Exploiting light chains for the scalable generation and platform purification of native human bispecific IgG

Abstract

Bispecific antibodies enable unique therapeutic approaches but it remains a challenge to produce them at the industrial scale, and the modifications introduced to achieve bispecificity often have an impact on stability and risk of immunogenicity. Here we describe a fully human bispecific IgG devoid of any modification, which can be produced at the industrial scale, using a platform process. This format, referred to as a κλ-body, is assembled by co-expressing one heavy chain and two different light chains, one κ and one λ. Using ten different targets, we demonstrate that light chains can play a dominant role in mediating specificity and high affinity. The κλ-bodies support multiple modes of action, and their stability and pharmacokinetic properties are indistinguishable from therapeutic antibodies. Thus, the κλ-body represents a unique, fully human format that exploits light-chain variable domains for antigen binding and light-chain constant domains for robust downstream processing, to realize the potential of bispecific antibodies.

Figure 1. Approaches for the generation of bispecific IgG based on light-chain diversity.

(a) Parallel discovery of two bispecific arms from a fixed VH library. (1) Phage-display scFv libraries containing a single VH and diversified VL are used for selection and screening of scFv specifically binding to two different proteins (A and B). The libraries containing κ and λ variable light-chain domains are kept separated. (2) scFv candidates are reformatted into IgG and characterized for binding and functional activity. (3) The common heavy chain and two light chains (one κ and one λ) are cloned into a single mammalian expression vector. (4) Co-expression of the three antibody chains leads to the expression and secretion of an antibody mixture with a theoretical distribution of 25% monospecific κ, 25% monospecific λ and 50% bispecific IgG with κ and λ light chains (κλ-body). (5) Bispecific κλ-bodies specific for target A and B are purified using affinity resins binding to constant regions of the heavy chains (either CH1 or Fc) and to the constant regions of the κ and λ chains. The affinity-purification process can be used for any arm combination (as described in Fig. 3). (b) Sequential discovery of a second arm compatible with an existing antibody. (1) The VH domain of an antibody directed against target A is combined with diversified variable light chains to build a scFv phage display library. If the first antibody contains a κ light chain, then diversified λ light chains are used to build the library, or vice versa. (2) The resulting library is used to identify scFv candidates against a second target, B, and are reformatted into IgG for characterization. Steps 3–5 are identical to those described for (a).

Figure 2. Common heavy-chain antibodies are specific and have neutralization potential.

(a) Flow cytometry profiles obtained with different IgGs selected against CD19, EpCAM, EGFR and CD47. The profiles on antigen-negative versus -positive cell lines are shown shaded and open, respectively (for CD19: Jurkat/Raji; EpCAM: HEK-293/MCF-7; EGFR: MCF-7/A431; CD47: CHO/CHO-CD47). (b) Dose–response inhibition of SIRPα binding to CD47 expressed on cells for the positive control mAb B6H12 (open circles), IgGK1 (closed circles) and IgGK2 (crosses) that were directly isolated from the fixed heavy-chain libraries. (c) Neutralization potential in the SIRPα/CD47-binding assay for variants IgGK3 and IgGK4 (closed and open squares, respectively) obtained via optimization of IgGK2 (crosses). (d) Neutralization activity in the MSLN/MUC16 interaction assay for IgGO1 (open circles) and IgGO3 (closed triangles) in comparison with a negative control mAb (closed circles) and the therapeutic antibody Amatuximab (closed squares). (e) Dose–response binding profile measured by flow cytometry on CD19-positive Raji cells for IgG1B7 (open circles) and variants IgGL7-1 (triangles), and IgGL7-2 (closed circles) that were generated by affinity maturation. Error bars in b, c and d represent s.e.m. of two replicates. The data are representative of three independent experiments.

Figure 3. Purification and characterization of κλ-bodies.

(a) Purification workflow. In the first step, total IgGs are recovered from the culture supernatant either using protein A or IgG-CH1 Capture Select affinity chromatography, resulting in the elimination of free light chains and other contaminants. In the second step, IgG containing a κ light chain are captured using KappaSelect affinity resin and the monospecific IgGλ are eliminated in the column flow through. In the final step, pure bispecific κλ-bodies are recovered using LambdaFabSelect affinity resin and separated from the monospecific IgGκ that do not bind to the resin. (b) Analysis on an Agilent Bioanalyzer 2100 of total IgG obtained after the first IgG-CH1 Capture Select affinity chromatography step (CH1) and after the third affinity chromatography steps (BiAb) for several κλ-bodies. The bands corresponding to the common heavy chain (H) and the κ and λ light chains are indicated. (c) Analysis of the same fractions by isoelectric focusing. The top and bottom bands observed after IgG-CH1 Capture Select affinity chromatography step correspond to the two mAbs, whereas the central bands (that is, with an intermediate isoelectric focusing point) correspond to the κλ-body.

Figure 4. Biological activity of bispecific κλ-bodies.

(a) Schematic representation of monovalent and bivalent binding of CD19xCD47 κλ-bodies on the surface of CD47⁺/CD19⁻DS-1 cells (top) and CD47⁺/CD19⁺Raji cells (bottom). (b) Binding of CD47 × EpCAM (left panels) and CD47 × CD19 κλ-bodies (right panels) to DS-1 (top panels) and Raji cells (bottom panels) monitored by FACS. κλ bodies were used at 0.01, 0.1 and 1 mg ml⁻¹, control staining with fluorescently labelled secondary antibody is indicated by the shaded area. (c) Inhibition of SIRPα binding to the CD47 on DS-1 (top) and Raji cells (bottom) by increasing concentrations of different antibodies: anti-CD47-positive control mAb B6H12 (grey open circles); CD47 × CD19 K3L7-1 (black open circles, dotted line); CD47 × CD19 K3L7-2 (closed circles); CD47 × EpCAM K3L6-1 (open triangles). Error bars represent s.e.m. of four replicates and the data are representative of three independent experiments. (d) T-cell-mediated killing of EpCAM-positive cells by CD3 × EpCAM κλ-bodies. MCF-7 cells were incubated with purified human T cells in the presence of increasing concentrations of different antibodies. CD3 × EpCAM κλ-bodies: L6-3L3-1 (closed circles), L6-4L3-1 (closed squares), L6-5L3-1 (closed triangles); anti-EpCAM mAbs: L6-3 (open circles), L6-4 (open squares), L6-5 (open triangles); anti-CD3 mAb L3-1 (crosses). Error bars represent s.d. of three replicates and the experiment was repeated with the blood of two independent donors. The isoelectric focusing (IEF) gels and electrophoresis profiles corresponding to the different κλ-bodies and mAbs are shown in Fig. 3 and Supplementary Fig. 1.

Figure 5. Product quality assessment during the platform purification process.

The 25 l wave and 100 l CHO fed-batch cultures expressing H2-1H1-1 and H2-1H1-2, respectively, as depicted in Fig. 5, were clarified to supply loading material for purification. (a) Reduced SDS–PAGE analysis of samples taken throughout the platform purification process of the two κλ-bodies H2-1H1-1 and H2-1H2-2. For both, the purification consisted of three consecutive chromatography steps, MabSelect SuRe followed by KappaSelect and LambdaFabSelect (H2-1H1-1), or the two latter steps in the opposite order (H2-1H2-2). Analysed samples were the supernatant unprocessed bulk (UB), the non-retained fraction of the protein A resin (protein A flow through, pA FT), the protein A eluate (pA elution), the KappaSelect flow through and eluate (κS FT and κS elution, respectively), the LambdaFabSelect flow through and eluate (λFS FT and λFS elution, respectively). (b) Analysis by HIC-HPLC of the relative abundance of the three IgG forms expressed in the CHO cell culture supernatants (top two panels) and present at the end of the purification process (bottom panels). Samples from the purification process of H2-1H1-1 and H2-1H2-2 are depicted as indicated. The small peak observed with the purified H2-1H1-1 sample does not correspond to a monospecific IgG1κ contamination but to a degradation product, as this peak increases with prolonged incubation at 40 °C (Supplementary Fig. 5). (c) Isoelectrofocusing gel analysis of purified κλ-bodies alongside their monospecific controls. (d) Determination of aggregate levels in purified κλ-bodies by SEC-HPLC.

Figure 6. In-vitro and in-vivo stability profile of κλ-bodies.

κλ-bodies H2-1H1-1 and H2-1H2-2 were purified as depicted in Fig. 5, and were assessed for stability in vitro and in vivo. (a) Densitometry scanning fluorimety profiles of κλ-bodies (green) in comparison with the monoclonal controls IgG1κ (blue) and IgG1λ (red). (b–e) Stability of both purified κλ bodies formulated in 25 mM Histidine, 150 mM NaCl, pH 6 at 10 mg ml⁻¹ was determined at 5, 25 and 40 °C for up to 12 (T12) weeks by SEC-HPLC (b), reduced SDS–PAGE (c), non-reduced SDS–PAGE (d) and gel electrofocusing (e). (f) Serum concentration in mice injected intravenously with 5 mg kg⁻¹ of either control mAb or κλ-bodies. Serum titres of H2-1H1-1 (open circles), K15L7-2 (closed triangles), hIgG1κ (open triangles) and hIgG1λ (closed circles) were determined by ELISA. (g) Mean (n=3) serum concentration ±s.e.m. in cynomolgus monkeys injected intravenously with either 0.5 mg kg⁻¹ (open circles) or 10 mg kg⁻¹ (closed circles) of K15L7-2.