Manufacturability and functionality assessment of different formats of T-cell engaging bispecific antibodies

ABSTRACT

T-cell-engaging bispecific antibodies (T-bsAbs) are promising immunotherapies for cancer treatment due to their capability of redirecting T-cells toward destroying tumor cells. Numerous T-bsAb formats have been developed, each with advantages and disadvantages in terms of developability, immunogenicity, effector functions, and pharmacokinetics. Here, we systematically compared T-bsAbs produced using eight different formats, evaluating the effect of molecular design of T-bsAbs on their manufacturability and functionality. These eight T-bsAb formats were constructed using antigen-binding fragments (Fabs) and single-chain variable fragments (scFvs) of antibodies linked to the crystallizable fragment (Fc) domain of immunoglobulin G. To ensure a fair comparison of growth and production data, we used recombinase-mediated cassette exchange technology to generate the T-bsAb-producing CHO cell lines. The produced T-bsAbs were assessed for their purification profile and recovery, binding capability, and biological activities. Our findings indicated that the manufacturability of bsAbs was adversely affected with increased number of scFv building blocks, while the functionality was affected by the combination of multiple factors, including the binding affinity and avidity of targeting moieties and the flexibility and geometry of formats. These results provide valuable insights into the impact of the format design on the optimal production and function of T-bsAbs.

KEYWORDS:

• CHO cell
• avidity
• bispecific antibody
• functionality
• manufacturability
• targeted integration
• thermal stability

Introduction

Bispecific antibodies (bsAbs) are a growing and promising class of next-generation antibodies for diagnostic and therapeutic applications. Unlike natural monoclonal antibodies (mAbs), bsAbs are designed to bind two different antigens simultaneously, allowing novel modes of action that cannot be achieved with the combination of the two mono-specific antibodies. These novel functions include recruiting cytolytic immune cells to target tumor cells for cancer treatment, mimicking cofactor function for treating the bleeding disorder, and blocking two receptors simultaneously to suppress autoimmunity.¹ Among the various types of bsAbs, T-cell-engaging bsAbs (T-bsAbs) are the most promising for cancer immunotherapy. Currently, 84% of all bsAbs are being evaluated for cancer treatment in clinical trials, with T-bsAbs being the most dominant class.² T-bsAbs have one paratope that specifically binds to a tumor-associated antigen (TAA) on tumor cells and another paratope that recognizes the CD3 moiety on T cells. By bringing T cells into close proximity with tumor cells, T-bsAbs activate T cells and eradicate the tumor cells via the formation of a cytolytic immune synapse.³ Out of seven bsAbs approved by the FDA so far, three are T-bsAbs. Blinatumomab, a bispecific CD19×CD3 T cell engager, was the first T-bsAb approved by the FDA for the treatment of acute B lymphocytic leukemia in 2014.⁴ More recently, Lunsumio, a CD20×CD3 T-bsAb,⁵ and Tecvayli, a BCMAxCD3 T-bsAb,⁶ were approved by the FDA for the treatment of follicular B cell lymphoma and multiple myeloma, respectively. The clinical success of these antibody drugs has driven growth in this research area, attracting the attention of many biotech companies and labs worldwide.⁷

To date, hundreds of bsAb formats have been developed, each with its own advantages and drawbacks in terms of developability, immunogenicity, effector functions, and pharmacokinetics (PK).⁸ The publicly available formats can be broadly categorized into three groups: 1) fragment-based, 2) appended-IgG, and 3) IgG-like bsAbs.⁹ The fragment-based bsAbs lack Fc region and consist of multiple single-chain variable fragments (scFvs), single-chain Fabs (scFabs) or single variable domains (sVDs) linked by short peptide linkers. This class of BsAbs has a simple structure and smaller size compared to conventional IgG monoclonal antibodies. However, they are challenging to express, prone to aggregate, and have a shorter half-life due to the absence of Fc region.¹⁰ The appended-IgG bsAbs are constructed by adding scFv, scFab, or sVD to either the N- or C- terminus of the light chain (LC) or heavy chain (HC) polypeptides of a conventional IgG via short peptide linkers.^11,12 This class of bsAbs has a half-life comparable to that of the conventional IgGs, but its non-natural IgG structures may result in increased aggregation and a higher risk of developing anti-drug antibodies in patients.¹³ The IgG-like bsAbs, the third class of bsAb, retain many of the favorable traits of the conventional mono-specific IgG antibodies, such as good developability, favorable PK properties, and low immunogenicity risk.¹⁴ However, the production of IgG-like bsAbs requires expression of two distinct LCs and two distinct HCs for binding to two different epitopes. The main issue with this class of bsAbs is the mispairing of the antibody chains, such as undesirable HC homodimer formation and LC/HC mispairing. Strategies such as ‘knobs-into-holes’ (KIH)¹⁵ or electrostatic complementarity^16–18 can mitigate the HC homodimerization. For instance, the widely-used ‘knobs-into-holes’ technology can enhance the HC heterodimerization rate by up to 90% and meet clinical and market needs.¹⁵ The LC/HC mispairing can also be addressed through various strategies, such as common LC,^17,19 orthogonal Fab interface^20,21 or Crossmab, which exchanges the LC and HC domains within the Fab of one arm of the bsAb.²² Replacing one or both Fab arms with scFvs or scFab also can be used to avoid LC/HC mispairing.²³

The potency, specificity, and toxicity of the T-bsAbs are determined by various molecular designs of the bsAb format. These include binding affinity and valency, flexibility, and the geometry of the targeting moieties. For instance, low-affinity, monovalent CD3-binding is preferred for improved safety, as it has been shown to maintain functional activities while reducing cytokine release and reduced side effects.^24–27 However, selecting the correct binding affinity and valence against the TAA is more complex. If TAA expression is specific to the tumor cells, a high binding affinity is desired for enhancing tumor cell-killing potency.²⁸ However, TAAs in most solid tumors are also present at low levels on normal cells. Increasing the avidity while lowering the affinity of the TAA-targeting moiety in the T-bsAbs has been shown to increase the selectivity to targeted tumor cells without compromising potency, while sparing cells that express low amounts of TAA in normal human tissues.^28–30 Additionally, the flexibility of the TAA- and CD3-binding arms plays a crucial role in T-bsAbs’ tumor cell-killing potency. The efficiency of cytolytic immunological synapse formation between the T and tumor cells can be increased with more flexible linkers, depending on the target densities on cell surfaces and the cell-to-cell distance.³¹ Detailed discussions regarding the impact of molecular designs on the functional activities of BsAb have been previously published.^28,32 It is recognized in this literature that comparisons across various formats of bsAbs are difficult due to the different targeting antibodies and testing methods used in different studies.

The success of a bsAb in clinical trials is not only determined by its functional properties, but also by its manufacturability, such as high yield, low aggregation, and high stability. IgG-like bsAbs are known to have better manufacturability than fragment-based and appended-IgG bsAbs. The format design, using different building blocks and linkers, can affect the flexibility and geometry of the targeting moieties and, therefore, the target-binding avidity. However, there is limited information on the impact of molecular design on the manufacturability and functionality of a bsAb. The use of scFv in the development of bsAbs is popular due to the compact structure that makes it highly suitable for engineering.³³ Several T-bsAb formats, including Fab-FcK/scFv-FcH,³⁴ scFv-FcK/scFv-FcH,^35–37 FcK/scFv2-FcH,^37,38 Fab-FcK/FabscFv-FcH,³⁴ (FabscFv-FcW)2^39,40 and (scFv2-FcW)2⁴¹ have been reported in previous studies and/or used in clinical trials (Figure 1). A previous study reported that the scFv-FcK/scFv-FcH had better expression in a cell-free expression system than the FcK/scFv2-FcH.³⁷ However, the manufacturability and functional properties of different formats has not previously been systematically compared.

Figure 1. Schematic diagram of the eight T-bsAb formats tested in the current study.

Illustration of the 8 bsAb formats using circles and lines. Each bsAb targets both HER2 and CD3 and are of different orientation and valency. Top left are 3 antibodies in the “1+1” format; Bottom left are 3 antibodies in the “2+1” format; Right are 2 antibodies in the “2+2” format.

Formats tested in this study include “1 + 1”, “2 + 2” and “2 + 1” mixed-valency bispecific antibodies enabled by either a homodimeric Fc or a heterodimeric Fc using the ‘Knob in hole’ technology. One arm of the scFv2 only contains the hinge, C_H2, and C_H3 domains. Green arms target HER2, and Blue arms target CD3.

To fill this knowledge gap, we designed and produced eight KIH-based IgG-like T-bsAbs targeting human CD3 and human epidermal growth factor receptor 2 (HER2), a membrane tyrosine kinase that is overexpressed on the plasma membrane of breast cancer cells. We made these HER2×CD3 T-bsAbs using Fab and scFv moieties of anti-HER2 or -CD3 antibodies as the building blocks and produced them in Chinese hamster ovary (CHO) cells. We evaluated the effect of different flexibility, geometry, and valency of the HER2- or CD3-targeting moieties on the manufacturability and functionality of the T-bsAbs. Our findings indicated that the manufacturability of bsAbs was adversely affected as the number of scFv building blocks increased, while the functionality was affected by the combination of multiple factors, including the binding affinities and valency of targeting moieties, flexibility, and geometry of the formats. These findings provide valuable insights into the effect of format design on the optimal manufacture and function of T-bsAbs.

Results

Design of the HER2×CD3 T-bsAb in different formats

The HER2×CD3 T-bsAbs were constructed in eight formats using HER2- and CD3-targeting Fab and scFv as building blocks. The HER2-targeting moiety was either Fab or scFv derived from the monoclonal antibody trastuzumab, which specifically binds HER2.⁴² The CD3-targeting moiety was scFv derived from the bispecific T-cell engaging antibody (BiTE) pasotuxizumab.⁴³ Each T-bsAb consists of one or multiple Fab or scFv linked to either wild-type Fc (FcW), mutated Fc with its CH3 domain engineered to form a knob (FcK) or a hole (FcH) (Figure 1). VH and VL in each scFv were connected through a 15-amino acids (G4S)₃ linker. A 5-amino acids G4S linker was used for Fab-scFv, scFv-Fv, and scFv-Fc connections. The eight T-bsAbs were classified into three categories: 1) bivalent “1 + 1”, 2) trivalent “2 + 1”, and 3) tetravalent “2 + 2” formats, based on the valency of HER2- and CD3-targeting. The “1 + 1” and “2 + 1” bispecific antibodies are asymmetrical, in which FcK and FcH pairing is efficiently achieved through the KIH technology. The “2 + 2” bispecific antibodies were symmetrical, with two FcWs forming a homodimer. For the Fab-FcK/scFv-FcH and scFv-FcK/scFv-FcH formats in the “1 + 1” group, the two targeting moieties were linked to the Fc independently, making the two arms flexible. On the other hand, the FcK/scFv2-FcH format had its two scFvs, which target HER2 and CD3 respectively, linked in tandem with the FcH, making the interdomain region more rigid. The three bsAb formats in the “2 + 1” group were constructed by linking one extra anti-HER2 Fab or scFv to the N-terminus of the anti-CD3 scFv, thus increasing the avidity toward HER2. All six T-bsAbs in the groups of “1 + 1” and “2 + 1” formats used monovalent engagement to CD3, while the two “2 + 2” formats were designed to engage both HER2 and CD3 bivalently.

Comparison of different formats of T-bsAbs for expression

The eight T-bsAbs were produced in stably transfected pools in 14-day fed-batch cultures. Each stably transfected pool was generated by integrating a targeting vector carrying genes for a specific T-bsAb into a CHO master cell line (MCL) through the recombinase-mediated cassette exchange (RMCE) (Figure 2). The results for growth and productivity were collected from four fed-batch cultures of four independent stable pools for each bsAb. They were calculated as the average and standard error of the mean (SEM) for comparison (Figure 3). Among the eight T-bsAbs, the Fab-FcK/scFv-FcH format gave rise to the highest antibody titer of 613 ± 24 mg/L, followed by FcK/scFv2-FcH (564 ± 25 mg/L), Fab-FcK/FabscFv-FcH (523 ± 7 mg/L), (scFv2-FcW)2 (520 ± 9 mg/L) and Fab-FcK/scFv2-FcH (516 ± 16 mg/L). scFv-FcK/scFv-FcH (442 ± 28 mg/L), (FabscFv-FcW)2 (418 ± 9 mg/L), and scFv-FcK/scFv2-FcH (387 ± 5 mg/L) had the lowest titers (Figure 3a). Cell growth analysis showed that all the eight stably transfected T-bsAb-producing pools had a similar integrated viable cell density (IVCD) (Figure 3b), resulting in a similar trend between the specific productivities (qP) and titers for the different T-bsAbs (Figure 3c). These results suggest that the growth of antibody-producing CHO cells is similar amongst the eight different T-bsAb designs, which only marginally affect the yield of antibodies produced by the CHO cells.

Figure 2. Schematic diagram of the RMCE process and targeting vector design.

(Top) Diagram of the RMCE process, showing the replacement of the landing pad with the gene of interest using complementary FRT sites. (Bottom) Diagram of the components of the 8 bsAb formats.

(a) Overview of RMCE targeted integration platform for expressing the BsAbs in CHO cells. CHO-K1 cells were stably tagged with an Flp/FRT RMCE cassette (landing pad) via random integration. Cells bearing single landing pad was selected and used as master cell line(MCL). The targeting vector was designed to be promoter-less and consist of the FRT3/FRT sites (black and white triangles), gene-of-interest (GOI), and an IRES element to activate puromycin gene expression upon successful RMCE. (b) Schematic representation of targeting vectors carrying bsAb LC and HC genes linked by multiple wild-type EMCV IRES elements. hCMV: Human cytomegalovirus; EGFP; enhanced green fluorescence protein; NPT; neomycin phosphotransferase; pA: polyA tail; Pur: puromycin resistant gene; FLPe: flippase IRES: internal ribosome entry site; VH: variable heavy chain; VL: variable light chain; CHK: heavy chain-knob; FcH: fab-hole; DHFR: dihydrofolate reductase.

Figure 3. Comparison of titer, IVCD and qP of eight different HER2×CD3 T-bsAbs produced in CHO cells.

Bar Chart showing the titer. IVCD and qP of the bsAb cultures on day 14. The charts showed that all 8 bsAbs have similar IVCD however titer and qP is different for each format.

Comparison of the (a) titer, (b) IVCD and (c) qP of the 8 bsAb cultures harvested on day 14. Error bars show the standard error of the mean (SEM) of quadruplicate fed-batch cultures of four independent stable pools. IVCD: Integral viable cell concentration; qP: specific productivity, calculated by using day 14 titer/day 14 IVCD.

Comparison of different formats of T-bsAbs for purification efficiency

Aggregation is a critical quality attribute of antibody therapeutics as it may result in decreased bioactivity and increased immunogenicity.^44,45 The eight T-bsAbs produced in fed-batch cultures were purified using protein A affinity chromatography followed by cation-exchange (CEX) chromatography. The purified products at each step were subjected to size exclusion chromatography (SEC) to determine the aggregate levels (Table 1). The SEC profiles indicated that the eight protein A-purified T-bsAbs contained greater than 90% of monomer. This apparent high percentage of monomer detected in all samples could be due to the detection limit of SEC, in which different species in a sample cannot be discriminated due to their slight size difference. To achieve a better resolution of these monomer peaks and to further increase the purity of these T-bsAbs, a second purification step via CEX was performed. The CEX profiles revealed multiple peaks in these Protein A-purified samples, indicating multiple species with different charges existed in these samples (Figure 4). The predominant fraction in each CEX-purified sample was collected and subjected to SEC analysis. The recovery of the CEX product is calculated as the amount of purified product collected from the predominant peak after CEX divided by the total amount of product before CEX purification. The SEC chromatogram for the post-CEX samples showed a sharper and narrower peak (green) compared to those for the post-Protein A samples (black), indicating higher purity of the CEX-purified product, although the percentages remained similar to those obtained after Protein A purification (Table 1). As such, the CEX-purified products were used for the subsequent in vitro cell-based functional assays. The products from formats containing a higher number of scFv had more peaks and were more heterogeneous, while those from formats containing a lower number of scFv were more homogeneous (Figure 4). The number of scFv in a T-bsAb appears to be reversibly correlated to the product recovery. The T-bsAb (scFv2-FcW)2, containing four scFvs, had the lowest product recovery (32.7%). Those T-bsAbs containing two scFvs, such as scFv-FcK/scFv-FcH, FcK/scFv2-FcH, Fab-FcK/scFv2-FcH and (FabscFv-FcW)2, had increased product recovery ranging from 38.8% to 46.9%. Further reducing the number of scFvs to one, as seen in the formats Fab-FcK/scFv-FcH and Fab-FcK/FabscFv-FcH, resulted in even higher product recovery at 67.8% and 74.0%, respectively. The scFv-FcK/scFv2-FcH format, which contained three scFvs, was an outlier. It had a product recovery of 55.2%, which was lower than the format containing one scFv (Fab-FcK/scFv-FcH) and higher than the format containing four scFvs (scFv2-FcW)2 as expected. However, its product recovery was also higher than those formats containing two scFvs (scFv-FcK/scFv-FcH, FcK/scFv2-FcH, and Fab-FcK/scFv2-FcH).

Figure 4. AKTA profile during CEX purification and SEC profile of eight different HER2×CD3 T-bsAbs after an initial protein a step.

AKTA and SEC chromatograms arranged according to valency in the following order: “1+1” (left), “2+1” (middle) and “2+2” (Right). AKTA chromatogram showed multiple peaks for all 8 bsAb. After CEX purification, all 8 bsAb SEC profiles displayed narrower peaks.

AKTA chromatograms were shown on the left of each graph. SEC profiles before CEX (black) and after CEX (green) were shown on the right of each graph.

Table 1. Aggregation profile of T-bsAbs of different formats post Protein a and post CEX.

BsAb Format	Valency	(Fab : scFv)	Post Protein A purification			Post-cation exchange
			Purity			Purity			CEX Recovery (%)
			HMW (%)	Mono (%)	LMW (%)	HMW (%)	Mono (%)	LMW (%)
Fab-FcK/scFv-FcH	1 + 1	1 : 1	1.55	97.84	0.61	1.8	98.2	0	67.8
scFv-FcK/scFv-FcH		0 : 2	4.09	95.64	0.27	5.22	94.78	0	44.4
FcK/scFv2-FcH		0 : 2	5.81	92.16	2.02	8.29	90.87	0.85	44.1
Fab-FcK/FabscFv-FcH	2 + 1	2 : 1	1.46	95.59	2.95	1.48	97.83	0.69	74
Fab-FcK/scFv2-FcH		1 : 2	4.86	94.11	1.03	6.99	92.66	0.35	46.9
scFv-FcK/scFv2-FcH		0 : 3	1.81	97.19	1.00	3.25	96.17	0.58	55.2
(FabscFv-FcW)2	2 + 2	2 : 2	0.77	95.82	3.41	5.81	92.44	1.75	38.8
(scFv2-FcW)2		0 : 4	3.00	95.82	1.18	3.70	95.90	0.39	32.7

Thermostability of different formats of T-bsAbs

Differential scanning calorimetry (DSC) is commonly used to study the thermostability of multi-domain proteins, allowing the thermal unfolding of individual domains to be observed.⁴⁶ Here, we used trastuzumab IgG1 as a control to map the melting curves in the thermogram to the protein domains found in our T-bsAbs (Table 2). A detailed thermogram of each T-bsAb can be found in Supplementary Figure 1. The trastuzumab contained three peaks in the thermogram, which could be due to the unfolding of the Fc (CH2 and CH3 regions) and Fab segments.^47,48 However, all T-bsAbs had an additional peak at ~ 61°C, likely due to the scFv melting. The melting temperatures for the eight T-bsAbs were similar and slightly lower than trastuzumab, indicating they have similar thermostabilities to the standard IgG1 mAbs.

Table 2. Effect of different format on Thermal Melting Temperatures (T_m1, T_m2, T_m3, and T_m4) of Trastuzumab and bispecific antibodies as measured by DSC. The T_m value indicated the midpoint temperature of the thermal unfolding transition of antibody.

BsAb format	Tm1 (°C)	Tm2 (°C)	Tm3 (°C)	Tm4 (°C)
Trastuzumab	-	71.5	80	81.6
Fab-FcK/scFv-FcH	61.3	67.5	78.7	80.6
scFv-FcK/scFv-FcH	61.6	67.2	78.7	81.2
FcK/scFv2-FcH	61.0	66.9	78.8	81.1
FabFcK-FabscFv-FcH	61.1	66.0	79.1	80.7
Fab-FcK-scFv2-FcH	61.2	66.3	78.7	80.3
scFv-FcKscFv2-FcH	62.6	67.6	78.8	81.4
(FabscFv-FcW)2	60.5	66.5	79.0	80.5
(scFv2-FcW)2	60.4	65.6	81.0	84.2

Antigen bindings of different formats of T-bsAbs

The binding of a T-bsAb to CD3 and a TAA is crucial for determining its capability of killing the tumor cells and the accompanied risk of side effects, such as uncontrolled cytokine release. The binding kinetics of the eight T-bsAb against HER2 and CD3 were first analyzed using bio-layer interferometry (BLI). The trastuzumab IgG1 monoclonal was included for comparison (Table 3). The binding affinity of pasotuxizumab to CD3 was not measured due to the unavailability of this product in our lab and on the market. Sensorgrams of each bsAb binding to CD3E/D and HER2 can be found in Supplementary Figures 3 and 4, respectively. Considering the accuracy of BLI experiments, which typically exhibited variations between duplicate measurements of 2 to 3 fold difference, the BLI analysis revealed that the binding affinities of the eight T-bsAb variants to CD3 and HER2, respectively, were similar. The equilibrium dissociation constants (K_D) for CD3 binding were around 3 nM, while for HER2 binding, the K_D ranged from approximately 0.7 to 1.58 nM (Table 3).

Table 3. Binding affinity of different antibody formats to CD3E/D and Her2 via Octet and cell-based assays.

Antibody	Valency	Specificity and modality on each arm		Binding affinity via Octet, KD (nM)		Cell binding by flow cytometry IC50 (nM)
		FcK	FcH	Binding to HER2	Binding to CD3E/D	Binding to HER2	Binding to CD3E/D
Trastuzumab	2	Fab-Her2	Fab-Her2	0.81	-	-	-
Fab-FcK/scFv-FcH	1 + 1	Fab-Her2	scFv-CD3	0.99	3.87	126.7	42.3
scFv-FcK/scFv-FcH		scFv-Her2	scFv-CD3	1.58	3.56	224.6	45.3
FcK/scFv2-FcH		-	scFv-Her2	0.76	3.86	188.2	61.9
			scFv-CD3
Fab-FcK/FabscFv-FcH	2 + 1	Fab-Her2	Fab-Her2	0.79	3.35	20.7	70.9
			scFv-CD3
Fab-FcK/scFv2-FcH		Fab Her2	scFv-Her2	0.78	2.87	31.5	69.0
			scFv CD3
scFv-FcK/scFv2-FcH		scFv- Her2	scFv Her2	0.64	3.26	150.6	72.6
			scFv-CD3
(FabscFv-FcW)2	2 + 2	Fab-Her2	Fab-Her2	0.84	3.25	31.3	20.1
		scFv-CD3	scFv-CD3
(scFv2-FcW)2		scFv-Her2	scFv-Her2	0.69	3.37	228.8	40.9
		scFv-CD3	scFv-CD3

In addition to BLI, we evaluated the CD3 and HER2 binding of the T-bsAbs by flow cytometry using CD3-expressing Jurkat T and HER2-expressing SK-OV-3 ovarian cancer cells, respectively (Table 3 and Figure 5). Our T-bsAbs possessed either one or two paratopes for CD3 or HER2 antigens, unlike the mAbs with bivalent binding for a single antigen. Thus, indirect staining with a secondary antibody, such as an anti-Fc antibody, could not accurately reflect their binding to the individual antigen. To accurately determine their binding to CD3 or HER antigens, we adopted a competitive staining strategy. We first tested 3 commercially available anti-CD3 mAbs and found that clone UCHT1 effectively competed for CD3 binding in a dose-dependent manner with the anti-CD3 antibody (CD3a), which was derived from the pasotuxizumab and used in our T-bsAbs in scFv format (Supplementary Figure 4(a, b)). Then, we stained Jurkat T cells with UCHT1 mAb conjugated with allophycocyanin (APC) in the presence of various T-bsAbs in different concentrations. The APC fluorescence intensity, which was reversely correlated to the binding capacities of T-bsAbs, was measured with flow cytometry. The three T-bsAbs within the “1 + 1” format group displayed comparable CD3 binding capacities with an IC₅₀ ranging from 42.3 to 61.9 nM (Table 3 and Figure 5a, upper). The slightly lower binding capacity of the FcK/scFv2-FcH T-bsAb (61.9 nM) may be due to its less “exposed” CD3 scFv compared with the other two formats (Table 3 and Figure 1a). Interestingly, the CD3 binding capacities of the three T-bsAbs in the “2 + 1” format group were very similar, with IC₅₀ around 70 nM, but lower than that of three T-bsAbs of the “1 + 1” format (Figure 5a, middle). This could be due to the “less exposed” geometry of the monovalent CD3 binding arm. In contrast, BLI analysis could not detect these geometric differences (Table 3). The two T-bsAbs in the “2 + 2” format had better CD3-binding capabilities, probably due to having two CD3-binding scFvs. The (FabscFv-FcW)2 format had the best CD3 binding, with an IC₅₀ of 20.1 nM (Table 3 and Figure 5a, lower), indicating that overall CD3 binding can be improved by increasing valency, for instance, by converting from “2 + 1” to “2 + 2” format.

Figure 5. Flow cytometric analysis of antigen binding capacities of eight different HER2×CD3 T-bsAbs.

Flow cytometric analysis of CD3 binding (left) and HER2 binding (right) of the 8 bsAbs. CD3 binding of “1+1”, “2+1” and “2+2” formats were similar within the same group, with the 2+2 formats having the highest binding to CD3. The “2+1” and “2+2” formats have higher binding to HER2 as compared to “1+1” format.

The binding capacities of various HER2×CD3 T-bsAbs for antigens CD3 (a) and HER2 (b) antigens were determined by a competitive binding assay using flow cytometry. Jurkat T or SK-OV-3 cells were stained with diluted anti-human CD3 (clone UCHT1) or HER2 (Trastuzumab) antibodies conjugated with APC in the presence of various serially diluted T-bsAbs. The intensities of APC fluorescence were assayed on an LSR II (BD Pharmingen) and used for calculating the IC₅₀ with a four-parameter analysis model of GraphPad Prism 6. The smallest and highest mean fluorescence intensity (MFI) for each data set was normalized as 0% and 100%, respectively.

We also attempted to identify commercially available anti-HER2 mAbs that could compete with trastuzumab for HER2 binding, but our efforts were unsuccessful. As a result, we adopted an alternative strategy by conjugating trastuzumab with APC. We then stained the HER2-expressing SK-OV-3 cells with the trastuzumab-APC in the presence of various T-bsAbs at different concentrations and measured the APC fluorescence intensity, which we expected to be reversely proportional to their HER2-binding capacities. It has been reported that trastuzumab binds to HER2 in a bivalent manner and removal of one binding arm resulted in reduction of binding affinity.⁴⁹ This is consistent with our flow cytometry results, which showed that HER2 binding was largely proportional to the number of anti-HER2 moieties in the T-bsAbs. For instance, T-bsAbs of the “2 + 1” or “2 + 2” format manifested better HER2 binding compared to those with “1 + 1” format, except for (scFv-FcW)2, which had lower HER2 binding (Table 3 and Figure 5b). We found that the Fab format of anti-HER2 moiety performed better than the scFv format within each format group. For example, in the “1 + 1” format group, Fab-FcK/scFv-FcH T-bsAb (IC₅₀ = 126.7 nM) had a better HER2-binding than the scFv-FcK/scFv-FcH (IC₅₀ = 224.6 nM) and FcK/scFv2-FcH (IC₅₀ = 188.2 nM) antibodies, which had HER2 binding moieties in the scFv format (Table 3 and Figure 5b, upper panel). Similarly, the Fab-FcK/FabscFv-FcH (IC₅₀ = 20.7 nM) and (FabscFv-FcW)2 (IC₅₀ = 31.3) T-bsAbs, which had two Fab anti-HER2 moieties, showed the best HER2 binding capacities within the “2 + 1” and “2 + 2” format groups, respectively (Table 3 and Figure 5b, middle and lower panels).

Functional characterization of different formats T-bsAbs

We then assessed the functionalities, including T-cell activation, killing efficacies, cytokine production of the eight T-bsAbs and summarized the results in Table 4. The up-regulation of CD25 and CD69 triggered by various T-bsAbs in the presence of tumor target cells was measured to indicate T-cell activation. Total CD3⁺ T cells purified from human peripheral blood mononuclear cells (PBMCs) were co-cultured with HER2⁺ SK-OV-3 tumor cells for 16 hours in the presence of different T-bsAbs at various concentrations. The surface expression of CD25 and CD69 was measured by flow cytometry. We found that the FcK/scFv2-FcH T-bsAb in the “1 + 1” format group, which has HER2- and CD3-targeting scFvs linked in tandem to the FcH, was the weakest in triggering CD25 upregulation, with an EC₅₀ of about 3.2 × 10⁻¹ and 4.3 × 10⁻¹ nM for CD4⁺ and CD8⁺ T cells, respectively (Table 4 and Figure 6(a,b); left panels). On the other hand, the symmetric (FabscFv-FcW)2 T-bsAb in the “2 + 2” format group was the strongest in inducing CD25 expression, with an EC₅₀ of 3.5 and 4.4 × 10⁻⁴ nM, respectively, for CD4⁺ and CD8⁺ T cells (Table 4 and Figure 6(a,b); right panel). Four T-bsAbs, including the (scFv2-FcW)2 in the “2 + 2” format group, the Fab-FcK/scFv-FcH and the scFv-FcK/scFv-FcH in the “1 + 1” format group, and the Fab-FcK/FabscFv-FcH antibodies in the “2 + 1” format group, were also potent in triggering CD25 expression in T cells, with their EC₅₀ at 10⁻³ nM level ranging from 1.4 to 9.0 × 10⁻³ nM. The other two T-bsAbs, in the Fab-FcK/scFv2-FcH and the scFv-FcK/scFv2-FcH in the “2 + 1” format group, were moderately potent in inducing CD25 expression, with their EC₅₀ between 1.0 and 2.5 × 10⁻² nM. A similar potency trend in triggering CD69 expression was also evident for the eight different T-bsAbs, with the FcK/scFv2-FcH and the (FabscFv2-FcW)2 having the weakest and the strongest CD69-inducing efficacies for T cells, respectively (Table 4 and Figure 6(c,d)).

Figure 6. T cell activation upon stimulation of various HER2×CD3 T-bsAbs in the presence of tumor cells.

Chart of CD25 and CD69 upregulation in CD4 and CD8 T cells. The symmetric (FabscFv-FcW)2 in the “2+2” format has the strongest CD25 and CD69-inducing efficacy among the bsAbs tested.

CD25 (a,c) and CD69 (b,d) cell surface expression in CD4⁺ and CD8⁺ T cells treated with different HER2×CD3 bsAbs at various concentrations and in the presence of SK-OV-3 cells. Purified human T cells were co-cultured with SK-OV3 cells at an E:T ratio of 1:1 in the presence of the indicated concentration of HER2×CD3 T-bsAbs for 16 h. The cells were harvested and stained with fluorescein-conjugated anti-human CD4, CD8, CD25, and CD69 antibodies and subjected to flow cytometric analysis. The percentages of CD25⁺ or CD69⁺ cells of the CD4⁺ or CD8⁺ T cells were analyzed by Flow Jo. The EC₅₀ of CD25 and CD69 upregulation of the individual T-bsAb was calculated with a four-parameter logistic sigmoidal dose-response curve using GraphPad Prism. The data shown are representative of more than three independent experiments.

Table 4. Summary of results for cell-based functional characterization assays.

Antibody	Valency	T Cell activation EC50 (nM)				Killing potency EC50 (nM)	Cytokine production
		CD4+CD25+	CD8+CD25+	CD4+CD69+	CD8+CD69+		IL-2	IFNγ
Fab-FcK/scFv-FcH	1 + 1	1.37×10^−3	1.77×10^−3	1.44×10^−3	2.06×10^−3	1.35×10^−3	+++	+++
scFv-FcK/scFv-FcH		4.92×10^−3	9.21×10^−3	3.35×10^−3	4.42×10^−3	3.79×10^−3	++	++
FcK/scFv2-FcH		3.18×10^−1	4.34×10^−1	2.03×10^−1	2.36×10^−1	1.88×10^−1	+	+
Fab-FcK/FabscFv-FcH	2 + 1	2.59×10^−3	2.25×10^−3	2.50×10^−3	2.91×10^−3	1.48×10^−3	+++	+++
Fab-FcK/scFv2-FcH		1.02×10^−2	2.47×10^−2	7.93×10^−3	1.32×10^−2	2.95×10^−3	+	+
scFv-FcK/scFv2-FcH		1.51×10^−2	1.98×10^−2	1.23×10^−2	1.53×10^−2	3.12×10^−2	+	+
(FabscFv-FcW)2	2 + 2	2.75×10^−4	4.40×10^−4	2.75×10^−4	3.95×10^−4	2.63×10^−4	+++	+++
(scFv2-FcW)2		3.50×10^−3	5.82×10^−3	3.50×10^−3	4.49×10^−3	6.29×10^−3	++	++

The potencies of different bsAbs relative to each other for inducing cytokine production were indicated as strong (+++), moderate (++) and weak (+).

We also evaluated the ability of the T-bsAbs to induce T cell-mediated target tumor cell-killing using a luciferase-based cytotoxicity assay. Human CD3⁺ total T cells were purified and co-cultured with firefly luciferase-expressing HER2⁺ SK-OV-3 cells for 48 h in the presence of different T-bsAbs at various concentrations. The luciferase activity of the cultured SK-OV-3 cells was measured after the addition of the substrate luciferin and used as a measure of cell viability. We first noticed that all eight T-bsAbs could achieve the maximum killing toward the SK-OV-3 cells (~80 to 90%), but at different concentrations (Figure 7a). Among the eight T-bsAbs, the FcK/scFv2-FcH in the “1 + 1” format group was the least potent antibody with the highest killing saturation concentration at ~10 nM (Figure 7a, left panel). The Fab-FcK/scFv-FcH and scFv-FcK/scFv-FcH in the “1 + 1” format group and the (FabscFv-FcW)2 and (scFv2-FcW)2 in the “2 + 2” format group were most potent in killing target tumor cells, with saturation concentrations at ~0.1 nM, which were approximately 100 times lower than that of the FcK/scFv2-FcH (Figure 7a, left and right panels). The other three T-bsAbs in the “2 + 1” format group had intermediate killing saturation concentrations at ~1 nM (Figure 7a, middle panel).

Figure 7. Tumor cell-killing and cytokine production by T cells induced by eight different HER2×CD3 T-bsAbs.

(Top) Chart of T cell-mediated killing showing maximum killing achieved at different bsAb concentrations. (Bottom) Bar chart showing IL2 and IFNγ production in T cells after stimulation with the 8 bsAbs in the presence of SK-OV-3 cells. (FabscFv-FcW)2 in the “2+2” format has the greatest killing efficacy and induced the highest production of IL2 and IFNγ in T cells.

(a) Analysis of T-bsAb-mediated killing of SK-OV-3 cells by T cells. Purified CD3⁺ human T cells were co-cultured with Her2⁺ SK-OV3-luciferase cells at an E:T ratio of 1:1 in the presence of the indicated concentration of various HER2×CD3 T-bsAb for 48 hours. The luciferase activity was used as the surrogate for cell viability and determined by measuring the luminescence in triplicates after adding the substrate luciferin. The EC₅₀ of the tumor cell-killing of the T-bsAb was calculated with a four-parameter logistic sigmoidal dose-response curve using GraphPad Prism. (b,c) Production of IL-2 (b) and INFγ (c) by T cells upon stimulation of T-bsAbs. Human T cells were co-cultured with HER2⁺ SK-OV-3-luc cells in the presence of the indicated concentrations of various HER2×CD3 T-bsAb for 48 h. IL-2 and IFNγ in the culture supernatant were determined using ELISA in duplicates. The data shown are representative of more than three independent experiments.

The eight T-bsAbs also exhibited distinct killing efficacies, with the FcK/scFv2-FcH having the lowest killing efficacy, with an EC₅₀ of ~0.188 nM (Table 4 and Figure 7a). On the other hand, the (FabscFv-FcW)2 in the “2 + 2” format group showed the greatest killing efficacy, with an EC₅₀ of 2.63 × 10⁻⁴ nM, which was ~ 700 times lower than that of the FcK/scFv2-FcH. Four T-bsAbs, including the Fab-FcK/scFv-FcH and the scFv-FcK/scFv-FcH in the “1 + 1” format group, the Fab-FcK/FabscFv-FcH T-bsAb in the “2 + 1” format group, and the (scFv2-FcW)2 T-bsAb in the “2 + 2” format group, also showed high killing efficacy with EC₅₀ ranging from 1.35 × 10⁻³ to 6.29 × 10⁻³ nM. While the other two T-bsAbs, the Fab-FcK/scFv2-FcH and scFv-FcK/scFv2-FcH, in the “2 + 1” format group had moderate efficacy, with EC₅₀ of ~ 3×10⁻² nM. Noticeably, the tumor cell-killing efficacy trend of the eight T-bsAbs is similar to that of the CD25 and CD69 upregulation (Table 4).

We also examined the cytokine production by T cells stimulated with the eight T-bsAbs in the presence of SK-OV-3 cells using ELISA. The ability of T-bsAbs to induce interleukin-2 (IL-2) and interferon gamma (IFNγ) production was strongly correlated with their efficacy in triggering T-cell activation and tumor cell killing. The FcK/scFv2-FcH T-bsAb, which showed the weakest efficacy in inducing CD25 and CD69 expression in T cells and mediating tumor cell-killing, induced the lowest IL-2 and IFNγ production at the concentrations of 10 nM, 1 nM, and 0.1 nM (Table 4 and Figure 7(b,c)). scFv-FcK/scFv2-FcH and Fab-FcK/scFv2-FcH in the “2 + 1” format group, which were moderately efficacious in triggering T cell activation and tumor cell-killing, also induced less IL-2 and IFNγ production compared to the other five T-bsAbs, particularly at 1 nM and 0.1 nM concentrations.

Our findings indicate that among the eight T-bsAbs examined, the asymmetric T-bsAb FcK/scFv2-FcH, which bears HER2- and CD3-targeting moieties in the scFv configuration and linked in tandem with the FcH, displays the weakest functionalities, such as inducing T cell activation, T cell-mediated tumor cell-killing and cytokine production by T cells. In contrast, the symmetric (FabscFv-FcW)2 in the “2 + 2” format group, which has two HER2-binding moieties in the Fab and two CD3-binding moieties in the scFv designs, displays the greatest functionalities, including inducing T cell activation and T cell-mediated tumor cell-killing. Notably, the three T-bsAbs in the “2 + 1” format group and the other two T-bsAbs in the “1 + 1” format group have relatively comparable efficacies in inducing T cell activation and T cell-mediated tumor cell killing, even though the former three T-bsAbs contain two HER2-targeting moieties.

Discussion

Both good manufacturability and desirable functional properties are essential for the development of bsAbs. While numerous studies have previously focused on the impact of molecular designs, such as the valency, binding affinity, flexibility, and geometry of the different antigen-targeting moieties on the functional properties of bsAbs, the impact of these factors on manufacturability has not been systematically explored. As the IgG-like bsAb format is the most favored format due to the conventional IgG structure advantage,⁵⁰ we designed eight IgG-like T-bsAbs using Fab and scFv as building blocks with different flexibility, geometry, and valency. These eight T-bsAbs were first evaluated for their manufacturability in CHO cells, followed by their functionalities using in vitro cell-based assays. We found no clear relationship between the molecular designs and expression levels. All eight T-bsAbs also showed similar thermal stability regardless of their different designs, suggesting that the property of individual building blocks may determine the entire molecule’s thermostability. However, the abundance of scFv in a T-bsAb appears to affect the heterogeneity of purified products. T-bsAbs containing more scFvs tend to have multiple peaks in the CEX profile. A similar observation has been reported in a previous study.⁵¹ This behavior could be due to the use of flexible linkers that allow the scFv to adopt either a closed or open configuration, resulting in the reversible interconversion of protein configuration that interacts differently with the chromatographic matrix. Although our unpublished results indicated the different fractions in the CEX-purified products had similar functional properties based on in vitro cell-based assays, a higher homogeneity of purified T-bsAbs is desirable for reducing batch-to-batch variations during manufacturing.

Functional properties are equally important for successful bsAb development. The binding capacity of the bsAbs to their antigens is critical for their functions. In our study, we used BLI assay and cell-based flow cytometry to measure the affinity and avidity of different HER2×CD3 T-bsAbs. Despite comparable affinities, the eight T-bsAbs exhibited different avidities, probably due to the different experimental setups of the two assays. In the BLI assay, antibodies were bound onto the chip probing the antigen of interest flowing over the surface of the chip. Also, the binding affinity was determined by measuring the monovalent binding to the antigen. However, the cell-based flow cytometry measured direct interaction between the antibodies with cell surface-expressed antigens, which might differ conformationally from the soluble antigens used in the BLI assay. In addition to affinity, the antigen-binding avidity of a T-bsAb is influenced by other factors, including valency and geometry.² Here, we observed that the CD3-binding avidities of the eight HER2×CD3 T-bsAb were similar, with the two bivalent anti-CD3 scFv-containing antibodies in the tetravalent “2 + 2” format group exhibiting slightly stronger binding than the other six monovalent anti-CD3 scFv-containing antibodies. These results suggest that the geometry of the anti-CD3 moiety might be more important than the valency in determining the CD3-binding avidity of T-bsAbs. Since all eight T-bsAbs have the same anti-CD3 moiety in the same scFv format, they have similar CD3-binding avidities.

Similarly, the geometry of anti-HER2 moiety also plays a more significant role than the valency in determining HER2 binding. The Fab moiety confers stronger binding avidity than the scFv molecule (Table 3 and Figure 6). For example, the HER2-binding avidity of (FabscFv-FcW)2, which has anti-HER2 moiety in the Fab configuration, is about 9-fold stronger than that of (scFv2-FcW)2, which has anti-HER2 moiety in the scFv geometry, despite both being bivalent for anti-HER2 moiety (Figure 6). Among the three T-bsAbs in the “2 + 1” formats, the HER2-binding avidity correlates with the number of Fabs they harbor. Additionally, the Fab-FcK/scFv-FcH format, with HER2-binding moiety in the Fab configuration, exhibited a stronger HER2-binding avidity than the other two formats that have HER2-binding moieties in the scFv design within the “1 + 1” format group. The reduced HER2-binding capability of the scFv-containing T-bsAbs that contain anti-HER2 scFv arms could be due to the loss of affinity when the complementarity-determining region from Fab was grafted onto a scFv scaffold.^52,53 This may explain why the (FabscFv-FcW)2 T-bsAb, which is a bivalent and in the favorable Fab configuration for HER2 moiety, exhibited the strongest HER2-binding avidity among all eight T-bsAbs.

We also find that the T-bsAbs’ antigen-binding avidity, not the affinity, has a greater impact on their functionality. Despite similar binding affinities for both antigens, the eight T-bsAbs exhibit distinct functionalities due to their varying avidities for CD3 and HER2 binding (Tables 3 and 5). Binding avidities for both targets are crucial for a T-bsAb’s functionality. For instance, the “2 + 1” format group’s three T-bsAbs have the same CD3-binding avidities but differ in HER2-binding, resulting in distinct tumor cell-killing and T-cell activation (Figures 6 and 7). In the tetravalent “2 + 2” format group, the (FabscFv-FcW)2 T-bsAb has stronger binding avidities for CD3 (2-fold) and HER2 (7-fold) than that of the (scFv2-FcW)2 T-bsAb, leading to its ~ 24-fold greater tumor cell-killing potency.

The molecular design and geometry of T-bsAbs also affect their functionalities. For example, the three T-bsAbs in the bivalent “1 + 1” format group have similar binding avidities for CD3 and HER2, but exhibit dramatically different potencies of inducing tumor cell-killing and T-cell activation (Tables 3 and 5). The FcK/scFv2-FcH T-bsAb’s tumor cell-killing potency is ~ 50- to 140-fold weaker than the other two T-bsAbs in the same group, possibly due to different geometry impacting the interaction between T and tumor cells and the formation of the immunological synapse.⁵⁴ Previous studies have demonstrated that bsAbs functionality is influenced by the size of target antigens and the distance between epitopes targeted by the bsAb.^55,56

In summary, we have designed eight different T-bsAbs incorporating various building blocks, valency, geometry, and flexibility. We found the bivalent Fab-FcK/scFv-FcH and the trivalent Fab-FcK/FabscFv-FcH T-bsAb to be the optimal formats for developing Ig-like T-bsAbs in terms of both manufacturability and functionality. Aggregates may result in hyper-potency due to their additional avidity. The stronger potency observed for these two formats should not be attributed to the aggregated molecules, as their purified products had the lowest percentage of high molecular weight (HMW) species among the eight T-bsAbs. Although the tetravalent (FabscFv-FcW)2 T-bsAb demonstrated the strongest potency in terms of T cell activation, tumor cell killing and cytokine production, it is not considered an optimal format due to its high level of heterogeneity after protein A purification and low CEX purification recovery efficiency (Figure 4). Our work provides a starting point for selecting the optimal format for T-bsAb development. Further optimization can consider additional factors such as the affinity of each single targeting antibody, the orientation of VL and VH in scFv, the order of the two scFvs in tandem scFv construct, linker length and the density of TAAs in tumor cells.^15,53,57,58 SK-OV-3 cells used in this study have approximately 1.38 × 10⁶ HER2 receptors on the cell surface and are considered to have high HER2 density relative to other breast cancer cells.⁵⁹ By using cancer cells with different HER2 densities, a previous study has demonstrated that the “2 + 1” format HER2×CD3 T-bsAb is more favorable than the “1 + 1” format for discriminating tumor cells with low TAA and normal cells.²⁹ Testing the eight T-bsAbs designed in this work using tumor cells with different HER2 densities may provide further insights for designing T-bsAbs with improved functionalities.

Materials and methods

Vector construction

The eight bispecific antibodies consisted of nine chains: a standard trastuzumab LC and a mutated trastuzumab HCK (containing engineered CH3 to form a knob), which were assembled to form Fab-FcK, trastuzumab scFv (VH-(G4S)3-VL) linked to FcK (containing engineered CH3 to form a knob) to form scFv-FcK, the anti-CD3 scFv (VH-(G4S)3-VL) linked to FcH (containing engineered CH3 to form a hole) to form scFv-FcH, FcK without the addition of any targeting moieties, trastuzumab Fab linked to the anti-CD3 scFv and then to either FcH or FcW to form FabscFv-FcH and FabscFv-FcW, respectively, and trastuzumab scFv linked to the anti-CD3 scFv and then to the FcH or the wild-type Fc (FcW) to form scFv2-FcH and scFv2-FcW, respectively. Engineering CH3 to form a knob or hole was designed based on the previous study.¹⁵ The G4S linker was used between Fab and scFv, scFv and scFv, and Fab/scFv/FabscFv/scFv2 and Fc. The engineered and wild-type Fc consisted of IgG1 positions 221 to 447 based on the Eu numbering system. A basic targeting vector carrying the expression cassette of FRT3-VL1CL-IRES-VH1CHK-IRES-VH2VL2FcH-IRES-FRT was constructed by Genescript. VL1 and VH1 represented the trastuzumab LC and HC variable regions, respectively. VL2 and VH2 represented the anti-CD3 LC and HC variable regions, respectively. The cDNAs of trastuzumab VL and VH and anti-CD3 VL and VH were designed based on the amino acid sequences of trastuzumab and pasotuxizumab published in the international ImMunoGeneTics information system (IMGT). CHK represented HC constant region with its CH3 domain engineered to form a knob. IRES represented the wild-type encephalomyocarditis virus (EMCV) internal ribosome entry site. Four unique restriction enzyme sites, EcoRV, NotI, NruI, and EcoRI, were included between different components. In addition, two restriction enzyme sites, MluI and MfeI, were created by mutating the six bases in front of the 10th ATG in the EMCV IRES upstream of VH1CHK and VH2VL2Fc, respectively, during gene synthesis. Targeting vectors expressing other bispecific antibodies were subsequently constructed by inserting the synthesized components between EcoRV and NotI, MluI and NruI, and MfeI and EcoRI. Synthesis of dihydrofolate reductase (DHFR), FcK, VH1CH1-VH2(G4S)3VL2-G4S-FcH, VH1(G4S)3VL1-G4S-VH2(G4S)3VL2-G4S-FcH, VH1CH1-(G4S)3-VH2(G4S)3VL2-G4S-FcW, VH1(G4S)3VL1-G4S-VH2(G4S)3-VL2-G4S-FcW and cloning them into the basic targeting vector were all done by Genscript. The sequences of the DHFR, neomycin-phosphotransferase (NPT), enhanced green fluorescence protein (EGFP), wild-type IRES, IRESv18, FRT, and FRT3 were described previously.^60–64 The vector expressing enhanced FLP recombinase (FLPe), mCMV-FLPe, was constructed in our previous study.⁶⁵

Production of T-bsAbs

The eight bispecific antibodies were produced in stably transfected pools of CHO K1 cells in fed-batch cultures. The stably transfected pools were generated through recombinase-mediated-cassette-exchange (RMCE) by co-transfection of the MCL with an appropriate targeting vector expressing a bispecific antibody and the vector expressing FLPe (Figure 2a). Detailed protocols for developing MCL, generation of stably transfected pools, and production in fed-batch cultures were described in our previous study.⁶⁶ In brief, 14-day fed-batch cultures were performed for each stable pool in 50 mL tube spins (TPP) in the humidified Kuhner shaker (Adolf Kühner AG) with 8% CO₂ at 37°C. To start the fed-batch, 30 mL of cultures with a viable cell density of 3 × 10⁵ cells/mL were inoculated at day 0. The growth medium was Protein free and consisted of HyQ PF (GE Healthcare Life Sciences) and CD CHO (Thermo Fisher Scientific) at 1:1 ratio and supplemented with 1 g/L sodium bicarbonate (Sigma), 6 mM glutamine (Sigma), 0.1% Pluronic F-68 (Thermo Fisher Scientific). Ex-Cell Advanced CHO Feed 1 (with glucose) (SAFC, Sigma) was added at 10% of the culture volume every other day from 3 to 11. Cell density, viability, and antibody titer of each culture were monitored on days 3, 5, 7, 9, 11, and 14. The integrated viable cell density (IVCD) was determined based on the trapezoidal method. The specific antibody productivity (qP) of the cultures was calculated as antibody concentration at day 14 divided by the IVCD.

Antibody purification

BsAbs in the culture supernatant were purified using MabSelect SuRe Protein A column (GE Healthcare) on a GE AKTA explore 100 (GE Healthcare). The Protein A purified BsAbs were further purified via CEX chromatography. All CEX chromatographic runs were performed at 2 min residence time on an ÄKTA^TM Avant 25 (Cytiva), using 1 mL SP Sepharose^TM High-Performance resin (Cytiva) packed in a Tricorn^TM 5 column (Cytiva) with a bed height of 5.1 cm. The column was equilibrated with 20 mM acetic acid, pH 5.0 (5 CV). The eluate obtained from Protein A chromatography was adjusted to pH 5.0 before loading onto the column, followed by a wash step with equilibration buffer (5 CV) and a linear gradient elution from 20 mM acetic acid, pH 5.0, 0 mM sodium chloride (NaCl) to 20 mM acetic acid, pH 5.0, 500 mM NaCl in 50 CV, with a 13 CV hold at the end. 0.5 CV fractions were collected, and the relevant fractions were pooled together for subsequent analysis.

The analysis of antibody concentration and purity was performed by HPLC-SEC using a TSK_gel G3000SW_XL column (7.8 mm i.d. x 30 cm; Tosoh Bioscience) at a flow rate of 0.6 mL/min, with a mobile phase comprising of 50 mM MES, 200 mM L-arginine, 5 mM EDTA, 0.05% sodium azide (w/w), pH 6.5. The resultant concentrations were determined by comparing the area of the peaks observed at 280 nm UV absorbance to a calibration curve obtained using standard samples. The percentage of HMW and low molecular weight (LMW) species present in the sample was determined based on the area of the peaks that eluted before and after the monomeric peak, respectively.

Differential scanning calorimetry

BsAb samples were diluted in phosphate-buffered saline, pH 7.4, to a concentration of 1 mg/ml and sent to Lakepharma (Ca, USA) for DSC thermal denaturation experiments. Each T-bsAb sample was submitted to the Nano DSC system (TA Instrument) for analysis. A temperature ramp of 1°C/min was performed with monitoring from 25°C to 100°C. The Tm values were calculated after deconvolution using the Nano DSC software.

Bio-layer interferometry assay

BsAb samples were sent to Lakepharma (CA, USA) for binding affinity assay. Binding experiments were performed on Octet Red at 25°C The antibodies were loaded onto Anti-hIgG Fc Capture (AHC) biosensors for 300 seconds. The ligand-loaded sensors were dipped into a 3-fold series dilution (starting at 300 nM) of the antigens (CD3 E/D and HER2) for 180 seconds, followed by dissociation for 400 seconds. Kinetic constants were calculated using a monovalent (1:1) binding model. The antigens used in this binding assay were purchased from ACROBiosystems: Human CD3 epsilon CD3 delta Heterodimer Protein, His Tag-Free (Cat# CDD-H52W1), Human HER2/ErbB 2 Protein, His Tag (Cat# HE2-H5225).

Competitive binding assay

The binding of CD3 and HER2 by various HER2×CD3 T-bsAbs was indirectly determined by a competitive binding assay using flow cytometry. To examine the binding of CD3, Jurkat T cells were stained with diluted mouse anti-human CD3 (clone UCHT1) mAb conjugated with APC in the presence of various serially diluted T-bsAbs. To assess HER2 binding, the anti-human HER2 mAb, trastuzumab, was first labeled using the LYNX Rapid APC Antibody Conjugation Kit (Biorad) following the manufacturer’s instruction. The SK-OV-3 cells were stained with the diluted anti-human HER2 (trastuzumab) mAb conjugated with APC in the presence of various serially diluted T-bsAbs. The intensities of APC fluorescence were assayed on an LSR II (BD Pharmingen) and analyzed with BD FACSDiva software V8. The IC₅₀ value was calculated using a four-parameter analysis model of GraphPad Prism 6. The smallest and highest mean fluorescence intensity (MFI) for each data set was normalized as 0% and 100%, respectively.

In vitro T cell activation

Human PBMCs were prepared from healthy donors using Ficoll-Paque PLUS (Cytiva) as described previously.⁶⁷ Total T cells were further isolated from the PBMCs using a Pan T cell isolation kit (Miltenyi Biotec) according to the manufacturer’s protocol. The purity of the purified T cells (CD3⁺CD56⁻) was routinely higher than 95%, as determined by flow cytometric analysis. The SK-OV-3 and purified T cells were co-cultured at effector (E): target (T) cell ratio of 1:1 with various anti-HER2×CD3 T-bsAbs at the indicated concentrations for 16 hours at 37°C, supplemented with 5% CO₂ in a humidified incubator. The cells were harvested and stained with fluorescein-conjugated anti-human CD4 (BD; SK3), CD8 (BD; SK1), CD25 (BD; M-A251), and CD69 (BD; FN50) antibodies for 30 minutes on ice in the dark. The upregulation of CD25 and CD69 was assessed by flow cytometric analysis. Dead cells were excluded by DAPI staining.

T cell-mediated killing of target tumor cells

The SK-OV-3 cells were transduced with pLenti-CMV-Puro-LUC (Addgene) lentivirus and then selected in RPMI 1640 medium containing 4 μg/mL puromycin (Invivogen) over 10 days to get the stably transfected luciferase-expressing SK-OV-3 (SK-OV-3-luc) cell line. The SK-OV-3-luc and the purified T cells were seeded in 96-well plates in duplicate at an E:T ratio of 1:1 in RPMI cell culture medium containing 10% fetal bovine serum in the presence of various anti-HER2×CD3 T-bsAbs at indicated concentrations and incubated for 48 hours at 37°C, supplemented with 5% CO₂ in a humidified incubator. The viability of the target cells was assessed using Bright Glow Assay System according to the manufacturer’s protocol. The percentage of killing target cells by T cells was calculated as described previously.⁶⁷

Cytokine production of T cells

The SK-OV-3-luc and purified T cells were cultured at an E:T ratio of 1:1 with the various bsAbs at the indicated concentrations at 37°C, supplemented with 5% CO₂ in a humidified incubator. The culture supernatant was collected after 48 hours. The T cell-secreted IFNγ and IL-2 were measured by the enzyme-linked immunosorbent assay (ELISA) using human IFNγ (Biolegend, 430115) and human IL-2 (Biolegend, 431816) detection kits according to the manufacturer’s protocol.

Abbreviation

Table

APC	Allophycocyanin
BiTE	Bispecific T-cell engaging antibody
BLI	Bio-Layer Interferometry
bsAb	Bispecific antibodies
CEX	Cation-exchange
CHO	Chinese hamster ovary
DHFR	Dihydrofolate reductase
DSC	Differential scanning calorimetry
EMCV	Encephalomyocarditis virus
EGFP	Enhanced green fluorescence protein
ELISA	Enzyme-linked immunosorbent assay
Fab	Fragment antigen-binding
Fc	Fragment crystallizable
FcH	Fc Hole
FcK	Fc Knob
FcW	Wildtype Fc
FLPe	Enhanced FLP recombinase
HC	Heavy chain
hCMV	Human cytomegalovirus
HER2	Human epidermal growth factor receptor 2
HMW	High molecular weight
IFNγ	Interferon gamma
IgG	immunoglobulin G
IL-2	Interleukin-2
IMGT	International ImMunoGeneTics information system
IRES	Internal ribosome entry site
KD	Equilibrium dissociation constant (KD
KIH	Knobs-into-holes
LC	Light chain
LMW	Low molecular weight
Mabs	Monoclonal antibodies
MCL	Master cell line
MFI	Mean fluorescence intensity
NPT	Neomycin-phosphotransferase
PBMC	Peripheral blood mononuclear cells
PK	Pharmacokinetics
qP	Specific productivity
RMCE	Recombinase-mediated cassette exchange
scFabs	Single-chain Fabs
scFv	Single-chain variable fragment
SEC	Size exclusion chromatography
SEM	Standard error of the mean
sVD	Single variable domains
TAA	Tumor-associated antigen
T-bsAbs	T-cell-engaging bispecific antibodies

Author contributions

Y.Y.S., S.L.X. and K.P.L. conceived and designed the project. H.P.L, F.B.M, S.W.C., Y.H., W.Z. and J.X.H carried out the experiments and analyzed the data. Y.Y.S., S.L.X., H.P.L, S.W.C., Y.H. and W.Z. wrote and reviewed the manuscript. All authors approved the final manuscript.

Acknowledgments

This work was supported by the Biomedical Research Council (BMRC) of the Agency for Science, Technology and Research (A*STAR), Singapore and Bioprocessing Technology Institute, A*STAR, Singapore.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2023.2231129.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.