the scC1q column occurs 2–3 min later than seen on the blank column, regardless of the level of NaCl present (Fig. 2A, 2B). We confirmed the run-to-run stability of the columns by including an IgG1 standard sample in each experiment. After an initial decrease from the first experimental run, the retention time was highly stable for both the blank column (Supplemental Fig. 2C, left) and the 1-mg scC1q columns (Supplemental Fig. 2C, right).

Because the C1q binding region of IgG is known to be located in its Fc domain, we sought to determine whether the Fc region of IgG was responsible for the increased binding of IgG on the scC1q column over the blank column. Recombinant human IgG1 Fc showed a 2-min shift of increased binding on the scC1q column versus the blank column (Fig. 2D), whereas Fab alone showed nearly identical binding on the two columns (Fig. 2C).

In order to further characterize the column in terms of its relevance to biology, an anti-CD20 Ab was tested on the column in each of the four IgG subclasses, as well as in cell-based assays of C1q recruitment and CDC. As would be expected, anti-CD20 IgG3 showed the greatest level of C1q recruitment to the cell surface of opsonized Ramos cells and induced the greatest CDC of opsonized Ramos cells in the presence of human serum (Fig. 3C, 3D). These data matched the observation that the anti-CD20 IgG3 was the strongest binder on the scC1q column, with the biggest specific window of binding over the blank column (Fig. 3B, left). The relative levels of C1q recruitment and CDC induction of anti-CD20 IgG1, IgG2, and IgG4 also matched their relative binding strengths on the scC1q column (Fig. 3A), giving a hierarchy of complement activation for the IgG subtypes of IgG3 > IgG1 > IgG2/IgG4.

IgG Fc glycosylation is known to affect multiple Ab functions, particularly Ab-dependent, cell-mediated cytotoxicity through the fucose residue and also complement-mediated cytotoxicity (38). To see if our scC1q column could detect differences in Ab–C1q interactions due to differences in glycosylation, we generated different glycovariants of three different clinical grade IgG1 Abs. The effect of these glycovariants was characterized in terms of C1q recruitment to the cell surface and CDC induction using an anti-CD20 Ab. As reported previously, increased CDC and C1q recruitment was seen for highly galactosylated anti-CD20 but also for anti-CD20 mAb bearing increased sialylation (Fig. 4E, 4F) (39, 40).

There are reports in the literature of sialylated IgG being anti-inflammatory and having reduced complement-activating properties when compared with galactosylated Ab (40). These observations conflict with the data presented here, where sialylated anti-CD20 IgG1 increased CDC and C1q binding over galactosylated and mock-treated anti-CD20 IgG1. These differences could be explained by the methods of glycosylation manipulation applied by different groups, the purities of sialylated IgG generated (and therefore the levels of galactosylated IgG), and the methods of studying CDC, with different treatment protocols and serum sources used.

Despite this apparent conflict with some previous reports, testing these Ab glycoforms on the scC1q and blank columns appeared to recapitulate the order of complement activation from the C1q recruitment and CDC assays, albeit with only subtle differences between the glycoforms (Fig. 4A4D). The order of samples was consistent across three different Ab scaffolds (anti-HER2, anti-IL-6R, and anti-CD20) and repeat measurements. Most notably, addition of galactose to agalactosylated IgG caused an increase in binding, and addition of sialic acid further increased binding, but only when in an α2-6 linkage (Supplemental Fig. 4). The specificity of the α2-6 linkage over the α2-3 linkage was also reported by SPR measurement of C1q binding to protein L captured IgG glycovariants (37). Although the shifts seen on the scC1q column are very subtle for all Ab backbones tested, a small increase in C1q binding for each Fc present at a target cell surface could cause a far larger increase in C1q binding to multiple Fcs through increased avidity. This may explain the larger differences between the glycoforms in CDC and C1q recruitment than on the scC1q column. For investigation of this effect, glycovariants may need to be tested in combination with IgG hexamerization engineering. Furthermore, although Fc glycans are largely buried within the core of the CH2 domains, sialic acid is believed to at least begin to protrude out of the Ab core and thus may have some effect on Fc–Fc or Fc–C1q interactions.

We selected several Fc mutations previously reported to alter C1q binding or complement activation to test on our column and demonstrate its utility for Ab screening. Three mutants were found to have greatly increased binding to the scC1q column, with no or little increase in binding to the blank column (Fig. 5A5C). Two of these mutants were also tested for their binding under higher NaCl levels and showed a reduced sensitivity to NaCl as compared with wild-type Ab (Fig. 5E). Interestingly, the P329G mutant (known to have highly abrogated Fc effector function) showed no difference in binding between the scC1q column and the blank column, further suggesting that the 2–3-min increased binding on the C1q column is a real Fc–C1q interaction (Fig. 5B, 5C, green line). There were three mutants that showed greatly increased binding to the scC1q column, with little or no change in binding to the blank column, most notably the RGY (E345R, E430G, and S440Y) triple mutant reported by Diebolder et al. (23). This mutant forms solution state hexamers in an ∼80/20 ratio of hexamers to monomers and has greatly enhanced CDC. Interestingly, two elution peaks were seen for the RGY mutant, likely indicating the hexameric and monomeric species as the peak areas matched the 80/20 ratio, the larger peak eluting later (Fig. 5D). This highlights the potential of this method to visualize subpopulations within a sample by their affinity for C1q, which is not possible by ELISA or SPR. It is notable that the monomeric peak also showed greater binding than the wild-type Ab, indicating that even in monomeric form, this mutant shows increased binding to C1q.

Finally, we tested other non-IgG ligands of C1q for binding to the columns. Both IgM and CRP were found to bind to the scC1q column with greater retention than IgG1 and a larger window of specificity between the retention on the blank column and the scC1q column (Fig. 6). This indicates that, as reported by Moreau et al. (35), scC1q is capable of binding to multiple C1q ligands and not just IgGs. Both of these ligands typically only bind to C1q after immobilization to a surface through binding to their ligand (antigen for IgM, phosphocholine for CRP), which induces a conformational change to permit C1q binding. These molecules are not bound to ligand when being passed through the columns. However, it is possible that the buffer conditions cause IgM and CRP to adopt a conformation permissive of C1q binding, allowing them to bind to the column.

In summary, we have developed an HPLC-based assay that is able to detect C1q binding to multiple ligands and make biologically relevant distinctions between different IgG subtypes, mutants, and glycoforms, as well as being able to detect subpopulations with different C1q binding properties. This method is useful for screening various mutated ligands to detect those with altered binding for potential therapeutic use or to probe the residues that are important for C1q interaction.

M.J.E.M. was a Biotechnology and Biological Sciences Research Council Industrial Collaborative Awards in Science and Engineering student with sponsorship from Roche and an employee of Roche during this work. A.K. and T.S. are listed as coinventors on a patent related to this work (36). The other authors have no financial conflicts of interest.

The online version of this article contains supplemental material.

CDC

complement-dependent cytotoxicity

cDMEM

complete DMEM

CRP

C-reactive protein

cRPMI

complete RPMI 1640

scC1q

single-chain monomeric form of C1q

SPR

surface plasmon resonance

1
Weiner
,
G. J.
2015
.
Building better monoclonal antibody-based therapeutics
.
Nat. Rev. Cancer
15
:
361
370
.
2
Faurschou
,
M.
,
D. R.
Jayne
.
2014
.
Anti-B cell antibody therapies for inflammatory rheumatic diseases
.
Annu. Rev. Med.
65
:
263
278
.
3
Lorefice
,
L.
,
G.
Fenu
,
J.
Frau
,
G.
Coghe
,
M. G.
Marrosu
.
2014
.
Monoclonal antibodies: a target therapy for multiple sclerosis
.
Inflamm. Allergy Drug Targets
13
:
134
143
.
4
Marshall
,
M. J. E.
,
R. J.
Stopforth
,
M. S.
Cragg
.
2017
.
Therapeutic antibodies: what have we learnt from targeting CD20 and where are we going?
Front. Immunol.
8
:
1245
.
5
Antibody Society
.
2018
.
Antibody therapeutics approved or in regulatory review in the EU or US
.
Available at: http://www.antibodysociety.org/news/approved-antibodies/. Accessed: April 27, 2021
.
6
Beurskens
,
F. J.
,
R. A.
van Schaarenburg
,
L. A.
Trouw
.
2015
.
C1q, antibodies and anti-C1q autoantibodies
.
Mol. Immunol.
68
:
6
13
.
7
Sopp
,
J.
,
M. S.
Cragg
.
2018
.
Deleting malignant B cells with second-generation anti-CD20 antibodies
.
J. Clin. Oncol.
36
:
2323
2325
.
8
Kennedy
,
A. D.
,
P. V.
Beum
,
M. D.
Solga
,
D. J.
DiLillo
,
M. A.
Lindorfer
,
C. E.
Hess
,
J. J.
Densmore
,
M. E.
Williams
,
R. P.
Taylor
.
2004
.
Rituximab infusion promotes rapid complement depletion and acute CD20 loss in chronic lymphocytic leukemia
.
J. Immunol.
172
:
3280
3288
.
9
Glennie
,
M. J.
,
R. R.
French
,
M. S.
Cragg
,
R. P.
Taylor
.
2007
.
Mechanisms of killing by anti-CD20 monoclonal antibodies
.
Mol. Immunol.
44
:
3823
3837
.
10
Bologna
,
L.
,
E.
Gotti
,
M.
Manganini
,
A.
Rambaldi
,
T.
Intermesoli
,
M.
Introna
,
J.
Golay
.
2011
.
Mechanism of action of type II, glycoengineered, anti-CD20 monoclonal antibody GA101 in B-chronic lymphocytic leukemia whole blood assays in comparison with rituximab and alemtuzumab
.
J. Immunol.
186
:
3762
3769
.
11
Mok
,
C. C.
,
C. S.
Lau
.
2003
.
Pathogenesis of systemic lupus erythematosus
.
J. Clin. Pathol.
56
:
481
490
.
12
Hiepe
,
F.
,
W.
Schössler
,
E.
Apostoloff
,
K.
Precht
.
1990
.
Development of a C1q-adsorbent for the selective removal of circulating immune complexes
.
Biomater. Artif. Cells Artif. Organs
18
:
529
534
.
13
Pfueller
,
B.
,
K.
Wolbart
,
A.
Bruns
,
G. R.
Burmester
,
F.
Hiepe
.
2001
.
Successful treatment of patients with systemic lupus erythematosus by immunoadsorption with a C1q column: a pilot study
.
Arthritis Rheum.
44
:
1962
1963
.
14
Berner
,
B.
,
A. K.
Scheel
,
V.
Schettler
,
K. M.
Hummel
,
M. A.
Reuss-Borst
,
G. A.
Müller
,
E.
Oestmann
,
H. P.
Leinenbach
,
M.
Hepper
.
2001
.
Rapid improvement of SLE-specific cutaneous lesions by C1q immunoadsorption
.
Ann. Rheum. Dis.
60
:
898
899
.
15
Kishore
,
U.
,
K. B.
Reid
.
2000
.
C1q: structure, function, and receptors
.
Immunopharmacology
49
:
159
170
.
16
Sellar
,
G. C.
,
D. J.
Blake
,
K. B.
Reid
.
1991
.
Characterization and organization of the genes encoding the A-, B- and C-chains of human complement subcomponent C1q. The complete derived amino acid sequence of human C1q
.
Biochem. J.
274
:
481
490
.
17
Kishore
,
U.
,
C.
Gaboriaud
,
P.
Waters
,
A. K.
Shrive
,
T. J.
Greenhough
,
K. B.
Reid
,
R. B.
Sim
,
G. J.
Arlaud
.
2004
.
C1q and tumor necrosis factor superfamily: modularity and versatility
.
Trends Immunol.
25
:
551
561
.
18
Kishore
,
U.
,
R.
Ghai
,
T. J.
Greenhough
,
A. K.
Shrive
,
D. M.
Bonifati
,
M. G.
Gadjeva
,
P.
Waters
,
M. S.
Kojouharova
,
T.
Chakraborty
,
A.
Agrawal
.
2004
.
Structural and functional anatomy of the globular domain of complement protein C1q
.
Immunol. Lett.
95
:
113
128
.
19
Bally
,
I.
,
V.
Rossi
,
T.
Lunardi
,
N. M.
Thielens
,
C.
Gaboriaud
,
G. J.
Arlaud
.
2009
.
Identification of the C1q-binding sites of human C1r and C1s: a refined three-dimensional model of the C1 complex of complement
.
J. Biol. Chem.
284
:
19340
19348
.
20
Hughes-Jones
,
N. C.
,
B.
Gardner
.
1978
.
The reaction between the complement subcomponent C1q, IgG complexes and polyionic molecules
.
Immunology
34
:
459
463
.
21
Sledge
,
C. R.
,
D. H.
Bing
.
1973
.
Binding properties of the human complement protein Clq
.
J. Biol. Chem.
248
:
2818
2823
.
22
Patel
,
R.
,
A.
Neill
,
H.
Liu
,
B.
Andrien
.
2015
.
IgG subclass specificity to C1q determined by surface plasmon resonance using protein L capture technique
.
Anal. Biochem.
479
:
15
17
.
23
Diebolder
,
C. A.
,
F. J.
Beurskens
,
R. N.
de Jong
,
R. I.
Koning
,
K.
Strumane
,
M. A.
Lindorfer
,
M.
Voorhorst
,
D.
Ugurlar
,
S.
Rosati
,
A. J.
Heck
, et al
.
2014
.
Complement is activated by IgG hexamers assembled at the cell surface
.
Science
343
:
1260
1263
.
24
Idusogie
,
E. E.
,
L. G.
Presta
,
H.
Gazzano-Santoro
,
K.
Totpal
,
P. Y.
Wong
,
M.
Ultsch
,
Y. G.
Meng
,
M. G.
Mulkerrin
.
2000
.
Mapping of the C1q binding site on rituxan, a chimeric antibody with a human IgG1 Fc
.
J. Immunol.
164
:
4178
4184
.
25
Roumenina
,
L. T.
,
M. M.
Ruseva
,
A.
Zlatarova
,
R.
Ghai
,
M.
Kolev
,
N.
Olova
,
M.
Gadjeva
,
A.
Agrawal
,
B.
Bottazzi
,
A.
Mantovani
, et al
.
2006
.
Interaction of C1q with IgG1, C-reactive protein and pentraxin 3: mutational studies using recombinant globular head modules of human C1q A, B, and C chains
.
Biochemistry
45
:
4093
4104
.
26
Liu
,
R.
,
R. J.
Oldham
,
E.
Teal
,
S. A.
Beers
,
M. S.
Cragg
.
2020
.
Fc-engineering for modulated effector functions—improving antibodies for cancer treatment
.
Antibodies (Basel)
9
:
64
.
27
Dangl
,
J. L.
,
T. G.
Wensel
,
S. L.
Morrison
,
L.
Stryer
,
L. A.
Herzenberg
,
V. T.
Oi
.
1988
.
Segmental flexibility and complement fixation of genetically engineered chimeric human, rabbit and mouse antibodies
.
EMBO J.
7
:
1989
1994
.
28
Klein
,
M.
,
N.
Haeffner-Cavaillon
,
D. E.
Isenman
,
C.
Rivat
,
M. A.
Navia
,
D. R.
Davies
,
K. J.
Dorrington
.
1981
.
Expression of biological effector functions by immunoglobulin G molecules lacking the hinge region
.
Proc. Natl. Acad. Sci. USA
78
:
524
528
.
29
Dall’Acqua
,
W. F.
,
K. E.
Cook
,
M. M.
Damschroder
,
R. M.
Woods
,
H.
Wu
.
2006
.
Modulation of the effector functions of a human IgG1 through engineering of its hinge region
.
J. Immunol.
177
:
1129
1138
.
30
Rayner
,
L. E.
,
G. K.
Hui
,
J.
Gor
,
R. K.
Heenan
,
P. A.
Dalby
,
S. J.
Perkins
.
2014
.
The Fab conformations in the solution structure of human immunoglobulin G4 (IgG4) restrict access to its Fc region: implications for functional activity
.
J. Biol. Chem.
289
:
20740
20756
.
31
Isenman
,
D. E.
,
K. J.
Dorrington
,
R. H.
Painter
.
1975
.
The structure and function of immunoglobulin domains. II. The importance of interchain disulfide bonds and the possible role of molecular flexibility in the interaction between immunoglobulin G and complement
.
J. Immunol.
114
:
1726
1729
.
32
Bruhns
,
P.
,
B.
Iannascoli
,
P.
England
,
D. A.
Mancardi
,
N.
Fernandez
,
S.
Jorieux
,
M.
Daëron
.
2009
.
Specificity and affinity of human Fcgamma receptors and their polymorphic variants for human IgG subclasses
.
Blood
113
:
3716
3725
.
33
Thomann
,
M.
,
T.
Schlothauer
,
T.
Dashivets
,
S.
Malik
,
C.
Avenal
,
P.
Bulau
,
P.
Rüger
,
D.
Reusch
.
2015
.
In vitro glycoengineering of IgG1 and its effect on Fc receptor binding and ADCC activity
.
PLoS One
10
:
e0134949
.
34
Reusch
,
D.
,
M.
Haberger
,
D.
Falck
,
B.
Peter
,
B.
Maier
,
J.
Gassner
,
M.
Hook
,
K.
Wagner
,
L.
Bonnington
,
P.
Bulau
,
M.
Wuhrer
.
2015
.
Comparison of methods for the analysis of therapeutic immunoglobulin G Fc-glycosylation profiles—part 2: mass spectrometric methods
.
MAbs
7
:
732
742
.
35
Moreau
,
C.
,
I.
Bally
,
A.
Chouquet
,
B.
Bottazzi
,
B.
Ghebrehiwet
,
C.
Gaboriaud
,
N.
Thielens
.
2016
.
Structural and functional characterization of a single-chain form of the recognition domain of complement protein C1q
.
Front. Immunol.
7
:
79
.
36
Knaupp
,
A.
,
L.
Lariviere
,
P.
Rueger
,
T.
Schlothauer
,
S.
Seeber
, inventors;
Hoffmann-La Roche, Inc.
, assignee.
Complement factor based affinity chromatography
. United States patent application US11440942B2, Publication No. 20200002391.
2022
Sep
13
.
37
Jovic
,
M.
,
F.
Cymer
.
2019
.
Qualification of a surface plasmon resonance assay to determine binding of IgG-type antibodies to complement component C1q
.
Biologicals
61
:
76
79
.
38
Yu
,
X.
,
M. J. E.
Marshall
,
M. S.
Cragg
,
M.
Crispin
.
2017
.
Improving antibody-based cancer therapeutics through glycan engineering
.
BioDrugs
31
:
151
166
.
39
Peschke
,
B.
,
C. W.
Keller
,
P.
Weber
,
I.
Quast
,
J. D.
Lünemann
.
2017
.
Fc-galactosylation of human immunoglobulin gamma isotypes improves C1q binding and enhances complement-dependent cytotoxicity
.
Front. Immunol.
8
:
646
.
40
Quast
,
I.
,
C. W.
Keller
,
M. A.
Maurer
,
J. P.
Giddens
,
B.
Tackenberg
,
L. X.
Wang
,
C.
Münz
,
F.
Nimmerjahn
,
M. C.
Dalakas
,
J. D.
Lünemann
.
2015
.
Sialylation of IgG Fc domain impairs complement-dependent cytotoxicity
.
J. Clin. Invest.
125
:
4160
4170
.

Supplementary data