1. Introduction
In the last two decades, bispecific antibodies (bsAbs) have emerged as a new class of therapeutic molecules. The key difference to a standard antibody is their capacity to simultaneously engage two antigens [
1]. This property enables unique modes of action such as directing cells of the immune system to cancer cells or improving trafficking across the blood brain barrier [
2,
3]. The recent approval of a third bsAb for the treatment of haemophilia A demonstrates that, beyond oncology, the potential of bsAbs can be applied to different therapeutic areas [
4]. Protein engineering efforts have led to a stunning variety of different possible approaches for the generation of bsAbs, with a current estimate of more than 100 different formats [
5,
6]. These can be classified in different ways such as size, valency of binding sites, overall molecular structure and similarity to a standard IgG format. Another important difference between these formats is the number of chains that are co-expressed to assemble the bsAb which can vary between 1 to 4 polypeptides [
7]. For example, Blincyto, the approved anti-CD3, anti-CD19 (CD3xCD19) bispecific T cell engager (BiTE) is a tandem single chain variable fragment scFv format composed of a single polypeptide. Another example is Hemlibra, which is the recently approved bsAb with Factor VIII mimetic activity that has a standard IgG structure and requires the co-expression of three polypeptides: two heavy chains and a single common light chain capable of pairing with both heavy chains [
8]. Another example requiring the expression of four different chains is the CrossMab technology, incorporating interface alterations and domain swapping leads to that preferential assembly of the bispecific antibody [
6]. When a bsAb format requires co-expression of three or four polypeptides, by-products resulting from incorrect chain pairing are produced, regardless of the format and engineering approach. Although these by-products may be considered as minor contaminants, it can be challenging to remove them during downstream processing [
5,
9].
In this context, differences in the expression level of each chain composing the bsAb can lead to significant increase in by-product levels, reducing bsAb yield and making downstream purification even more challenging. Obtaining cell lines with appropriate expression of each chain of the bsAb is particularly important for large scale manufacturing in order to maximize yield and purity of the desired bispecific molecule. Although the stoichiometric expression of each polypeptide could
a priori be considered as the optimal situation for assembly, this is not always the case. Indeed, even in the simpler context of standard antibody expression, the level of free light chains in the culture supernatant correlates with antibody titer, indicating that an excess of light chain is required to achieve high IgG secretion [
10,
11].
An important characteristic of the κλ body format is its native human IgG structure as it does not incorporate any mutations or foreign sequences [
12]. A κλ body is composed of two different light chains, one kappa and one lambda, that pair with two copies of the same heavy chain. As a result, in addition to assembly of bispecific κλ bodies (IgG κλ), two monospecific by-products incorporating either two kappa chains (IgGκκ) or two lambda chains (IgGλλ) are also generated upon co-expression of the common heavy chain and the two light chains. Monospecific and bispecific antibody formation is dependent on the random assembly of heavy and light chains. It is theoretically anticipated that the equivalent expression and random association of a common heavy chain, a kappa light chain and a lambda light chain leads to the secretion of three antibody forms, IgG κλ, IgGκκ and IgGλλ, following a 2:1:1 ratio. Indeed, based on the expression and purification of over 100 different κλ bodies, we found that the bsAb represents 40 to 50% of secreted IgG when the expression of each light chain falls within a range of 30 to 70% of total light chain content. Thus it is only in rare cases that unbalanced expression of the two light chains is such that the percentage of bsAb is significantly affected. However, beyond discovery and bsAb characterization, when considering large scale expression and manufacturing, maximizing yield via fine tuning the expression of different chains might become particularly important. We previously demonstrated that altering the expression of one light chain could significantly shift the distribution between bsAb and by-products [
13]. Interestingly, reducing the expression of the most abundant light chain was significantly more effective than optimizing the codon usage to increase expression of the underrepresented light chain. This was achieved by introducing codons having a lower frequency in mammalian cells into the DNA sequence, encoding the variable and constant regions of the over expressed light chain. This strategy allowed a more balanced expression of the light chains resulting in significant improvement in yield for a suboptimal producing candidate.
Here we further applied codon de-optimization to other candidates already considered as suitable in terms of product yield. Fine tuning the expression of the polypeptides both maximized the yield and minimized by-product contamination facilitating downstream processing. Importantly, we demonstrate that modifying the expression by restricting sequence alterations to the constant region of the light chains will simplify optimization of multiple candidates in parallel.
2. Materials and Methods
2.1. Codon Optimization and De-Optimization
Non-optimal codon sequences for mammalian cells were generated by GeneArt® (GeneOptimizer® software, Regensburg, Germany) on the gene encoding of the lambda constant light chain. Inhibitory motifs (such as possible splice sites) have been removed in all sequences.
2.2. Plasmid Generation
The common heavy chain, the kappa and the lambda light chains were cloned into a single pNOVI expression vector containing three expression cassettes under the transcriptional control of the human cytomegalovirus promoter.
First, second and third promoter drove, respectively, the expression of the K2 kappa light chain which binds to human cluster of differentiation 47 (hCD47), the common heavy chain, and the different lambda light chains (O30, O35 and O41) which target human mesothelin (hMSLN). Different variants were obtained by cloning these variable lambda domains upstream of either wild type or de-optimized constant lambda domains (
Figure S1).
2.3. Transient IgG Expression in Mammalian Cells
Transformed Human Embryo Kidney monolayer epithelial cells (PEAK cells; Edge Bio, La Jolla, CA, USA) were maintained in 5% CO2 at 37 °C in a humidified atmosphere in DMEM (Invitrogen, Carlsbad, CA, USA) containing 10% FCS (Sigma-Aldrich, St Louis, MO, USA) and supplemented with 2 mM glutamine (Sigma-Aldrich, St Louis, MO, USA), referring to complete DMEM. One day before transfection, confluent cells were splited 1:4 in T175 flasks (Nunc™ Cell Culture Treated Flasks with Filter Caps, Thermo Scientific, Waltham, MA, USA) in order to have cells in the exponential growth phase. Transient transfections were performed using a mix containing 30 µg of DNA and 42 µL of Lipofectamine 2000 transfection reagent (Invitrogen, Carlsbad, CA, USA) in 1.4 mL of DMEM for 107 cells per T175 flask in 50 mL of complete DMEM.
IgG expression was measured using the Octet RED96 instrument with protein A-coated biosensors (Pall ForteBio, Menlo Park, CA, USA). According to antibody concentration, supernatants were harvested 7 to 10 days after transfection.
2.4. Stable Pool Generation in CHO-S Cells
CHO-S (Thermo Fisher Scientific, Waltham, MA, USA) cells were routinely cultured in suspension at 2 × 105 cells/mL in CD CHO medium (Sigma-Aldrich, St Louis, MO, USA) and supplemented with 6 mM l-glutamine (Sigma-Aldrich, St Louis, MO, USA) in Erlenmeyer flasks. Cell culture was performed at 37 °C with 5% CO2 and 85% relative humidity at 140 rpm. For stable transfection, CHO-S cells in the exponential growth phase were resuspended at 1.43 × 107 cells/mL in CD CHO medium without glutamine and mixed with 40 mg of linearized DNA (Sigma-Aldrich, St Louis, MO, USA) in an electroporation cuvette of 0.4 cm (Bio-Rad, Hercules, CA, USA). After electroporation by single pulse of 300 V, 900 mF and infinite resistance using Gene Pulser XcellTM Electroporation Systems (Bio-Rad, Hercules, CA, USA), cells were immediately added in 50 mL of CD CHO medium without glutamine and distributed in two 6-well plates at 4 mL/well or in two T-75 flasks (Nunc™ Cell Culture Treated Flasks with Filter Caps, ThermoScientific, Waltham, MA, USA)and placed in humidified 10% CO2 incubator (Thermo Scientific, Waltham, MA, USA) set at 37 °C. The following day, l-methionine sulfoximine (MSX) (Sigma-Aldrich, St Louis, MO, USA) was added to the culture at 50 mM final concentration for transfected cell selection. After 4 to 5 weeks of growth under selection pressure, pools were assessed for their IgG productivity, and transferred to Erlenmeyer flasks and amplified in suspension.
2.5. Productivity Evaluation by Fed-Batch Overgrowth Cells (FOG)
The productivity of CHO pools was assessed by FOG evaluation. Cells were inoculated at 3 × 105 cells/mL in 50 mL of CD CHO medium supplemented with 300 mg/L of l-cysteine (Sigma-Aldrich, St Louis, MO, USA), 120 mg/L of l-tyrosine (Sigma-Aldrich, St Louis, MO, USA), 50 µM MSX and fed with 15 mL CHO CD EfficientFeed™ B Liquid Nutrient Supplement (Invitrogen, Carlsbad, CA, USA) at day 1. The glucose level was monitored using the GlucCell Glucose Monitoring System (CESCO Bioproduct, Atlanta, GA, USA), and adjusted when necessary. Cells were harvested at day 15 post inoculation or when viability had dropped below 75%. IgG quantitation was measured by Octet.
2.6. IgG Purification
After 7–10 and 15 days of antibody production respectively for PEAK cells and CHO cells, the supernatant was harvested, clarified by centrifugation 10 min at 2000 rpm and filtered on a 0.22 µm membrane (Millipore, Burlington, MA, USA). Total IgGs were purified by one affinity chromatography step using the FcXL resin (Life Technologies, Carlsbad, CA, USA) or the MabSelect SuRe resin (GE Healthcare, Chicago, IL, USA) for PEAK and CHO supernatant respectively. A serum-containing medium is used for production in PEAK cells; bovine IgGs do not bind to the FcXL resin, while they bind to the MabSelect SuRe resin. Then, two additional affinity chromatography steps were required to isolate the κλ body and eliminate the two monospecific mAbs: one purification on the KappaSelect resin (GE Healthcare, Chicago, IL, USA) to eliminate the lambda mAbs and one purification on the LambdaFabSelect resin (GE Healthcare, Chicago, IL, USA) to get rid of kappa mAbs.
An appropriate amount of MabSelect SuRe or FcXL resin was washed three times with (phosphate-buffered saline) PBS (Sigma-Aldrich, St Louis, MO, USA) and resuspended in PBS. The resin was added to the supernatant and the mix was incubated overnight at 4 °C and 15 rpm. Samples were centrifuged at 2200 rpm for 10 min to collect the resin and the flow through was discarded. The resin was washed with PBS and transferred on SigmaPrepTM spin column (Sigma-Aldrich, St Louis, MO, USA). Elution was performed with glycine (Sigma-Aldrich, St Louis, MO, USA) 50 mM at pH 3.0 (Sigma-Aldrich, St Louis, MO, USA). Following purification, the total IgG and the κλ body were formulated into 25 mM histidine (Sigma-Aldrich, St Louis, MO, USA), 125 mM NaCl (Sigma-Aldrich, St Louis, MO, USA) at pH 6.0, by desalting on Amicon Ultra-4 centrifugal filters with membrane Ultracel 50 kDa (Merck Millipore, Burlington, MA, USA) previously equilibrated with formulation buffer (25 mM histidine, 125 mM NaCl at pH 6.0).
The final antibody concentration was evaluated by Nanodrop® (Thermo Scientific, Waltham, MA, USA)
2.7. Characterization of Purified Antibodies
Monospecific antibodies and κλ body distribution and integrity were assessed by electrophoresis, isoelectric focusing gel analysis (IEF), hydrophobic interaction high performance liquid chromatography (HIC-HPLC) and size exclusion high performance liquid chromatography SEC-HPLC. Purified IgGs were analyzed by electrophoresis in denaturing and reducing conditions. The Agilent 2100 Bioanalyzer was used with the Protein 80 kit as described by the manufacturer (Agilent Technologies, Santa Clara, CA, USA). The distribution of the different formats of IgG (monospecific lambda, kappa and bispecific antibody) was determined by isoelectric focusing (Cambrex pH 7–11 IsoGel agarose plates) and HIC-HPLC analysis using ProPac HIC-10 column (Dionex, Sunnyvale, CA, USA). A gradient of mobile phase A (0.01 M sodium phosphate dibasic buffer (Sigma-Aldrich, St Louis, MO, USA) + 1.5 M ammonium sulphate (Sigma-Aldrich, St Louis, MO, USA), pH 3.5) from 100 to 10% and a growing gradient of mobile phase B (0.01 M sodium phosphate dibasic buffer + 10% acetonitrile (Merck KGaA, Darmstadt, Germany), pH 3.5) from 0 to 90% were applied. A blank was performed with mobile phase A, pH 7.0. Aggregate and fragment levels were determined by SEC-HPLC with a Biosep-SEC-s3000 column (Phenomenex, Torrance, CA, USA) using a 200 mM sodium phosphate dibasic buffer, pH 7.0 mobile phase.
4. Discussion
Over the last decade, many options have become available to generate bsAbs with very different formats. In most cases, the objective is to achieve the production of the bsAb from a single cell while minimizing the generation of by-products such as miss-paired chains or partially assembled molecules. In order to achieve this goal, elegant protein engineering approaches have been developed to remodel protein-protein interfaces and promote correct chain pairing [
6,
7]. When the κλ body format was developed, the main objective was to avoid the introduction of any protein sequence modification to preserve the properties of monoclonal antibodies, such as stability, prolonged circulating half-live and low immunogenicity, that contributed to their success as therapeutic agents. Our approach relies on effective affinity purification of the IgGκλ, that is the only molecule binding to all affinity media used in the downstream purification process, ensuring both high purity and integrity of the purified material. As for other formats that rely on the co-expression of multiple chains, κλ body yield as well as abundance of by-products is affected by the expression and assembly rates of the different polypeptides. Thus, it is important to aim at controlling their relative expression levels, which can be achieved by altering gene copy number, promoter strength or by DNA sequence optimization [
13,
18,
19]. We previously described that the introduction of lower frequency codons into the sequence encoding both the variable and constant domains of a highly expressed light chain could significantly improve bsAb yield. Here we demonstrated that limiting the modification of codon content to a part of the sequence (i.e., the constant domain of the light chain) was also effective in tuning down expression. The first benefit of this modular approach is that it facilitates the evaluation of multiple de-optimized domains in parallel on several bsAb candidates. Our data also indicates that the effect of de-optimized sequences can be different depending on the candidate, reinforcing the need to test multiple combinations. Modular de-optimization is in principle applicable to other bsAb formats or even more widely to the expression of other protein complexes. Furthermore, it has been shown that the removal of lower frequency codons can significantly alter co-translational folding of some proteins by removing pause sites during translation [
19,
20]. Although we saw no difference in aggregate levels between the different bsAb variants, we cannot exclude that newly introduced low frequency codons might have an effect on translation rate and folding. Thus, another benefit of restricting codon alteration to a single domain is to limit such risks.
We have shown in our previous study that the effects on light chain content and IgG form distribution obtained with different de-optimized variants were consistent between transient transfections using PEAK cells and expression from stable CHO pools [
13]. In this study, the results obtained with both systems were also similar for K2O41 but appeared to be different for K2O30. This finding suggests a candidate-dependent effect but might also reflect differences between the cell types used for expression. To improve consistency in future studies, CHO cells can be used for both transient expression and stable cell line generation. Yield, integrity and purity are critical parameters when expressing and purifying therapeutic proteins. BsAbs are complex molecules and their large scale production has been, and remains, far more challenging than standard monoclonal antibodies. In particular, removal of product-related impurities (or by-products) can be difficult [
9,
17]. We have shown that tuning the expression of one polypeptide enabled the identification of stable transfected CHO cells not only secreting bsAb titers of more than 1 g/L in shake flasks, but also having an unbalanced content of monospecific by-products that can simplify their elimination during downstream processing. Thus, although bsAb titer remains the most important parameter for cell line development, modulating polypeptide expression can also improve other parameters that can have a significant impact when considering large scale manufacturing operations.