Phage-derived fully human monoclonal antibody fragments to human vascular endothelial growth factor-C block its interaction with VEGF receptor-2 and 3

Abstract

Vascular endothelial growth factor C (VEGF-C) is a key mediator of lymphangiogenesis, acting via its receptors VEGF-R2 and VEGF-R3. High expression of VEGF-C in tumors correlates with increased lymphatic vessel density, lymphatic vessel invasion, sentinel lymph node metastasis and poor prognosis. Recently, we found that in a chemically induced skin carcinoma model, increased VEGF-C drainage from the tumor enhanced lymphangiogenesis in the sentinel lymph node and facilitated metastatic spread of cancer cells via the lymphatics. Hence, interference with the VEGF-C/VEGF-R3 axis holds promise to block metastatic spread, as recently shown by use of a neutralizing anti-VEGF-R3 antibody and a soluble VEGF-R3 (VEGF-C/D trap). By antibody phage-display, we have developed a human monoclonal antibody fragment (single-chain Fragment variable, scFv) that binds with high specificity and affinity to the fully processed mature form of human VEGF-C. The scFv binds to an epitope on VEGF-C that is important for receptor binding, since binding of the scFv to VEGF-C dose-dependently inhibits the binding of VEGF-C to VEGF-R2 and VEGF-R3 as shown by BIAcore and ELISA analyses. Interestingly, the variable heavy domain (V(H)) of the anti-VEGF-C scFv, which contains a mutation typical for camelid heavy chain-only antibodies, is sufficient for binding VEGF-C. This reduced the size of the potentially VEGF-C-blocking antibody fragment to only 14.6 kDa. Anti-VEGF-C V(H)-based immunoproteins hold promise to block the lymphangiogenic activity of VEGF-C, which would present a significant advance in inhibiting lymphatic-based metastatic spread of certain cancer types.

Figure 1. Amino acid sequences of ΔNΔC-VEGF-C and ΔNΔC-VEGF-D variants used in the study.

The region of 100% identity within used ΔNΔC-VEGF-C variants is shown within the black frame, the possible epitope regions of VC2.2.2 anti-VEGF-C scFv on ΔNΔC-VEGF-C and the corresponding regions on ΔNΔC-VEGF-D are shown within grey frames. P.p, P. pastoris-derived ΔNΔC-VEGF-C; R&D, commercially available mammalian cell-derived ΔNΔC-VEGF-C or ΔNΔC-VEGF-D, respectively.

Figure 2. Binding specificities of anti-VEGF-C scFv.

(A) ELISA screening of random clones obtained after 2 or 3 rounds of panning against ΔNΔC-VEGF-C. (B) ELISA analysis of representative anti-VEGF-C scFv clones for the 4 different amino acid sequences obtained. Maxisorp or streptavidin-precoated (SA) plates were coated with his-tagged human ΔNΔC-VEGF-C derived from P. pastoris or biotinylated his-tagged human ΔNΔC-VEGF-C from mammalian cells or P. pastoris, respectively. Control surfaces were left untreated. Antibody fragments and control antibodies were subsequently added and the ELISA was developed as described in Materials and Methods. (C) Cross-reactivity tested by ELISA. Human ΔNΔC-VEGF-C orΔNΔC-VEGF-D (both from mammalian cells) were coated on a maxisorp plate. Anti-VEGF-C scFv clone VC2 or a negative control (PBS only) was added and the ELISA was developed as described in Materials and Methods. (D) BIAcore profiles from the 4 different anti-VEGF-C scFv clones. Different concentrations of protein-A purified scFv were injected on a streptavidin-precoated sensorchip coated with ca. 2000 RU biotinylated mammalian cell-derived ΔNΔC-VEGF-C.

Figure 3. Affinity matured anti-VEGF-C scFvs possess a higher affinity.

(A, B) ELISA analysis of bacterial supernatant from randomly picked affinity matured clones after 1 to 3 rounds of selection on biotinylated (A) P. pastoris-derived or (B) mammalian cell-derived ΔNΔC-VEGF-C. (C) BIAcore profiles of monomeric affinity matured anti-VEGF-C scFvs. Monomeric fractions of protein-A purified scFv were prepared by FPLC and injected as 2-fold dilution series on a streptavidin-sensorchip coated with 2000 RU biotinylated ΔNΔC-VEGF-C derived from mammalian cells.

Figure 4. VC2.2.2 blocks binding of VEGF-C to VEGF-R2 and VEGF-R3 as measured by SPR.

VEGF-R3-Fc was bound to a CM5-sensorchip coated with anti-human IgG antibody. 10 nM ΔNΔC-VEGF-C was then preincubated with a 9 to 900 times molar excess of (A) anti-VEGF-C scFv or (B) control scFv and injected on the VEGF-R3-Fc surface and the amount of binding of ΔNΔC-VEGF-C was measured by SPR. (C) VEGF-R2-Fc was bound to a CM5 sensorchip coated with anti-human IgG antibody. 10 nM of ΔNΔC-VEGF-C or VEGF-A with or without preincubation together with anti-VEGF-C scFv were then injected on the VEGF-R2 surface and bound VEGF-A or VEGF-C was measured by SPR.

Figure 5. VC2.2.2 blocks binding of VEGF-C to VEGF-R2 and VEGF-R3 as measured by competitive ELISA.

2 nM biotinylated ΔNΔC-VEGF-C was preincubated with varying amounts of anti-VEGF-C scFv or control scFv and added on a VEGF-R2 or VEGF-R3 coated microtiter plate. Plotted datapoints are means from 4 replicates ± SEM. The datapoints were fitted to a sigmoidal dose-response curve model using GraphPad Prism 4. Inhibition of VEGF-C binding by anti-VEGF-C scFv reached significance vs the control scFv (*, p<0.05, Student's t-test) at molar excess of 100× more anti-VEGF-C scFv vs VEGF-C. #: p = 0.067 for anti-VEGF-C/VEGF-R2 vs control at molar excess of 33× more anti-VEGF-C than VEGF-C.

Figure 6. Possible VC2.2.2 anti-ΔNΔC-VEGF-C epitope-localization within or near the receptor-binding region on ΔNΔC-VEGF-C.

(A) The VEGF-C residues contacting VEGF-R2 as reported in are represented in yellow (N-terminal helix), red (loop 1), orange (loop 2) and brown (loop 3). The two epitope stretches identified in the peptide scan are colored in blue. From epitope B, only SCRCMSKL is shown, the C-terminal end is missing in the reported structure. Overlaps of epitope A (FFKPPCVSVYRC) and receptor-contacting residues in loop 1 are colored in purple. Residues in loop 1 found to affect (B) VEGF-R2-binding or (C) VEGF-R3-binding by mutational analysis are shown with the same colors as in (A). The localization of the epitopes within the VEGF-C dimer is shown in (D) and their interference with the boxed interface of VEGF-R2 (cyan) is shown in (E), with magnifications in “side” view (F) and “top view” (G). The pdb file 2X1X was used for the representation (www.pdb.org).

Figure 7. The VC2.2.2 V_H is sufficient to bind ΔNΔC-VEGF-C.

Binding of the single V_H-domain of VC2.2.2 to ΔNΔC-VEGF-C and an unrelated antigen (alpha-2-macroglobulin, a2MG) was tested by ELISA using either anti-myc and anti-mouse HRP or protein-A-HRP as detection compounds.

Figure 8. SEC gel-filtration profiles of anti-VEGF-C scFvs.

Protein-A purified scFv were injected on a Superdex 75 10/300 GL size-exclusion gel-filtration column. Markers represent the major elution peaks of the molecular mass standards ovalbumin (43 kDa) and ribonuclease A (13.7 kDa).

Figure 9. SDS-PAGE and immunoblot analysis as well as SEC gel-filtration profiles of anti-VEGF-C VC2.2.2 V_H and anti-GST V_H.

Protein-A purified V_H under non-reducing (nr) and reducing (red) conditions were separated by SDS-PAGE and (A) stained using Coomassie-Blue or (B) immunoblotted using an anti-myc antibody followed by anti-mouse HRP. Protein-A purified (C) anti-VEGF-C VC2.2.2 V_H or (D) anti-GST V_H were injected on a Superdex 75 10/300 GL size-exclusion gel-filtration column. Markers represent the major elution peaks of the molecular mass standards ovalbumin (43 kDa) and ribonuclease A (13.7 kDa).