Bispecific FpFs: a versatile tool for preclinical antibody development

Abstract

We previously described FpFs 1̲ (Fab–PEG–Fab) as binding mimetics of IgGs. FpFs are prepared with di(bis-sulfone) conjugation reagents 3̲ that undergo disulfide rebridging conjugation with the accessible disulfide of each Fab (Scheme 1). We have now prepared bispecific FpFs 2̲ (bsFpF and Fab₁–PEG–Fab₂) as potential bispecific antibody mimetics with the intent that bsFpFs could be used in preclinical antibody development since sourcing bispecific antibodies may be challenging during preclinical research. The di(bis-sulfone) reagent 3̲ was first used to prepare a bsFpF 2̲ by the sequential conjugation of a first Fab and then a second Fab to another target (Scheme 2). Seeking to improve bsFpF synthesis, the asymmetric conjugation reagent, bis-sulfone bis-sulfide 1̲6̲, with different thiol conjugation reactivities at each terminus (Scheme 4) was examined and the bsFpFs appeared to be formed at similar conversion to the di(bis-sulfone) reagent 3̲. To explore the advantages of using common intermediates in the preparation of bsFpF families, we investigated bsFpF synthesis with a protein conjugation–ligation approach (Scheme 5). Reagents with a bis-sulfone moiety for conjugation on one PEG terminus and a ligation moiety on the other terminus were examined. Bis-sulfone PEG trans-cyclooctene (TCO) 2̲8̲ and bis-sulfone PEG tetrazine (Tz) 3̲0̲ were used to prepare several bsFpFs targeting various therapeutic targets (TNF-α, IL6R, IL17, and VEGF) and tissue affinity targets (hyaluronic acid and collagen II). Surface plasmon resonance (SPR) binding studies indicated that there was little difference between the dissociation rate constant (k_d) for the unmodified Fab, mono-conjugated PEG–Fab and the corresponding Fab in a bsFpF. The Fab association rate (k_a) in the bsFpF was slower than for PEG–Fab, which may be because of mass differences that influence SPR results. These observations suggest that each Fab will bind to its target independently of the other Fab and that bsFpF binding profiles can be estimated using the corresponding PEG–Fab conjugates.

We previously described FpFs 1̲ (Fab–PEG–Fab) as binding mimetics of IgGs.

Introduction

We previously described an antibody mimetic that is called FpF (Fab–PEG–Fab) 1̲, where a poly(ethylene glycol) (PEG) linker mimics the IgG hinge.^1,2 The IgG hinge enables flexibility of the two antigen-binding fragments (Fabs) to provide enhanced binding that can be derived from bivalency to facilitate rebinding to slow dissociation rates.^3,4 The hinge region comprises single polypeptide chains that can be susceptible to degradation.⁵ FpFs 1̲ (structures 1) share comparable solution sizes, binding characteristics,¹ functional activity² and enhanced stability⁶ with IgG antibodies using PEG linkers with molecular weights in the range of 5–20 kDa. FpFs were designed to maintain the Fab topology and flexibility that has evolved in IgG antibodies with increased stability by substituting the hinge region with a PEG linker that has been stably covalently conjugated to each Fab. With these characteristics of FpFs 1̲ in mind, we decided to examine the synthesis and binding properties of bispecific antibody mimetics called bsFpFs 2̲ (Fab₁–PEG–Fab₂).

FpFs are prepared using the di(bis-sulfone) reagent 3̲ to conjugate 2 Fabs that can be obtained by the proteolytic digestion of IgGs (Scheme 1).¹ Each bis-sulfone conjugation moiety in the di(bis-sulfone) reagent 3̲ undergoes site specific bis-alkylation conjugation with the two thiols from the accessible Fab disulfide furthest from the target binding region by a sequence of addition–elimination reactions to give thiol ethers that are more stable than the initial disulfide (Scheme S1, ESI†).^7,8 Di(bis-sulfone) 3̲ has also been used to dimerise other proteins by bis-alkylation of cysteine thiols^9,10 including an analogous Fc-fusion mimetic¹⁰ which involved conjugating the extracellular receptor binding domains of an Fc-fusion protein. The Fc-fusion mimetic is called a RpR for receptor–PEG–receptor and displayed better binding characteristics compared to the corresponding parent Fc-fusion, aflibercept.

Scheme 1. IgG representation showing how the two Fabs topologically are shown to be at the two ends of a flexible linear molecule (note the red dotted line). Proteolytic digestion of the IgG provides the corresponding Fab after purification which can be conjugated to each end of the di(bis-sulfone) reagent 3̲ to give FpF 1̲.

Bispecific antibody-based medicines are clinically proven modalities used to treat cancer, heamophilia and neovascular retinal diseases^11–16 with many clinical candidates undergoing development with encouragement from the regulatory authorities. Faricimab is a bispecific antibody that binds to 2 ligands in the vitreous cavity to treat retinal diseases. Intravitreal injections are administered in a small volume (50 μL) and are difficult for patients to endure. Increasing residence time in the vitreous is broadly proportional to the concentration of an antibody-based medicine in the formulation.¹⁷ Utilising a high concentration of a single (bispecific) antibody in an intravitreal injection provides better clinical benefits than multiple intravitreal injections of a combination of antibodies.

Bispecific antibodies can also bring two targets together to cause an enhanced biological function not possible by using a combination of 2 separate antibodies. Such spatial–temporal properties have been shown to bring (i) two cells together (e.g. blinatumomab and epcoritamab)^13,14 to enhance immune recognition and (ii) two proteins together that are necessary to maintain the coagulation cascade (e.g. emicizumab).¹⁶ BsFpFs utilise PEG linkers of a sufficient molecular weight designed to optimise spatial–temporal relationships.

IgGs are multifunctional molecules that can also exert immune-related effector functions through Fc. Many clinically used mono-specific antibodies exert an effector function which is important for their clinical efficacy. Some Fc effector functions can cause inadvertent immune-driven agonism,^18,19 but IgG₄ subtypes are sometimes used to abrogate unwanted Fc-induced effector functionality.¹⁸ Many applications where the spatial–temporal properties of a bispecific antibody can be exploited do not require Fc effector functions.

The concept of bispecific antibodies has been considered for decades²⁰ as well as their preparation by chemical conjugation.^21–23 Purely recombinant strategies are used for the preparation of clinically used bispecific antibodies,²⁴ but challenges remain to optimise the structure and format of bispecific antibodies for researchers in early preclinical research.^24–26 The use of chemical conjugation and modification strategies to aid in the development of therapeutic proteins including antibody-based molecules is widespread, clinically proven and is being driven by much creative research studies.^27–34 There is also intense interest in developing protein–protein and protein–drug conjugates^35–38 as the combined use of recombinant and conjugation technologies may yield complex molecules for study and potential development.

Our goal is to develop a practical method to prepare bsFpFs, making these valuable tools readily accessible to researchers interested in a wide range of preclinical research including drug target development, drug delivery, tissue engineering and immunocytochemistry. We first explored the di(bis-sulfone) reagent 3̲ to prepare bispecific FpFs 2̲ (Scheme 2) in a conjugation-only approach. Since the preparation of bsFpFs with the di(bis-sulfone) reagent 3̲ requires the sequential addition of 2 different Fabs, there can be limitations due to the formation of small amounts of homodimeric FpF during the addition of the first Fab, so we also examined a conjugation–ligation strategy for the preparation of bispecific FpFs (Scheme 6). The conjugation–ligation strategy enabled the synthesis of an extensive panel of seventeen different bsFpFs, targeting a diverse range of therapeutic molecules implicated in ocular inflammation (TNF-α, IL6R, IL17, and VEGF) and ocular drug delivery (hyaluronic acid and collagen II). To the best of our knowledge, this represents the first synthesis of such a broad spectrum of bsFpFs which would be more costly and difficult to achieve by recombinant means alone.

Scheme 2. Preparation of a bsFpF 2̲ by the conjugation of Fab₁ with an excess di(bis-sulfone) reagent 3̲. The intermediate conjugate 4̲ can be purified by IEX and in some cases by spin filtration to remove the excess starting reagent 3̲ and unreacted Fab₁. Conjugation of Fab₂ to the intermediate conjugate 4̲ gives the bsFpF 2̲.

Scheme 6. A conjugation–ligation strategy using the reagents, bis-sulfone–PEG–TCO 2̲8̲ and bis-sulfone–PEG–Tz 3̲0̲, to prepare bsFpFs 2̲. The PEG–Fab conjugate intermediates 2̲9̲ and 3̲1̲ are prepared by site-specific conjugation of two different Fabs through the bis-sulfone functional group. Ligation of these conjugate intermediates results in the preparation of the bsFpF 2̲.

Experimental section

Materials

Bevacizumab (Avastin®, 25 mg mL⁻¹, anti-VEGF IgG), infliximab (Remicade®, 10 mg mL⁻¹, anti-TNFα IgG), tocilizumab (Avtemra®, 20 mg mL⁻¹, anti-IL6R IgG) and ranibizumab (Lucentis®, 10 mg mL⁻¹, anti-VEGF Fab) were obtained from the pooled remaining contents of vials that had been used clinically. Secukinumab (Cosentyx®, 150 mg) was purchased commercially. Phosphate buffered saline (PBS; 0.16 M NaCl, 0.003 M KCl, 0.008 M Na2HPO4 and 0.001 M KH2PO4) was prepared with tablets purchased from Oxoid. Acetate buffer A (100 mM sodium acetate, pH 4.0) and acetate buffer B (100 mM sodium acetate, 1 M NaCl, pH 4.0) were prepared for ion-exchange chromatography. Novex bis–tris 4–12% gels, sharp blue standard protein markers, NuPAGE MOPS running buffer, NuPAGE LDS sample buffer and SilverXpress silver staining kit were purchased from Invitrogen. InstantBlue was purchased from Expedeon Ltd. Perchloric acid (0.1 M) and barium chloride (5.0%) solutions for barium iodide staining were prepared in the lab. A PD-10 column, cation exchange columns (HiTrap SP HP 1.0 mL) and a Superdex 200 prep grade size exclusion column (34.0 μm particle size) along with Biacore consumables were all purchased from GE Healthcare. Anti-human IgG (Fab specific)-peroxidase, 3,3′,5,5′-tetramethylbenzidine (TMB) and human vascular endothelial growth factor (hVEGF₁₆₅) were purchased from Sigma Aldrich.

Methods

Fab isolation

Proteolytic digestion of bevacizumab, infliximab, tocilizumab and secukinumab was performed. The digestion method for digesting these IgGs has been optimised previously in our lab.³⁹ In brief, a pre-digestion buffer (200 mL, 20 mM NaH₂PO₄, 10 mM EDTA) with pH 7.0 was prepared. The pH of the solution was checked to ensure that it remained at pH 7.0 ± 0.5 and was adjusted if necessary. An IgG antibody (15 mg, 1 mL) was prepared using a digestion buffer. Lyophilised papain (150 μL at 5 mg mL⁻¹) was then added to prepare a 1 : 20 papain:IgG ratio. The digestion mixture was placed in an incubator at 37 °C for 30 minutes. The crude digestion mixture was then purified using protein L chromatography (Hitrap Protein L) using buffer A (100 mM sodium phosphate, 150 mM sodium chloride, 500 mL, pH 7.2) and buffer B (100 mM glycine, 500 mL, pH 2.5) and an elution gradient.

Eluted fractions were monitored using SDS–PAGE and further purified using an SEC (Superdex 200 Increase 10/300 GL, flow rate of 0.5 mL min⁻¹ and PBS as the mobile phase). Eluted fractions were collected and monitored using SDS–PAGE.

Representative bsFpF preparation by the di(bis-sulfone) reagent 3̲

Fab_VEGF (1.0 mg in 1.0 mL of conjugation buffer; 20 mM sodium phosphate, 10 mM EDTA, pH 7.6) was first incubated with DTT (1.0 mg) at ambient temperature without shaking for 30 min. DTT was removed by elution over a PD-10 column. Into 1.0 equivalent of reduced-Fab_VEGF was added the di(bis-sulfone) PEG₁₀ reagent 3̲¹ (5 equivalents, 1 mg) and the solution was incubated for 1 h at ambient temperature. The PEGylation reaction mixture was then purified using a Macrocap SP cation exchange column (Macrocap SP, 5 mL). The IEX-purified bis sulfone–PEG₁₀–Fab_VEGF was then incubated with the pre-DTT treated Fab_HER2 (1.0 mg in 3.3 mL after the PD-10 column). The formation of bispecific Fab_VEGF–PEG₁₀–Fab_HER2 was then monitored for 2, 15 and 48 hours at ambient temperature using SDS–PAGE. The Fab_VEGF–PEG₁₀–Fab_HER2 was purified using single-step size-exclusion chromatography (SEC), and SEC fractions were analysed by SDS–PAGE.

Representative bsFpF preparation by conjugation–ligation

Fab_VEGF (1.0 mg in 1.0 mL of conjugation buffer; 20 mM sodium phosphate, 10 mM EDTA, pH 7.6) was incubated with DTT (1.0 mg) at ambient temperature without shaking for 30 min. DTT was removed by elution over a PD-10 column. Bis-sulfone–PEG–TCO 2̲8̲ (0.1 mL, 4.0 mg mL⁻¹ in distilled water, 1.0 eq.) was then added to the reduced Fab_VEGF solution (1.0 mg in 3.3 mL). The PEGylation solution was incubated at ambient temperature for approximately 6 h without shaking. In a separate vial, Fab_IL6R (1.0 mg in 1.0 mL of conjugation buffer; 20 mM sodium phosphate, 10 mM EDTA, pH 7.6) was incubated with DTT (1.0 mg) and the DTT was removed by elution over a PD-10 column. Bis-sulfone–PEG–Tz 3̲0̲ (0.1 mL, 2.0 mg mL⁻¹ in distilled water) was then added to the reduced Fab_IL6R solution (1.0 mg in 3.3 mL). The PEGylation solution was incubated at ambient temperature for approximately 6 h without shaking. Both PEG–Fab conjugates (Fab_VEGF–PEG TCO 2̲9̲ and Fab_IL6R–PEG Tz 3̲1̲) were purified using a Macrocap SP cation exchange column (Macrocap SP, 5 mL). Fractions (1.0 mL) were analysed by SDS–PAGE. Purified intermediate molecules 2̲9̲ and 3̲1̲ were then mixed for approximately 18 hours at 4 °C to undergo ligation, forming a bsFpF molecule 2̲. bsFpFs were then purified using cation exchange chromatography followed by SEC. The purity of the bsFpF was assessed using silver staining and the concentration of the purified bsFpF was calculated using a micro-BCA assay.

The products are denoted using subscript ‘n’ on PEG_n to indicate the PEG molar weight, e.g. FabVEGF–PEG15–FabTNF_α is derived from the FabVEGF from bevacizumab, FabTNF_α from infliximab, the 10 kDa for the PEG₁₀ reagent 2̲8̲ and the 5 kDa for the PEG5 reagent 3̲0̲.

Comparative binding determined by ELISA

A flat bottom 96-well plate was coated with VEGF (100 μL of a 1 μg mL⁻¹ VEGF solution, 0.1 μg VEGF per well) and incubated overnight at 4 °C. The next day, the VEGF solutions were removed and without washing, blocking buffer (300 μL, PBS with 1% BSA and 0.05% Tween 20) was added into each well and incubated at ambient temperature for 2.0 h. After this, the blocking buffer was aspirated and the plate was washed once with washing buffer (300 μL, PBS with 0.05% Tween 20) and then tested compounds, which were prepared in PBS at a range of concentrations, were added into each well (100 μL). The plate was incubated for 2.0 hours at ambient temperature. After 2.0 h incubation, the protein solutions were removed and wells were washed with washing buffer (300 μL) three times. Then, anti-human Fab₂ (Fab specific)-peroxidase (100 μL, 1/5000 dilution) was added into each well and incubated for 1.0 hours at ambient temperature. The solutions were then removed, and the plate was washed off with a washing buffer three times. TMB (100 μL) was then added, and the development of the blue colour was monitored. After approximately 5 min, when the blue colour was visible enough for each well, HCl (50 μL, 1.0 M) solution was added to produce a constant yellow colour. The plate was then read using a plate reader at 450 nm wavelength. The data were processed using a graph pad prism (V9) and affinities were generated using one site-specific binding fitting method. No adjustments to the methodology were made for IL-6R, and ligand concentrations, incubation times and washing steps were identical.

Comparative binding determined by SPR (Biacore)

For binding molecules, human recombinant VEGF₁₆₅ (38 kDa) was immobilised on a CM3 chip at an immobilisation level of 95 RU using standard carbodiimide-mediated coupling (NHS/EDC, 50/50) and ethanolamine (pH 8.5). Samples were prepared in HBS-EP running buffer (10 mM HEPES, pH 7.4, 150 mm NaCl, 3.0 mM EDTA, 0.005% surfactant P20). All kinetic measurements were conducted at 25 °C at a flow rate of 30 μL min⁻¹ with an association time of 180 s and dissociation rate of 1200 s. Chip regeneration was accomplished by exposure to 10.0 mM glycine–HCl (pH 2.0) for 1200 s. Double-referencing was performed to account for bulk effects caused by changes in the buffer composition or nonspecific binding. For IL-6R binding molecules, human his-tagged recombinant IL-6R (42 kDa) was captured onto a NTA chip by first activating the chip surface with a nickel (Ni²⁺) solution. Samples were prepared in HBS-EP running buffer (10 mM HEPES, pH 7.4, 150 mm NaCl, 3.0 mM EDTA, 0.005% surfactant P20). For a TNFα binding molecule, human his-tagged recombinant TNFα was captured onto a NTA chip using a similar procedure described for his-tag IL6R. All kinetic measurements were conducted at 25 °C at a flow rate of 30 μL min⁻¹ with an association time of 120 s and a dissociation rate of 1200 s. The NTA chip surface was regenerated with EDTA solution in between each analysis cycle. Data were evaluated with the BIAevaluation software (version 2.1) and the best fit (lowest Chi²) was obtained using a 1 : 1 binding model. The sensorgram was fitted globally over the association and dissociation phases. Equilibrium dissociation constants (affinity) were calculated from the rate constants (K_D = k_off/k_on).

Microscale thermophoresis of antibodies and antibody conjugates

Antibodies and bispecific antibody mimetics were labelled as per the instructions found in the monolith amine reactive protein labelling kit. An MST assay buffer (50 mM Tris, 150 mM NaCl, 10 mM MgCl₂, pH 7.4, and 0.05% w/v Tween 20) was used to dilute the labelled antibody or bispecific antibody mimetic (bsFpF) to a concentration of 5 nM. Serial dilutions of relevant ligands (VEGF, IL6R, 1.0 × 10⁻⁶ to 1.52 × 10⁻¹¹ M) and combined ligands (VEGF + IL6R) were mixed with the labelled antibodies or antibody bsFpF in glass capillaries for measurements. Assays were performed using a LED power of 15% and with a laser power set to high. A Monolith Pico system was used for all MST measurements.

Results

bsFpFs prepared by conjugation

The di(bis-sulfone) reagent 3̲ derived from PEG with a molecular weight of 10 kDa was used to prepare a bsFpF from Fabs targeted to VEGF and HER2 (Scheme 2). Fabs can be obtained by thiol protease digestion of IgGs using either immobilised or soluble papain.³⁹ Bevacizumab was the IgG that was proteolytically digested to give the Fab targeted to VEGF and trastuzumab was the IgG source for the Fab targeted to HER2. The accessible disulfide of Fab_VEGF was reduced with dithiothreitol (DTT) (Fig. 1A, lanes 2 and 3).

Fig. 1. (A) Representative SDS–PAGE analysis for preparation of bsFpF 2̲ using the di(bis-sulfone) reagent 3̲ and the SDS-PAGE was stained with instant blue for protein staining (lanes 1–15) and silver staining for detecting any trace of impurity (lane 14), lane 1: Novex pre-stained protein marker, lane 2: Fab_VEGF (no DTT), lane 3: Fab_VEGF + DTT, lane 4: reaction mixture between PEG reagent 3̲ (4–5 eq.) and reduced-Fab_VEGF (1 eq.) after 1 h incubation at ambient temperature, lane 5: Fab_VEGF–PEG–bis sulfone 4̲ was obtained using ion-exchange chromatography, lane 6: Fab_HER-2 + DTT, lane 7: reaction mixture between Fab_VEGF–PEG–bis sulfone 4̲ and reduced Fab_HER-2 (crude PEG reagent 3̲ was used for first conjugation), lane 8: reaction mixture between PEG–Fab_VEGF6̲ and reduced Fab_HER-2 (HPLC purified PEG reagent 3̲ was used for the conjugation), lanes 9–13, purification fraction using size-exclusion chromatography, lane 14: purified Fab_VEGF–PEG₁₀–Fab_HER-22̲ and lane 15: purified Fab_VEGF–PEG₁₀–Fab_TNFa2̲ appeared at the 110 kDa MW band. (B) SDS–PAGE analysis for scouting reactions at 2, 15 and 48 h for the conjugation of the second Fab (Fab_HER2) to the intermediate conjugate 4̲ (Fab_VEGF–PEG–bis sulfone). (C) Synthesis of PEG–Fab 6̲. (D) Average kinetic rate constants and parameters achieved for Fab_VEGF, PEG₁₀–Fab_VEGF6̲ and bispecific Fab_VEGF–PEG₁₀–Fab_HER-22̲ using a CM3 chip immobilised with VEGF (55 RU). The numbers of replicates were 2 for bispecific conjugate and 3 for other samples. (E) Dissociation profiles overlayed for Fab_VEGF, PEG₁₀–Fab_VEGF6̲ and Fab_VEGF–PEG₁₀–Fab_HER-22̲.

Fig. 1 After removal of DTT using a PD-10 column,^1,7 the reduced Fab (1.0 mg in 3.3 mL) was allowed to incubate with an excess of the di(bis-sulfone) reagent 3̲ (5 equivalents, 1 mg) in the PEGylation buffer with pH 7.8 for 1 hour at ambient temperature (Fig. 1A, lane 4). The reaction solution essentially comprised a mixture of the desired intermediate conjugate Fab_VEGF–PEG₁₀–bis-sulfone 4̲, unreacted di(bis)-sulfone reagent 3̲, trace starting Fab_VEGF and the undesired homodimer FpF, Fab_VEGF–PEG₁₀–Fab_VEGF. The excess di(bis-sulfone) reagent 3̲ was used to minimise the formation of the homodimer, Fab_VEGF–PEG₁₀–Fab_VEGF (Fig. 1A, lane 4). The reaction mixture was then eluted over an ion exchange column (a MacroCap SP) to give predominantly a band at 70 kDa, thought to be the desired intermediate conjugate Fab_VEGF–PEG₁₀–bis-sulfone 4̲ (Fig. 1A, lane 5). The excess di(bis-sulfone) reagent 3̲ was removed from the reaction mixture preferably by ion exchange chromatography or by centrifugal filtration to prevent suppression of bsFpF formation.

In a separate vial, Fab_HER2 (1.0 mg in 1.0 mL of the PEGylation buffer) was incubated with DTT (1.0 mg), then DTT was removed and reduced-Fab_HER2 (Fig. 1A, lane 6) was incubated with the intermediate conjugate, Fab_VEGF–PEG₁₀–bis-sulfone 4̲ for 12 hours to give the desired bsFpF 2̲ (Fab_VEGF–PEG₁₀–Fab_HER2). Scouting reactions (Fig. 1B) indicated that the conjugation of Fab_HER2 to the intermediate Fab_VEGF–PEG₁₀–bis-sulfone 4̲ required a longer incubation time than conjugation of the first Fab (Fab_VEGF) to the starting di(bis-sulfone) reagent 3̲.

Purification of the di(bis-sulfone) reagent 3̲ by HPLC (Fig. S1, ESI†) resulted in better conversion to the desired Fab_VEGF–PEG₁₀–Fab_HER22̲ (Fig. 1A, lanes 7 and 8) which was purified by size exclusion chromatography (SEC) (Fig. 1A, lanes 9–13). The purity of the bsFpF 2̲, Fab_VEGF–PEG₁₀–Fab_HER2, was confirmed by silver staining (Fig. 1A, lane 14).

Binding of Fab_VEGF–PEG₁₀–Fab_HER22̲ was evaluated by surface plasmon resonance (SPR) with each ligand immobilised on separate CM3 chips at low response units to allow kinetic studies to be conducted (VEGF (55 RU) and HER2 (65 RU)). Ligand binding was first confirmed with the parent antibodies and bevacizumab for VEGF and trastuzumab for HER2. Additionally, there was no non-specific binding observed when bevacizumab was incubated with the immobilised HER2 chip and when trastuzumab was incubated with the immobilised VEGF chip (Fig. S2A and B, ESI†). The bsFpF 2̲, Fab_VEGF–PEG₁₀–Fab_HER2, displayed concentration-dependent binding to each immobilised ligand (Fig. S2C and D, ESI†). Both VEGF and HER2 were immobilised to a single CM3 chip and binding of bevacizumab and trastuzumab was observed (control) as well as the concentration-dependent binding of bsFpF, Fab_VEGF–PEG₂₀–Fab_HER2 (Fig. S2E and F, ESI†).

SPR kinetic studies were performed with Fab_VEGF, PEG₁₀–Fab_VEGF6̲ and the bsFpF 2̲, Fab_VEGF–PEG₁₀–Fab_HER2 using VEGF immobilised to a CM3 chip. PEG₁₀–Fab_VEGF6̲ was prepared from the PEG₁₀ bis-sulfone reagent 5̲ used for protein PEGylation as previously described (Fig. 1C).^1,2,40 The SPR data indicated that Fab_VEGF exhibited a faster association rate constant (k_a) compared to both PEG₁₀–Fab_VEGF6̲ and Fab_VEGF–PEG₁₀–Fab_HER22̲. This is likely attributed to the smaller molecular weight of Fab_VEGF. However, no discernible difference was observed in the dissociation rate constant (k_d) between Fab_VEGF, PEG₁₀–Fab_VEGF6̲ and Fab_VEGF–PEG₁₀–Fab_HER22̲ upon dissociation from immobilised VEGF (Fig. 1D and E).

A second bsFpF 2̲ (Fab_VEGF–PEG₁₀–FabTNF_α) derived from Fab_VEGF and a Fab targeted to tumour necrosis factor alpha (TNF-α) was prepared using the di(bis-sulfone) reagent 3̲ (Fig. 1A, lane 15). Fab_VEGF–PEG₁₀–FabTNF_α also displayed concentration-dependent binding as observed by SPR to both VEGF and TNF-α (Fig. S3, ESI†).

During the preparation of bsFpFs 2̲ using the di(bis-sulfone) reagent 3̲, it was thought that a reagent with more different conjugation reactivity at each PEG terminus would allow more efficient preparation of a bsFpF 2̲ by a conjugation only approach. It was thought that a reagent with reduced conjugation reactivity on one terminus would require a lower excess of the FpF reagent for the first Fab conjugation while producing a less homodimer.

The bis-sulfone conjugation moiety functions by a sequence of addition–elimination reactions (Scheme S1, ESI†). Initial elimination of one equivalent of the toluene sulfinic acid leaving group 8̲ is necessary to generate the α,β-unsaturated carbonyl moiety (e.g. structure 9̲, Scheme 3A). The initial elimination reaction is driven by the pK_a of the α-proton to the carbonyl electron-withdrawing group in the bis-sulfone conjugating moiety 7̲ (Scheme 3A). If the pK_a value of the α-proton was increased slightly as in the bis-sulfide precursor 1̲0̲ (Scheme 3A), this would reduce the rate of the initial elimination step to potentially slow conjugation compared to the bis-sulfone moiety.

Scheme 3. (A) Elimination reactions of the bis-sulfone 7̲ and bis-sulfide 1̲0̲ conjugation moieties to yield enones capable of alkylation by a thiolate from a reduced disulfide. The pK_a value of the α-proton in bis-sulfone 7̲ is thought to be slightly less than that of the α-proton in bis-sulfide 1̲0̲. (B) Estimates of leaving group pK_as.

The conjugation reactivity of the bis-sulfone moiety 7̲ is also due to the pK_a of the toluene sulfinic acid leaving group 8̲ (Scheme 3B) which we estimate to be about ∼1.6 and is much lower than that for the cysteine thiol (pK_a ∼ 10) to drive conjugation. Increasing the pK_a of the leaving groups (e.g. structures 1̲1̲ and 1̲3̲–1̲5̲, Scheme 3) relative to toluene sulfinic acid 8̲ would potentially contribute to reduced conjugation reactivity. We therefore sought to examine the asymmetric conjugation reagent 1̲6̲ (Scheme 4). A bsFpF could potentially be made by conjugation to the bis-sulfone moiety with the first Fab (Fab₁) and then conjugation of the second Fab (Fab₂) to the less reactive bis-sulfide moiety in the intermediate Fab–PEG–bis sulfide intermediate 1̲7̲ (Scheme 4).

Scheme 4. Preparation of a bsFpF 2̲ by the conjugation of Fab₁ with the bis-sulfone bis-sulfide reagent 1̲6̲. Conjugation is intended to occur on the bis-sulfone moiety of reagent 16 to give the intermediate conjugate 1̲7̲ can be purified by IEX or by spin filtration to remove the excess starting reagent 1̲6̲ and unreacted Fab₁. Conjugation of Fab₂ to the intermediate conjugate 1̲7̲ gives the bsFpF 2̲.

We prepared the bis-sulfide PEGylation reagents 1̲8̲, 1̲9̲, 2̲0̲, and 2̲1̲ (Structures 2) to examine relative conjugation reactivities with the PEG–bis sulfone 5̲ (Fig. 1C). PEG₁₀–bis-sulfide 1̲8̲ with the unsubstituted tolyl thiol leaving group was not reactive enough under the mild conditions normally employed for Fab conjugation (Figures S4A and B, ESI†). In contrast, the aryl tetra-fluoro thiol leaving group in PEG₁₀–bis sulfide 2̲1̲ appeared to have comparable conjugation reactivity with the bis-sulfone moiety in PEG–bis sulfone 5̲ (Figure S4A, ESI†).

The ortho-fluoro and di-fluoro PEG₁₀–bis sulfides 1̲9̲ and 2̲0̲ underwent Fab conjugation (Fig. 2A and B). Reagents 1̲9̲ and 2̲0̲ appeared to have sufficient conjugation reactivity while being less reactive than that of the bis-sulfone moiety. An experiment to examine the sequential conjugation reactions of Fab_VEGF to ortho-fluoro bis-sulfide bis-sulfone reagent 2̲2̲ did appear to give a good conversion to homodimeric FpF 1̲ (Fig. 2C). An analogous conjugation was conducted with the di-fluoro bis-sulfide bis-sulfone reagent 2̲3̲ where formation to the desired FpF (Fig. S4C, ESI†) was also observed. These preliminary efforts indicated that considerable Fab₂ remained which made the IEX purification difficult necessitating SEC purification. Fab₂ conjugation also appeared to be slower with the intermediate conjugate Fab₁–PEG–bis sulfide intermediate 1̲7̲ than Fab conjugation with PEG–fluoro-bis-sulfone (e.g.1̲9̲).

Fig. 2. Conjugation of fluorodisulfide reagents (novex BisTris 4–12% gel stained with oomassie blue). (A) Scouting conjugation of Fab_VEGF to PEG₁₀–bis-sulfide ortho-fluoride 1̲9̲ (2 eq.) from pH 4 to 8 at different incubation times. Lane 1: protein markers, lanes 2–7: 3-hour incubation time, lane 8: blank, lanes 9–14: 18-hour incubation time, lanes 15–20: 23-hour incubation time. (B) Scouting conjugation of Fab_TNFα to PEG₁₀–bis-sulfide-difluoride 2̲0̲ (2 eq.) from pH 5 to 8 at different incubation times. Lane 1: protein markers, lanes 2–5: 5-hour incubation time, lane 6: protein markers, lanes 7–10: 18-hour incubation time, lane 11: protein markers, lanes 12–15: 23-hour incubation time. (C) Scouting reactions to prepare a FpF in two steps using the ortho-fluoro bis-sulfide bis-sulfone reagent 2̲2̲. The same Fab_VEGF was used in both conjugation steps, so FpF 1̲ was prepared rather than a bsFpF 2̲: (i) first Fab_VEGF conjugation reaction, lane 1: protein markers, lanes 2–3: 2 batches of reduced Fab_VEGF before PD10 elution, lanes 4 and 5: reduced Fab_VEGF after PD10 elution, lanes 6 and 7: first conjugation of Fab_VEGF to the reagent 2̲2̲. (ii) Second Fab_VEGF conjugation reaction, lanes 2 and 3: 2 batches of reduced Fab_VEGF before PD10 elution, lanes 4 and 5: reduced Fab_VEGF after PD10 elution, lanes 6–10: second conjugation of increasing stoichiometries of Fab_VEGF with the corresponding intermediate conjugate, Fab_VEGF–PEG₁₀–ortho bis sulfone 1̲7̲, which had not been eluted over an IEX column, lanes 11 and 12: second conjugation of increasing stoichiometries of Fab_VEGF with the corresponding intermediate conjugate, Fab_VEGF–PEG₁₀–ortho bis sulfone 1̲7̲, which had been eluted over an IEX column. (iii) Fractions from SEC purification of the FpF 1̲, Fab_VEGF–PEG₁₀–Fab_VEGF after a first step of IEX elution.

bsFpFs prepared by conjugation–ligation

Considering the encouraging binding properties of Fab_VEGF–PEG₁₀–Fab_HER2, we wanted to evaluate a larger number of bsFpF molecules. The di(bis-sulfone) 3̲ and the asymmetric conjugation bis-sulfide bis-sulfone reagents 1̲6̲ are good reagents and can be used to prepare a bsFpF or other heterodimeric protein conjugates. To prepare several bsFpFs that could be used in early preclinical research, we decided to examine preparing these molecules by a conjugation–ligation strategy (Scheme 5). Conjugation–ligation strategies to prepare protein–protein and protein–drug conjugates are well known.^41–43 In principle, conjugation ligation reagents 2̲4̲ and 2̲5̲ are first used to site-specifically modify separate proteins by disulfide rebridging conjugation to give intermediate conjugates 2̲6̲ and 2̲7̲ that can be ligated to give the desired bsFpF 2̲. Different Fabs can be used to provide stock solutions of the intermediate conjugates 2̲6̲ and 2̲7̲ that can be ligated to form a family of bsFpFs. Such an approach minimises the number of protein conjugation reactions needed and the need to purify bsFpFs from the unmodified protein.

Scheme 5. A conjugation–ligation strategy to prepare bsFpFs 2̲. The conjugate intermediates (e.g.2̲6̲ and 2̲7̲) can be used with different Fabs in different combinations to prepare families of bsFpFs 2̲.

The ligation step requires reactive moieties that will not undergo a reaction with the conjugated protein. Many ligation strategies and chemical moieties have been described, and we focus here on ligation moieties capable of undergoing cycloaddition.^41,43 To prepare bsFpFs by a conjugation–ligation approach, we examined the bis-sulfone trans-cyclooctene (TCO) and tetrazine (Tz) reagents 2̲8̲ and 3̲0̲ (Scheme 6). The bis-sulfone–PEG–TCO reagent 2̲8̲ was derived from a mono-Boc protected PEG di-amine precursor with a molecular weight of either 5 and 10 kDa and the bis-sulfone–PEG–Tz reagent 3̲0̲ was derived from a PEG precursor of 5 and 10 kDa (ESI†). The TCO and Tz bis-sulfone reagents 2̲8̲ and 3̲0̲ readily underwent ligation at pH values ranging from 5 to 9 (ESI,† Fig. S5). Several Fabs were conjugated to the TCO and Tz bis-sulfone reagents 2̲8̲ and 3̲0̲ using our standard conjugation conditions with 1.5 to 2.0 equivalents of the conjugation reagent^40,44 to produce intermediate Fab conjugates 2̲9̲ and 3̲1̲ for ligation (Scheme 6).

A representative SDS–PAGE gel is shown in Fig. 3 displaying the intermediate conjugates 2̲9̲ and 3̲1̲ (lanes 4 and 5). Ligation to produce the bsFpF 2̲ (or the homodimeric FpF 1̲) was accomplished by first eluting the individual conjugation reaction mixtures over an ion exchange column to remove any un-conjugated TCO and TZ bis-sulfone reagents 2̲8̲ and 3̲0̲. It was also possible to remove unconjugated reagents by viva spin. A representative ligation reaction mixture is shown in the gel in Fig. 3 (lane 6). Purification of the final bsFpFs 2̲ was achieved by IEX and SEC with examples shown in lanes 7 to 12 (silver staining was used for detection in lanes 9–12) with the bsFpFs displaying a band at about 115 kDa (Fig. 3).

Fig. 3. Representative SDS–PAGE analysis for preparation of bsFpF 2̲ using the bis-sulfone conjugation–ligation reagents 2̲8̲ and 3̲0̲, and the SDS–PAGE was stained with instant blue for protein staining (lanes 1–8) and silver staining for detecting any trace of impurity (lanes 9 to 12), lane 1: protein markers, lane 2: Fab_VEGF, lane 3: reduced Fab_VEGF, lane 4: Fab_VEGF conjugated to the bis-sulfone–PEG₁₀–TCO reagent 2̲8̲, lane 5: Fab_VEGF conjugated to the bis-sulfone–PEG₅–Tz reagent 3̲0̲, lane 6: reaction mixture for the ligation reaction between conjugate intermediates Fab_VEGF–PEG₁₀–TCO 2̲8̲ and Fab_VEGF–PEG₅–Tz 3̲0̲, lanes 7–12: purified bsFpFs 2̲ (Fab₁–PEG₁₅–Fab₂, silver stain used for lanes 9–12 to assess purity). Different Fabs are used for conjugation; lane 7: homodimer Fab_VEGF–PEG₁₅–Fab_VEGF, lane 8: Fab_VEGF–PEG₁₅–Fab_HA, lane 9: Fab_VEGF–PEG₁₅–Fab_IL6R, lane 10: Fab_VEGF–PEG₁₅–FabTNF_α, lane 11: Fab_IL6R–PEG₁₅–FabTNF_α, and lane 12: Fab_VEGF–PEG₁₅–Fab_COL2.

Fig. 3 Several clinically approved IgGs including bevacizumab (for anti-VEGF Fab), tocilizumab (for anti-IL6R Fab), infliximab (for anti-TNFα Fab) and secukinumab (for anti-IL17 Fab) were digested to provide the Fabs that were used. We have previously described scaling of the papain digestion process to accommodate 100 mg of IgG resulting in the isolation of 50 mg of pure and stable Fab.³⁹Table 1 lists the bsFpFs that were prepared by conjugation–ligation using the TCO and Tz bis-sulfone reagents 2̲8̲ and 3̲0̲. To facilitate binding comparisons, the homodimer FpF_VEGF was also prepared using the same reagents 2̲8̲ and 3̲0̲, with Fab_VEGF sourced from digestion of bevacizumab. Additionally, another homodimer FpF_VEGF construct was prepared using the di(bis-sulfone) reagent 3̲. The concentration of both FpFs 1̲ and bsFpFs 2̲ was determined by the micro-BCA-assay. Isolated yields varied between 15 and 20% at the reaction scales used and typically about 0.2 mg of purified bsFpF could be obtained starting from 1 mg of each Fab.

List of all the bsFpFs, prepared using conjugation–ligation reagents 3̲2̲ and 3̲4̲ targeting pro-inflammatory targets (TNFα, IL6R and IL17) and pro-angiogenic target (VEGF) and affinity targets (HA and collagen-2) along with their isolation yields.

Fab1	Fab2	bsFpF 2̲ Isolated yields %
FabVEGF	FabIL6R	20
FabVEGF	FabTNFα	16
FabTNFα	FabIL6R	14
FabVEGF	FabIL17A	11
FabTNFα	FabIL17A	11
FabVEGF	FabCOL2	13
FabVEGF	FabHA	12

A conjugation–ligation–conjugation strategy was also examined (Scheme 7). The Fab_VEGF–PEG₁₀–TCO conjugate 2̲9̲ was ligated with the Tz–PEG₅–bis-sulfone reagent 3̲0̲ (1.5 equivalents) to give the ligation intermediate, Fab_VEGF–PEG₁₅–bis-sulfone 3̲2̲ (Fig. 4, lane 6). The Fab targeted to an interleukin-6 receptor (Fab_IL6R) was then conjugated to intermediate 3̲2̲ to give the final bsFpF 2̲ (Fab_VEGF–PEG₁₅–Fab_IL6R) (Fig. 4, lane 7). The conversion to give this bsFpF, Fab_VEGF–PEG₁₅–Fab_ILR6 appeared to be about the same as when prepared by ligation of the separate conjugate intermediates 2̲9̲ and 3̲1̲ (Scheme 6).

Scheme 7. A conjugation–ligation–conjugation strategy to prepare the bsFpF 2̲. The Fab–PEG–TCO conjugate intermediate 2̲9̲ undergoes ligation with the bis-sulfone–PEG–Tz reagent 3̲0̲ to give the intermediate Fab–PEG–TCO 3̲2̲. A second Fab is then conjugated to the intermediate 3̲6̲ to give the final bsFpF 2̲.

Fig. 4. The SDS–PAGE gel displaying preparation of a bsFpF 2̲ by conjugation–ligation–conjugation; Fab is first conjugated to bis-sulfone–PEG₁₀–TCO 2̲8̲ to give the Fab–PEG₁₀–TCO conjugate intermediate 2̲9̲ which is then ligated to bis-sulfone–PEG₅–Tz 3̲0̲ to form the bis-sulfone terminated ligation intermediate 3̲2̲ (note Scheme 7) which can undergo conjugation with another Fab to give the bsFpF. Lane 1: protein markers, lane 2: Fab_VEGF, lane 3: Fab_VEGF + DTT, lanes 4 and 5: conjugation of reduced Fab_VEGF to bis-sulfone–PEG₁₀–TCO 2̲8̲, and then purified, lane 6: Fab_VEGF–PEG₁₀–TCO ligation with bis-sulfone–PEG₅–Tz 3̲0̲ to give the ligation intermediate 3̲2̲ which then underwent conjugation with reduced Fab_IL6R in lane 7: resulting in the formation of the Fab_VEGF–PEG₁₅–PEG_IL6R bispecific molecule.

We also prepared bis-sulfone–PEG₃-N₃3̲3̲ and bis-sulfone–PEG₅–DBCO 3̲5̲ (ESI,† Fig. S6) to examine the well-known strain-promoted alkyne–azide cycloaddition which uses azide (N₃) and dibenzocyclooctyene (DBCO) moieties for ligation.^42,45,46 Good conjugation conversion with Fab_TNFα was achieved with 1.5 to 2.0 equivalents of the bis-sulfone reagents 3̲3̲ and 3̲5̲ after a 5-hour incubation time at pH 8 to give the respective conjugate intermediates 3̲4̲ and 3̲6̲ (ESI,† Fig. S6A and B). The conjugation solutions were centrifuged using a viva spin column to remove excess starting reagents. Upon mixing the intermediate conjugates 3̲4̲ and 3̲6̲ (ESI,† Fig. S6C), the desired homodimeric FpF 1̲ was formed but at apparently lower conversion than with the corresponding conjugate intermediates derived from the TCO and Tz bis-sulfone reagents 2̲8̲ and 3̲0̲. Similar ligation results were observed for preparation of the homodimeric FpF derived from Fab_VEGF (ESI,† Fig. S7A) and ligation could not be improved when an excess reagent was removed by IEX chromatography instead of viva spin. Commercially available methoxy PEG₁₀–DBCO and bis-sulfone PEG₃-N₃ did undergo ligation as would be expected without the presence of the conjugated Fabs (ESI,† Fig. S7B, lane 4).

The presence of the conjugated protein in each ligation intermediate (e.g. Fab–PEG-N₃3̲4̲ and Fab–PEG–DBCO 3̲6̲, ESI,† Fig. S6) may have resulted in conformational masking of the hydrophobic ligation moiety in the relatively large PEG linker element when one terminus of the PEG is conjugated to a 50 kDa protein (i.e. Fab). The TCO and Tz moieties are known to undergo faster ligation than the azide and DBCO moieties⁴³ and this may allow more efficient ligation of the Fab conjugate intermediates. Shorter PEG linkers would be expected to allow more facile ligation.

The ELISA was first used to evaluate the binding affinity of Fab_VEGF–PEG₁₅–Fab_IL6R. Separate plates were coated with VEGF and IL6R. There was no non-specific binding observed when infliximab (anti-TNFα IgG) was incubated with the VEGF or IL6R coated plates. The binding affinities (K_D) determined by the ELISA for the parent anti-IL6R antibody (tocilizumab) and the associated Fab_IL6R obtained by proteolytic digestion were 0.13 nM for tocilizumab and 1.50 nM for Fab_IL6R (ESI,† Fig. S8). As expected, tocilizumab exhibited a lower K_D and higher binding affinity due to its bivalent nature as an IgG compared to the monovalent Fab_IL6R. ELISA affinities represent average values from two replicates. The ELISA binding affinity of the bispecific Fab_VEGF–PEG₁₅–Fab_IL6R to VEGF (K_D 1.80 nM) and IL6R (K_D 2.55 nM) is shown in Fig. 5. For comparison, the ELISA derived binding of PEG₁₀–Fab_VEGF (K_D 2.25 nM) and PEG₅–Fab_IL6R (K_D 3.20 nM) was determined (Table 2).

Fig. 5. ELISA results for (A) Fab_VEGF–PEG₁₅–Fab_IL6R (green data point) and mono PEG–Fab_VEGF (red data point) with a concentration range of 1.3 × 10⁻⁷ to 1.3 × 10⁻¹³ M over wells coated with VEGF (0.1 μg). (B) Fab_VEGF–PEG₁₅–Fab_IL6R (green data point) and mono PEG–Fab_IL6R (red data point) with a concentration range of 1.3 × 10⁻⁷ to 1.3 × 10⁻¹³ M over wells coated with IL6R (0.1 μg).

Binding affinities (K_D) for mono-PEG–Fabs and Fab_VEGF–PEG₁₅–Fab_IL6R against VEGF and IL6R using the ELISA technique.

Conjugates	Ligand	K D (nM)
PEG10–FabVEGF	VEGF	2.25
FabVEGF–PEG15–FabIL6R	VEGF	1.80
PEG5–FabIL6R	IL6R	3.20
FabVEGF–PEG15–FabIL6R	IL6R	2.55

The binding affinity of Fab_VEGF–PEG₁₅–Fab_IL6R was also evaluated by SPR. VEGF₁₆₅ was immobilised to a CM3 chip (95 RU) and his-Tag IL6R was immobilised to a nitrilotriacetic acid (NTA) chip. The precursor Fabs and PEG–Fab conjugates were used for comparison (i.e. Fab_VEGF, Fab_IL6R, PEG₁₀–Fab_VEGF and PEG₅–Fab_IL6R). The concentration-dependent binding of the bispecific Fab_VEGF–PEG₁₅–Fab_IL6R was observed for both VEGF and IL6R (ESI,† Fig. S9). The kinetic rate constants and affinities were calculated (Table 3) using a 1 : 1 binding model. ESI,† Table S1 summarises the SPR derived kinetic rate constants and affinities that were obtained for bsFpFs prepared by conjugation–ligation using the TCO and Tz bis-sulfone reagents 2̲8̲ and 3̲0̲.

Average kinetic constant rates of mono-PEG–Fabs and Fab_VEGF–PEG₁₅–Fab_IL6R (N = 3) using the 1 : 1 binding model. Chi² values, a fitting quantitative measure, were 0.12 for Fab_VEGF–PEG₁₅–Fab_IL6R and 0.52 for mono-PEG–Fabs. The optimal Chi² value is within a 10% range of the R_max value.

Conjugates	Ligand	k a (1/Ms) × 104	SD ka × 104	k d (1/s) × 10−4	SD kd × 10−4	K D (nM)	SD KD
PEG10–FabVEGF	VEGF	1.30	0.06	1.4	0.35	10.9	3.0
FabVEGF–PEG15–FabIL6R	VEGF	0.84	0.31	1.0	0.4	12	7.8
PEG5–FabIL6R	IL6R	3.90	1.83	12.5	4.94	38.1	29.6
FabVEGF–PEG15–FabIL6R	IL6R	2.80	1.30	5.7	0.56	23.3	14.4

The homodimer FpF_VEGF prepared by the di(bis-sulfone) PEG reagent 3̲ appeared to have lower K_D due to a faster association rate (k_a) constant and a slower dissociation rate (k_d) compared to the homodimer FpF_VEGF synthesised by a conjugation–ligation (reagents 2̲8̲ and 3̲0̲) approach, studied by both ELISA and SPR (Fig. 6).

Fig. 6. Binding analysis of the homodimer FpF_VEGF prepared using the conjugation reagent 3̲ and conjugation–ligation reagents 2̲8̲ and 3̲0̲ using (A) ELISA, plate coated with VEGF₁₆₅ (0.1 μg) and (B) SPR technique, and the CM3 chip immobilised with VEGF (95 RU).

Immunoblotting (dot blot assays) is also useful for qualitative assessment of binding interactions; for example, Fab_VEGF–PEG₁₅–Fab_COL2 displayed binding to its respective antigens using a dot blot assay (Fig. S10, ESI†). A limitation of ELISA, SPR and dot blot assays is that these techniques require immobilisation of one of the binding partners which may not fully represent the dynamic interactions of the binding partners in solution.⁴⁷ Microscale thermophoresis (MST) enables the practical measurement of binding interactions in solution. Using MST, the binding analysis of bispecific Fab_VEGF–PEG₁₅–Fab_IL6R was performed on VEGF, IL6R and a mixture of VEGF/IL6R (Fig. S11, ESI†). The binding traces revealed that the bispecific Fab_VEGF–PEG₁₅–Fab_IL6R maintained its binding towards both VEGF and IL6R, both as individual targets and when presented as a mixed target (VEGF/IL6R) in solution.

Discussion

There is a rich history of investigating dimeric low molecular weight molecules and proteins linked by PEG, e.g.^48–51 As a dimeric antibody-based molecule, the FpF motif is a good IgG binding mimetic.^1,2 Each Fab is covalently bound to a terminus of a linear, flexible PEG in the region where the Fab is naturally anchored in an IgG. The PEG linker appears to mimic some of the conformational properties of the IgG hinge to allow flexible Fab–epitope interactions during binding, which are important for antibody functions.³ FpFs evaluated to date appear to have comparable binding thermodynamics to the corresponding IgG antibody with the same Fabs.^1,2 FpF affinity is achieved by a slower association rate (k_a) that is then compensated by a correspondingly slower dissociation rate (k_d). A slower k_d may have important therapeutic potential to help maintain localised therapeutic concentrations in tissues.^52–54

FpFs do not have the extended solution structure of a PEG–Fab conjugate where there is a protein only on one terminus of the PEG molecule.¹ FpFs appear to have a similar solution size to IgGs broadly independent on the PEG linker size in the range of 5–20 kDa.¹ It is thought that the Fab moieties in a FpF may self-associate to reduce the size of PEG in solution compared to what is observed for PEGylated proteins with a protein conjugated at only one PEG terminus. FpFs in solution may possess some of the conformational properties associated with A–B–A block copolymers^55,56 where the Fabs (A block) can self-associate. Since bsFpFs have a Fab at each terminus of a PEG molecule like the homo-dimeric FpFs, it would be expected that the bsFpFs will have a similar size to the homodimeric FpFs.

It will be important to examine the differences in the solution structures for the FpFs made by the conjugation-only approach and by the conjugation–ligation approach where there is a ligation element located within the PEG linker. Although binding appears broadly similar, the ligation elements may exert conformational influences (Fig. 6). It is also clear from other studies that the PEG linker length in dimeric protein conjugates can influence binding, especially below a threshold length that is shorter than the ligand or the epitope distance.^49,50,57 We recognise that there is also potential to optimise the bsFpF linker length depending on the specific application.

Since the IgG hinge region is susceptible to degradation, the use of a PEG linker and site-specific bis-alkylation conjugation at the accessible Fab disulfide contribute to the stability and reduced propensity for aggregation of FpFs.¹ The thiol–ether bonds conjugating each Fab to the flexible PEG linker acting as the surrogate hinge are more stable than the unmodified accessible disulfide in the Fab. Much effort remains focused on developing antibody-based molecules with optimal physicochemical properties,⁵⁸ some of which FpFs may have the potential to display.

Multifunctional protein conjugates including bispecific antibodies can be made using proteins that are readily accessible by recombinant means (e.g. Fabs and single chain fragments) which are then conjugated by selective chemical strategies. Outstanding studies have been published exemplifying this approach,^35,59–61 including strategies to conjugate proteins or peptides at each terminus of a functionalised PEG linker analogous to what we have described for FpFs.^9,10,62 FpFs comprise elements (e.g. Fabs and PEG) which separately have been clinically proven viable for use.

The di(bis-sulfone) reagent 3̲ was examined to prepare the heterodimeric bsFpFs 2̲ because this is an effective reagent to prepare the homodimeric FpFs 1̲. We anticipated that conjugation of the first Fab to the di(bis-sulfone) 3̲ would be faster than the second Fab due to possible hydrophilic steric shielding effects of the second bis-sulfone conjugation moiety after conjugation of Fab₁. Modification of the leaving groups in the bis-sulfone conjugation moiety to increase hydrophilicity was considered,⁶³ although an excess reagent would still be required for the first conjugation step in a conjugation-only approach with a di(bis-sulfone) reagent. The addition of excess di(bis-sulfone) 3̲ during the conjugation of Fab₁ predominantly gave the desired Fab₁–PEG–bis sulfone 4̲ intermediate (e.g. Fab_VEGF–PEG₁₀–bis-sulfone 4̲, Fig. 1A). The use of excess di(bis-sulfone) 3̲ necessitates its removal prior to the conjugation of Fab₂. Ion exchange chromatography (IEX) effectively removed the excess reagent along with trace remaining Fab₁ and the homodimer (i.e. Fab₁–PEG–Fab₁).

Conjugation of Fab₂ required longer incubation times (Fig. 1B), which may not be ideal for maintaining Fab₂ stability and preventing reoxidation of the Fab₂ accessible disulfide. The di(bis)-sulfone 3̲ was purified by precipitation,⁶⁴ which works well for reagents designed to undergo a single conjugation reaction, e.g. bis-sulfone PEGylation reagents 5̲ (Fig. 1C).⁶⁵ The use of pure PEG precursors and the chromatographic purification of the di(bis-sulfone) 3̲ does give a reagent with less possible dead chain ends (Fig. S1, ESI†)⁹ and does appear more effective for the second conjugation step (Fig. 1A, lane 8). Although we are keen to utilise reagents that could be prepared without a tedious purification to isolate the reagent, it is important to acknowledge that minimisation of dead chain ends in all conjugation reagents is generally preferred.

The asymmetric FpF reagents 1̲6̲ (Schemes 3 and 4) indicate that we could reduce the excess of the reagent needed for the conjugation of Fab₁ and reduce the formation of the undesired homodimer. The fluoro-substituted bis-sulfide conjugation moieties are less reactive than the bis-sulfone conjugation moiety; however, all these bis-alkylation moieties yield the same conjugate product (e.g. PEG–Fab 6̲, compare Fig. 1C and 2A). There are few if any di-conjugation reagents that undergo the same site-specific conjugation reaction at each moiety but with varied reactivity.

Our preliminary experiments indicated that the combination of reduced reactivity for the bis-sulfide conjugation moiety for Fab₂ and the reduced reaction rate for the second conjugation to the Fab–PEG–bis-sulfide intermediate 1̲7̲ (Scheme 4) meant that there was often remaining Fab₂ present with the conditions that were examined. Considering the advantages of a cleaner reaction for Fab₁, more work is required to optimise the conjugation of the second protein (e.g. Fab₂).

The ligation approach to prepare bsFpFs is to utilise bis-alkylation conjugation that site-specifically rebridges the two cysteine thiols from the accessible native disulfide of a Fab to give conjugate intermediates that are then ligated via an orthogonal cycloaddition reaction. Many ligation strategies have been described (e.g. ref. 66–68). Ligations are often accomplished by orthogonal reactions (e.g. ref. 69), with cycloaddition reactions dominating in recent years (e.g. ref. 45, 46 and 68) to make multifunctional proteins including bispecific antibody mimetics.^35,45 Protein modification strategies involving conjugation and/or ligation will inevitably evolve in the context of engineering protein structure recombinantly to best match a given modification strategy. Although our use of reagents with PEG molecular weights greater than 3 kDa is designed to yield bsFpFs that can be optimised to exploit spatial–temporal relationships that exist with IgGs, the use of these relatively long PEG linkers may also result in conjugation intermediates (e.g. Fab–PEG₁₀–TCO, Fab–PEG₅–Tz) that have reduced ligation reactivity due to conformational masking of the ligation moiety.

Conjugation–ligation (Scheme 6) allows for a combinatorial approach to prepare bsFpFs 2̲. For example, one conjugate intermediate (e.g.2̲9̲) could undergo ligation with many different versions of its partner conjugate intermediate (e.g.3̲1̲) to give a small family of bsFpFs 2̲. Reaction orthogonality to the protein in the ligation step also reduces the potential for non-selective protein conjugation which is possible by the conjugation only approach when longer incubation times are used (e.g. >1 day). A conjugation–ligation approach also avoids generation of homo-dimeric protein conjugates that can result from a conjugation only approach, minimising the number of protein conjugation reactions needed and the need to purify bsFpFs from the unmodified protein.

Ligation using trans-cyclooctene (TCO) and tetrazine (Tz) moieties is faster than the cycloaddition between azide and DBCO moieties.^42,43,70 Bis sulfone–PEG–TCO 2̲8̲ and bis sulfone–PEG–Tz 3̲0̲ were prepared and used to make several bsFpFs 2̲ (Scheme 6) that were isolated (Fig. 3 and Table 1). We found that the isolated yields of purified proteins modified by different thiol specific conjugation strategies at a small scale (∼0.5 to 2.0 mg) often give moderate to low yields due to the loss of protein during purification.^71,72 It is possible with high conversion that disulfide rebridging PEGylation reactions at these or slightly higher scales give isolated yields of 45–65% of the modified protein (e.g. ref. 40). Although the conjugation–ligation reagents 2̲8̲ and 3̲2̲ were precipitated 3–4 times during isolation (ESI†), additional chromatographic purification of the reagent (e.g. Fig. S1, ESI†) would be expected to increase overall conversion and isolated yields of the bsFpFs 2̲ at the scales we examined.

Challenges exist in the assay development of dual-targeting molecules.^73–76 Using Fab_VEGF–PEG₁₅–Fab_IL6R as a representative example, we examined the binding affinity by dot blot, ELISA, SPR and MST (Tables 2 and 3). The K_D value obtained from the binding assays varies across these different experiments (compare the K_D value for the same bsFpF molecule between Tables 2 and 3). ELISA experiments indicate each Fab in Fab_VEGF–PEG₁₅–Fab_IL6R has a similar affinity (K_D) as the corresponding PEG–Fab (Table 2). PEG conjugation reduces protein activity generally, and certainly Fab affinity compared to the unmodified Fab due to steric shielding effects of PEG. Disulfide-rebridging conjugation at the accessible Fab disulfide is near the region of the Fab that is connected to the hinge in a native IgG and is maximally distal to the complementarity-determining region (CDR) responsible for Fab binding to its respective target. Conjugation at this accessible disulfide is thought to cause less reduction in Fab affinity than conjugation elsewhere in the Fab.^1,40

The binding kinetics to give the association (k_a) and dissociation (k_d) rate constants for three bsFpFs 2̲ determined by SPR are summarised in Table 3 and ESI,† Table S1. Low-density ligand immobilisation and a high flow rate (30 μL min⁻¹) were used to minimise mass transfer limitations and re-binding effects to the immobilised ligand.⁷⁷ Considering binding to immobilised VEGF₁₆₅, the apparent K_D and k_d values for mono-PEG₁₀–FabVEGF 6̲ (K_D 10.9 nM; k_d 1.4 × 10⁻⁴ 1/s) were similar to the bsFpFs 2̲ that with one Fab_VEGF; (i) Fab_VEGF–PEG₁₅–FabIL6R (K_D 12 nM and k_d 1.0 × 10⁻⁴ 1/s) (Table 3) and (ii) Fab_VEGF–PEG₁₅–FabTNF_α (K_D 10 nM and k_d 0.81 × 10⁻⁴ 1/s) (ESI,† Table S1).

We and others have previously shown that the dissociation rate constant of PEG modified Fabs remains broadly similar compared to the unmodified Fab.^40,78 This trend was also observed for the dissociation of Fab_beva in PEG₁₀–Fab_beva, and Fab_VEGF–PEG₁₅–Fab_IL6R from VEGF and Fab_IL6R in PEG₅–Fab_IL6R and Fab_VEGF–PEG₁₅–Fab_IL6R from IL6R. Similar trends in the dissociation rate constants were observed for Fab_VEGF–PEG₁₀–Fab_HER2 prepared using the PEG di(bis-sulfone) reagent 3̲ (Fig. 1D). The data indicate that there was a little difference between the dissociation rate constants for Fab_VEGF, PEG₁₀–Fab_VEGF and Fab_VEGF–PEG₁₀–Fab_Her-2 conjugates (Fig. 1D). There is potential that the preparation of PEG–Fab₁ will be a good surrogate for the binding of Fab₁ in the corresponding bsFpF (i.e. Fab₁–PEG–Fab₂).

The association rate constants (k_a) were smaller for the FabVEGF–PEG₁₅–FabIL6R products to both immobilised VEGF and IL6R compared to the corresponding mono-PEG–Fab constructs (Table 3). The association rate constant of Fab_VEGF–PEG₁₅–Fab_IL6R is slower probably because the bsFpF is larger in molecular weight than a PEG–Fab which results in a slower rate of transport from bulk solution to the chip surface. A similar trend was observed for Fab_VEGF–PEG₁₀–Fab_HER2 to have a slower association rate constant compared to PEG₁₀–Fab_VEGF. Reduction in binding affinities (K_D) for Fab conjugates compared to the corresponding unmodified Fab tends to be due to reductions in the association rate constants (k_a).^1,40,78 All bsFpFs molecules have a larger mass than PEG₁₀–Fab conjugates which results in a slower rate of transport over the chip surface.

ELISA, SPR and dot blot techniques are useful to confirm the binding of each Fab element in a bsFpF 2̲ but immobilisation of binding targets is required on a chip or the surface of a plate. Such assays do not fully replicate the dynamic nature of in vivo interactions where the binding partners interact in solution while moving freely.^79,80 Isothermal calorimetry can be used to evaluate binding properties in solution, but MST is also a real-time, solution-based method to measure binding interactions. MST requires that one protein be non-specifically labelled with a fluorescent dye. We elected to label the bsFpF in these experiments to utilise the same labelled molecule to evaluate both binding moieties in the bsFpF. Using MST, the binding analysis of bispecific Fab_VEGF–PEG₁₅–Fab_IL6R was performed on VEGF, IL6R and a mixture of VEGF/IL6R (Fig. S11A–C, ESI†). MST binding curves can appear in both directions as observed in Fig. S11A–C (ESI†) depending on the diffusion coefficients of the labelled protein and complex.⁸¹ The binding curves indicate that the bispecific Fab_VEGF–PEG₁₅–Fab_IL6R maintained its binding towards both VEGF and IL6R separately and as a mixture.

To consider the presence of the ligation element in the PEG linker, we evaluated a homodimer FpF_VEGF prepared by PEG di(bis)-sulfone 3̲ and homodimer FpF_VEGF prepared by conjugation–ligation (reagents 2̲8̲ and 3̲0̲). While the dissociation rate constant (k_d) appeared broadly similar, the binding affinity (K_D) was smaller for the FpF_VEGF prepared by the ligation approach (Fig. 6). The lower binding affinity for the ligated FpF_VEGF appears to be due to a slightly slower association rate constant (k_a). The ligation element may slightly restrict the freedom for the association of the second Fab to the immobilised ligand. It will be worthwhile to explore the conformational influences of the presence and absence of the ligation element, specifically with the PEG linker molecular weight.

A range of bispecific FpFs targeting different antigen epitopes with therapeutic properties for pro-inflammatory targets (tumour necrosis factor alpha or TNF-α, interleukins 6R, and interleukins17) and a pro-angiogenic target (vascular endothelial growth factor or VEGF) were prepared. We also explored the preparation of 2 bsFpFs, Fab_VEGF–PEG–Fab_HA and Fab_VEGF–PEG–Fab_COL2 where one Fab functions to bind to a specific non-therapeutic tissue (affinity targeting) and the other Fab functions to bind to a therapeutic target. Hyaluronic acid (HA) and collagen-II (COL2) are two endogenous tissue targets that are envisaged for increasing the biological residence time of a molecule within the vitreous cavity of the eye.^82–85 If the relevant Fab can be sourced, then bsFpFs may have utility as bispecific antibody mimetics for early preclinical studies designed to explore new therapeutic strategies.

Conclusions

FpFs are IgG mimetics with the potential to act as IgG surrogates in a range of applications.¹ Bispecific FpFs were prepared as possible mimetics of bispecific IgGs. Although each element of the bsFpF (e.g. Fab and PEG linker) has been used clinically or can be sourced (e.g. Fab_COL2 and Fab_HA), the potential of bsFpFs may be in preclinical research to develop strategies for (i) drug delivery where one Fab targets a specific anchor point (e.g. hyaluronic acid or collagen II in the vitreous cavity of the eye) to achieve the enhanced residence time in the tissue or organ of interest, (ii) tissue regeneration where one moiety would bind to a scaffold^86–88 while the other moiety would provide a required biological function or for cell immobilisation⁸⁹ and (iii) drug target development to optimise the spatial–temporal aspects^90,91 of binding kinetics of a bispecific antibody and where it is not possible to use multiple antibodies in a single dosage form (e.g. intraocular indications).

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts to declare.