Introduction

Angiogenesis is the formation of new blood vessels, which is an essential process in growth and development as well as in wound healing1,2. However, it is also a key step in tumor development3 and important in several other diseases, such as age-related macular degeneration of the eye and diabetic retinopathy4. Some of the main regulators of angiogenesis are the members of the vascular endothelial growth factor (VEGF) protein family, including VEGF-A, VEGF-B, VEGF-C, VEGF-D and placental growth factor (PlGF)1,2,5. The receptors that mediate the angiogenic activities of these ligands include VEGF receptor 1 (VEGFR1), VEGFR2 and VEGFR31. VEGFR2 plays the most important role in angiogenesis and is activated by different isoforms of VEGF-A, -C and -D. The extracellular domain of the receptors consists of seven Ig-homology domains, of which domains 2 and 3 contain the ligand-binding site6. Binding of the dimeric VEGF to VEGFR2 mediates receptor dimerization, which induces phosphorylation of the intracellular kinase domains and activates downstream signaling pathways resulting in endothelial cell proliferation, cell migration and vascular permeability2.

A number of therapeutic strategies have been developed to interfere with angiogenesis2. Examples include small molecules acting as antagonists7,8, or inhibiting the tyrosine kinase signaling9,10, as well as monoclonal antibodies and fusion proteins targeting the VEGF ligands and receptors2. However, although several agents are in the clinic11,12, lack of response in some patients and development of resistance has been observed2,13. In order to quickly identify such problems, better biomarkers and diagnostic tools are needed14. IMC-1121B (Ramucirumab, ImClone Systems)15,16 is a human anti-VEGFR2 mAb, which was approved by the FDA in 2014 as a treatment for advanced gastric cancer (www.clinicaltrials.gov). Targeting of the receptor rather than the ligand is potentially preferable, since activation of the receptor by other species of VEGF (VEGF-C and –D) is also inhibited11. VEGFR2 is also a promising target for in vivo imaging of angiogenesis, which can be used as a companion diagnostic tool for monitoring response as well as distinguishing responders from non-responders to various therapies14,17,18,19. However, mAbs are in general unsuitable for molecular imaging due to their very long circulation time in blood, which results in poor contrast20. Still, monoclonal antibodies are the traditional affinity proteins for targeted therapeutics. In for example oncology, the Fc-mediated effects often result in important mechanisms of action, such as ADCC and complement activation. However, in therapeutic applications aimed at blocking a protein/protein interaction for inhibition of for example signaling, the large and relatively complex mAb format is not necessarily required. Numerous studies have reported that much smaller affinity proteins based on non-immunoglobulin scaffolds can be engineered for extremely high affinity and specificity21,22. Moreover, such alternative scaffolds have typically many additional valuable properties compared to antibodies. Development of VEGFR2-specific Nanobodies have recently been reported23 and an Adnectin denoted CT-322 (based on a protein scaffold derived from the 10th type III of human fibronectin), targeting human and murine VEGFR2, showed promising results in Phase I clinical trials24 and is currently in Phase II clinical trials (www.clinicaltrials.gov). Affibody molecules are a class of alternative scaffold proteins that have been thoroughly investigated for both therapeutic and diagnostic applications25. A HER2-specific Affibody molecule is currently evaluated in the clinic for molecular imaging of breast cancer26 and an Affibody-based inhibitor of complement protein C5 entered phase I clinical trials in 2014 (http://clinicaltrials.gov). Affibody molecules are around 6.5 kDa and are efficiently produced in prokaryotic hosts as well as by chemical peptide synthesis. The sequence is devoid of disulphides and unique cysteines can be incorporated for site-directed conjugation of various compounds, such as cytotoxic payloads. The fast and independent folding of the small domain makes engineering of various multimer formats relatively straightforward. The small size also results in improved tissue penetration as well as rapid renal clearance, leading to excellent contrast in molecular imaging27,28,29,30. For therapy, several strategies are available for prolonging the half-life in the circulation, including an in-house developed approach of fusing a small albumin-binding domain for non-covalent association to serum albumin. Depending on the format, the same Affibody molecule might hence be used for both diagnostic and therapeutic purposes. Taken together, the properties of Affibody molecules are well suited for targeting of angiogenesis.

In this study, we selected two antagonistic and cross-reactive anti-VEGFR2 Affibody molecules with phage display. Interestingly, competition experiments revealed that the candidates targeted two non-overlapping epitopes on VEGFR2. We used an in-house developed bacterial display method and FACS for affinity maturation of both binders. To potentially increase the affinity even further, we engineered new biparatopic Affibody molecules. Our hypothesis was that if the two domains were able to bind simultaneously to a single receptor molecule, the dimeric construct might result in even higher affinity by an avidity effect that would be independent of receptor density. The new biparatopic format was successful and the heterodimer demonstrated an off-rate that was around two orders of magnitude slower compared to the monomeric binders. We also included an albumin-binding domain (ABD) in the fusion protein, which might be used for extending the circulatory half-life in future in vivo studies. Biosensor assays demonstrated that the biparatopic Affibody molecules, comprising ABD, indeed could bind the receptor while simultaneously interacting with human serum albumin. The Affibody molecules were also able to target VEGFR2 on human 293/KDR cells and inhibit VEGF-A induced phosphorylation. We believe they have potential for both future therapeutic purposes (as fused to ABD for long residence in the circulation) and diagnostic purposes (as non-ABD fused for rapid clearance and improved imaging contrast), in cancer as well as in ophthalmic diseases.

Results

Phage display selection and screening of binders to VEGFR2

An Affibody library displayed on M13 filamentous phage was used to select binders to human VEGFR2. After four selection rounds, candidate clones were analyzed using ELISA for binding to the receptor. Two clones binding to both human and murine VEGFR2 were identified and denoted ZVEGFR2_1 and ZVEGFR2_2, respectively. Comparison of the sequences of these two variants revealed a low level of similarity, indicating that the two Affibody molecules might be interacting with different epitopes on VEGFR2 (Supplementary Table S1).

Specificity and epitope analysis

The specificity of the two VEGFR2-binding clones was analyzed using ELISA. Both clones showed binding to human and murine VEGFR2, but not to the closely related receptors VEGFR1 and VEGFR3 (Supplementary Fig. S1). Next, we analyzed if the selected binders were recognizing an epitope that was overlapping with the binding site of VEGF-A. Human or murine VEGFR2 was pre-incubated with a 15-fold molar excess of human VEGF-A. Pre-incubation dramatically reduced the signal for both variants (Supplementary Fig. S1), indicating that the Affibody molecules recognized the same or a partially overlapping binding site as VEGF-A.

Analysis of heat stability and refolding of VEGFR2-binding Affibody molecules

The genes encoding ZVEGFR2_1 and ZVEGFR2_2 were subcloned into an expression vector for production of soluble proteins. Proteins were purified by IMAC and eluates were analyzed using SDS-PAGE, demonstrating single bands of correct length with no detectable contaminants (data not shown). The heat stability and refolding capacity of ZVEGFR2_1 and ZVEGFR2_2 was analyzed using circular dichroism (CD) spectroscopy. Both Affibody molecules had an alpha-helical content that was similar to previously reported binders and refolded after heat-induced denaturation at 91°C. Melting temperatures were estimated to around 45°C for ZVEGFR2_1 and around 50°C for ZVEGFR2_2 (Supplementary Fig. S2).

Affinity determination of VEGFR2-binding Affibody molecules

The kinetics of the binding of the two Affibody molecules to human and murine VEGFR2 were analyzed in a surface plasmon resonance (SPR)-based biosensor assay. Binding of ZVEGFR2_1 and ZVEGFR2_2 to human and murine VEGFR2 was detected by injecting the Affibody molecules over human or murine VEGFR2, immobilized on the chip surface. Data was fitted using non-linear regression to a monovalent binding model and the equilibrium dissociation constants (KD) were calculated from the obtained association and dissociation rates to 162 ± 19 nM for ZVEGFR2_1 and 78 ± 0 nM for ZVEGFR2_2 for human VEGFR2, as well as to 258 ± 22 nM for ZVEGFR2_1 and 91 ± 18 nM for ZVEGFR2_2 for murine VEGFR2 (Fig. 1a; Supplementary Fig. S3).

Figure 1
figure 1

Characterization of VEGFR2-binding Affibody molecules from phage display selection.

(a) Sensorgrams from SPR analysis of ZVEGFR2_1 and ZVEGFR2_2 binding to immobilized human VEGFR2. Affibody molecules were injected at concentrations ranging from 50 nM to 500 nM. Data is double referenced by subtraction of simultaneous responses from reference surface and a buffer injection. The experiment was performed in duplicates. (b) Representative results from SPR-based competition assay. Sensorgrams were obtained from a double injection, where a first injection of 1 μM of ZVEGFR2_1 (I) was immediately followed by a second injection (II) of either a combination of 1 μM of ZVEGFR2_1 and 1 μM of ZVEGFR2_2, or 2 μM of ZVEGFR2_1, over immobilized human VEGFR2. The experiment was performed in duplicates. (c) Representative results from SPR-based analysis of human VEGF blocking. 40 nM of human VEGFR2, which had been pre-incubated for 40 min with a 25 × molar excess of ZVEGFR2_1, ZVEGFR2_2 or PBS (control), was injected over a surface of immobilized human VEGF-A. The experiment was performed in duplicates.

Full size image

SPR-based competition assays

An SPR-based approach was used to investigate whether ZVEGFR2_1 and ZVEGFR2_2 could bind simultaneously to the receptor. First, a saturating concentration of ZVEGFR2_1 was injected over immobilized human or murine VEGFR2, directly followed by an injection of a mix of ZVEGFR2_1 and ZVEGFR2_2 or only ZVEGFR2_1 for comparison. A substantial additional increase in response signal was observed for the mix, but not for ZVEGFR2_1 alone (Fig. 1b). The experiment was also conducted in the reverse order and again, there was a large increase in signal upon the second injection of ZVEGFR2_1 and ZVEGFR2_2 (Supplementary Fig. S4). The experiments were also repeated using murine VEGFR2, resulting in a similar response (Supplementary Fig. S4). The results demonstrated that ZVEGFR2_1 and ZVEGFR2_2 were able to bind simultaneously to VEGFR2 and, consequently, that the Affibody molecules interacted with distinct and non-overlapping epitopes on VEGFR2.

Moreover, an additional SPR-based competition experiment was conducted to verify the results from the ELISA, which indicated that the Affibody molecules competed with VEGF-A for binding to VEGFR2. Human and murine VEGFR2 was preincubated with ZVEGFR2_1, ZVEGFR2_2 or PBS, respectively and injected over surfaces with immobilized human and murine VEGF-A, respectively. Injection of the control sample, containing VEGFR2 preincubated with PBS, resulted in an increase in signal, as expected (Fig. 1c). Injection of the samples containing VEGFR2 preincubated with the Affibody molecules resulted in a dramatically lower response, showing that the Affibody molecules blocked the interaction between VEGFR2 and VEGF-A (Fig. 1c). The experiments were also repeated using murine VEGFR2, demonstrating similar results (Supplementary Fig. S5).

Alanine scanning of VEGFR2-binding Affibody molecules ZVEGFR2_1 and ZVEGFR2_2

With the aim to increase the affinity of ZVEGFR2_1 and ZVEGFR2_2 for VEGFR2, a combinatorial-based affinity maturation strategy using display on the surface of S. carnosus was applied. First, alanine scanning mutagenesis was used to analyze the individual contributions from 13 residues (same positions as randomized in the library) in the Affibody molecules ZVEGFR2_1 and ZVEGFR2_2 to the interaction with VEGFR2. The 26 mutants were expressed and displayed on the surface of staphylococci and binding to VEGFR2 was analyzed by flow cytometry (Fig. 2). For ZVEGFR2_1, the majority of the mutations lead to a substantial decrease in VEGFR2 affinity, indicating that these positions were important for VEGFR2-binding. Three positions (position 11, 14 and 25), were not affected or showed an increased binding signal when mutated to alanine. For ZVEGFR2_2, the effect of the mutations varied more among the positions. Mutations in position 10, 11, 14, 18 and 25 had the smallest effect on the binding.

Figure 2
figure 2

Alanine scanning of first-generation VEGFR2-specific Affibody molecules.

(a) ZVEGFR2_1 (b) ZVEGFR2_2. The 13 residues in the VEGFR2-binding Affibody molecules that were substituted with alanine are represented on the X axis and the fold change in normalized binding signal (a ratio of FL-1 fluorescence intensity, corresponding to VEGFR2 binding and FL-6 fluorescence intensity, corresponding to surface expression level) compared to the corresponding non-mutated binder (ZVEGFR2_1 or ZVEGFR2_2) is represented on the Y axis. Binding to human (green bars) or murine (purple bars) VEGFR2 is shown. The error bars show the standard deviation of two independent experiments.

Full size image

Design and construction of affinity maturation libraries

Two affinity maturation libraries were designed and the mutation frequency in each position was based on the results from the alanine scan of ZVEGFR2_1 and ZVEGFR2_2, in principle as previously described by Malm et al.21. Briefly, each selected position was randomized with 18 codons corresponding to all amino acids except cysteine and proline. The original amino acid residues from the sequences of the VEGFR2 binders ZVEGFR2_1 and ZVEGFR2_2 were included at a higher proportion, in order to generate an average mutation frequency of approximately three mutations per molecule. The randomization frequency in each position was also normalized with the results from the alanine scan, resulting in a lower degree of mutation in positions important for VEGFR2 binding and a higher degree of mutation in positions of less importance (Table 1 and 2). The libraries of DNA fragments were subcloned into the staphylococcal display vector and transformed into S. carnosus generating a diversity of approximately 4 × 107 individual transformants for each library. Sequence analysis of individual library members verified a distribution of codons in accordance with the theoretical design (data not shown).

Table 1 Library design of affinity maturation library ZVEGFR2_1matlib
Full size table
Table 2 Library design of affinity maturation library ZVEGFR2_2matlib
Full size table

Selection of affinity-matured VEGFR2-specific Affibody molecules using FACS

Fluorescence-activated cell sorting (FACS) was used for isolation of staphylococcal cells displaying Affibody molecules with increased affinity for VEGFR2. Four rounds of sorting was performed, using a starting concentration of 50 nM human VEGFR2 and decreasing to 20 nM in the third round (Fig. 3). Surface expression level was monitored using fluorescently labeled HSA as described previously31. In the fourth sorting round, an off-rate selection strategy was applied. The bacterial displayed library was incubated with 50 nM of labeled VEGFR2, followed by washing and incubation for 30 min or 4.5 h with an excess of unlabeled VEGFR2 to minimize rebinding of dissociated labeled target. In addition to the off-rate selection in the last round, the selection stringency was also increased throughout the four sorting rounds by using more stringent gating and sorting parameters. Sequencing of the isolated output identified 50 unique variants from ZVEGFR2_1matlib and 17 unique variants from ZVEGFR2_2matlib. Interestingly, all except three variants from ZVEGFR2_2matlib contained a mutation in a non-randomized position (K33N). Position 11, 14 and 25 were the most frequently mutated in the output from ZVEGFR2_1matlib, which was in agreement with the alanine scan data. Position 11 was the most frequently mutated position in the output from ZVEGFR2_2matlib. The average number of mutated amino acids among the selected clones was 2.9 from ZVEGFR2_1matlib and 2.1 from ZVEGFR2_2matlib, thus also in good agreement with the library design.

Figure 3
figure 3

Sorting of the affinity maturation libraries.

(a) ZVEGFR2_1matlib (b) ZVEGFR2_2matlib displayed on staphylococcal cells. The VEGFR2 binding signal (monitored by the binding of fluorescently labeled anti-Fc antibody to human VEGFR2/Fc) is represented on the Y axis and the surface expression level (monitored by fluorescently labeled HSA binding) is represented on the X axis. The dot plots show staphylococcal cell populations from the original unsorted library as well as cell populations isolated in the 1st, 2nd, 3rd and 4th selection round, respectively. For the 4th selection round, dot plots are shown for selections including an off-rate experiment of 0 min, 30 min or 4.5 h.

Full size image

On-cell affinity ranking of second-generation VEGFR2-specific Affibody molecules

The binding of individual variants to human VEGFR2 was compared by flow-cytometric analysis of recombinant staphylococci. The 22 most frequently abundant binders from ZVEGFR2_1matlib and the 17 unique binders from ZVEGFR2_2matlib were included in the assay (Supplementary Fig. S6). In order to obtain a ranking influenced by the off-rates, the samples were subjected to 30 min or 4.5 h incubations with unlabeled VEGFR2 after incubation with labeled target. All analyzed clones showed slower dissociation compared to the original binders ZVEGFR2_1 and ZVEGFR2_2. The top ten binders from ZVEGFR2_1matlib and the top eight binders from ZVEGFR2_2matlib were also analyzed using murine VEGFR2, showing retained cross-reactivity to murine VEGFR2 (Supplementary Fig. S6). Four candidates from each library (ZVEGFR2_11, ZVEGFR2_16, ZVEGFR2_19 and ZVEGFR2_22 from ZVEGFR2_1matlib and ZVEGFR2_33, ZVEGFR2_38, ZVEGFR2_40 and ZVEGFR2_41 from ZVEGFR2_2matlib) were produced as His6-tagged soluble proteins for further downstream characterizations.

Analysis of secondary structure content, thermal stability and refolding capacity

Secondary structure content of the eight selected candidates was analyzed by CD spectroscopy, in principle as described above (Supplementary Fig. S7). All clones showed an alpha-helical content, although ZVEGFR2_11 had a relatively lower degree of alpha-helicity. All candidates demonstrated a similar secondary structure content as for non-heat treated samples when lowering the temperature back to 20°C after heating to 91°C, indicating efficient refolding (Supplementary Fig. S7). The melting temperatures of the clones obtained from ZVEGFR2_1matlib were around 45°C (Table 3). Notably, the three clones from the ZVEGFR2_2matlib containing the K33N mutation (ZVEGFR2_41, ZVEGFR2_38, ZVEGFR2_33) showed a lower melting temperature than both the first generation binder ZVEGFR2_2 and the second generation binder ZVEGFR2_40, lacking the K33N mutation (Table 3; Supplementary Table S1; Supplementary Fig. S7), indicating that the non-intended mutation in position 33 had a negative influence on the thermal stability. ZVEGFR2_41, ZVEGFR2_38 and ZVEGFR2_33 were therefore excluded from further characterizations. ZVEGFR2_11 was also excluded, due to the relatively lower degree of alpha-helical structure content.

Table 3 Equilibrium dissociation constants (KD), dissociation rate constants (kd) and estimated melting temperatures (Tm) for affinity-matured Affibody molecules
Full size table

Affinity determination of second-generation binders

Binding kinetics of ZVEGFR2_22, ZVEGFR2_19, ZVEGFR2_16 and ZVEGFR2_40 were analyzed using surface plasmon resonance (Fig. 4a; Supplementary Fig. S8). Data was fitted using non-linear regression to a monovalent binding model and the equilibrium dissociation constants (KD) were calculated from the obtained association and dissociation rate constants. Equilibrium dissociation constants for human VEGFR2 were determined to 5.0 nM–10.9 nM for the four candidates (Table 3), representing around a 30-fold and a 7-fold increase in affinity compared to the original binders ZVEGFR2_1 and ZVEGFR2_2, respectively. The affinities for murine VEGFR2 were determined to 7.8 nM–11.9 nM, representing around a 30-fold and an 8-fold increase in affinity compared to first-generation binders (Table 3). The affinity-matured variants ZVEGFR2_22 and ZVEGFR2_40 were selected for further studies.

Figure 4
figure 4

Characterization of affinity matured VEGFR2-binding Affibody molecules selected by staphylococcal display.

(a) Representative sensorgrams from SPR analysis of affinity-matured Affibody molecules (ZVEGFR2_22 and ZVEGFR2_40) binding to immobilized VEGFR2, showing the response signal. Affibody molecules were injected at concentrations of 5, 10 and 20 nM. Data is referenced by subtraction of simultaneous responses from reference surface. The experiment was performed in duplicates. (b) Representative results from SPR-based competition assay. Sensorgrams were obtained from a double injection, where 1 μM of ZVEGFR2_22 was injected (1), immediately followed by a second injection (2) of either a combination of 1 μM of ZVEGFR2_22 and 1 μM of ZVEGFR2_40, or 2 μM of ZVEGFR2_22, over immobilized human or murine VEGFR2. The experiment was performed in duplicates. (c) Representative results from SPR-based assay of VEGF blocking. 40 nM of human VEGFR2, which had been pre-incubated for 40 min with a 25 × molar excess of ZVEGFR2_22, was injected over a surface of immobilized human VEGF-A. The experiment was performed in duplicates.

Full size image

SPR-based competition assays

In order to confirm that the affinity-matured variants had retained the ability to bind simultaneously to VEGFR2 and compete with VEGF-A, two SPR-based competition assays were conducted. In the first assay, human VEGFR2 was immobilized on the sensor chip. Analysis of potential simultaneous binding of ZVEGFR2_22 and ZVEGFR2_40 to the receptor was performed using a double injection, as described above. As for the previous experiments, injection of the mix resulted in a substantially higher signal, confirming that the affinity-matured candidates had retained the ability to bind simultaneously to VEGFR2 (Fig. 4b; Supplementary Fig. S9).

The second competition assay was performed to verify that the affinity-matured Affibody molecules could block VEGF-A from binding to VEGFR2. Human or murine VEGF-A was immobilized on the chip surface. VEGFR2 was preincubated with ZVEGFR2_22, ZVEGFR2_40 or PBS, followed by injection over the surface. Injection of the control sample, containing VEGFR2 pre-incubated with PBS, resulted in an increase in signal, showing binding of VEGFR2 to VEGF. In contrast, injection of the samples containing VEGFR2 preincubated with the Affibody molecules resulted in a substantially lower response, showing that the Affibody molecules blocked the interaction between VEGFR2 and VEGF-A (Fig. 4c; Supplementary Fig. S10). The results from the competition assays hence demonstrated that the affinity-matured Affibody molecules retained both the ability to simultaneously interact with VEGFR2 and block human as well as murine VEGF-A from binding to the receptor.

Design and production of dimeric VEGFR2-binding Affibody molecules

Next, we investigated whether engineering of biparatopic constructs, comprising the two different VEGFR2-specific Affibody molecules, might result in a further increase in VEGFR2 affinity due to potential avidity effects as well as an increase in VEGFR2-binding surface area. Two heterodimeric fusion proteins in different orientations were designed: i) ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40 and ii) ZVEGFR2_40-(S4G)4-ABD035-(S4G)4-ZVEGFR2_22 (Fig. 5a). Moreover, two homodimeric constructs (ZVEGFR2_22-(S4G)4-ABD035-(S4G)4- ZVEGFR2_22 and ZVEGFR2_40-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40) were included as controls (Fig. 5a). Serine/glycine-based linkers (20 aa) were included between the three domains to increase the possibility for simultaneous binding to both epitopes. The albumin-binding domain (ABD035) domain has previously been engineered to femtomolar affinity for HSA32 and was intended to serve a potential means for future in vivo half-life extension33,34. In addition, ABD would function as an additional linker, as well as a detection tag or as a purification tag for convenient recovery by affinity chromatography. The four constructs were expressed in E. coli and purified by affinity chromatography using HSA-sepharose. SDS-PAGE and MALDI-time of flight (TOF) mass spectrometry confirmed that the proteins were pure and had the correct molecular weights (data not shown).

Figure 5
figure 5

Characterization of biparatopic Affibody constructs.

(a) Schematic overview of the design of the dimeric constructs. (b) Schematic overview of SPR-based off rate analysis assay. HSA was immobilized on the chip surface. A first injection of dimeric Affibody constructs resulted in a negligible off rate due to the femtomolar affinity of ABD for HSA. VEGFR2 binding was analyzed by subsequent injection of monomeric VEGFR2. The experiments were performed in duplicates. (c) Representative sensorgrams obtained from the SPR-based off-rate analysis assay, showing the injection of 40 nM monomeric human VEGFR2 over each of the four dimeric Affibody molecules. Data is double referenced by subtraction of simultaneous responses from reference surface and a buffer injection. (d) Flow-cytometric analysis of binding of dimeric Affibody constructs to VEGFR2-expressing 293/KDR cells. Binding of the Affibody constructs is monitored by the binding of fluorescently labeled HSA to the ABD tag. A higher shift in mean log fluorescence intensity compared to the negative control construct ZTaq-ABD-ZTaq or cells labeled with secondary reagent only was observed for the heterodimeric constructs than for the homodimeric constructs upon binding to VEGFR2-expressing cells. The experiment was performed in duplicates. (e) Inhibition of VEGF-A induced phosphorylation of VEGFR2 on 293/KDR cells. Cells were pre-treated with the biparatopic binder ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40, a combination of 30 nM or 3 nM of each of the monomeric binders ZVEGFR2_22 and ZVEGFR2_40, 30 nM or 3 nM of the negative control construct ZTaq-ABD035-ZTaq, or PBS, followed by stimulation with VEGF-A. VEGFR2 phosphorylation was determined by ELISA. The biparatopic binder and the combination of monomers both resulted in a decrease in phosphorylation level compared to the controls and a more potent inhibition was observed for the biparatopic binder. The data is presented as the OD450 for each sample normalized against the OD450 of untreated cells. The experiment was performed in duplicates.

Full size image

SPR-based analysis of dimeric constructs

The binding of the four dimeric constructs to human and murine VEGFR2 was analyzed in a biosensor assay. We needed to immobilize the dimeric Affibody molecules on the chip surface and inject soluble monomeric VEGFR2 in order to be able to reveal any potential avidity effects. Moreover, we also wanted to assess whether the new constructs could interact with the receptor while simultaneously binding to albumin, which would be essential for future in vivo experiments. HSA was therefore covalently immobilized on the chip and the dimeric constructs were injected over respective surfaces, resulting in a directed non-covalent immobilization of the Affibody constructs on the chip (Fig. 5a). The dissociation of the Affibody constructs from the surface was practically negligible, due to the femtomolar affinity of ABD035 for HSA32. Monomeric human VEGFR2 was thereafter injected over the surfaces (Fig. 5b). The complex mode of interaction resulted in difficulties to fit the association phase to a 1:1 binding model and assessment was focused on the dissociation phase. It should be noted that any avidity effects due to simultaneous binding for the heterodimers were expected to mainly influence the off-rate and potential improvements in dissociation should hence reflect improvements in equilibrium binding. The heterodimeric constructs demonstrated a substantially slower dissociation compared to the homodimeric controls (Fig. 5c; Supplementary Fig. S11). For the best-performing of the heterodimeric constructs (ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40), the dissociation rate constant was estimated to around 8.4 × 10−5 s−1 for human VEGFR2. The respective dissociation rate constants for the monomeric ZVEGFR2_22 and ZVEGFR2_40 were around 5.8 × 10−3 s−1 and 1.5 × 10−2 s−1 (Table 3). The improvements achieved by formatting the Affibody molecules as biparatopic binders corresponded to around 70-fold and 180-fold slower off-rate, respectively. In comparison to the respective homodimeric constructs, the improvement in dissociation rate constants corresponded to around a 14-fold and a 20-fold. Dissociation rate constants for the dimeric Affibody constructs are summarized in Table 4. The slower dissociation rates for the heterodimeric constructs verified that ZVEGFR2_22 and ZVEGFR2_40 could bind simultaneously to VEGFR2 also when linked together in one fusion protein. Consequently, the assay showed that the two epitopes are non-overlapping and located in relatively close proximity on the receptor. The results also revealed that the multimeric binders could target VEGFR2 and at the same time bind to HSA, which thus demonstrated that including ABD in the biparatopic constructs is a promising means for prolonging the circulatory half-life in vivo.

Table 4 Dissociation rate constants (kd) for dimeric Affibody constructs
Full size table

Flow-cytometric analysis of mammalian cell binding

In order to verify that the dimeric Affibody constructs could bind to VEGFR2 expressed on the surface of mammalian cells, a flow-cytometry based assay was performed. HEK 293 cells transfected with human VEGFR2 (293/KDR cells) (Sibtech Inc.), were incubated with the four dimeric construct. To assess whether the Affibody constructs could target the mammalian cells while simultaneously being bound to albumin, fluorescently labeled HSA was used as a secondary reagent. An Affibody construct with non-relevant specificity (ZTaq-(S4G)4-ABD035-(S4G)4-ZTaq, specific for Taq polymerase) was included as negative control. In addition, non-transfected HEK293 cells were included as controls in the assay to verify that potential cell binding was specific for VEGFR2. An anti-human VEGFR2 monoclonal antibody (R&D Systems) was included as a positive control. A shift in fluorescence intensity was observed for the 293/KDR cells incubated with all the VEGFR2-specific Affibody constructs compared to the negative control construct or cells labeled with only secondary reagent (Fig. 5d). Moreover, the samples with non-transfected HEK293 cells were negative in the assay, suggesting that the Affibody molecules specifically targeted VEGFR2 on the surface of the mammalian cells (Supplementary Fig. S12). The positive control antibody confirmed that VEGFR2 was present at the cell surface of 293/KDR but not on HEK293 cells (Supplementary Fig. S12). Incubation with heterodimeric constructs resulted in higher fluorescence signal intensity compared to the homodimeric constructs, confirming that the heterodimeric constructs had higher affinities for VEGFR2 also expressed on the cell surface.

Inhibition of VEGF-A induced phosphorylation of VEGFR2

In order to investigate if the VEGFR2-specific Affibody molecules could inhibit VEGFR2 phosphorylation, an ELISA-based phosphorylation assay was performed. 293/KDR cells were incubated with 3 nM or 30 nM of the biparatopic binder ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40, a combination of 30 nM or 3 nM of each of ZVEGFR2_22 and ZVEGFR2_40, 30 nM or 3 nM of the negative control construct ZTaq-(S4G)4-ABD035-(S4G)4-ZTaq, or PBS only. After incubation with the Affibody molecules, VEGFR2 phosphorylation was stimulated by incubation with human VEGF-A. Cells were lysed and the degree of phosphorylation on VEGFR2 Tyr1175 was analyzed by ELISA.

The results revealed a decrease in signal for cells that were treated with VEGFR2-specific Affibody molecules, demonstrating inhibition of VEGFR2 signaling (Fig. 5e). The biparatopic binder resulted in a more potent inhibition of phosphorylation than the combination of the two monomeric binders.

Discussion

Activation of angiogenesis is a critical step in tumor development and other diseases3,4. With the aim to target angiogenesis, we generated biparatopic VEGFR2-specific Affibody molecules, with high affinities for human as well as murine VEGFR2. Phage display was used for the selection of two Affibody molecules, which were cross-reactive to human as well as murine VEGFR2. The cross-reactivity to murine VEGFR2 is valuable for preclinical studies in mouse models. Due to the complex regulatory role of VEGFR1 in angiogenesis35, cross-reactivity to this receptor might complicate future in vivo studies. Cross-reactivity to VEGFR3 should preferably also be avoided due to potential unwanted effects on lymphatic vessels35. Results from the ELISA-based specificity experiments demonstrated that the two variants did not cross-react with VEGFR1 or VEGFR3. The results also indicated that the Affibody molecules recognized the same or partially overlapping epitopes on VEGFR2 as the natural ligand VEGF-A, which is promising for future therapeutic approaches with the aim to block the interaction between VEGF and the receptor.

To improve the two candidates even further, we initiated an affinity maturation effort. Staphylococcal surface display has previously been successful for generating high-affinity Affibody molecules from affinity maturation libraries31,36. An advantage of cell surface display is the possibility to sort libraries using flow cytometry, which enables real-time monitoring of the selection process and efficient affinity discrimination using fine-tuned sort gates. Isolation of second-generation candidates by FACS was successful and the best variants had an affinity in the single-digit nanomolar range.

SPR experiments indicated that the two Affibody molecules bind to different epitopes on VEGFR2. Both binders also competed with VEGF, suggesting that the two different binding sites might be situated relatively close to each other. Based on these results, we reasoned that a possible strategy for obtaining an additional increase in affinity would be to engineer so-called biparatopic Affibody constructs, i.e. dimers with two different paratopes, recognizing two distinct sites on VEGFR2. We also included a high-affinity albumin-binding domain to get a first indication whether ABD-fusions might be a promising strategy for prolonging circulation time of the biparatopic Affibody molecules in future in vivo experiments. The biparatopic format was successful and connecting the two Affibody molecules with flexible linkers resulted in dramatically slower dissociation kinetics. Although the association rate constants were difficult to determine, the off-rate was in the same range as for the high-affinity HER2-specific Affibody molecule (KD of 22 pM)22. Interestingly, the off-rate was also similar as for the full-length bivalent IMC-1121B (Ramucirumab)16, suggesting a comparable affinity as the much larger antibody. Data also revealed that the format allowed for simultaneous interaction with VEGFR2 and albumin, both in biosensor assays on recombinant receptor as well as in flow-cytometry experiments with VEGFR2-positive mammalian cells. Furthermore, one of the biparatopic binders was tested for the ability to inhibit VEGF-A induced phosphorylation of VEGFR2 expressed on mammalian cells. A substantial decrease in phosphorylation level was observed after treatment with the biparatopic binder, suggesting that this Affibody construct had an antagonistic effect on VEGFR2 signaling, which is promising for future therapeutic applications. The results also demonstrated a more potent inhibition by the biparatopic binder than a combination of each of the monomeric binders, confirming the higher affinity of the heterodimeric format.

Although the trivalent Affibody molecules with ABD comprised three domains and two linkers, the size was still relatively small (around 22 kDa), which is smaller than an scFv. The small size of the Affibody makes it potentially well-suited for applications where an alternative route of administration (e.g. eye drops) is necessary or beneficial. This could be an important advantage compared to larger molecules, such as antibodies, for treatment of neovascular eye conditions37.

In future studies, additional optimizations of for example orientation of the domains, linker length and linker composition might improve the properties even further. Competition with VEGF-A suggests that the Affibody molecules also recognize domain 2 and/or domain 3 on the extracellular part of VEGFR2. Mapping the epitopes in fine detail would likely result in valuable insights regarding the modes of binding and potential new strategies for further protein engineering. Moreover, as mentioned above, Affibody molecules are excellent for construction of various bi- and multispecific variants and combining the biparatopic VEGFR2-binder with Affibody molecules for other targets or payloads might allow for design of even more potent agents. Recently, several studies have reported on arming monoclonal antibodies with various Affibody molecules to efficiently achieve bi- or even multispecific mAbs (so-called AffiMabs38,39. Such AffiMabs retain the properties of the original antibody, but with an additional specificity mediated by the Affibody molecules. We believe that our new Affibody molecules would be excellent for adding VEGFR2-specificity to already existing mAbs or other affinity proteins, a strategy that has previously been demonstrated to generate promising results40,41,42. Finally, this is the first report on the engineering of biparatopic Affibody molecules and the successful results showed that it is an efficient approach for increasing the affinity and it might become a general future strategy for improving the potency of various Affibody molecules.

Methods

Selection of Affibody molecules using phage display

A combinatorial library of Affibody molecule variants displayed on bacteriophage was subjected to four rounds of selection using human VEGFR2/Fc (hVEGFR2/Fc, R&D Systems, Minneapolis, MN, USA), essentially as described by Grönwall et al43. Selection was followed by ELISA screening for analysis of VEGFR2-binding, specificity and VEGF-A blocking. Details are provided in Supplementary methods and Supplementary Table S1 online.

Subcloning, protein production and purification of ZVEGFR2_1 and ZVEGFR2_2

Subcloning of the genes encoding ZVEGFR2_1 and ZVEGFR2_2 into the pET-26b(+) vector followed by soluble protein production and purification was performed as previously described21 and the buffer was exchanged to PBS using PD-10 columns (GE Healthcare, Uppsala, Sweden).

Circular dichroism analysis of ZVEGFR2_1 and ZVEGFR2_2

The purified proteins were analyzed by circular dichroism (CD) spectroscopy as previously described21.

Binding kinetics analysis of ZVEGFR2_1 and ZVEGFR2_2

Surface plasmon resonance (SPR) experiments were performed using a ProteOn XPR36 instrument (Biorad Laboratories, Hercules, CA, USA) using phosphate-buffered saline supplemented with 0.1% Tween 20 (PBST 0.1) as running buffer and 10 mM HCl or PBST 0.1 for regeneration. Human and murine VEGFR2/Fc (R&D systems) were immobilized by amine coupling on two surfaces of a GLM sensor chip (Biorad Laboratories). Binding of ZVEGFR2_1 and ZVEGFR2_2 to both human and murine VEGFR2 was analyzed by injections of five different concentrations of the Affibody molecules (50, 100, 200, 350 and 500 nM) over the immobilized VEGFR2/Fc. PBST 0.1 was used as a running buffer and 10 mM HCl for regeneration. The experiment was performed in duplicates using freshly prepared reagents.

Surface plasmon resonance-based competition assays

Competition assays were performed using a BIAcore™ 3000 instrument (GE Healthcare). All experiments were performed in duplicates using freshly prepared reagents, with PBST 0.1 as a running buffer and 10 mM NaOH for regeneration. For analysis of potential simultaneous binding of ZVEGFR2_1 and ZVEGFR2_2 to VEGFR2, human and murine VEGFR2/Fc (R&D Systems) was immobilized on a CM-5 sensor chip (GE Healthcare). First, 1 μM of ZVEGFR2_1 was injected, directly followed by an injection of a mixture of 1 μM ZVEGFR2_1 and 1 μM ZVEGFR2_2, or 2 μM ZVEGFR2_1 as a control. In a second assay, ZVEGFR2_2 was injected first, followed by a mixture of ZVEGFR2_1 and ZVEGFR2_2, or 2 μM ZVEGFR2_2.

For analysis of competition with VEGF-A, human or murine VEGF-A (R&D Systems) was immobilized on a sensor chip surface. 40 nM of human or murine VEGFR2/Fc (R&D Systems), which had been preincubated with a 25-fold molar excess of ZVEGFR2_1 or ZVEGFR2_2 (or PBS, as a control) for 40 min, was injected over the surface.

Construction of single-point mutations for alanine scan of ZVEGFR2_1 and ZVEGFR2_2

Single-point mutations of ZVEGFR2_1 and ZVEGFR2_2 with substitution to codon for alanine in residues 9, 10, 11, 13, 14, 17 (not in ZVEGFR2_1), 18 (not in ZVEGFR2_2), 24, 25, 27, 32 or 35 were constructed by PCR amplification using oligonucleotides encoding for each mutation. Residues Ala17 in ZVEGFR2_1 and Ala18 in ZVEGFR2_2 were mutated to valine. Cloning of the mutant sequences into the staphylococcal display vector pSCZ1 and transformation to S. carnosus was performed essentially as previously described21, but using restriction enzymes XhoI and SacI.

Flow-cytometric analysis of alanine scan mutants

Flow-cytometric analysis of each alanine (or valine) mutant was performed essentially as previously described21, using 100 nM human or murine VEGFR2/Fc (R&D Systems). Details are provided in Supplemetary methods online.

Construction of affinity maturation libraries

Two affinity maturation libraries were designed and randomized oligonucleotides encoding helix 1 and 2 of the Affibody molecules were purchased from Isogenica (Isogenica, Essex, UK). The libraries were amplified by 15 cycles of PCR using Phusion DNA polymerase (Finnzymes) and the PCR products were purified using QIAquick PCR purification kit (Qiagen) and digested with restriction enzymes XhoI and SacI, followed by purification using preparative gel electrophoresis (2% agarose gel) and a QIAquick gel extraction kit (Qiagen). The staphylococcal display vector pSCZ1 was digested with the same restriction enzymes and purified by preparative gel electrophoresis as described above. The libraries were ligated into pSCZ1 using T4 DNA ligase (New England Biolabs), according to the supplier's recommendations. Plasmids were purified and concentrated using phenol chloroform extraction and ethanol precipitation and subsequently transformed to E. coli SS320 (MC1061 F′) electrocompetent cells (Lucigen, Middleton, WI, USA). Plasmids were prepared and transformed to S. carnosus by electroporation as previously described44. 192 colonies were analyzed by sequencing.

FACS from bacteria-displayed Affibody libraries

Aliquots of the libraries containing 1.2 × 109 cells (30-fold the estimated library size) were inoculated to 500 ml TSB + YE (Merck) with 10 μg/ml chloramphenicol (Sigma-Aldrich) and incubated at 150 rpm and 37°C for 24 h. Around 109 cells were harvested by centrifugation and labelled for flow cytometry essentially as described above for the alanine scan mutants, using 1.5 ml of 20 nM hVEGFR2/Fc (R&D Systems) in a first incubation of 2 h and 800 μl of the secondary Antibody and HSA in a second incubation of 45 min. Before sorting, cells were washed three times in ice-cold PBSP and re-suspended in 1.4 ml ice-cold PBSP. Four rounds of sorting of the libraries were performed using a MoFlo Astrios flow cytometer (Beckman Coulter, Indianapolis, IN, USA). In sorting round 2–4, approximately 10 × the library size was labeled for sorting. Around 0.5% of the population, with the highest VEGFR2-binding signal, was gated and sorted to TSB + YE (Merck) and grown overnight with 10 μg/ml chloramphenicol (Sigma-Aldrich). In the first round, the libraries were incubated with 50 nM hVEGFR2/Fc and in the second and third round with 5 nM hVEGFR2/Fc. In the fourth round, an off-rate selection strategy was applied. First, cells were incubated with 20 nM hVEGFR2/Fc (R&D Systems) as described above. After washes, cells were incubated with 100 nM of unlabeled VEGFR2 (Sino Biological Inc., Beijing, China) for 30 min or 4 h, before incubation with labeled secondary antibody and HSA, as described above. Colonies from the fourth sorting round were analyzed by sequencing.

On-cell ranking of second-generation VEGFR2-binding Affibody molecules

Individual clones from the fourth sorting round that were selected for on-cell ranking were cultivated and labeled for flow-cytometry essentially as described above for the off-rate selection approach in the fourth sorting round and analyzed using a Gallios flow cytometer (Beckman Coulter). Cells displaying first-generation binders ZVEGFR2_1 and ZVEGFR2_2 were included for comparison and cells displaying a TNF-binding Affibody molecule were included as a negative control.

The top 10 ranked clones from ZVEGFR2_1matlib and the top 8 ranked clones from ZVEGFR2_2matlib were also analyzed for binding to murine VEGFR2, by labeling with mVEGFR2/Fc (R&D Systems) and analyzed using flow cytometry as described above. All experiments were performed in duplicates on different days using freshly prepared reagents.

Subcloning, expression and purification of second-generation binders

Four clones from ZVEGFR2_1matlib (ZVEGFR2_11, ZVEGFR2_16, ZVEGFR2_19 and ZVEGFR2_22) and four clones from ZVEGFR2_2matlib (ZVEGFR2_33, ZVEGFR2_38, ZVEGFR2_41 and ZVEGFR2_40) were sub-cloned to pET26-b(+), transformed to E. coli BL21 star (DE3) cells (Invitrogen) and soluble proteins were produced and purified as described above for the first generation binders ZVEGFR2_1 and ZVEGFR2_2.

Circular dichroism analysis of second-generation VEGFR2-binding Affibody molecules

Circular dichroism experiments were performed as described previously21.

Binding kinetics analysis of second-generation binders using surface plasmon resonance

Human VEGFR2/Fc was immobilized on a CM-5 sensor chip (GE Healthcare). Binding kinetics were analyzed by injecting different concentrations of the Affibody molecules (5, 10 and 20 nM) using a BIAcore™ 3000 instrument (GE Healthcare). Binding kinetics for ZVEGFR2_40 binding to murine VEGFR2 were determined using a BIAcore™ 3000 instrument (GE Healthcare) as described for human VEGFR2. For determination of the affinity constants of ZVEGFR2_22, ZVEGFR2_19 and ZVEGFR2_16 for murine VEGFR2, the Affibody molecules were immobilized on a GLC sensor chip (Biorad Laboratories) using a ProteOn XPR36 instrument (Biorad Laboratories). 5, 10 and 20 nM murine VEGFR2 (Sino Biological) was injected over the surfaces. PBST 0.1 was used as a running buffer and 10 mM HCl for regeneration. The experiment was performed in duplicates using freshly prepared reagents.

SPR-based competition assay

Competition assays were performed essentially as described for the first-generation binders ZVEGFR2_1 and ZVEGFR2_2. Details are provided in Supplementary Methods online.

Subcloning, expression and purification of dimer constructs

The genes encoding ZVEGFR2_22 and ZVEGFR2_40 were subcloned into the plasmid pET26b-Z02891-ABD035-ZTaq45 to generate four constructs: ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40, ZVEGFR2_40-(S4G)4-ABD035-(S4G)4-ZVEGFR2_22, ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_22 and ZVEGFR2_40-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40. Subcloning and cultivation was performed as previously described21.

Affibody molecules were purified using affinity chromatography with HSA immobilized on a sepharose matrix and the buffer was exchanged to PBS using PD-10 columns (GE Healthcare, Uppsala, Sweden). The purity and size of the purified proteins were analyzed by SDS-PAGE and MALDI-time of flight (TOF) mass spectrometry using a LaserToF LT3 Plus instrument (SAI, Manchester, UK).

SPR analysis of dimeric Affibody constructs

The binding of the four dimeric constructs to human and murine VEGFR2 was analyzed by SPR. HSA (Sigma-Aldrich) was immobilized on a GLM sensor chip (Biorad Laboratories) using a ProteOn XPR36 instrument (Biorad Laboratories). The apparent dissociation constants (kd) were estimated by injections of 200 nM of each of the homodimeric and heterodimeric constructs over the immobilized HSA, immediately followed by injections of 5, 15 and 45 nM human or murine VEGFR2 (for the heterodimers), or 15, 30 and 60 nM of human or murine VEGFR2 (for the homodimers). The dissociation constants were obtained from sensorgrams using a monovalent binding equation and non-linear regression. The experiment was performed in duplicates using freshly prepared reagents.

Cell-binding analysis

293/KDR cells (Sibtech Inc., Brookfield, CT, USA) and HEK293 cells (ATCC, Manassas, VA, USA) were cultivated in Dulbecco's Modified Eagle Medium (DMEM, Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS) and (for 293/KDR) 2 mM L-glutamine. Cells growing on a petri dish were washed with PBS and harvested by pipetting. Cells were incubated for 20 min on ice with 500 nM of dimeric Affibody construct, the negative control construct ZTaq-(S4G)4-ABD035-(S4G)4-ZTaq or the positive control mouse anti-hVEGFR2 IgG (R&D Systems), in PBS + 1% BSA (Sigma-Aldrich) (PBSB). Thereafter, cells were pelleted by centrifugation at 300 rpm for 4 min and re-suspended in 150 nM HSA (Sigma-Aldrich) conjugated to Alexa Fluor 647 (Invitrogen). After incubation on ice in the dark for 20 min, cells were pelleted and re-suspended in PBSB and analyzed using a Gallios flow cytometer (Beckman Coulter). The experiment was performed in duplicates using freshly prepared reagents.

Inhibition of VEGFR2 phosphorylation

293/KDR cells (Sibtech Inc.) were cultured as described above and seeded in DMEM (Sigma-Aldrich) supplemented with 10% FSB and 2 mM L-glutamine. 6 h after seeding, the medium was exchanged to starvation media (DMEM supplemented with 0.5% FBS) and cells were incubated overnight. Thereafter, cells were treated with 30 nM or 3 nM of the biparatopic binder ZVEGFR2_22-(S4G)4-ABD035-(S4G)4-ZVEGFR2_40, a combination of 30 nM or 3 nM of each of the monomeric binders ZVEGFR2_22 and ZVEGFR2_40, 30 nM or 3 nM of the negative control construct ZTaq-(S4G)4-ABD035-(S4G)4-ZTaq, or PBS. After incubation for 1 h at 37°C, 1 nM human VEGF-A (R&D systems), or PBS, was added and cells were incubated for 10 min at 37°C. Thereafter, cells were washed with PBS containing 1 mM sodium orthovanadate (Sigma) and scraped from the plate. Cell lysates were prepared by sonication for 1 min in Cell Lysis Buffer (Cell Signaling Technology, Danvers, MA, USA), supplemented with 1 mM sodium orthovanadate (Sigma) and 1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma) and analyzed by ELISA using a PathScan® Phospho-VEGFR-2-Tyr1175 Sandwich ELISA kit (Cell Signaling Technology), according to the supplier's recommendations. The experiment was performed in duplicates.