RUBY® – a tetravalent (2+2) bispecific antibody format with excellent functionality and IgG-like stability, pharmacology and developability properties

ABSTRACT

Despite the large number of existing bispecific antibody (bsAb) formats, the generation of novel bsAbs is still associated with development and bioprocessing challenges. Here, we present RUBY, a novel bispecific antibody format that allows rapid generation of bsAbs that fulfill key development criteria. The RUBY^TM format has a 2 + 2 geometry, where two Fab fragments are linked via their light chains to the C-termini of an IgG, and carries mutations for optimal chain pairing. The unique design enables generation of bsAbs with mAb-like attributes. Our data demonstrate that RUBY bsAbs are compatible with small-scale production systems for screening purposes and can be produced at high yields (>3 g/L) from stable cell lines. The bsAbs produced are shown to, in general, contain low amounts of aggregates and display favorable solubility and stress endurance profiles. Further, compatibility with various IgG isotypes is shown and tailored Fc gamma receptor binding confirmed. Also, retained interaction with FcRn is demonstrated to translate into a pharmacokinetic profile in mice and non-human primates that is comparable to mAb controls. Functionality of conditional active RUBY bsAbs is confirmed in vitro. Anti-tumor effects in vivo have previously been demonstrated, and shown to be superior to a comparable mAb, and here it is further shown that RUBY bsAbs penetrate and localize to tumor tissue in vivo. In all, the RUBY format has attractive mAb-like attributes and offers the possibility to mitigate many of the development challenges linked to other bsAb formats, facilitating both high functionality and developability.

KEYWORDS:

• Antibody format
• bispecific antibodies
• IgG
• immuno-oncology
• RUBY®
• tetravalent

Introduction

Bispecific antibodies (bsAbs) are a promising drug modality capable of simultaneously binding two different epitopes either on the same or different antigens.¹ BsAbs are regarded as promising options with the potential to provide biological advantages over monospecific monoclonal antibodies (mAbs). They hold the capacity to induce novel biology by, for example, recruitment of effector cells to cancer-associated epitopes, such as with CD3 bsAbs,^2–4 dampening of cancer cell escape mechanisms via concurrent regulation of two unique signaling pathways^5–8 and tumor conditionally active immunomodulators.^9,10 To date, 14 bsAb drugs have been granted marketing approvals^11,12 of which 9 are approved in cancer indications, and importantly, 3 of these are approved for solid tumor indications, further demonstrating the potential of bsAb drugs. Outside of oncology, there is also an increasing interest in developing bsAbs in other areas of immunomodulating therapies, including antibacterial and antiviral, chronic inflammatory and autoimmune diseases.¹³

Despite these successes and the many formats engineered to date,¹⁴ generation of bsAbs with high purity and stability can be challenging.¹⁵ This often results in the need for multi-dimensional engineering approaches, requiring not only a large number of binders against each target, but often also more than one bsAb format to generate a development candidate with the appropriate functional, physicochemical, pharmacokinetic (PK) and pharmacodynamic characteristics.^1,16,17 Alternatively, substantial efforts are put into lead optimization or optimization of various parameters within chemistry, manufacturing and controls (CMC) upstream and downstream processes in order to obtain developable and stable bsAbs.¹⁸

Following the six criteria critical for clinical development and commercial manufacturing of a good therapeutic bsAbs described by Nie et al.,¹⁹ we identified attributes that should be inherent for an optimal bispecific antibody format. These are presented in Table 1 and have formed the basis for a novel bsAb format that we named RUBY^TM. The main aim was to generate a format that contains the characteristics of the full-length IgG, in terms of stability, manufacturability, half-life and functionality, while having a 2 + 2 geometry, that is, enabling engagement of two targets bivalently at the same time.

Table 1. The six criteria critical for clinical development and commercial manufacturing of a good therapeutic bispecific antibody as highlighted by Nie et al.^19.

The 6 criteria for advancing compounds to clinical development and commercial manufacturing	Preferred format intrinsic features/properties of the bispecific format to best meet the criteria
Clinical efficacy	Target binding valency, possibility to make use of avidity
	Fc function opportunities
Appropriate safety profile	Good physiological stress, low aggregation
Favorable PK/PD properties	mAb like serum half-life, FcRn interaction
	Good tumor/tissue penetration
Appropriate physicochemical properties	Good thermostability
	Good colloidal stability
	Low viscosity
	No additional post-translational modification sites
Scalable manufacturing	High expression
	mAb like CLG and upscaling process
	Protein A purification possible
	Good freeze/thawing stability
	Good shear stress stability
	Good stability at low pH
Minimal or no immunogenicity	Consists of naturally existing building blocks
	Framework mutations carry low risk for immunogenicity
	Immune-inert linkers

RUBY bsAbs belong to the appended IgG-like class of bsAb formats. The Fab domains are linked through their light chains to the C – terminal end of IgG molecules and mutations in the interfaces between light and heavy chains are introduced to promote heterodimerization of heavy and light chains and thus minimize formation of aggregates and soluble antigen-binding fragments (Fabs) (Figure 1). In addition, mutations have been introduced to minimize the inherent risk for immunogenicity. In all, our results demonstrate that RUBY is a tetravalent bispecific antibody format with many IgG-like properties, such as stability, manufacturability, and functionality.

Figure 1. RUBY format structure. RUBY bsAbs comprise of an IgG (dark green, light green and gray) coupled to two fabs (dark and light turquoise). Three separate chains make up a construct, (1) a long chain that consists of the IgG heavy chain (dark green and gray) and the light chain of the additional fab fragment (light turquoise), (2) a light chain (light green) that binds to the VH and CH1 domains of the IgG part, and (3) a short heavy chain (dark turquoise) that binds to the light chain appended to the IgG.

Results

Setting the unique structure of RUBY bsAb

The architecture of RUBY bsAbs (Figure 1) consists of an IgG connected via polypeptide linkers (3×G4S) to two Fab domains positioned at the C-terminal end of the Fc part. RUBY bsAbs consist thus of six polypeptide chains, which are double copies of three different polypeptide chains: 1) a long chain that consists of the IgG heavy chain and the light chain of the additional Fab, 2) a light chain that binds to the VH and CH1 domains of the IgG part, and 3) a short heavy chain that binds to the light chain appended to the IgG. Since the Fabs are appended via their light chains, the classical light chain mispairing problems cannot occur in the sense that in RUBY correctly paired product will contain all the correct chains. This is a valuable feature, as light chain mispairing can cause a severe change in the drug potency and is problematic to analyze, and wrongly paired product-related impurities are difficult to remove.

A second critical feature of the RUBY bsAb format is the use of mutations in the interfaces between the VH-CH1 and VL-CL pairs, which create surfaces that limit formation of aggregates and by-product soluble Fabs, as well as mutations to reduce the risk of immunogenicity by decreasing risk for formation of MHC II binding peptides.

Mutations for heterodimerization of heavy and light chains

The mutations introduced to promote heterodimerization of heavy and light chains can, for example, cause steric hindrance or charge incompatibility of wrong chains or promote interactions between correct chains, for example by creating salt or disulfide bridges. A total of 20 different mutational variants were generated and their properties in regard to stability, binding, and function were evaluated. Descriptions of the mutations used and how these differed between the two different VH-CH1 and VL-CL pairs are included in Supplementary Table S1 and S2.

These bsAb variants were produced and analyzed for yield and degradation. The amount of high molecular weight species (HMWs) after transient small-scale production in HEK cells and purification on protein A columns varied between bsAb variants that differed only in the mutations introduced between the CH1-CL and VH-VL interfaces, as shown in Supplementary Table S3. Variant 1 constructs, carrying no mutations, generated high amounts of HMWs (29.8 and 48.7%) during expression and purification. All mutations introduced improved the purity and quality of the expressed bispecific antibodies, lowering the levels of HMWs. The levels of HMWs at modified chain transfection conditions for several of the variants were very low, 2.5% or less. Based on this data as well as an in silico immunogenicity analysis (further discussed below), mutational variant 9 was chosen for further studies. Although selected as the prioritized variant based on data generated with a limited set of bsAbs, its favorable profile has been shown to be generally applicable on subsequently generated larger sets of diverse RUBY bsAbs (outlined below).

Minimizing format inherent immunogenicity risk

A common challenge for many bispecific antibodies is the risk of inducing immunogenic responses, including induction of neutralizing anti-drug antibodies due to the presence of non-self peptides.²⁰ In addition to the intrinsic presence of such non-self peptides in the complementarity-determining region loops of synthetic antibodies, additional risk sites can be created when introducing mutations to framework regions or the constant domains of bsAbs, or when introducing non-natural polypeptide linkers and junctions between these and antibody domains. Therefore, in silico assessment of the immunogenic potential of the different RUBY variants was performed using the AbEpiAnalyzer tool that predicts binding of peptides to a large set of major histocompatibility complex (MHC) class II alleles.²¹ Mutational variant 9 had low impact on the predicted immunogenic potential of the bsAb yet, to further decrease the inherent immunogenic potential of the RUBY format, bsAbs generated using variant 9 were subjected to a round of deimmunization (Supplementary Information 2 “Deimmunization studies”). Using in silico peptide MHC class II-binding predictions and in vitro evaluation of MHC II display in an MHC Associated Peptide Proteomics (MAPPs) assay, two potential risk sites encompassing RUBY-specific mutations were identified and subsequently deimmunized by introduction of one additional single amino acid substitution per risk site. Successful deimmunization was confirmed using in silico predictions and an additional round of the MAPPs assay, demonstrating a low predicted inherent immunogenic potential of the RUBY format. The final set of deimmunizing mutations in combination with the variant 9 mutations (hereafter referred to as RUBY) is summarized in Table 2.

Table 2. Summary of RUBY variant 9 and deimmunizing (in bold) mutations.

RUBY element	RUBY mutations*
	VH	VL	CH1	Cκ/Cλ
IgG	X44R	X44E	H168A, F170G,	L135Y, S176W
Appended Fab	X44E	X44R	T187E, S183T**,	T109P**, S/T114A, N/S137K

*X refers to any amino acid in the specific position according to IMGT numbering.

**Deimmunizing mutations only apply to antibodies of kappa (κ) light chain origin.

Importantly, none of the mutations, included either to promote correct chain heterodimerization or reduce risk for immunogenicity, introduce liable sequence motifs that are known to lead to aspartic acid isomerization,²² asparagine deamidation²² or N-glycosylation.²³

In all, the RUBY format presents a novel approach to the generation of bsAbs. Introducing charged-pair and knob-into-hole mutations to modify the interface of VH-CH1 and VL-CL is not a new solution to generate bsAbs,^24–26 but combining this strategy with the 2 + 2 architecture from fusing Fab domains via their light chain to the IgG C-terminal end and including mutations to minimize sequence-related immunogenic risk is what makes the engineering of RUBY unique.

RUBY bsAbs were successfully produced using both transient and stable expression systems

Development of novel bispecific drug candidates often rely on generation and screening of large (>100) sets of compounds, making the expression of bsAbs in transient high-throughput systems critical. Productivity of RUBY bsAbs was assessed by transient expression of a set of 150 unique RUBY bsAbs in a 96-well high throughput Expi293 system. The set contained binding domains with diverse immunoglobulin heavy (n = 7) and light (n = 9) germline gene origins and all carried RUBY v9 as well as deimmunizing mutations. Most (82%) of a set of 150 unique RUBY bsAbs could be produced at levels sufficient for early screening activities with average yield at levels (30 mg/L) comparable to internal control mAb, demonstrating compatibility of RUBY bsAbs with an HT production system. Expression of a smaller set of RUBY bsAbs in transient 30 mL Expi293 and ExpiCHO systems again demonstrated that these bsAbs are compatible with conventional transient antibody expression systems. Although generated with a limited set of RUBY bsAbs, thus not fully capturing the potential variability in behavior of larger antibody sets, this is in line with the encouraging data obtained for the larger set using an HT production system and shows that RUBY bsAbs can have capacity to be produced using such systems with comparable yields and quality similar to that of mAbs (Supplementary Table S6 and Supplementary Table s7) and could be purified using standard IgG purification procedures.

Appropriate productivity in stable expressing cells is a key requirement for further development into clinical programs. Therefore, assessment of whether the results translate from transiently transfected Expi293 HEK cells to stable CHO cell lines was performed. Stable cell line clones that produce RUBY bsAbs were generated by co-transfecting CHOZn cells with three plasmids encoding for chains 1, 2, and 3 (Figure 1). Transfected cells were seeded into glutamine-free medium containing puromycin and hygromycin for selection of plasmid transfection. After three weeks the mini-pools were assayed for total antibody and bispecific binding titers. The highest ranking mini-pools were subjected to single-cell cloning. 96 individual clones with high IgG expression levels were selected for fed-batch production assay in deep-well plates.

After 14 days, the IgG titers were measured and protein A-captured RUBY bsAbs were analyzed for aggregation and fragmentation. The cell line development work yielded multiple clones with satisfying protein quality attributes. The unoptimized fed-batch titers for the top three clones after 14 days of culturing ranged from 0.5 to 3 g/L.

Our results demonstrate that the random co-integration-based cell line development process indeed enables selection of clones with the desired properties for productivity and protein attributes. While these experiments were performed with CHO cells grown in deep-well plates, a scale-up fed-batch study of one clone was done in shaker cultures of working volumes of 200 mL to 1.2 L. Comparable productivities were observed between deep-well and shake-flask conditions. These data suggest that suitable clones can be identified for RUBY-type molecules for further clinical development. Purification of the RUBY bsAbs was done using a single capture step on Protein A.

High stability was shown in various assays, indicating good inherent format developability of RUBY bsAbs

Selected RUBY bsAbs were evaluated for developability, including studies to determine temperature, shear stress, colloidal, freeze-thawing, serum, and pH stability. The bsAbs displayed good thermal stability with melting temperatures between 64°C and 74°C (Table 3). In addition, the bsAbs showed good shear stress as demonstrated by low protein concentration differences before and after agitation (Figure 2(a)). The molecules also displayed good colloidal stability (Figure 2(b)) where PEG concentrations above 9% were required to induce 50% protein loss.

Figure 2. Stability characterization of RUBY v9 bsAbs. (a) Ability of the molecules to withstand shear stress determined by measuring A280 before and after rapid agitation. (b) Colloidal stability of the molecules was determined by incubating them at various PEG concentrations. (c) Molecules were incubated in human serum or BSA for seven days after which a dual ELISA was performed. (d) Molecules were incubated at various temperatures for up to two weeks and dual ELISA was performed thereafter. (e) Stability of the molecules in 8 different buffers was analyzed by incubation at 40°C for 4 weeks. Representative results of RUBY v9 #1 after SEC-HPLC analysis are shown for HMWs (f) and LMWs (g). (h) RUBY bsAbs were incubated at pH 3.5 up to 2 hours. SEC-HPLC analysis was performed after neutralization of the proteins at 30, 60, 90 or 120 minutes.

Table 3. Measurements of melting temperatures (Tm) for different RUBYv9 bsAbs.

Construct	Antigen binding domains (IgG – Fab orientation)	Tm (°C)
RUBY v9 #1	αCD137 - α5T4	69
RUBY v9 #2	αCD40#1 - αEpCAM	66
RUBY v9 #3	αCD40#2 - α5T4	64
RUBY v9 #4	α5T4 - αCD40#2	66
RUBY v9 #5	αCD40#2 - αEpCAM	65
RUBY v9 #6	αCEACAM5 - αCD40#2	64
RUBY v9 #7	α5T4 - αCD40#1	72
RUBY v9 #8	αCD40#1 - α5T4	74

To further investigate the stability of RUBY bsAbs, serum stability studies were performed followed by dual ELISA measuring intact bsAb. Retained binding was observed for RUBY bsAbs incubated in both 50% human serum (HS)/50% phosphate-buffered saline (PBS) or PBS/0.1% bovine serum albumin (BSA). Data (Figure 2(c)) showed no difference in dual target binding between the samples, indicating good stability for these molecules at physiological conditions. Furthermore, we assessed storage stability of RUBY bsAbs at 2–8°C, room temperature (RT), and 40°C for 4 weeks and potential degradation after 1 or 3 rounds of freeze/thawing. SEC-HPLC analysis (Table 4), SDS-PAGE (not shown) and dual ELISA (Figure 2(d)) performed thereafter showed no degradation occurred either at elevated temperatures or after freeze/thawing. Similarly, samples at 10 mg/mL of RUBY v9 #1 in eight buffer formulations (Figure 2(e)) were incubated at 40°C for 2 or 4 weeks and analyzed by SEC-HPLC. Little degradation was observed for RUBY v9 #1, with less than 2% change in the amount of high or low molecular weight species (LMWs) in seven of the eight tested buffers (Figure 2(f,g)), indicating high stability for RUBY bsAbs also at higher protein concentrations. Minimal degradation was also observed following similar analysis of a RUBY bsAb after storage at 2–8°C for 1 year (data not shown). We also investigated the ability of RUBY bsAbs to withstand incubation at low pH since this is a critical virus inactivation step performed in standard antibody purification processes. Very minimal degradation was observed for these proteins, indicating that the RUBY bsAb format is inherently not sensitive to low pH (Figure 2(h)). Taken together, these results further confirm the potential of RUBY as a versatile format suitable for generating stable bsAbs.

Table 4. Changes in degradation after different treatments to evaluate the stability of RUBY v9 bsAbs. Protein concentrations ranged between 0.5 and 3 mg/ml in non-optimized buffer (PBS).

Description	Room temperature, 4 weeks		40°C, 4 weeks		Freeze/thaw, 1 round		Freeze/thaw, 3 rounds
	Δ% HMWs	Δ% LMWs	Δ% HMWs	Δ% LMWs	Δ% HMWs	Δ% LMWs	Δ% HMWs	Δ% LMWs
RUBY v9 #1	0	0	0	2	0	0	0	0
RUBY v9 #2	0	0	0	1	0	0	0	0

HMWs-high molecular weight species, LMWs-low molecular weight species.

Freeze thaw, 1 round- SEC-HPLC analysis performed after 24 hours.

Freeze thaw, 3 round- SEC-HPLC analysis performed after 72 hours.

Antigen binding properties of RUBY bsAbs

The monomeric binding affinity of selected RUBY bsAbs and corresponding monospecific mAb variants to their respective antigens was determined by bio-layer interferometry (BLI) measurements with an Octet instrument (Table 5). Only minor differences in 1:1 affinity between the RUBY and corresponding mAb versions were observed and positioning of the antigen binding domains, in the IgG or appended Fabs, had little impact on the antigen binding for the evaluated set of antibodies.

Table 5. 1:1 binding kinetics of RUBY bsAbs and corresponding mAb controls.

Sample ID	Antigen target	KD (nM)	kon (1/Ms)	kdis(1/s)
CD40-CEA RUBY#9	CD40	110	4E + 05	5E–02
CEA-CD40 RUBY #10	CD40	90	4E + 05	4E–02
CD40 mAb	CD40	80	3E + 05	3E–02
4-1BB-TAA RUBY #11	4-1BB	10	2E + 05	2E–03
TAA-4-1BB RUBY #12	4-1BB	26	1E + 05	3E–03
4-1BB mAb	4-1BB	8	3E + 05	2E–03
HER2-GFP RUBY #13	HER2	1	2E + 05	2E–04
GFP-HER2 RUBY #14	HER2	0.3	3E + 05	6E–05
HER2 mAb	HER2	0.7	4E + 05	3E–04

In addition, to assess dual target binding capacity, dual ELISA was performed, which confirmed that tested RUBY bsAbs were able to engage both their respective targets simultaneously as shown in Figure 3(a). Studies with a larger set of RUBY bsAbs further indicate that, although 1:1 target binding affinity is generally not dependent on whether the binder is situated in the IgG or as the appended Fab, overall performance in ELISA and FACS studies (data not shown) was reduced for a subset of the binders when situated in the appended Fabs compared to the IgG. Such impact on target binding by altered avidity should be somewhat expected because the 3×G4S polypeptide linkers used in RUBY bsAbs do not fully mimic the geometry provided by the hinge region in the IgG. Available data demonstrating variations in avidity between the IgG and appended Fab position for a subset of binders is not sufficiently comprehensive enough to allow for any conclusions regarding why certain binders are affected, but not all. Future studies are needed to assess whether this is entirely epitope dependent or if antibodies targeting antigens of certain classes or sizes are more likely to be affected than others.

Figure 3. Assessing binding and in vitro functionality of RUBY v9 bsAbs. (a) Dual ELISA showing simultaneous binding of RUBY v9 bsAbs to their respective antigen targets. ELISA plates were coated with one antigen, RUBY bsAbs were added followed by detection using a biotinylated second antigen. (b) The activity of a CD40-EpCAM RUBY v9 bsAb in a CD40 reporter assay. CD40 reporter cells (Promega) were incubated with EpCAM-transfected CHO cells or wild type CHO cells in the presence of the bsAb. CD40 activation was measured according to the manufacturer’s instructions. (c) Demonstration of the agonistic function of the CD137-5T4 RUBY bsAbs v9 bsAb on human CD8 T cells where a dose response dependent IFN-gamma production (absolute values) by human CD8 T cells from one representative individual donor activated with the bispecific constructs in the presence or absence of immobilized 5T4-Fc was observed. Obtained mean (and SD) IFN-gamma levels from one representative individual donor from experiment set 2 is shown.

FcγR binding properties of RUBY bsAbs

Binding of RUBY bsAbs of different Fc isotype variants (IgG1, IgG1 LALA and IgG2) and monoclonal isotype controls toward soluble Fc gamma receptors (FcγR) was evaluated with BLI. RUBY bsAbs and their respective mAb isotype controls showed similar interactions with human Fcγ receptors (Table 6). High affinity to human FcγRI was observed for RUBY IgG1 constructs with dissociation constants similar to those measured between mAb IgG1 and FcγRI, and to values previously reported.²⁷ As expected, no binding could be detected between IgG2 constructs, either RUBY bsAbs or mAbs, and any Fcγ receptors. In addition, IgG1 constructs showed comparable binding to FcγRIIa and hFcγRIIIa 176 V. As expected, no quantitative binding to the low affinity FcγRIIb and hFcγRIIIA-V176F was detected (data not shown).

Table 6. Kinetic measurements for RUBY bsAbs of different Fc isotype variants and monoclonal isotype controls after binding to hFcγri, hFcγriia and hFcγriiia 176 V.

Construct	hFcγRI			hFcγRIIa			hFcγRIIIa 176 V
	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)
IgG1 RUBY v9 #15	<1E–12	5E + 05	<1E–07	2E–07	2E + 06	4E–01	3E–08	1E + 06	3E–02
IgG1 mAb	<1E–12	3E + 05	<1E–07	3E–07	1E + 06	4E–01	5E–08	6E + 05	3E–02
IgG1 LALA RUBY v9 #2	4E–08	7E + 05	3E–02	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD
IgG1 LALA mAb	5E–08	5E + 05	2E–02	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD
IgG2 RUBY v9 #16	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD
IgG2 mAb	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD

LOD – Limit of detection.

Additional evaluations were also performed using mouse FcγRs and, as expected, IgG1 RUBY bsAbs and mAb constructs exhibited similar binding to mFcγRI. Moreover, comparable binding to mFcγRIII was observed for both IgG1 and IgG2 constructs (Table 7). No binding to mFcγRIIb was detected for any of the constructs, as expected using the current BLI-based assay setup.²⁸ Affinity for mFcγRIV was detected for RUBY IgG1 bsAb and the reference IgG1mAb. These results indicate that RUBY bsAbs retain interactions with FcγRs with characteristics comparable to mAbs, thus allowing for exploitation of biological functions dependent on such interactions.

Table 7. Kinetic measurements for RUBY bsAbs of different Fc isotype variants and monoclonal isotype controls after binding to mFcγri, mFcγriib, mFcγriii and mFcγr IV.

Construct	mFcγRI			mFcγRIIb			mFcγRIII			mFcγR IV
	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)
IgG1 RUBY v9 #15	4E–08	7E + 05	3E–02	<LOD	<LOD	<LOD	1E–08	7E + 05	1E–02	4E–08	7E + 05	3E–02
IgG1 mAb	4E–08	6E + 05	2E–02	<LOD	<LOD	<LOD	2E–08	4E + 05	6E–03	5E–08	4E + 05	2E–02
IgG1 LALA RUBY v9 #2	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD
IgG1 LALA mAb	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD
IgG2 RUBY v9 #16	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	1E–08	6E + 05	6E–03	<LOD	<LOD	<LOD
IgG2 mAb	<LOD	<LOD	<LOD	<LOD	<LOD	<LOD	1E–08	7E + 05	8E–03	<LOD	<LOD	<LOD

LOD – Limit of Detection.

Evaluation of in vitro functionality using CD40 reporter and T cell activation assays

The functionality of RUBY bsAbs was evaluated in a variety of assays. It was shown that a CD40-EpCAM RUBY bsAb displayed conditional CD40 activation in a reporter assay, similar to what has been shown by Hägerbrand et al.²⁹ The bsAb induced activation of CD40-expressing cells only in the presence of EpCAM expressed on CHO cells whereas no activation was observed in a similar set up with CHO cells lacking EpCAM expression (Figure 3(b)). Additionally, a CD137-5T4 RUBY bsAb induced potent T cell activation as measured by a dose-dependent increase in interferon (IFN)-gamma release in the presence of 5T4. In the absence of 5T4, there was no dose-dependent increase in T cell activity driven by the bsAb construct (Figure 3(c)).

The T cell activation was mediated by crosslinking of CD137 induced by a 5T4-dependent clustering triggered by the bsAb. These data demonstrate the strength of the RUBY format in generating conditionally functional bsAbs aimed at linking immune cells to tumor cells and limiting activity to target tissues, minimizing the risk of systemic toxicities.

RUBY bsAbs display unaltered FcRn binding and mAb-like PK profile in mice and cynomolgus monkeys

Evaluations were performed to determine the interactions between RUBY bsAbs and the neonatal Fc receptor (FcRn), which is important for antibody recycling and half-life.³⁰ Interaction between recombinant human FcRn and RUBY bsAbs carrying different Fc isotypes was assessed with BLI and the results confirmed that RUBY bsAbs efficiently bind human FcRn at levels comparable to mAb controls (Table 8 and Figure 4(a,b)). Further, binding to recombinant mouse FcRn for a RUBY bsAb carrying an IgG1 Fc with LALA mutations (L234A, L235A) was confirmed to be at levels expected for human antibodies of IgG1 isotype (Table 8).³¹

Figure 4. RUBY bsAbs interaction with human FcRn and pharmacokinetics. BLI binding kinetics measurements of (a) RUBY bsAb and (b) monoclonal antibody control displaying similar interaction with human FcRn. (c) C57bl/6 mice were intravenously (i.v.) administered with molar equivalent doses of a human IgG1 LALA-mutated RUBY v9 bsAb (167 µg) or a corresponding mAb control (100 µg). Blood was collected 1, 2, 4, 24, 72 and 168 hours post dosing (n = 3/timepoint) and serum concentration of human IgG was measured using a human heavy chain-specific ELISA. (d) Human CD40 transgenic mice were administered with molar equivalent doses i.v. of two CD40-TAA RUBY bsAbs (360 µg) and a CD40 mAb (216 µg). Blood was collected 0.5, 1, 2, 4, 6, 24, 48 and 72 hours post dosing (n = 3/timepoint) and serum concentration of human IgG was measured using a human heavy chain-specific ELISA. (e) CEA-CD40 RUBY bsAb #2 was administered i.v. at 10 or 37.5 mg/kg to one female and one male cynomolgus monkey per dose level. Blood was collected 0.08, 0.5, 2, 4, 8, 24, 48, 72, 96 and 168 hours post dosing and serum concentration of human IgG was measured using a human CD40 and human CEACAM5 dual target ELISA. The ratio of full RUBY molecules over total human IgG in the serum of (f) human CD40 transgenic mice following i.V. administration of 417 µg CD40-EpCAM RUBY v9 #2, or (g) cynomolgus monkeys administered 10–37.5 mg/kg CEA-CD40 RUBY bsAb #2 was determined using ELISA. The graphs show the mean (±SD) of 3 mice/timepoint (c, c, and f), measured values over time for individual animals (e) or the mean (±SD) of four animals (g).

Table 8. Kinetic measurements of controls and RUBY bsAbs of different fc isotypes binding to FcRn receptors of both human and mouse origin.

Constructs	hFcRn			mFcRn
	KD (M)	kon (1/Ms)	kdis (1/s)	KD (M)	kon (1/Ms)	kdis (1/s)
IgG1 RUBY v9 #15	6E–07	1E + 05	8E–02	N/A	N/A	N/A
IgG1 mAb	5E–07	1E + 05	5E–02	N/A	N/A	N/A
IgG1 LALA RUBY #10	6E–07	1E + 05	8E–02	7E–8	1E + 05	7E–03
IgG1 LALA mAb	5E–07	2E + 05	1E–01	N/A	N/A	N/A
IgG2 RUBY v9 #16	1E–06	2E + 05	2E–01	N/A	N/A	N/A
IgG2 mAb	6E–07	2E + 05	1E–01	N/A	N/A	N/A

N/A – Data not available.

To further confirm that RUBY bsAbs display unaltered binding to FcRn of different species and assess whether there are other format specific attributes that impact the half-life of these bsAbs, we evaluated the PK profile of a RUBY bsAb carrying a LALA silenced IgG1 Fc in mice. C57Bl/6 wild type mice were dosed intravenously (i.v.) with equimolar amounts of either the RUBY bsAb or a monospecific antibody corresponding to the mAb part of the RUBY and the concentration in serum was followed over 7 days (Figure 4(c)). Levels of the RUBY bsAb were strikingly similar to those of the CD40 monospecific antibody over the evaluated time span, indicating minimal differences in half-life between the RUBY bsAb and corresponding monospecific antibody. These results were further confirmed in a setting affected by target-mediated drug disposition, where similar serum concentration profiles of two different CD40-targeting RUBY bsAbs and a CD40 targeting mAb were observed following i.v. dosing of human CD40 transgenic mice with equimolar amounts of the above-mentioned compounds (Figure 4(d)). The serum concentration profile observed of a CD40-TAA RUBY bsAb after i.v. administration to cynomolgus monkeys further supports a mAb-like PK profile (Figure 4(e)). To assess whether the RUBY bsAbs remain intact in vivo over time, ELISA detecting dual target binding was used to measure the RUBY concentration in mouse and non-human primate serum samples, and the ratio of RUBY concentration over concentration obtained in ELISA detecting only the IgG part of the bsAb was calculated. No decrease in RUBY/IgG ratio could be observed over time, demonstrating that the C-terminal Fab remains attached in vivo both in mice (Figure 4(f)) and non-human primate (Figure 4(g)). It can thus be concluded that the addition of C-terminal Fabs and introduction of mutations to variable as well as constant domains of RUBY bsAb have no or very little impact on the PK profile of such bsAb compared to monospecific antibodies.

RUBY bsAbs efficiently localize to tumors in vivo

For applications such as tumor-directed immunotherapies using bsAbs, the drug must efficiently leave circulation and reach the tumor to be effective. To assess whether antibodies in the RUBY format meet this demand, mice bearing tumors expressing EpCAM were treated with CD40-EpCAM targeted RUBY bsAb and the frequency of IgG-positive cells 24 hours post treatment measured (Figure 5(a)). The results clearly demonstrate that RUBY bsAbs efficiently reach and are enriched in tumors expressing the relevant antigen, while treatment with a monospecific antibody targeting CD40 renders significantly fewer tumor-derived cells IgG positive.

Figure 5. Tumor localization and hydrodynamic sizes of RUBY bsAbs and controls. (a) RUBY v9 bsAb efficiently localized to the tumor area, where they bound EpCAM expressing tumor. Mice were inoculated s.c. with MB49 tumors expressing human EpCAM or the same cell line lacking the EpCAM expression. The mice received a single CD40-EpCAM RUBY v9 #2 bsAb, CD40 mAb or vehicle i.p. injection and 24 hours later the tumors were collected and the frequency of IgG+ cells were analyzed by flow cytometry. (b) A RUBY bsAb and (c) Monoclonal antibody control displayed similar hydrodynamic sizes as determined by DLS.

The ability of antibodies, mono- or bispecific, to localize to tumors depends on several factors, such as size, affinity, and the expression pattern of the target antigen. One concern when using bispecific antibodies for tumor-directed therapies is that additional domains, such as the Fab domains of RUBY, could render them too big to efficiently migrate into tissue and reach the tumor. We therefore evaluated the hydrodynamic diameter of RUBY bsAbs using dynamic light scattering (DLS) and compared the results to monospecific antibodies with corresponding specificities (Table 9, Figure 5(b,c)). The results show that there is very little difference in hydrodynamic size between a RUBY bsAb and the evaluated monospecific antibodies, with the average diameter of the monomeric fraction of RUBY bsAbs being 13 nm (radius 6.5 nm) compared to 10 nm for the control mAb, which is as expected. These results are in line with efficient tumor localization of a RUBY bsAb previously seen in vivo in tumor-bearing mice.²⁹

Table 9. Hydrodynamic diameter of monomeric fraction of RUBY bsAbs and control mAb measured by DLS.

Protein	Description	Z-Ave. Dia (nm)
RUBY #17	bispecific	14
RUBY #18	bispecific	13
RUBY #19	bispecific	12
RUBY #20	bispecific	13
RUBY #21	bispecific	13
RUBY #22	bispecific	13
IgG1 mAb	monospecific	10

These results demonstrate that the RUBY format is very well suited for development of bispecific antibodies intended for localization to tumors, or potentially other relevant tissues.

Discussion

BsAbs are attractive biological therapeutics, and their first report has been followed by generation of a plethora of formats.¹⁴ These formats often vary in their attributes, including molecular weight, geometry, number of antigen-binding sites, intrinsic affinity of individual arms and PK.¹⁴ Considering that therapeutic utilization of bsAbs is often impeded by poor stability, PK, and manufacturing challenges,³² we set out to engineer a novel format with IgG-like properties and natural bivalent target interaction. Criteria, in regard to the inherent format properties for the new format, were set based on the 6 critical criteria for bsAbs postulated by Nie et al.¹⁹ Here, we describe development of RUBY, a unique bsAb format with outstanding biophysical and biological properties.

An evaluation of the biophysical and physicochemical properties of antibody-based drugs forms an integral part of the drug development process. Several larger studies evaluating marketed antibodies have shown that no definitive physicochemical descriptor for developability success for antibodies drugs exists.^33–35 Instead, a holistic approach that balances multiple biophysical and physicochemical properties is necessary to select candidate drugs with higher chance to make it to the market and patients. A combination of in silico analyses of the primary sequence and cross-examination of measurements for key features, including aggregation propensity, thermal stability, colloidal stability/solubility viscosity, and shear stress stability, are often used during such assessments.³⁶ Obtaining antibodies with optimal biophysical and physicochemical properties comparable to those of conventional IgGs remains a notable challenge.³⁷ In this study, a battery of combined assays, often included during early antibody drug development, were performed to obtain an indication of the developability of RUBY bsAbs. The results showed good stability, solubility, and low aggregation for tested RUBY bsAbs. RUBY bsAbs displayed melting temperatures (Tm) ranging from 64–74°C, which places them in the same range for what has been observed for human mAbs (57–82°C).³⁸ The mAb-like thermostability could be attributed to the use of Fabs in RUBY, instead of scFv or other variable domain fragments as is the case for other tetravalent bsAb formats. It is known that the lack of the CH1 and CL domains renders scFv unstable¹ and that scFv have a propensity to form aggregates under thermal stress.^39,40 Indeed, a negative correlation between manufacturability and amount of scFv in a bsAb has been observed in previous studies.⁴¹ Another feature that indicates good overall inherent format developability for RUBY bsAbs is the relatively low level of aggregation, measured after only protein A purification, observed from transient and stable cultures. This is beneficial for large-scale manufacturing and functional screening of several binder combinations where a “one-size-fits-all” automated approach is preferred to generate research material for hundreds of novel bsAbs as opposed to the need for multi-dimensional generation or optimization of bsAbs reported elsewhere.^1,16

The success of any therapeutic drug also depends on the establishment of scalable processes for production, purification, and formulation of good yields of the active compound. In this study we evaluated whether RUBY bsAbs could withstand various stress conditions (low pH, agitation, and freeze/thaw) that are part of the mAb CMC processes³⁶ and demonstrated that RUBY bsAbs remain stable after shear stress, low pH incubation or freeze/thawing. This indicates that the RUBY bsAb format may have inherent stability similar to mAbs and that biophysically challenging steps in the mAb CMC, such as agitation, virus inactivation and freezing, may not be a general issue for RUBY bsAbs.

Another important parameter to evaluate in a new bsAb format is the production yields, as bsAbs commonly produce at yields lower than what is reported for mAbs,⁴² although there are reports of bsAbs produced at relatively high yields.⁴³ Reported production yield values for bsAbs from stable CHO cell cultures for BiTEs, DART, TandAb and asymmetric IgG-like have been below 1.5 g/L.^25,44–46 Our data indicate that RUBY bsAbs can be produced from stable CHO cells at 3 g/L, which is encouraging and shows potential for scalable manufacturing. In addition, with RUBY, due to the presence of an Fc region, most manufacturing CMC processes that are well-established for mAbs can be used. The small-scale evaluations performed in this study show the benefits of having an Fc and possibility of protein A purification. Future.Future investigations focused on the downstream process of RUBY bsAbs in large scale, as well as the viscosity, stability, and solubility at very high concentrations (>100 mg/ml),) will broaden the understanding of the RUBY format’s manufacturability and applicability for subcutaneous administration.

The immunogenicity and safety of a drug candidate is highly compound specific and depends partly on factors that are not format inherent, such as binding target(s) and mechanism of action. Some general aspects were nevertheless considered in the generation of the RUBY bsAb format concerning immunogenicity and safety. Bispecific antibodies have been shown to cause immunogenicity and anti-drug antibodies in clinical studies,^47,48 which can limit their therapeutic use.⁴⁹ When engineering RUBY we sought to minimize the sequenced-based format attribute-related immunogenic risk. Generation of neo-epitopes were minimized by the use of naturally existing building blocks and immune-inert linkers. In addition, potential neo-epitopes as predicted using in silico tools were removed by mutagenesis and the absence of immunogenic sequence-related motifs in the RUBY bsAb format was demonstrated in in vitro MAPPs assays. Immunogenicity is, however, a complex process to predict and has been shown to be influenced by other factors in addition to sequence neo-epitopes, including impurities, geometry, mechanism of action, and dosing regimens.⁴⁷ Inherent features, such as the geometry of the bsAb format, and generation of immunogenic impurities can also affect immunogenicity. The impact on immunogenicity due to the 2 + 2 geometry was not evaluated in this study and will have to be analyzed on a compound-to-compound basis.

The safety of RUBY has been partly evaluated in a study by Hägerbrand et al.²⁹ where non-human primates were administered up to 37.5 mg/mL repeated doses of a CEAxCD40 RUBY, which showed no pathological findings related to either format or compound. Additional prerequisites for appropriate safety and immunogenicity profiles are good physiological stress and low presence of product-related impurities, as it is known that these can increase the drug toxicity and patient immunogenic responses.^50,51 The data presented herein demonstrates that RUBY bsAbs have high stability and that stress conditions, including elevated temperatures, result in limited increase of HMWs. The current data showing minimal introduction of neo-epitopes in RUBY bsAbs, and the high physiological stability are encouraging, and the safety studies will be further supported when RUBY bsAbs are tested in patients.

BsAbs can display inferior in vivo PK properties compared to their parental mAbs.³² The studies shown in this report and in Hägerbrand et al.²⁹ demonstrate that the PK profile in mice and non-human primates of RUBY bsAbs is similar to IgG controls and similar to published reports showing that the half-life of IgG in circulation in mice is 6–8 days.⁵² Mab-like half-life for RUBY bsAbs correlates with the mAb-like FcRn measurements that further demonstrated that affinity toward both human and mouse FcRn is similar for monoclonal IgGs and RUBY bsAbs of various isotypes, and in line with what has been reported previously.³¹ The relationship between FcRn binding and half-life is well documented⁵³ and it is known that binding to FcRn prevents endosomal degradation and leads instead to the recycling of antibodies back to the serum.⁵⁴ The longer half-life is an important therapeutic factor since it enables such molecules to effectively accumulate at the tumor site and act on tumor cells.⁵⁵ Collectively, these findings indicate that it is possible to generate RUBY bsAbs with PK features that resemble observations with mAbs. Additional studies are needed to understand if RUBY bsAbs also display mAb-like PK in patients.

This study and findings reported in Hägerbrand et al.²⁹ also indicate that tumor penetration occurs with RUBY bsAbs. Briefly, in Hägerbrand et al.,²⁹ mice bearing tumors established from MC38 human CEACAM5 (CEA) expressing cells were treated with a CD40×CEA-targeting RUBY bsAb or a CD40 mAb. Staining of collected tumors, analyzed with immunohistochemistry for the presence of human Fc, showed a higher presence of human Fc in tumors from mice treated with CD40×CEA RUBY compared to CD40 mAb, demonstrating that tumor tissue uptake occurred with the RUBY bsAb. Tumor uptake is an intricate process affected by tumor and drug-specific factors. Serum half-life, protein size and target, which to some degree counteract each other’s influence, are drug-specific factors important for tumor uptake.⁵⁶ The molecular size of a protein influences several transport parameters and it is known to be inversely proportional to the rate of tumor diffusion,⁵⁷ leading sometimes to the erroneous misconception that antibody fragments have a higher tumor distribution compared to full-length IgG. Schmidt and Wittrup⁵⁸ showed, however, based on data from literature on PK studies, that if taking into account tumor and drug-specific factors (including plasma clearance rate) a mathematical model can be established that can predict tumor uptake, which indicates a maximum tumor uptake for proteins with hydrodynamic radii of around 6.5 nm. In addition, Li et al.,⁵⁹ comparing trastuzumab with an FcRn-nonbinding trastuzumab variant and trastuzumab-derived fragments (scFv, Fab, F(ab)2), demonstrated that, if one isolates the impact of protein size on tumor uptake, 100 kDa is the optimal protein size. This, however, only applied for non-FcRn binding proteins, as the same study also demonstrated that trastuzumab outperformed both the FcRn-nonbinding variant and all trastuzumab fragments, including the F(ab)2 variant of 100 kDa size. Combined, these previously reported findings and the herein presented measurements on the hydrodynamic radii of RUBY bsAbs of 6.5 nm and the FcRn binding data showing mAb-like FcRn affinity indicate that RUBY bsAbs may behave as mAbs in regard to tumor uptake. Examination using other in vivo tumor models would be valuable to further assess the biodistribution of RUBY bsAbs in different tissues.

The most intriguing potential of bsAbs is their capacity to enable new therapeutic concepts and mechanisms of action otherwise not possible with mAbs, including the generation of novel cell synapses such as the T cell redirection achieved by blinatumomab,² conditional activation of immune cells observed for ALG.APV-527¹⁰ or ATOR-4066²⁹ and spatial approximation of given receptors by emicizumab.⁶⁰ As shown in this report and Hägerbrand et al.,²⁹ the RUBY format generates conditionally active agonistic antibodies that can stimulate several cell types, including T cells, B cells and dendritic cells (DCs), as demonstrated with primary cell in vitro assays. In addition, Hägerbrand et al.²⁹ also showed that RUBY can be used for the generation of compounds with a unique mechanism of action termed Neo-X-Prime™ where bsAbs binding to CD40 and tumor-associated antigens (TAAs) can deliver tumor debris potentially carrying neoantigens to DCs. This results in the activation of neoantigen specific T cells and importantly increased anti-tumor effect compared to a monoclonal CD40 antibody, as demonstrated in vivo using human transgenic mice inoculated with TAA-expressing cancer cell lines. Combined these studies show that the RUBY format can be used to generate bsAb drug candidates with high in vitro and in vivo functionality.

Which other functionalities the RUBY format is particularly well-suited for remains to be studied. As a general rule, biological needs dictate the demands on the bsAb format in regard to features such as geometry, valency, and the presence of an Fc. However, the field is still determining how these features impact the general performance and functionality of bsAbs and no format is believed to be optimal for all mechanisms of action. Findings on variables such as the impact of target affinity, toxicity,^61,62 potency,^63,64 functionality,⁶⁵ and tumor penetration⁶⁶ are changing our collective view on how to engineer novel bsAbs, and prompting the evaluation of modest binding affinities and/or the use of avidity. RUBY, being a 2 + 2 format, is suited for high-avidity solutions. Indeed, the conditionally functional CD40×CEA RUBY bsAb reported in Hägerbrand et al.²⁹ has been specifically developed to recognize cell bound CEA as opposed to serum-secreted CEA. This feature was engineered using high-avidity CEA binding domains and would not have been possible in a format with only one binding arm toward CEA. There are likely other applications for a 2 + 2 geometry as demonstrated by Loh et al.⁴¹ in a systematic study that included the evaluation of the effects of valency, avidity, and geometry on function. HER2×CD3 bsAbs were constructed in eight formats and results showed the highest cell killing for the tetravalent, symmetric, 2 + 2 format.

Another feature of the RUBY format that may affect its biological suitability is its geometry and the distance between the two targets engaged, created by the Fc portion, which will need to be investigated in coming studies. Studies with other bsAb engineered using appended fragments to the Fc of an IgG have shown that this geometry is beneficial for generation of CD3 bsAbs that are more potent than BiTEs despite the smaller size of the latter, which can bring the TAA and T cells to a closer proximity, and it was shown to be the preferred geometry for Her2 × 4-1BB anticalin bsAbs.¹⁹

In summary, we demonstrated that the RUBY format provides a unique solution for construction of stable and manufacturable bsAbs that in many aspects show IgG-like properties, including IgG-like PK profile, stability, and production yields from stable CHO cultures. In addition, our data, including findings from Hägerbrand et al.,²⁹ demonstrate that the RUBY format can generate obligatory bsAbs with excellent in vitro and in vivo functionality, including higher anti-tumor efficacy compared to mAb and demonstrated tumor uptake. All in all, this warrants further evaluation of RUBY for other biological applications and in clinical studies.

Materials and methods

Design

RUBY bsAb variants were generated using different mutation combinations in the interface between VH-CH1 and VL-CL. The mutations used are well known in the public domain.

Nomenclature

The large set of RUBY bsAbs with different combinations of mutations evaluated during the design process are referred to as RUBY variants 1–20 (or RUBY v1–20). For convenience, the established format based on RUBY variant 9 with deimmunizing mutations is thereafter referred to as RUBY. For trackability, all unique RUBY bsAbs used in these studies have been numbered RUBY #1–24. When relevant, bsAb targets are stated in connection with the unique RUBY number, starting with the target of the IgG part, followed by the target of the appended Fabs.

Manufacturability in transient cultures

Bispecific antibodies were expressed using transient HEK293 and Expi293 HEK (Life Technologies) cultures at different volumes ranging from 600 μL − 2 L according to manufacturer’s instructions. Fed-batch productions from stable transfected CHO cell lines were performed in deep-well plates (2–25 mL) or in shaker flasks (200 ml − 1.2 L). Different feed strategies and media were tested. Purification of bispecifics from supernatants was made on protein A using the NGC system (BioRad), the ÄKTA Avant system (GE Healthcare) or Predictor MabSelectSure 50 μl 96-well plates (GE Healthcare). Cells were transfected with three different vectors encoding separately for each of the three polypeptides chains (i.e., the immunoglobulin heavy chain linked to the Fab light chain, the immunoglobulin light chain and the Fab heavy chain). Different transfection ratios of the three vectors were tested. Aggregation was measured with SE-HPLC in a 1260 Infinity II system (Agilent Technologies) using a TSK gel Super SW mAB HTP 4 μm, 4.6 × 150 mm column (TOSOH Bioscience) and 100 mM sodium phosphate, pH 6.8, 300 mM NaCl as mobile phase at ambient temperature and a flow rate of 0.35 ml/min. Reduced and non-reduced SDS assays were performed using a Caliper GXII with LabChip GX software.

Generation of stable CHO production cell lines

For the generation of stable CHO production cell lines, the glutamine synthetase deficient CHOZn-GS cell platform (Merck-SAFC) was used. The heavy-light fusion H1 chain , the light L1 chain or the heavy H2 chain genes were cloned into separate expression vectors each encoding a different selection marker. The cytomegalovirus (CMV) promoter was used to control expression of the chains. CHOZn-GS host cells were co-transfected with the three expression plasmids and stable mini-pools were selected in glutamine-free medium supplemented with hygromycin and puromycin. High ranking RUBY expressing mini-pools were subjected to single cell cloning by use of the Single cell printer (SCP, Cytena). Automated microscopy was performed using a CellMetric imaging system (Solentim) to ensure monoclonality. Based on cell growth and productivity, the best performing monoclonal cell lines were further expanded from microplates to deep-well plates and shaker flask cultures in glutamine-free medium. After upscaling, cell lines were grown routinely in shaker flasks in an orbital, humidified incubator shaker at 37°C, 5% CO2 and agitation at 110 rpm (50 mm orbit). Cell cultures were seeded at 3.0 × 10⁵ viable cells per mL and passaged every 2–3 days. Viable cell concentration was assessed using a ViCellXR (Beckman Coulter) by means of trypan blue exclusion. Antibody titers were quantified by use of Octet Protein A assay.

Dual ELISA

Plates were coated with 0.5 μg/mL antigen, human 5T4 (#ID95677.07, Innovagen) or human CD40 (1493-CD, R&D Systems), in PBS over night at 4°C. After washing in PBS/0.05% Tween 20 (PBST), the plates were blocked with PBS/0.2% bovine serum albumin (BSA) (#1.12018.0100, Merck) for at least 30 minutes at room temperature before being washed again. Samples serially diluted in PBS/0.5% BSA were then added and allowed to bind for at least 1 hour at room temperature. After washing, plates were incubated with 0.5 μg/mL biotinylated CD137 or EpCAM (# 10694 H02H-B, SinoBiological) for at least 1 hour at room temperature. Dual complexed bsAb with respective antigens were detected with horseradish peroxidase (HRP)-labeled streptavidin (#21126, Pierce). SuperSignal Pico Luminescent (#37069, Thermo Fischer Scientific) was used as substrate and luminescence signals were measured using Fluostar Optima (BMG Labtech)

Octet

Kinetic measurements were performed using the Octet RED96 platform (Sartorius). RUBY bsAb or corresponding mAb versions were coupled to anti-human IgG Fc Capture (AHC) Biosensor tips (Sartorius). Monomeric antigens were 1:2 serially diluted in 1× kinetic buffer (Sartorius) starting at 500 nM or 100 nM. Binding kinetics was studied in 1× kinetic buffer where association was allowed for 100 sec, 300 sec or 600 sec followed by dissociation for 100 sec, 300 sec or 3600 sec. Sensor tips were regenerated with 10 mM Glycine pH 1.7. Data generated were referenced by subtracting blank or parallel buffer blank, the baseline was aligned to the y-axis, inter-step correction by alignment against dissociation was performed and the data was smoothed by Savitzky-Golay filter in the data analysis software (v9.0.0.14). The processed data was fitted using a 1:1 Langmuir binding model with R² or X²as a measurement of fitting accuracy.

Binding to Fcγ receptors

RUBY bsAb and monoclonal isotype controls were diluted to 200 nM in 1× Kinetic buffer and captured on 8 parallel sensors for 300 sec. After setting a new baseline, the captured antibodies were assayed against any of the FcγR (hFcγRI (1257-FC-050, R&D Systems), hFcγRIIa (1330-CD-050, R&D Systems), hFcγRIIb (1875-CD-050, R&D Systems), hFcγRIIIa (4325-FC-050, R&D Systems), hFcγRIIIa V176F (8894-FC-050, R&D Systems), mFcγRI (2074-FC-050, R&D Systems), mFcγRIIb (1460-CD-050, R&D Systems), mFcγRIII (1960-FC-050, R&D Systems) mFcγRIV (50036-M27H-50, Sino Biologicals) for 60 seconds followed by dissociation for 60 seconds in kinetic buffer. The FcγRs were diluted in seven 1:2 dilutions starting at 100 nM. Sensor regeneration using 10 mM glycine pH 1.7 was performed before capturing of the next bsAb/ mAb. Data generated were referenced by subtracting a parallel buffer blank, the baseline was aligned with the y-axis, inter-step correlation by alignment against dissociation was performed and the data were smoothed by a Savitzky – Golay filter in the data analysis software (v.9.0.0.14). The processed data were fitted using a 1:1 Langmuir binding model.

Binding to Fc neonatal receptors

Binding toward soluble neonatal human and mouse Fc receptors (FcRn) for both RUBY bsAb of different Fc isotype variants and monoclonal isotype controls was also evaluated. Molecules were diluted to 2.0 μg/ml in 1× kinetic buffer and captured on 8 parallel sensors for 300 sec. A new baseline was set, and the captured antibodies were assayed against any of the FcRns (hFcRn (ITF02–200, Immunitrack) or mFcRn (6775-FC-050, R&D Systems)) for 60 seconds followed by dissociation for 60 seconds in sample buffer pH 6. The FcRns were run in seven 1:2 dilutions starting at 1.6 μM. Sensor regeneration using 10 mM glycine pH 1.7 was performed before capturing of the next RUBY bsAb/ mAb. Data generated were referenced by subtracting a parallel buffer blank, the baseline was aligned with the y-axis, inter-step correlation by alignment against dissociation was performed and the data were smoothed by a Savitzky – Golay filter in the data analysis software (v.9.0.0.14). The processed data were fitted using a 1:1 Langmuir binding model.

Thermostability and hydrodynamic size determination

The melting temperatures and hydrodynamic sizes were measured with the UNcle system (UNchained labs). The intrinsic fluorescence was measured during linear temperature ramping from 20°C to 95°C at a rate of 0.4°C/minute. The data analysis was performed with the UNcle Analysis software version 2.0 using default settings.

Temperature stability

Samples in non-optimized buffer (PBS) at low (<1 mg/mL) or high (10 mg/mL) protein concentration were incubated at 2–8°C, room temperature (RT) and 40°C for 1, 2 and 4 weeks or subjected to 3 rounds of freeze-thawing. Protein degradation was analyzed with SEC-HPLC, SDS-PAGE, A280, dual ELISA and by visual inspection under bright light against a dark background for any visible particles.

Shear stress stability

Samples in duplicates were subjected to shear stress in the form of heavy agitation at 2000 rpm on the 96-well plate shaker MixMate (Eppendorf) for 30 minutes. Protein precipitates were removed by centrifugation at 3000 g for 10 minutes and the absorbance at 280 nm was measured using a CLARIOstar (BMG LabTech).

Colloidal stability

Samples were diluted to 0.4 mg/ml in PBS and A280 measurements taken. The samples were then mixed with PBS/PEG solutions to a final concentration of 0.2 mg/ml with different PEG 8000 (#P0E040A, AppliChem) concentrations ranging from 8%-36%, added to buffer pre-wetted 96-well filter plates in triplicates and incubated over night at room temperature. Filtrates were obtained by centrifugation at 12,000 g for 15 minutes followed by absorbance measurements at 280 nm to determine protein loss.

Low pH stability

Stability at low pH was analyzed in conjunction with protein A purification by either exposing samples to low pH (3.5) during or after elution from protein A. Intermediate neutralization was performed after 30, 60, 90 or 120 minutes followed by analysis with SEC-HPLC for protein degradation.

Serum stability

Stability in serum was tested by incubating samples in 50% serum (#H4522, Sigma)/50% PBS or PBS/0.1% BSA at 37°C for 2 hours, 1, 2 and 7 days before analyzing binding in dual ELISA.

Solubility

The solubility in PBS or 20 mM histidine/150 mM arginine/pH 6.0 at 10 mg/mL was measured by concentrating the samples in ultrafiltration centrifugal units (Vivaspin 6, 10 kDa MWCO, GE Healthcare). Degradation was analyzed with HPLC, SDS-PAGE, A280 and by visual inspection under bright light against a dark background for any visible particles.

Reporter assays

A CD40 bioassay (Promega) was used according to the manufacturer’s protocol. Briefly 15,000 CD40 effector cells were plated in a 96-well plate (Thermo Scientific 136,101) and incubated in a 37°C, 5% CO2 incubator overnight. Next, the supernatant was aspirated, and serially diluted CD40-EpCAM RUBY bsAb was added into the wells followed by addition of CHO wild type cells or CHO cells expressing EpCAM and co-cultured with the CD40 effector cells for an additional 6 hours in a 37°C, 5% CO2 incubator. Thereafter, the plate was taken out to room temperature to equilibrate for 10–15 min. Finally, Bio-Glo Reagent was added into each well and the luminescence signal was measured using FLUOstar Optima plate reader (BMG Labtech). The data was analyzed with GraphPad Prism V.9.4.0 software (GraphPad software) and displayed as relative luminescence units (RLU).

T cell activation assay

The functional activity of the CD137-5T4 bsAb was evaluated in a CD8+ T cell assay, where cells were cultured in microtiter plates coated with 5T4-Fc and CD3 antibody. Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation using Ficoll-Paque (ρ 1.077 g/ml) (GE Healthcare #17-1440-02) from leucocyte concentrates obtained from healthy donors (Clinical Immunology and Transfusion Medicine, Labmedicin Region Skåne, Lund Sweden). CD8+ T cells were enriched by negative selection using the CD8+ T cell isolation kit (Miltenyi 130-096-495). Plates were coated overnight at 4°C with 3 μg/ml αCD3, clone OKT3 (Affymetrix eBioscience #16-0037-85), washed and coated with 5 μg/ml 5T4-Fc (#ID95677.07, Innovagen) for 2 h at 37°C. After the 5T4-Fc coating, plates were washed and blocked for a minimum of 30 minutes with RPMI (Gibco # 61870010) containing 10% fetal calf serum (FCS) (Heat inactivated, Gibco # 10270–106 lot 41Q9248K) and 10 mM HEPES (Gibco # 15630056). CD137-5T4 bsAb was diluted in RPMI containing 10% FCS and 10 mM HEPES and added to the plates 30 minutes before addition of CD8+ T cells (0.07 × 10⁶ cells/well). Assay plates were incubated for 68 h at 37°C, and culture supernatant harvested. IFN-γ levels in the supernatants were measured by ELISA (BD OptiEIA #555142).

Mice

Human CD40 transgenic mice (hCD40tg)⁶⁷ and human O×40 transgenic mice⁶⁸ were bred and maintained at the Medicon Village animal facility (Lund, Sweden). C57BL/6 mice were obtained from Janvier. Mice used for experiments were between 8 and 14 weeks of age. All experiments were performed after approval from the Malmö/Lund Animal Ethics Committee.

PK assessment in mice

For in vivo PK assessment, mice were injected i.v. with either formulation buffer (20 mM L-histidine/150 mM L-arginine) or a single dose of antibody. Blood was drawn via vena saphena or at termination via vena cava into serum collection tubes at either 0.5, 1, 2, 4, 6, 24, 48 and 72 hours post dosing for wild type mice, or at 0.5, 1, 2, 4, 6, 24, 48 and 72 hours post dosing for human CD40 transgenic mice (n = 3/timepoint). Each mouse was sampled at two of the timepoints.

The presence of human IgG in serum samples was analyzed by ELISA using an anti-human IgG Heavy chain coated to the assay plates (0.3 µg/well, #31118, Thermo Fisher Scientific,) to capture the antibody and an HRP-conjugated anti-human IgG kappa light chain (0.125 μg/ml, #STAR127P, Bio-Rad) for detection. Luminescence was measured using Clariostar. Standard curves were set up for the respective antibodies and used to interpolate concentrations in diluted serum samples with GraphPad Prism software.

For the CD40-EpCAM RUBY v9 #2, the ratio of RUBY and total human IgG in serum samples was determined as the ratio between the concentration of RUBY measured in an EpCAM-CD40 Dual ELISA and the concentration of the CD40 IgG part of the RUBY measured in a CD40 mono ELISA. Two sandwich ELISAs were developed, using mouse Fc conjugated human CD40 (#504–820, Ancell, 0.25 and 0.5 µg/ml, respectively, for mono and dual ELISA) coated in white high binding microplates (Greiner). HRP conjugated goat anti-human IgG1 Fc (0.004 µg/ml, #109-035-098, Jackson ImmunoResearch) was used as detection in the mono ELISA and biotinylated human Fc conjugated human EpCAM (0.2 µg/ml, #10694-H02H-B, Sino Biological) followed by HRP conjugated streptavidin (#21126, Pierce) were used as detection in the dual ELISA. SuperSignal Chemiluminescent substrate (#37069, Thermo Scientific) was used as substrate, generating a luminescent response measured in the Fluostar Optima (BMG). The responses from the standard were fit in a 4PL plot in GraphPad Prism from which the concentrations of the samples were extracted.

PK assessment in non-human primates

A CEA-CD40 RUBY bsAb was administered at 10 or 37.5 mg/kg by i.v. infusion in the tail vein over 60 minutes to one female and one male cynomolgus monkey per dose level. Blood was taken for preparation of serum pre dose (0 h) and 0.08, 0.5, 2, 4, 8, 24, 48, 72, 96 and 168 hours post dosing. The procedure was performed by Charles River Laboratories Edinburgh.

The bsAb concentration in serum was analyzed using a human CD40 and human CEACAM5 dual ELISA. Human CEACAM5 (0.5 μg/ml, #4128 CM, R&D systems) was coated in the plate for capture of the bsAb, and detection was mediated via 0.2 ug/ml of biotinylated CD40 (#504–030, Ancell) followed by HRP-conjugated streptavidin (#21126, Pierce) at 0.17 μg/ml. Luminescence was measured using Clariostar, and concentrations in serum were interpolated from a standard curve of the bsAb prepared in assay buffer using GraphPad Prism software.

To determine the ratio of RUBY over human IgG in non-human primate serum, a CEACAM5 mono ELISA was used to detect the IgG part of the RUBY. Human CEACAM5 (0.5 µg/ml, #4128 CM, R&Dsystems) was coated in white high-binding microplates (Greiner). Detection of the IgG part of the RUBY was done with HRP-conjugated goat anti-human IgG1 Fc (0.004 µg/ml, #109-035-098, Jackson ImmunoResearch). SuperSignal Chemiluminescent substrate (#37069, Thermo Scientific) was used as substrate, generating a luminescent response measured in the Fluostar Optima (BMG). The responses from the standard calibrators were fit in a 4PL plot in GraphPad Prism from which the concentrations of the samples were extracted.

Tumor localization

Murine MB49 bladder cancer cells, obtained from Millipore, were transfected to stably express human EpCAM, generating a MB49-hEpCAM single cell clone. MB49-wt cells were cultured in DMEM supplemented with 10% FCS. 0.25 ug/ml puromycin was added for culturing MB49-EpCAM cells. Mice were inoculated subcutaneously on the right flank with MB49-hEpCAM cells on day 0, followed by intraperitoneal administration of equal molar amounts of an anti-CD40 mAb (200 µg), a CD40×EpCAM RUBY v9 bsAb (333 µg), or vehicle control (PBS) on day 10. 24 hours later, tumors were collected, cut into small pieces, and incubated with 0.38 mg/ml Liberase TL (#05401020001, Roche) and 0.1 mg/ml DNase I (#10104159001, Roche) for 30 min. After incubation, the supernatants were collected and remaining tissue pieces were passed through a 100 µm cell strainer to generate a single cell suspension. Samples were subsequently incubated with mouse BD Fc block (#553142, BD Biosciences) for 30 minutes, followed by 30 minutes staining with anti-murine CD45-PerCP-Cy5.5 (#550994, BD Biosciences) and anti-human IgG-PE (#109-115-098, Jackson ImmunoResearch). FVS450 stain (#562247, BD Biosciences) was used according to the manufacturer’s instructions to identify viable cells. The presence of human IgG⁺ CD45⁻ tumor cells in samples was analyzed by flow cytometry using a FACSVerse and FlowJo software (BD Biosciences).

Acknowledgments

The authors would like to thank our former colleague Maria Hult for her contribution to protein characterization studies and the rest of the team at Alligator Bioscience for all their technical support with protein expressions, in vitro and in vivo studies.

Disclosure statement

At the time the research was performed all authors were employed by Alligator Bioscience.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2024.2330113

Additional information

Funding

The work in this study was funded by Alligator Bioscience. In addition, Barnabas Nyesiga receives funding for industrial PhD studies from the Knowledge Foundation, Sweden, under grant number 20180114.