Comparison of the ligand‐binding properties of fluorescent VEGF‐A isoforms to VEGF receptor 2 in living cells and membrane preparations using NanoBRET

Background and Purpose Vascular endothelial growth factor A (VEGF‐A) is a key mediator of angiogenesis. A striking feature of the binding of a fluorescent analogue of VEGF165a to nanoluciferase‐tagged VEGF receptor 2 (VEGFR2) in living cells is that the BRET signal is not sustained and declines over time. This may be secondary to receptor internalisation. Here, we have compared the binding of three fluorescent VEGF‐A isoforms to VEGFR2 in cells and isolated membrane preparations. Experimental Approach Ligand‐binding kinetics were monitored in both intact HEK293T cells and membranes (expressing nanoluciferase‐tagged VEGFR2) using BRET between tagged receptor and fluorescent analogues of VEGF165a, VEGF165b, and VEGF121a. VEGFR2 endocytosis in intact cells expressing VEGFR2 was monitored by following the appearance of fluorescent ligand‐associated receptors in intracellular endosomes using automated quantitative imaging. Key Results Quantitative analysis of the effect of fluorescent VEGF‐A isoforms on VEGFR2 endocytosis in cells demonstrated that they produce a rapid and potent translocation of ligand‐bound VEGFR2 into intracellular endosomes. NanoBRET can be used to monitor the kinetics of the binding of fluorescent VEGF‐A isoforms to VEGFR2. In isolated membrane preparations, ligand‐binding association curves were maintained for the duration of the 90‐min experiment. Measurement of the k off at pH 6.0 in membrane preparations indicated shorter ligand residence times than those obtained at pH 7.4. Conclusions and Implications These studies suggest that rapid VEGF‐A isoform‐induced receptor endocytosis shortens agonist residence times on the receptor (1/k off) as VEGFR2 moves from the plasma membrane to the intracellular endosomes.

Alternative mRNA splicing of the Vegfa gene leads to a number of endogenous VEGF-A isoforms of varying length, such as "prototypical" VEGF 165 a or VEGF 121 a, as well as variants (e.g., VEGF 165 b) that have distinct carboxy-terminus substitutions of exon 8a for exon 8b (Peach, Mignone, et al., 2018;Woolard et al., 2004). Additionally, programmed translational readthrough can lead to VEGF-Ax, an isoform containing both exon 8a-and 8b-encoded residues (Eswarappa et al., 2014). The residues present within each VEGF-A isoform determine whether they can interact with other membrane proteins (e.g., neuropilin 1; Cébe Suarez et al., 2006;Parker, Xu, Li, & Vander Kooi, 2012;Guo & Vander Kooi, 2015; and extracellular matrix components (Krilleke et al., 2007;Vempati, Popel, & Mac Gabhann, 2014). This causes isoforms to vary in their bioavailability and signalling outcomes with many isoforms acting as partial agonists relative to VEGF 165 a (Peach, Mignone, et al., 2018). VEGF-A isoforms also have distinct expression profiles in health and disease, such as down-regulation of VEGF 165 b in numerous cancer types (Bates et al., 2002;Pritchard-Jones et al., 2007). VEGF-A/VEGFR2 signalling has been targeted by a number of clinically approved inhibitors used to treat cancer, such as receptor TK inhibitors (RTKIs) that target the intracellular ATP-binding domain (Ferrara & Adamis, 2016).
The development of fluorescence-based technologies has advanced our pharmacological understanding of GPCRs, RTKs, and other classes of membrane proteins (Stoddart, Kilpatrick, & Hill, 2018;Stoddart, White, Nguyen, Hill, & Pfleger, 2016). For example, real-time ligand binding can be quantified in living cells using BRET (Stoddart et al., 2015;Stoddart et al., 2016). This proximity-based assay monitors energy transfer between a receptor tagged on its N-terminus with a bright 19-kDa nanoluciferase (NanoLuc) and a suitable fluorophore acceptor. We previously developed fluorescent variants of VEGF 165 a, VEGF 165 b, and VEGF 121 a labelled at a single site with tetramethylrhodamine (TMR; Kilpatrick et al., 2017;. Despite similar affinities, VEGF 165 a-TMR had distinct binding kinetics at VEGFR2 and its coreceptor neuropilin 1 expressed in living HEK293T cells . VEGF 121 a-TMR and VEGF 165 b-TMR were also shown to bind to VEGFR2 but not to neuropilin 1 using both NanoBRET and live-cell fluorescence imaging techniques . A striking feature of the binding of VEGF 165 a-TMR to VEGFR2 in intact living cells is that the BRET signal obtained with higher concentrations of the fluorescent probe declines over longer incubation times, after reaching a peak between 15 and 20 min (Kilpatrick et al., 2017). One possible explanation is that this is a consequence of receptor internalisation and uncoupling of ligand-receptor complexes within intracellular endosomes (Kilpatrick et al., 2017). Due to the complex spatiotemporal dynamics of VEGFR2, kinetic profiles of ligand binding to VEGFR2 in intact living cells are likely to contain components that represent the initial ligand-binding interaction and also components that reflect receptor endocytosis, whereby the endosomal environment can impact upon the stability of these ligand-receptor complexes.
In order to isolate the ligand-binding profiles of fluorescent VEGF-A isoforms to VEGFR2 from the potential influences of agonist-induced receptor endocytosis, the present study was undertaken with VEGF 165 a-TMR, VEGF 165 b-TMR, and VEGF 121 a-TMR to: (a) investigate the concentration-dependence and temporal profile of ligand-induced VEGFR2 endocytosis, (b) the influence of VEGFR2 phosphorylation on endocytosis and ligand binding, and (c) the kinetics of the ligand-receptor interactions in isolated membrane preparations, where the potential for parallel receptor endocytosis is not present.

What is already known
• VEGF-A is a key mediator of angiogenesis.
• The binding of VEGF-A to VEGFR2 is not sustained in living cells.

What this study adds
• The power of NanoBRET approaches to study real-time ligand-binding kinetics in membranes and cells.
• Fluorescent VEGF-A isoforms produced a rapid and potent translocation of ligand-bound VEGFR2 into intracellular endosomes.
• Endocytosis shortens agonist residence times as VEGFR2 moves from the plasma membrane to intracellular endosomes.

What is the clinical significance
• New insights into the impact of cellular location on the kinetics of VEGFR2 ligand-receptor interactions. At 70-80% confluency, cells were passaged using PBS (Lonza, Switzerland) and trypsin (0.25% w/v in versene; Lonza). Stable cell lines expressing VEGFR2 were generated using FuGENE HD (Promega Corporation, USA) at a 3:1 ratio of reagent to cDNA. As described previously by , N-terminal NanoLuc-tagged VEGFR2 (NM_002253) was cloned into a pFN31K vector encoding the secretory IL-6 signal peptide fused to the N-terminus of NanoLuc, followed by a GSSGAIA linker before the receptor. Additionally, VEGFR2 was cloned into a pFN21A vector with the IL-6 signal peptide followed by a sequence encoding HaloTag and an EPTTEDLYFQSDNAIA linker at the receptor N-terminus, as described in . Fluorescent VEGF 165 a, VEGF 165 b, and VEGF 121 a were labelled at a single N-terminal cysteine residue with TMR using the HaloTag mammalian protein detection and purification system (G6795; Promega Corporation, USA) as described previously (Kilpatrick et al., 2017;. Fluorescent ligands were characterised in terms of labelling efficiency, dimerisation, and function as described in Kilpatrick et al. (2017) and . Binding affinities of fluorescent VEGF isoforms are approximately an order of magnitude lower than those of their native counterparts . Ligands were stored at −20°C in 2.5 mg·ml −1 protease-free BSA (Millipore, USA). Cediranib was purchased from Sequoia Research Products (Pangbourne, UK), and unlabelled recombinant human VEGF isoforms were purchased from R&D Systems (Abingdon, UK). Furimazine was bought from Promega Corporation (Madison, USA), and other tissue culture reagents were purchased from Sigma-Aldrich (Gillingham, UK).

| Generation of a tyrosine phosphorylation-deficient variant of HaloTag-VEGFR2 and NanoLuc-VEGFR2
To generate tyrosine phosphorylation-deficient (VEGFR2-TPD) variants, site-directed mutagenesis was performed for HaloTag-VEGFR2 and NanoLuc-VEGFR2 using the following forward and The altered nucleotides are shown in lower case. Mutagenesis was performed sequentially with the above primers to generate the Y951F, Y1054, Y1059F, Y1175F, and Y1214F tyrosine phosphorylationdeficient mutant of VEGFR2 (VEGFR2-TPD; Figure 4d) using Pfu DNA polymerase (Promega Corporation, USA), followed by sequencing of the full plasmid to check for off-target SNPs. All HEK293T cells also stably expressed a Firefly luciferase reporter downstream of an NFAT response element to monitor NFAT-induced gene transcription (NFAT-RE-luc2P; Promega Corporation, USA), as in Carter, Wheal, Hill, and Woolard (2015).

| Membrane preparations
HEK293T cells stably expressing wild-type or tyrosine phosphorylation-deficient NanoLuc-VEGFR2 were grown in DMEM/10% FCS to 80-90% confluency in 148-mm 2 culture dishes (Corning, USA). The media was then replaced with PBS, cells were removed from the dish by scraping and then transferred into a 50-ml tube. Cells were centrifuged at 378 × g for 12 min at 4°C, the supernatant was removed, and the remaining pellet was stored at −80°C.
Thawed pellets resuspended in PBS were homogenised using an electronic handheld IKA T10 Ultra Turrax homogeniser in 10 × 3 s bursts at 15,000 rpm. Unbroken cells and nuclei were removed by centrifugation at 1,500 × g for 20 min (4°C). The supernatant was then centrifuged at 41,415 × g for 30 min at 4°C to pellet the remaining membranes. The pellet was resuspended in 1-ml PBS, transferred to a borosilicate glass homogeniser mortar, and homogenised 15 times using an IKA RW16 overhead stirrer attached to a serrated pestle (Kartell) at 1,000 rpm. Protein concentration was determined using a bicinchoninic acid assay (Pierce™ BCA Protein Assay; Thermo Fisher Scientific), and absorbance was measured using a Dynex Technologies 4.25 platereader. Membrane preparations were stored at −80°C, and retained their luminescence emissions following addition of furimazine and their ability to bind VEGF-TMR for at least 10 months. For experiments investigating the influence of membrane concentration on NanoBRET signals, increasing concentrations of membranes prepared from wild-type VEGFR2 cells were incubated in the presence and absence of 5-nM VEGF 165 b-TMR. Following incubation for 60 min in the dark, the NanoLuc substrate furimazine (10 μM) was added to each well and equilibrated for 5 min to enable NanoLucmediated furimazine oxidation and resulting bioluminescence emission. Emissions were recorded using the PHERAstar FS platereader (BMG Labtech), and BRET ratios were calculated as fluorescence over luminescence emissions from the second of three cycles.

| Measuring ligand binding using NanoBRET
For kinetic experiments, wells were pretreated with the NanoLuc substrate, furimazine (10 μM), for 5 min to enable NanoLuc-mediated furimazine oxidation and resulting bioluminescence emission. BRET ratios were then measured per well using the PHERAstar FS platereader (BMG Labtech) using filters measuring NanoLuc emissions at 450 nm (30-nm bandpass) and TMR emissions using a longpass filter at 550 nm. Following four initial measurements, intact cells 10-μM furimazine was added to each well and equilibrated for 5 min. Emissions were recorded using the PHERAstar FS platereader (BMG Labtech), and BRET ratios were calculated from the second of three cycles.
Both kinetic and saturation experiments were repeated in wildtype NanoLuc-VEGFR2 membranes to investigate the effect of pH on ligand binding, using VEGF 165 b-TMR as a representative fluorescent VEGF-A isoform. The pH of the assay buffer, HeBSS/0.1% BSA (pH 7.4), was lowered to 5.8-6.2 with concentrated hydrochloric acid on the day of the experiment and measured using a pH Meter PB-11 (Sartorius, Germany). Although EPES buffer has a pK a suited to physiological pH (6.8-8.2), control experiments confirmed the assay buffer remained within 0.2 of the initial pH in conditions replicating the experimental set-up (50 μl at 37°C for 2 hr). were added in duplicate wells (60 min; 37°C/5% CO 2 ). Alternatively, cells were stimulated with 10-nM VEGF 165 a-TMR, VEGF 165 b-TMR, or VEGF 121 a-TMR for between 5 and 120 min as a retrospective timecourse. Cells were washed with PBS (100 μl per well) and fixed with 3% paraformaldehyde in PBS for 15 min at room temperature (RT). Following another PBS wash, nuclei were stained with 2 mg·ml −1 H33342 in PBS for 15 min at RT and then washed and stored in PBS at 4°C. The following day, cells were imaged using an ImageXpress Micro widefield high-content platereader (Molecular Devices, USA).

| High-content widefield imaging quantifying endocytosis of fluorescent VEGF-A isoforms
Plates were imaged at four sites per well using a 20× ELWD (extra long working distance) objective, with a TRITC filter to image VEGF xxx x-TMR (560-nm excitation, 607-to 634-nm emission, and 2,000-ms exposure time) and a DAPI filter for nuclei (405-nm excitation, 447-to 460-nm emission, and 25-ms exposure time). Images were analysed using a granularity algorithm (Molecular Devices), whereby nuclei were identified based upon their size (5-to 25-μm diameter) and pixel depth in grey levels, which was kept consistent between experimental replicates. A nuclear mask defining cell nuclei was then placed over the nuclei within the acquired image forming the basis for automated image segmentation. Fluorescent granules were defined based upon size (2-to 15-μm diameter) and pixel depth in grey levels. Granules were then assigned to specific nuclei based upon proximity using the aforementioned segmented image.

| Immunofluorescent labelling of Rab5
HEK293T cells stably expressing wild-type or tyrosine phosphorylation-deficient HaloTag-VEGFR2 were seeded at 300,000 cells per well onto poly-D-lysine-coated coverslips (18 × 18 mm, 1.5H; Zeiss, Germany) in six-well plates. Following a 24 hr incubation at 37°C/5% CO 2 , coverslips were transferred to humidified wells of a six-well plate lined with parafilm and maintained in PBS to retain moisture. Receptors were labelled with 0.5-μM membraneimpermeant HaloTag-Alexa Fluor 647 (Promega Corporation, USA) in assay buffer (serum free DMEM containing 0.1% BSA). Following 30 min at 37°C/5% CO 2 , coverslips were washed twice with assay buffer and then incubated with 10-nM VEGF 121 a-TMR for 5 or 60 min (37°C/5% CO 2 ). Cells were washed with PBS and fixed with 3% paraformaldehyde in PBS for 20 min at RT. Following numerous wash steps in PBS (3 × 5 min), cells were permeabilised with Triton X-100 (0.025% in PBS). To minimise non-specific antibody labelling, To determine non-specific immunofluorescent labelling, this was repeated using a secondary antibody only in the absence of the primary antibody. Cells were then washed (3 × 5-min PBS), and nuclei were stained with 2 mg·ml −1 H33342 in PBS (15 min, RT) and washed (2 × 5-min PBS) before coverslips were mounted onto slides using ProLong Diamond (Thermo Fisher Scientific) and sealed for storage at 4°C. Coverslips were imaged using a Confocal Zeiss LSM880 fitted with a 63× Pan Apochromat oil objective (1.4 numerical aperture) using a 1-μm slice. Wavelengths were imaged in separate tracks with Rab5 immunolabelling imaged with an Argon488 laser (491-to 571-nm bandpass; 3% power); VEGF 121 a-TMR was imaged with a DPSS 561-10 laser (571-615 nm, 3% power); and HaloTag-Alexa Fluor 647 was imaged using a HeNe633 laser (638-747 nm; 15% power). Images were obtained at 1,024 × 1,024 pixels with eight averages and similar gains per replicate. The immuno-related procedures used comply with the recommendations made by the British Journal of Pharmacology.

| NFAT luciferase reporter gene assay
HEK293T cells stably expressing wild-type or tyrosine phosphorylation-deficient HaloTag-VEGFR2, as well as NFAT-RE-luc2P, were grown to 70-80% confluency. Cells were seeded at 25,000 cells per well in white 96-well plates pre-coated with poly-Dlysine. Following 24 hr at 37%/5% CO 2 , cell culture media were replaced with serum-free DMEM for another 24 hr. Cells were then stimulated with increasing concentrations of VEGF 165 a (R&D Systems) or vehicle (serum-free DMEM/0.1% BSA). Following stimulation for 5 hr at 37%/5% CO 2 , media were replaced with 50 μl per well assay buffer and 50 μl per well ONE-Glo Luciferase reagent (Promega Corporation, USA). Cells were incubated for 5 min to allow luciferase to react with the added reagent and for the background luminescence to subside, and then luminescence emissions were measured using a TopCount platereader (PerkinElmer, UK). Data were normalised to their respective vehicle (0%) and response of wild-type HaloTag-VEGFR2 to 10-nM VEGF 165 a (100%) per experiment. Data were pooled from five independent experiments with duplicate wells. Saturation binding curves were fitted simultaneously for total (fluorescent VEGF-A ligand alone) and non-specific binding (obtained in the presence of 100 nM of unlabelled VEGF-A) using the equation:

| Data and statistical analysis
where B max is the maximal specific binding, were shared across all data sets.
Binding affinities (K i ) of the unlabelled ligands were calculated using the Cheng-Prusoff equation:

BJP
where [L] is the concentration of fluorescent ligand used (nM). K D values (nM) were derived from saturation binding curves. IC 50 is the molar ligand concentration that will inhibit 50% of the specific binding of the fluorescent ligand concentration [L] and was calculated using the equation: where [A] is the concentration of competing drug used.
Fluorescent ligand-binding association kinetic data were fitted to the following mono-exponential association function: where Y max equals the level of specific binding at infinite time, t is the time of incubation, and k obs is the rate constant for the observed rate of association.
k on and k off values were determined by simultaneously fitting ligand-binding association kinetic curves obtained at different fluorescent ligand concentrations (L) to the above equation with the following relationship between k obs and two kinetic binding constants k on and k off : Residence time was calculated as the reciprocal of k off . Kinetically determined K D values were calculated from these kinetic parameters using the following equation: For VEGF 165 a-TMR concentration-response curves for internalisation of ligand-VEGFR2 complexes, the mean granule number per cell was calculated from four images per well. Data are expressed as a percentage of the responses obtained using 100-nM VEGF 165 a-TMR (100%) or vehicle (0%). Data were fitted using non-linear least squares regression using GraphPad Prism with the following equation: where E max is the maximal response and EC 50 is the concentration of agonist required to produce 50% of the maximum response.
The kinetic data for receptor internalisation were fitted to the following mono-exponential association function: where Y max equals the level of receptor internalisation at infinite time, t is the time of incubation, and k obs is the rate constant for the observed rate of receptor internalisation.
Statistical analyses of differences between Mander's overlap coefficients obtained at 5 and 60 min were performed using an unpaired non-parametric Mann-Whitney test. Statistical analysis of the difference of the Mander's overlap coefficient obtained from zero were performed using a Wilcoxon signed-rank test. Statistical analysis of differences between fitted kinetic ligand-binding parameters, in the presence and absence of cediranib, was performed using a paired t test.
AUC analysis (GraphPad Prism 7.02) was used to determine the effect of cediranib on NanoBRET ligand-binding times courses (Figure 4), and the statistical significance of these changes was determined by twoway ANOVA. Differences between kinetic binding constants in cells and membranes were performed using one-way ANOVA with post hoc Tukey test. Statistical significance was taken as P < .05.  (Figure 1a,b). The specific binding observed with 20-nM VEGF 165 b-TMR or VEGF 121 a-TMR, however, peaked at~20 min and then declined substantially towards baseline over the next 70 min (Figure 1a,b).

| Agonist-mediated internalisation of VEGFR2 in HEK293T cells
We have previously suggested that the fall in the VEGFR2-associated NanoBRET signal observed with VEGF 165 a-TMR over longer incubations periods is a consequence of VEGFR2 internalisation and the dissociation of VEGF 165 a-TMR from its receptor within intracellular endosomes (Kilpatrick et al., 2017). To look into the timecourse and and Table 1). In the presence of 10 nM of VEGF 165 a-TMR, VEGF 165 b-TMR, or VEGF 121 a-TMR, the appearance of ligand-bound VEGFR2 in endosomes was rapid with a t 1/2 of between 15 and 20 min (Figure 2d and Table 1).

| Presence of VEGF-TMR and VEGFR2 in Rab5+ endosomes
Immunofluorescent labelling of Rab5 in fixed cells was used to confirm whether VEGF-TMR and VEGFR2 complexes were localised in early Rab5-positive endosomes, using confocal microscopy for enhanced axial resolution (z) compared with high-content widefield imaging in Figure 2. As a representative VEGF-TMR ligand, 10-nM VEGF 121 a-TMR was significantly colocalised with Rab5 (P < .05; Wilcoxon signed-rank test; Figure 3a,b) at both intracellular sites and regions of the plasma membrane by 5 min, whereas it was largely intracellular by 60 min (Figure 3a  Following 5 min pretreatment with NanoLuc substrate furimazine, ligand was added (x = 0) at both a saturating (20 nM) concentration of fluorescent ligand and a concentration that was approximately equal to its K D value (3 nM). BRET ratios were calculated every 30 s at 37°C using the PHERAstar FS platereader. Data were baseline corrected to vehicle to adjust for background emissions. Data are shown as mean ± SEM from five independent experiments with duplicate wells after the initial peak at 20 min. Both wild-type and tyrosine phosphorylation-deficient NanoLuc-VEGFR2 receptors were expressed at the same level, as determined by the NanoLuc luminescence emissions (data not shown).

| NanoBRET ligand binding in membrane preparations
To determine the influence of membrane concentrations on NanoBRET ligand binding in membranes prepared from NanoLuc-

FIGURE 2
Quantifying endocytosis of VEGF 165 a-TMR, VEGF 165 b-TMR, and VEGF 121 a-TMR ligand-receptor complexes using high-content imaging. (a) HEK293T cells expressing NanoLuc-VEGFR2 were stimulated with 10-nM fluorescent VEGF 165 a-TMR (60 min, 37°C). Cells were fixed and incubated with nuclear stain (H33342). The following day, cells were imaged with an ImageXpress Micro widefield platereader (20× ELWD objective; four sites per well) using filter settings for TRITC (left; VEGF 165 a-TMR) and DAPI (middle; nuclei). Images were analysed using a granularity algorithm (Molecular Devices) whereby nuclei stained with H33342 (blue, middle panel) were identified based upon their size (5-to 25-μm diameter) and pixel depth in grey levels. A nuclear mask defining cell nuclei was then placed over the nuclei within the acquired image (right panel; green spots), forming the basis for automated image segmentation. Fluorescent granules detected in the TRITC channel (right; white spots) were defined based upon size (2-to 15-μm diameter) and pixel depth in grey levels. Granules were then assigned to specific nuclei based upon proximity using the aforementioned segmented image. Scale bars represent 20 μm. (b) To confirm receptor specificity, non-transfected wild-type HEK293T cells or stably expressing HaloTag-VEGFR2 cells were stimulated with 10-nM VEGF 165 a-TMR (n = 5), VEGF 165 b-TMR (n = 5), or VEGF 121 a-TMR (n = 5) for 60 min (37°C). Cells were fixed, stained, and imaged, as in (a). Data were normalised to mean vehicle (0%) or VEGF 121 a-TMR stimulation (100%) in HaloTag-VEGFR2 cells per experiment. (c) Cells were stimulated with increasing concentrations of fluorescent VEGF-A variants (60 min, 37°C). Data were normalised to mean vehicle (0%) and 100-nM VEGF 165 a-TMR-stimulated response (100%) per experiment. Cells were fixed, stained, and imaged as above. (d) Timecourse of internalisation of VEGF-TMR complexes, whereby NanoLuc-VEGFR2 cells were stimulated with 10-nM VEGF 165 a-TMR, VEGF 165 b-TMR, or VEGF 121 a-TMR for 0-120 min at 37°C. Cells were fixed, stained, and imaged as above and then normalised to vehicle (0%) and 60-min VEGF 165 a-TMR stimulation (100%) per experiment. Data from (b-d) are expressed as mean ± SEM and pooled from five independent experiments unless stated otherwise with duplicate wells imaged at four sites per well   2.03, 9.53, and 5.54 nM;.
However, non-specific binding was low in all cases (Figure 6a-c).
Competition binding experiments were also conducted for each

| Kinetics of the binding of fluorescent VEGF-A isoforms to VEGFR2 in membrane preparations
To determine the kinetic constants (k on and k off ) for the binding of  Table 2 were similar to those previously reported for these three ligands in intact cells, where the kinetic analysis was limited to the first 20 min of agonist stimulation .
Membrane kinetic experiments were also repeated in the presence of the RTKI cediranib (Table 3). There was a significant decrease in k off in membrane preparations when compared with cells (Table 3), but the k off values were not altered by cediranib treatment. There was a small significant decrease in the k on value for  (Table 3). To quantify the kinetics of binding at the tyrosine phosphorylation-deficient NanoLuc-VEGFR2 using membrane preparations, a stable cell line was generated. Kinetic parameters of FIGURE 4 VEGFR2 phosphorylation is not required for the decline in NanoBRET signal. (a) Schematic representing the mechanism of cediranib at the intracellular ATP-binding site of VEGFR2. (b, c) HEK293T cells stably expressing NanoLuc-VEGFR2 were pretreated with 0.01% DMSO (control) or 1-μM cediranib (30 min, 37°C). After the addition of furimazine for 5 min, the real-time binding of (b) 5-or (c) 20-nM VEGF 165 b-TMR was monitored every 30 s for 90 min at 37°C. Data are shown as mean ± SEM from five independent experiments with duplicate wells. AUC analysis provided the following areas (mean ± SEM): (b) 0.71 ± 0.01 and 0.77 ± 0.01* and (c) 1.80 ± 0.02 and 2.04 ± 0.02* for control and cediranib-treated cells respectively. *P < .05 significant difference between control and cediranib-treated cells (two-way ANOVA of every time point). (d) Schematic of tyrosine mutations in the tyrosine phosphorylation-deficient VEGFR2-TPD. (e) HEK293T cells stably expressing HaloTag-VEGFR2-TPD were labelled with membrane-impermeant HaloTag-Alexa Fluor AF647 and stimulated with 10-nM VEGF 121 a-TMR (60 min, 37°C). Following cell fixation and permeabilization, Rab5-positive endosomes were labelled with a monoclonal rabbit antibody and secondary chick antirabbit Alexa Fluor 488. Cells were imaged using a Zeiss LSM880F Confocal (63× oil objective) with representative images from five independent experiments. Scale bars show 20 μm. (f) NFAT luciferase production in response to 5-hr stimulation with increasing concentrations of VEGF 165 a. Stable cell lines were then used for Rab5+ confocal imaging figures to confirm cell surface expression. Data are shown as mean ± SEM from five independent experiments in duplicate wells, expressed as a percentage of response per experiment to 10-nM VEGF 165 a at wild-type HaloTag-VEGFR2 (100%) or respective vehicle (0%). (g, h) Time course of VEGF 121 a-TMR binding to tyrosine phosphorylation-deficient NanoLuc-VEGFR2 expressed transiently in HEK293T cells. Cells were pretreated with furimazine for 5 min, and then (g) 3-or (h) 20-nM ligand was added (x = 0). BRET ratios were monitored every 30 s at 37°C and baseline corrected to vehicle. Data are expressed as mean ± SEM from eight independent experiments with duplicate wells. AUC analysis provided the following areas (mean ± SEM): (g) 2.61 ± 0.06 and 3.54 ± 0.07* and (h) 7.28 ± 0.10 and 9.79 ± 0.11*, for wild-type VEGFR2 and VEGFR2-TPD cells respectively. *P < .05 significant difference between VEGFR2 and VEGFR2-TPD cells (two-way ANOVA of every time point) VEGF 121 a-TMR in VEGFR2-TPD membranes (Table 3) were similar to control VEGF 121 a-TMR kinetic parameters derived in wild-type VEGFR2 membranes (Tables 2 and 3).
Taking advantage of our binding assay in membrane preparations, we investigated the effect of an acidic pH on ligand-receptor binding to reflect early endosomes (pH 6.0) compared with physiological  (Table 3). The K D affinity estimated from kinetic parameters (k off /k on ) also yielded a similar binding affinity at both pH conditions, despite the difference in residence times.

| DISCUSSION
Initial kinetic studies of the ligand binding of VEGF 165 b-TMR and VEGF 121 a-TMR, in intact HEK293T cells, recapitulated previous findings obtained with VEGF 165 a-TMR in intact living cells (Kilpatrick   A isoforms bind with nanomolar affinity to VEGFR2 (Kilpatrick et al., 2017;. Thus, at concentrations close to their respective K D values, the specific binding of each ligand (monitored by NanoBRET) increased rapidly and was then relatively well maintained over the course of the experiment. In contrast, specific binding observed with higher concentrations (e.g., 20 nM) of VEGF 165 b-TMR or VEGF 121 a-TMR peaked at~20 min and then declined substantially towards baseline over the next 70 min. Previous studies with VEGF 165 a-TMR alone suggested that this was a consequence of VEGFR2 endocytosis and the dissociation of the fluorescent ligand from the receptor within intracellular endosomes, followed by subsequent recycling of ligand-free VEGFR2 back to the plasma membrane (Kilpatrick et al., 2017). Published studies in primary endothelial cells using immunofluorescence antibody labelling or biochemical techniques agreed that VEGFR2 internalised within 30-60 min of VEGF-A stimulation (Bruns et al., 2010;Ewan et al., 2006;Jopling et al., 2009).
To determine the agonist potency (EC 50 values) and kinetic profile of agonist-induced VEGFR2 endocytosis, we monitored the appearance of fluorescent ligand-associated receptors in intracellular endosomes using high-content quantitative imaging. This approach was able to quantify the internalisation of VEGF-A ligand isoforms independently of the known constitutive endocytosis of VEGFR2 (Basagiannis et al., 2016;Ewan et al., 2006;Jopling et al., 2009;Jopling et al., 2011). All three fluorescent VEGF-A isoforms stimulated comparable internalisation of VEGFR2 (EC 50 = 12-30 nM), with 10 nM of each fluorescent ligand eliciting a rapid appearance of ligand-bound VEGFR2 in endosomes (t 1/2 of between 15 and 20 min). This contrasted with previous findings in HUVEC that VEGF 121 a was less able to induce VEGFR2 endocytosis compared with VEGF 165 a . However, it is notable that the HEK293T cells used in the present study have minimal expression of the co-receptor neuropilin 1 , whereas human umbilical vein endothelial cells also express the co-receptor neuropilin 1. Neuropilin 1 does not interact with VEGF 121 a    (Carter et al., 2015).
Cediranib (1 μM) produced no significant effect on the kinetic rate constants (k on and k off ) or K D value of VEGF 165 b-TMR determined from binding to VEGFR2 in HEK293T cell membranes. In intact cells, however, cediranib produced a small elevation in the NanoBRET signal similar to that reported previously for VEGF 165 a-TMR (Kilpatrick et al., 2017). This was accompanied by a small decrease in k on but no significant difference in k off . The combination of these two effects was a small increase in K D (Table 3). It is also worth emphasising, however, that very small changes in kinetic parameters may also be a consequence of the need to fit association curves for four or more concentrations of fluorescent ligand simultaneously with shared values for k on and k off . To explore further the influence of receptor phosphorylation, we generated a tyrosine phosphorylation-deficient VEGFR2 (VEGFR2-TPD) where key intracellular phosphotyrosine residues were mutated to phenylalanine (Y951F, Y1054F, Y1059F, Y1175F, and Y1214F; Figure 4d). This mutant form of VEGFR2 was unable to stimulate NFAT signalling but mimicked the effect of cediranib on VEGF-TMR binding and produced a significant increase in the amplitude of the NanoBRET binding signal obtained with 20-nM VEGF 121 a-TMR. This is likely to be due to interference with endocytosis pathways that are dependent on agonist-induced VEGFR2 activation and phosphorylation. However, what is clear from both the experiments with VEGFR2-TPD and cediranib is that additional pathways can mediate VEGFR2 endocytosis. For example, endocytosis of both VEGFR2-TPD and VEGF 121 a-TMR was still observed using confocal microscopy ( Figure 4).
Our data suggest that substantial agonist-induced VEGFR2 endocytosis is occurring in intact HEK293T cells over the concentration range (1-20 nM) and timecourse (20 min) used previously to assess ligand-binding kinetics in live cells (Kilpatrick et al., 2017;. Furthermore, this internalisation could not be following a 1-hr incubation were slightly more potent than those previously reported in intact cells (Table 2; Kilpatrick et al., 2017;. This was particularly the case for The k on and k off values determined in membrane preparations were of a similar order to those previously reported for these three fluorescent ligands in intact cells when the kinetic analysis was restricted to the first 20 min of agonist stimulation, in order to reduce the influence of receptor endocytosis . k on rate constants were slightly smaller (1.2-fold to 2.0-fold) in membranes than the corresponding values obtained in intact cells. However, k off rate constants were approximately fourfold lower (Table 2) than those determined in HEK293T cells, indicative of a slower agonist dissociation rate. In matched experiments ( Figure 6 and Table 3), these differences were significant (P < .05). These data suggest that the impact of agonist-induced VEGFR2 endocytosis on kinetically derived k on , k off , and K D values is not large if the analysis in cells is restricted to early time points. However, the process of receptor endocytosis does lead to an underestimation of the equilibrium off-rate kinetic constant and receptor residence time (1/k off ) of VEGF 165 b-TMR on the receptor at the plasma membrane (e.g. the calculated residence time in cells for VEGF 165 b-TMR was 12.3 min whilst that in membranes preparations was 43.5 min; Table 3). This is consistent with our earlier suggestion that rapid dissociation of VEGF 165 a-TMR from VEGFR2 occurs in the environment of intracellular endosomes (as a consequence of the lower pH or the presence of endosomal proteases) and allows rapid recycling of ligand-free VEGFR2 back to the cell surface (Kilpatrick et al., 2017). Interestingly, in the tyrosine phosphorylation-deficient VEGFR2-TPD, the kinetic constants determined in membranes for VEGF 121 a-TMR were within a factor of two of those derived in membranes for the wild-type receptor.
The ability to define the pH at which ligand-binding studies are undertaken in membrane preparations provided an opportunity to study the influence of pH on ligand binding at the acidic pH normally found in intracellular endosomes. These data showed that the association and dissociation rate constants of fluorescent VEGF-A at NanoLuc-VEGFR2 were faster at pH 6.0 than those observed at pH 7.4. Indeed, the parameters approached those obtained in intact cells (Table 3). These data suggest that a proportion of the ligand-binding characteristics observed in intact cells is a consequence of receptor endocytosis and the influence of the lower pH environment.
In summary, the present study has shown for the first time that NanoBRET can be used to monitor the kinetics of the binding of fluorescent VEGF-A isoforms to VEGFR2 in isolated membrane preparations. Equilibrium measurements in membranes produced binding parameters that were of a similar order to those determined in live cells. However, in contrast to previous studies in intact cells where the NanoBRET signal falls towards baseline values after reaching a peak, kinetic experiments in membranes produced classic ligandbinding association curves that were maintained for the duration of the 90-min experiment. Automated imaging allowed a quantitative analysis of the effect of fluorescent VEGF-A isoforms on VEGFR2 endocytosis in intact cells. These studies confirmed that all three fluorescent ligands produced a rapid and potent translocation of ligand-bound VEGFR2 to intracellular endosomes.
Our data suggest that the largest impact of this rapid agonistinduced VEGFR2 endocytosis on ligand-binding parameters was on the equilibrium off-rate kinetic constant (k off ) and receptor residence time (1/k off ). Thus, rapid VEGFR2 endocytosis into intracellular endosomes receptor in intact cells shortened the measured residence time of VEGF 165 b-TMR on the receptor from 43.5 min (in membranes) to 12.3 min (in cells). These data suggest that the ligand-binding kinetics of VEGF-A isoforms differ between plasma membrane and intracellular endosomes and that agonist-induced receptor endocytosis can change both local signalling environment and ligand-binding kinetic properties of the receptor. These data provide important new insights into the impact of cellular location and pH on the kinetics of ligandreceptor interactions for a receptor that is a key mediator of both angiogenesis and vascular permeability, and an important drug target for the treatment of cancer.