DutaFabs are engineered therapeutic Fab fragments that can bind two targets simultaneously

Abstract

We report the development of a platform of dual targeting Fab (DutaFab) molecules, which comprise two spatially separated and independent binding sites within the human antibody CDR loops: the so-called H-side paratope encompassing HCDR1, HCDR3 and LCDR2, and the L-side paratope encompassing LCDR1, LCDR3 and HCDR2. Both paratopes can be independently selected and combined into the desired bispecific DutaFabs in a modular manner. X-ray crystal structures illustrate that DutaFabs are able to bind two target molecules simultaneously at the same Fv region comprising a VH-VL heterodimer. In the present study, this platform is applied to generate DutaFabs specific for VEGFA and PDGF-BB, which show high affinities, physico-chemical stability and solubility, as well as superior efficacy over anti-VEGF monotherapy in vivo. These molecules exemplify the usefulness of DutaFabs as a distinct class of antibody therapeutics, which is currently being evaluated in patients.

Fig. 1. The concept of DutaFab paratopes.

a The DutaFab concept of separating paratopes on a single Fab. The structure shows a top-down view of 133 overlaid Fab homology models of late-stage clinical antibodies from Jain et al. It is visible that CDRs 1 and 3 from one chain plus CDR 2 from the other chain each form a separate, contiguous surface. The H-side paratope shown in red comprises HCDR1, HCDR3, and LCDR2, while the L-side paratope shown in blue comprises LCDR1, LCDR3 and HCDR2. The structurally diverse HCDR3 is restricted in length and conformation in DutaFab library designs to ensure modularity. b Discovery principle of DutaFabs. Individual paratopes are isolated by separate panning of monospecific Fab phage display libraries, wherein either the H-side or the L-side is diversified in a subset of residues of the relevant CDRs. DutaFabs are generated by gene synthesis and expression of Fab fragments that comprise all the target-specific paratope residues against both targets.

Fig. 2. DutaFab libraries.

a DSC data of different human VH3-Vk1 DutaFab dummy scaffolds. Left side: Up to a Fab thermostability of 99 °C, co-operative melting of all 4 domains in the Fab is observed (blue trace). Above 99 °C, a left shoulder starts to appear, representing the unfolding of CH1-Ck constant domains in the presence of folded Fv (red, green and brown traces of Fab unfolding data). Right side: In case of high Fv stabilities, constant domain unfolding can be analyzed independently; the red trace represents the data for a Fab fragment with an Fv TM of 105.7 °C, the green trace the non-2-state fit for the earlier melting constant domain unfolding event and the blue trace the non-2-state fit for the later melting Fv unfolding event (fitted with Origin software). b Top-down view on a DutaFab scaffold showing conserved structural residues that maintain stability or conformation. DutaFab heavy and light chains are colored dark and light gray, respectively. Residues conferring high stability or restricting HCDR3 conformation are colored according to panel c. The structure is based on PDB 6T9D, which is discussed in further detail below. c Typical DutaFab library design, showing H-side diversified residues (dark red X), optionally diversified H-side enhancing residues (light red X), L-side diversified residues (dark blue X), optionally diversified L-side enhancing residues (light blue X), stability residues (pink #) and HCDR3 conformational residues (green $). Underlined: all positions designated as CDRs in either the Kabat, Chothia, and Contact definitions^–.

Fig. 3. Affinity maturation of the VEGF-PDGF DutaFab.

The potency of the original bispecific DutaFab VP was significantly increased towards both PDGF-BB and VEGFA-121 by affinity maturation of both paratopes. Affinity maturation of either paratope did not negatively impact potency on the opposite paratope. a Competition ELISA with VEGFR1-Fc and 5 pM VEGFA-165. b Competition ELISA with VEGFR1-Fc and 8 pM VEGFA-121. c Competition ELISA with PDGFRβ-Fc and 5 pM PDGF-BB. d Tabulated curve fit parameters. SD = standard deviation. A sample size of n = 2 independently prepared samples was used.

Fig. 4. Structure analysis and co-binding.

a X-ray structure of the VP mat DutaFab in complex with PDGF-BB dimer. The Fab is shown in dark gray (heavy chain) and light gray (light chain). The PDGF-BB dimer (beige) binds to the H-side of the DutaFab with the interacting paratope depicted as red surface. The Fab-target complex is always shown in a view from the top and a view from the side. b X-ray structure of the VP mat DutaFab in complex with VEGFA dimer. The VEGFA dimer (green) binds to the L-side of the DutaFab with the interacting paratope depicted as blue surface. c Model of the ternary complex of the VEGF-PDGF DutaFab bound to its two targets simultaneously. The model was constructed by superpositioning the individual structures (a) and (b) via their Fab domains. d Concomitant binding demonstrated by SPR.

Fig. 5. Structure-based engineering of VEGF potency.

a Side-view and bottom-view of the YHE co-operativity motif engineered into the Vk domain of the VP mat DutaFab. For clarity only the Vk domains (gray) are presented in complex with VEGF dimer (green). Residues involved in co-operative Fab binding are shown as blue sticks. Positions of sidechains have been modeled in silico using the Discovery Studio modeling suite based on the crystal structure of a DutaFab containing the co-operativity motif that is not presented in this study. b VEGF baseline assay with both VEGFA-121 and VEGFA-165 isoforms, showing complex stability of the VEGF dimer, fully blocked on both receptor binding epitopes by drug molecule. A low signal indicates complete blocking of the VEGF dimer. A high signal far away from baseline shows that a large fraction of VEGF has lost the blocking drug during the incubation with the competing receptor. The signal inflection seen at 1 nM drug paratope concentration is due to stoichiometric limitation. c HEK293-NFAT reporter gene assay and HUVEC proliferation assay with and without co-operativity motif. A sample size of n = 3 independently prepared samples was used in the experiments shown in panels b and c.

Fig. 6. In vivo efficacy of the VEGF-PDGF DutaFab.

The in vivo efficacy of the most potent bispecific DutaFab against VEGFA and PDGF-BB, clone VP mat YHE, was analyzed using a laser-induced choroidal neovascularization (LCNV) model performed in pigmented Brown Norway rats. Each data point shown represents the average fluorescence angiography (FA) score for the lesions in one eye 6 days post laser treatment. a Result for antibody treatment by a single intravitreal administration on the day of laser induction. b Result for antibody treatment by a single intravitreal administration three days before laser induction. Sample size was n = 7 for anti-DIG, n = 16 for anti-VEGF_A and n = 14 for VP mat YHE. The whiskers represent the range from the lowest to the highest value in the set; the horizontal line within each box represents the median; boxes range from the 25th to the 75th percentile. Numbers over brackets represent adjusted p-values obtained from one-way ANOVA followed by the Holm–Šídák multiple comparison method.

Fig. 7. CMC and biophysical properties of DutaFabs.

a Viscosity of clone VP mat YHE (25 °C, 20 mM His pH 6.0). b Long-term RAC after PBS incubation, measured by SEC-HPLC after 4 weeks storage of clone VP mat YHE at 200 mg/mL. c Aggregation onset analysis of DutaFab clone VP mat YHE, 1 mg/mL, 20 mM His pH 6.0.