Use of NanoBiT and NanoBRET to monitor fluorescent VEGF‐A binding kinetics to VEGFR2/NRP1 heteromeric complexes in living cells

VEGF‐A is a key mediator of angiogenesis, primarily signalling via VEGF receptor 2 (VEGFR2). Endothelial cells also express the co‐receptor neuropilin‐1 (NRP1) that potentiates VEGF‐A/VEGFR2 signalling. VEGFR2 and NRP1 had distinct real‐time ligand binding kinetics when monitored using BRET. We previously characterised fluorescent VEGF‐A isoforms tagged at a single site with tetramethylrhodamine (TMR). Here, we explored differences between VEGF‐A isoforms in living cells that co‐expressed both receptors.


| INTRODUCTION
Angiogenesis involves the growth of new blood vessels from existing vascular networks (Carmeliet, 2005). This important physiological process can also be dysregulated in numerous pathologies, such as in tumour development (Chung & Ferrara, 2011). VEGF-A is a key mediator of angiogenesis that primarily signals via its cognate receptor tyrosine kinase (RTK), the VEGF receptor 2 (VEGFR2) (Peach, Mignone, et al., 2018;Simons et al., 2016). VEGF-A binds across immunoglobulin-like domains 2 and 3 of VEGFR2 (Leppanen et al., 2010;Ruch et al., 2007). Agonist binding results in conformational changes throughout the VEGFR2 dimer that lead to auto-and trans-phosphorylation of key intracellular tyrosine residues. This triggers numerous signalling cascades that ultimately initiate endothelial cell proliferation, migration, and survival, as well as increased vascular permeability (Koch et al., 2011).
VEGF-A is an anti-parallel, disulphide-linked homodimer. Alternative splicing of VEGF-A mRNA leads to a number of distinct VEGF-A isoforms (Peach, Mignone, et al., 2018;Woolard et al., 2009). VEGF-A isoforms have different signalling properties in physiological systems with distinct expression profiles in health and disease (Vempati et al., 2014). VEGF-A isoforms differ in length, such as pro-angiogenic VEGF 165 a or the shorter VEGF 121 a isoform. A major site of splicing occurs at exon 8, where proximal splicing results in VEGF xxx a isoforms that contain exon 8a-encoded residues (CDKPRR) and VEGF xxx b isoforms that instead contain exon 8b-encoded residues (SLTKDD).
Fluorescence-based technologies have been used to advance our pharmacological understanding of GPCRs, RTKs, and other classes of membrane protein (Stoddart et al., 2017). For example, bioluminescence resonance energy transfer (BRET) is a proximity-based assay that can quantify real-time binding at 37 C in living cells (Stoddart et al., 2015). A receptor is tagged at the N-terminus with a 19-kDa NanoLuciferase (NanoLuc) such that NanoLuc emits luminescence upon oxidation of the furimazine substrate. This can excite a nearby fluorophore in close proximity (<10 nm), such as a compatible fluorescent ligand bound at the receptor's orthosteric site. We previously developed fluorescent VEGF-A isoforms that were single site labelled with tetramethylrhodamine (TMR), to monitor ligand binding at fulllength VEGFR2 or NRP1 tagged with NanoLuc Peach et al., 2019). Despite having a similar nanomolar binding affinity, VEGF 165 a-TMR binding kinetics were significantly faster at NRP1 than VEGFR2 . VEGFR2 and NRP1 were also subject to distinct subcellular trafficking in the absence or presence of ligand when expressed alone. These techniques were limited to quantifying protein-protein interactions at NanoLuc-tagged VEGFR2 or NRP1 expressed in isolation; however, endothelial cells and tumour cells endogenously express both VEGFR2 and NRP1 in the same cell (Fantin et al., 2013;Koch et al., 2014;Lee-Montiel et al., 2015;Prahst et al., 2008;Whitaker et al., 2001). As these receptors have distinct ligand binding dynamics and subcellular localisation, approaches are required that isolate the pharmacology of VEGF-A ligand binding to distinct complexes involving both VEGFR2 and NRP1.
NanoLuc Binary Technology (NanoBiT) uses a modified NanoLuc split into a large fragment (LgBiT; 156 amino acids) and a small What is already known

• Endothelial cells and tumour cells express both VEGFR2
and its co-receptor NRP1.
• VEGFR2 and NRP1 have distinct ligand binding kinetics and receptor localisation when expressed alone.

What this study adds
• Real-time assay quantifying fluorescent VEGF-A binding at defined heteromeric complexes in living cells at 37 C.

What is the clinical significance
• Aberrant VEGFR2 signalling in cancer is up-regulated by NRP1; however, existing drugs only target VEGF-A/ VEGFR2.
• NRP1 is a promising target in oncology due to its high expression localised to tumours. 11-amino-acid tag (HiBiT or SmBiT; Dixon et al., 2016). Complementation of fragments is required for luminescence emission. Numerous variants were developed of the small tag with different intrinsic affinities for complementation with the LgBiT fragment, including the "higher affinity" HiBiT fragment (K d $ 0.7 nM) and the lower affinity SmBiT fragment (K d $ 190 μM). Used in combination with a fluorescent ligand, interactions between the ligand and a particular protein pairing can be monitored using NanoBiT and BRET. Here, we have used this technology to investigate the kinetics of ligand binding of VEGF 165 a-TMR  and VEGF 165 b-TMR   HEK293T-NFAT-ReLuc2P cells did not emit luminescence in response to furimazine alone that interfered with NanoBiT or NanoBRET assays. Cells were passaged at 70-80% confluency using PBS (Lonza, Switzerland) and trypsin (0.25% w/v in versene; Lonza).

| NFAT luciferase reporter gene assay
HEK293T-NFAT-ReLuc2P cells stably expressed LgBiT-VEGFR2, HiBiT-VEGFR2, or SmBiT-VEGFR2. Cells were seeded at 25,000 cells per well in white 96-well plates pre-coated with poly-D-lysine in DMEM containing 10% FBS. Following incubation for 24 h at 37%/5% CO 2 , medium was replaced with serum-free DMEM, and cells were incubated for a further 24 h. On the day of experimentation, medium was replaced with serum-free DMEM containing 0.1% BSA. Cells were stimulated with increasing concentrations of VEGF 165 a (R&D Systems) for 5 h at 37%/5% CO 2 . Medium was replaced with 50 μl per well serum-free DMEM/0.1% BSA and 50 μl per well ONE-Glo Luciferase reagent. Following a 5-min delay to allow reagent to react with luciferase and background luminescence to subside, luminescence emissions were measured using a TopCount platereader (Perkin Elmer, UK).
Following incubation for 24 h, cells were transfected with a mixture of HaloTag-VEGFR2 and SnapTag-NRP1. Control wells were also transfected with a single construct and empty vector, such as SnapTag-NRP1 and pcDNA3.1/Neo. Transient transfections used FuGENE® HD at a 3:1 ratio of reagent to cDNA with a total 100-ng cDNA per well, with receptors transfected at equal amounts of 50-ng cDNA per well. Transfection solutions were made up in serum-free DMEM and added as 11 μl per well. Cells were incubated for a further 24 h at 37 C/5% CO 2 . Receptors were then labelled with a solution of serum-free DMEM/0.1% BSA containing both 0.5-μM membraneimpermeant HaloTag-AlexaFluor488 substrate (G1002; Promega Corporation, USA) and 0.5-μM membrane-impermeant SNAP-Surface AlexaFluor647 (S9136S; New England BioLabs). These were incubated for 30 min (37 C/5% CO 2 ). Cells were washed twice with 200 μl per well HBSS/0.1% BSA and then replaced with a final volume of 225 μl per well. Cells were incubated with vehicle, 10-nM unlabelled VEGF 165 b, or 10-nM unlabelled VEGF 165 a for 60 min at 37 C, adding 25 μl to a total volume of 250 μl. Cells were imaged live using a temperature-controlled LSM710 confocal microscope fitted with a 40× water objective (Pan Apochromat objective, NA 1.2).

| Bioluminescence imaging of NanoBiT complexes
HEK293T-ReLuc2P cells were plated in to poly-D-lysine (0.01 mgÁml −1 in PBS) coated four-chamber 35-mm dishes (10-mm glass coverslip; CellVis Greiner, 627871) at 100,000 cells per quadrant in DMEM/10% FBS. On day 2, cells were transfected using FuGENE HD at a 3:1 ratio of reagent to cDNA with a total 700-ng cDNA per chamber made up in OptiMEM (ThermoFisher). Cells were transfected with equal amounts of LgBiT-VEGFR2 (350-ng cDNA per chamber) and HiBiT-NRP1 WT (350-ng cDNA per chamber). Alternatively, NanoLuc-VEGFR2 or NanoLuc-NRP1 was transfected at 350-ng cDNA per chamber with an equal amount of pcDNA3.1/Zeo (350-ng cDNA per chamber). On day 3, medium was replaced with HBSS/0.1% BSA containing furimazine (26 μM). Following incubation for 10 min to allow for substrate oxidation, cells were imaged live at 37 C using the inverted Olympus LV200 Bioluminescence Imaging System, fitted with a 60× oil immersion objective (super Apochromat UPLSAPO 60×O objective; NA 1.35) with a 0.5× tube lens to focus the image; therefore, images had a final magnification of 30×. Luminescence was collected using a Hamamatsu Image EMx2 Electron Multiplying Charge Coupled Device (EMCCD) camera. Transmitted light images were collected using the camera in conventional CCD mode with a 250-ms exposure time. Luminescence emissions from the full-length NanoLuc or the NanoBiT complex were measured for 10-s exposure with a gain of 15-30. Images were taken as 8-bit images with 512 × 512 pixels per frame.

| BRET between NanoLuc-VEGFR2 and fluorescent NRP1
HEK293T-NFAT-ReLuc2P cells were plated in white 96-well plates pre-coated with poly-D-lysine (0.01 mgÁml −1 in PBS) at 25,000 cells per well in DMEM containing 10% FBS. Following 24 h, cells were transiently transfected with a total 125-ng cDNA per well using FuGENE HD at a 3:1 ratio of reagent to cDNA. Cells were transfected with a constant amount of NanoLuc-VEGFR2 (25-ng cDNA per well).
Additional wells only contained NanoLuc-VEGFR2. These transfection solutions were made up to equivalent to 125 ng per well using empty pcDNA3.1/Zeo vector in serum-free DMEM. Cells were incubated for another 24 h at 37 C/5% CO 2 . On the day of the experiment, cells were treated with 0.2-μM membrane-impermeant HaloTag-AlexaFluor488 substrate or 0.2-μM SNAP-Surface AlexaFluor488 substrate in serum-free DMEM/0.1% BSA. Cells were incubated for 30 min at 37 C/5% CO 2 . They were then washed twice with 100 μl per well HBSS/0.1% BSA and replaced with a final volume of 50 μl per well HBSS/0.1% BSA. At this stage, fluorescence emissions were quantified using the PHERAstar FS platereader using filters for excitation at 485 nm and emission at 520 nm. Cells were then incubated with the NanoLuc substrate furimazine (10 μM) for 5 min. Emissions were recorded using the PHERAstar FS platereader using filters simultaneously measuring NanoLuc emissions at 475 nm (30-nm bandpass) and AlexaFluor488 emissions at 535 nm (30-nm bandpass). BRET ratios were calculated as fluorescence over luminescence emissions from the second of three cycles. were quantified using the PHERAstar FS platereader using filters for excitation at 485 nm and emission at 520 nm. Cells were incubated with 10-μM furimazine for 10 min; then luminescence and fluorescence emissions were recorded using PHERAstar FS platereader.

| Luminescence from NanoBiT complementation
Emissions were simultaneously measured for NanoLuc at 475 nm (30-nm bandpass) and AlexaFluor488 at 535 nm (30-nm bandpass). For saturation experiments, increasing concentrations of VEGF 165 a-TMR or VEGF 165 b-TMR (0.5-20 nM) were added in the presence or absence of a high concentration of corresponding unlabelled ligand (100 nM, $100-fold greater than the estimated K d value). Following incubation for 60 min in the dark at 37 C, the NanoLuc substrate furimazine (10 μM) was added to each well and equilibrated for 5 min to enable NanoLuc-mediated furimazine oxidation and resulting luminescence emissions. Emissions were recorded using the PHERAstar FS platereader (BMG Labtech) using a filter simultaneously measuring NanoLuc emissions at 450 nm (30-nm bandpass) and TMR emissions using a longpass filter at 550 nm. BRET ratios were calculated as fluorescence over luminescence emissions from the second of three cycles.

| Fluorescent VEGF-A binding at a VEGFR2/ NRP1 NanoBiT complex
For kinetic experiments, cells were pretreated with furimazine (10 μM) for 5 min to enable NanoLuc-mediated furimazine oxidation and resulting luminescence emissions. BRET ratios were then measured per well using the PHERAstar FS platereader using the filters above. Following four initial measurements, intact cells were stimulated with 0.5-20 nM of VEGF 165 a-TMR or VEGF 165 b-TMR.
Emissions were recorded every 30 s for 20 or 90 min, using the temperature control function of the PHERAstar FS platereader to maintain conditions at 37 C.

| Data analysis
Data were analysed using GraphPad Prism 7.02 (GraphPad Software, La Jolla, CA, USA; RRID:SCR_002798). Data are presented as mean ± SEM. All experiments were performed in three to six independent experiments with duplicate or triplicate wells (see figure legends for details). Drug additions were randomly allocated to wells within each 96-well plate. Statistical significance was defined as P < .05. Confocal images were collected using Zen 2010 software (Zeiss, Germany). Confocal images were processed and analysed using ImageJ Saturation binding curves were fitted simultaneously for total (VEGF 165 a-TMR or VEGF 165 b-TMR alone) and non-specific binding (obtained in the presence of 100 nM of unlabelled VEGF-A) using the equation: Association kinetic studies were performed with four concentrations of ligand simultaneously for global fitting in order to determine the k on and k off . Kinetic studies of fluorescent ligand binding measured over time were fitted to a mono-exponential association function: This utilised the following relationship with k obs : further describing association rate, k on , in units of min −1 ÁM −1 ; and dissociation rate, k off , in min −1 . GraphPad Prism was used to fit each association curve to the above equations with the parameters for k on and k off shared between the fits for the four different concentrations of fluorescent ligand used in each experiment. This allowed values for k on and k off values to be determined for each experiment. These kinetic data were also used to estimate the binding affinities, due to the relationship between dissociation and association rates within an equilibrium: HaloTag-VEGFR2 and SnapTag-NRP1 were labelled with membraneimpermeant HaloTag-AlexaFluor488 and SnapTag-AlexaFluor647

| Complementation of NanoBiT fragments using N-terminal tagged VEGFR2 and NRP1
We then applied a split NanoBiT approach to isolate luminescence emissions from a defined VEGFR2/NRP1 heteromeric complex. Enzymic luciferase activity requires complementation between the large fragment (LgBiT) and the short 11-amino-acid tag (HiBiT or SmBiT).
To determine the optimal configuration for luminescence emissions, each NanoBiT fragment was appended to the N-terminus of both full-length VEGFR2 and NRP1. Luminescence emissions were higher for the combination with LgBiT-tagged VEGFR2 and the short fragment attached to NRP1 (Figure 3a). Emissions from the HiBiT complex were approximately 10-fold higher than the SmBiT complex.
NanoBiT-tagged receptors expressed independently emitted minimal luminescence in the presence of furimazine relative to the comple-

| Influence of NanoBiT tags on VEGFR2-mediated signalling
We confirmed that the NanoBiT fragments did not interfere with VEGFR2 signalling using an NFAT reporter gene assay . Concentration-response curves for VEGF 165 a were compared between cells stably expressing VEGFR2 tagged at the N-terminus with LgBiT, HiBiT, or SmBiT ( Figure 5) 3.4 | Nanomolar affinity of fluorescent VEGF-A at a defined VEGFR2/NRP1 complex  (Figure 7a).
Fitted to a global association curve (Table 1), VEGF 165 b-TMR had a slightly slower association rate constant (k on ) for the VEGFR2/ NRP1 complex (2.29 × 10 6 ± 0.30 × 10 6 min −1 ÁM −1 ) compared to F I G U R E 3 Complementation of a VEGFR2/NRP1 NanoBiT complex. (a) To determine the optimal orientation of labelling with NanoLuc Binary Technology (NanoBiT) fragments, each receptor was tagged with the 18-kDa fragment (LgBiT) and a smaller 11-amino-acid fragment.
HiBiT has a higher intrinsic affinity to complement with LgBiT compared to SmBiT (Dixon et al., 2016). HEK293T cells were transiently transfected in 96-well plates with equal amounts of LgBiT-tagged receptor (50-ng cDNA per well) and HiBiT-or SmBiT-tagged receptor (50-ng cDNA per well). Cells were incubated with 10-μM furimazine in HBSS/0.1% BSA for 10 min (37 C). Data were normalised to untransfected cells (0%) and HiBiT-NRP1/LgBiT-VEGFR2 (100%) per experiment. Data are expressed as mean ± SEM from five independent experiments (LgBiT-VEGFR2) or three independent experiments (LgBiT-NRP1), each with triplicate wells. (b) To compare emissions from individual NanoBiT-tagged receptors relative to a complemented NanoBiT complex, HEK293T cells were transiently transfected in 96-well plates with LgBiT-VEGFR2, HiBiT-NRP1, or SmBiT-NRP1 (50-ng cDNA per well). Dual expression cells expressed a complemented NanoBiT complex (filled bars) whereas single constructs (empty bars) were transfected with 50-ng cDNA per well of empty pcDNA3.1/Zeo vector for 100-ng total cDNA per well. Raw luminescence emissions were plotted as mean ± SEM from five independent experiments. (c) Cells expressed a single NanoBiT-tagged construct, with or without purified HiBiT or LgBiT (20-nM). Raw emissions were plotted as mean ± SEM from five independent experiments. (d) Prevention of NanoBiT complex formation by co-expression of increasing amounts of competing VEGFR2 or NRP1. HEK293T cells were transfected with equal amounts of LgBiT-VEGFR2 (50-ng cDNA per well) and either HiBiT-NRP1 or SmBiT-NRP1 at 50-ng cDNA per well. Cells were also transfected with increasing amounts of HaloTag-NRP1 (0-to 200-ng cDNA per well), as well as with pcDNA3.1/Zeo empty vector (for 300-ng total cDNA per well). Data were normalised to untransfected cells (0%) and the complemented NanoBiT complex in the absence of competing receptor (100%) per experiment. Data are expressed as mean ± SEM from three independent experiments, each with triplicate wells. In each experiment (a-d), cells were incubated with furimazine (10 μM) in HBSS/0.1% BSA for 10 min (37 C). Luminescence emissions (475-505 nm) were measured by the PHERAstar FS platereader HiBiT complex (Figure 7c). Association binding curves were globally fitted to kinetic data from the initial 20 min due to this decline (Table 1).
VEGF 165 a-TMR association kinetics at the NanoBiT complex in the initial 20 min were more comparable to NanoLuc-VEGFR2 than NanoLuc-NRP1 (NanoBiT k obs = 0.33 ± 0.04 min −1 , NanoLuc-VEGFR2 k obs = 0.31 ± 0.03 min −1 , NanoLuc-NRP1 k obs = 0.93 ± 0.09 min −1 ; n = 5 per group). These observed rate constants were significantly slower at the complex than NRP1 alone (repeated-measures ANOVA and Holm-Šidák's multiple comparisons; P < .05, n = 5 for each). These data suggest that the ligand binding profile for VEGF 165 a-TMR at the NanoBiT complex reflected VEGFR2 binding kinetics, as opposed to the faster binding observed at NRP1. NanoBiT complex can therefore be defined as a first-order reaction.

| Fluorescent VEGF-A kinetics were similar for the SmBiT complex
Binding kinetics were also monitored at the SmBiT complex using four concentrations of VEGF 165 a-TMR (Figure 8c). From the fitted data from the initial 20-min period using a global fit, there were no differences between the association kinetic parameters derived for VEGF 165 a-TMR for the HiBiT and SmBiT complexes. There was a linear relationship between the derived observed association rate (k obs ) constants and VEGF 165 a-TMR concentration (Figure 8d). Despite having the potential to bind to both receptors within the complex, the interaction between VEGF 165 a-TMR and the NanoBiT complex could also be defined by a first-order reaction.
F I G U R E 6 Saturation binding of VEGF 165 b-TMR and VEGF 165 a-TMR at a HiBiT complex of VEGFR2 and NRP1. (a) Fluorescent VEGF-A ligand binding was monitored at a defined complex of LgBiT-VEGFR2 and HiBiT-NRP1. Two distinct fluorescent VEGF-A isoforms were investigated, denoted by the x; the a isoform can engage both VEGFR2 and NRP1 whereas the b isoform cannot interact with NRP1. In the presence of furimazine, individual receptors do not emit luminescence in isolation. Upon NanoBiT complementation, luminescence emissions can excite the tetramethylrhodamine (TMR) in proximity. NanoBiT therefore only acts as a luminescent donor when VEGFR2 and NRP1 are in complex. (b,c) HEK293T cells were transfected in six-well plates with equal amounts of LgBiT-VEGFR2 (750-ng cDNA per well) and HiBiT-NRP1 (750-ng cDNA per well). Following 24 h, transfected cells were seeded in 96-well plates. On the day of experimentation, cells were incubated with increasing concentrations of VEGF 165 b-TMR (b) or VEGF 165 a-TMR (c). This was performed in the presence or absence of 100-nM VEGF 165 b (b) or VEGF 165 a (c) to determine non-specific binding. This is shown with fluorescent ligand only or with unlabelled ligand. Following 60 min at 37 C, 10-μM furimazine was added for 10 min (37 C). Emissions were measured on the PHERAstar platereader. BRET ratios are expressed as mean ± SEM from three independent experiments with duplicate wells 3.7 | Similar complex pharmacology using a binding-dead mutant of NRP1 Alternatively, cells were transfected with equal amounts of NanoLuc-VEGFR2 or NanoLuc-NRP1 (750-ng cDNA per well) and empty pcDNA3.1/ Zeo vector (750-ng cDNA per well). Following 24 h, transfected cells were seeded in 96-well plates. On the day of experimentation, cells were pretreated with furimazine (10 μM) and left to equilibrate at 37 C for 10 min. (a) Cells expressing the NanoBiT complex (LgBiT-VEGFR2/HiBiT-NRP1) were stimulated with four concentrations of VEGF 165 b-TMR added at x = 0. Kinetic data were fitted to a global association model with an unconstrained k on from the 90-min time course. (b) On the same plate, the real-time binding profile of 20-nM VEGF 165 b-TMR was monitored in cells only expressing either NanoLuc-VEGFR2 or NanoLuc-NRP1 (left hand Y-axis). This was directly compared to binding of the same concentration of VEGF 165 b-TMR at the LgBiT-VEGFR2/HiBiT-NRP1 NanoBiT complex (right hand Y-axis). (c) Cells expressing the HiBiT complex were stimulated with four concentrations of VEGF 165 a-TMR. Kinetic data were fitted to a global association model without a constrained k on from the initial 20 min due to the latter decline in BRET ratio. (d) The real-time binding profile of a saturating concentration of VEGF 165 a-TMR (10 nM) was compared between cells expressing LgBiT-VEGFR2/HiBiT-NRP1 (right hand Y-axis) to cells only expressing NanoLuc-VEGFR2 or NanoLuc-NRP1 (left hand Y-axis). For each experiment, emissions were simultaneously measured on the PHERAstar FS platereader every 30 s for 90 min at 37 C. BRET ratios were baseline corrected to vehicle at each time point per experimental replicate. In (b) and (d), the x axis was split to highlight the initial association (20 min) and long-term BRET signal (90 min). Data represent mean ± SEM from five independent experiments with duplicate wells. Derived k on , k off , and kinetic pK d parameters are in Table 1 binding from this VEGFR2/NRP1 Y297A complex, VEGF 165 a-TMR exhibited saturable binding at the NanoBiT complex ( Figure 9c). This was displaced by a high concentration of unlabelled VEGF 165 a, confirming that there was low non-specific binding. Derived equilibrium dissociation constants were in the nanomolar range and similar to the wild-type NanoBiT complex (VEGF 165 a-TMR/NanoBiT Y297A K d = 1.55 ± 0.38; pK d 8.84 ± 0.11; n = 3). Binding kinetics at the mutant NanoBiT complex were then monitored using four concentrations of VEGF 165 a-TMR (Figure 9d). This had a profile identical to that of VEGF 165 a-TMR at the wild-type HiBiT complex (Figure 7c), in that there was a small decline in BRET ratio following 30-60 min.

| DISCUSSION
NanoBiT technologies were used to quantify the real-time binding of two fluorescent VEGF-A isoforms at a defined receptor/co-receptor complex between VEGFR2 and NRP1 in living cells at 37 C. Previous work identified differences between VEGFR2 and NRP1 pharmacology in terms of their binding kinetics and localisation when expressed on their own . VEGFR2 and NRP1 are, however, endogenously co-expressed together in endothelial cells and tumour cells (Fantin et al., 2013;Koch et al., 2014;Lee-Montiel et al., 2015;Prahst et al., 2008;Whitaker et al., 2001). We first demonstrated that full-length VEGFR2 and NRP1 constitutively formed a heteromeric complex in living HEK293T cells. To then probe how this specific receptor/co-receptor heteromer interacted with ligand, we established a novel approach to quantify fluorescent VEGF-A binding at a defined complex using split NanoBiT fragments (Dixon et al., 2016). VEGFR2 and NRP1 tagged at their N-terminus with HiBiT and LgBiT tags led to NanoBiT complementation with minimal luminescence when each was expressed alone.  (Gelfand et al., 2014;Prahst et al., 2008;Whitaker et al., 2001) and proximity ligation assays using antibodies in situ on tumour tissue . Förster resonance energy transfer (FRET) has also been used to demonstrate complex formation using truncated VEGFR2 and full-length NRP1 tagged with fluorophores at their C-terminus (King et al., 2018). Here, we initially used BRET between full-length VEGFR2 and NRP1 tagged at their N-terminus with NanoLuc or a fluorophore to confirm complex formation in the absence of added VEGF-A. The approach monitored complex formation that originated at the cell membrane, because membrane-impermeant fluorophore-conjugated HaloTag or SnapTag substrates were used. Basal VEGFR2/NRP1 complex formation was also confirmed using both HiBiT-VEGFR2 and LgBiT-NRP1 complementation and the reverse LgBiT-VEGFR2 and HiBiT-NRP1 orientation.
Following the discovery that VEGF 165 a had faster binding kinetics for binding to NRP1 than to VEGFR2 when expressed on their own , it was proposed that the presence of NRP1 might enhance VEGF 165 a binding to the heteromeric complex.
The application of both NanoBiT technology and NanoBRET to monitor exclusively VEGF 165 a-TMR binding to VEGFR2/NRP1 complexes allowed us to test this hypothesis directly. Interestingly, the initial association kinetics (during the first 20 min) for VEGF 165 a-TMR binding to the VEGFR2/NRP1 heteromeric complex were closer to those observed at NanoLuc-VEGFR2 in isolation than to NanoLuc-NRP1.
This was evident from quantification of the observed rate constant T A B L E 1 Summary of binding parameters derived at the NanoBiT complex for VEGF 165 b-TMR and VEGF 165 a-TMR, compared to published values from receptors expressed alone Kinetic pK d k on (min −1 ÁM −1 ) k off (min −1 ) VEGF 165 b-TMR HiBiT complex 7.81 ± 0.10 (5) 2.29 × 10 6 ± 0.30 × 10 6 (5) 0.037 ± 0.007* (5) SmBiT complex 8.43 ± 0.17 (5) 2.94 × 10 6 ± 0.55 × 10 6 (5) 0.012 ± 0.003 (5) VEGF 165 a-TMR HiBiT complex 8.83 ± 0.12 (5) 3.12 × 10 7 ± 0.43 × 10 7 (5) 0.046 ± 0.007 (5) SmBiT complex 8.83 ± 0.31 (5) 2.83 × 10 7 ± 0.69 × 10 7 (5) 0.046 ± 0.020 (  were no differences observed at the level of ligand binding to F I G U R E 8 Fluorescent VEGF-A binding kinetics at a NanoBiT VEGFR2/NRP1 complex using split tags with lower intrinsic affinity. HEK293T cells were transfected in six-well plates with equal amounts of LgBiT-VEGFR2 (750-ng cDNA per well) and SmBiT-NRP1 (750-ng cDNA per well). Following 24 h, transfected cells were seeded in 96-well plates. Cells were pretreated with furimazine (10 μM) and left to equilibrate at 37 C for 10 min. (a) Cells were stimulated with four concentrations of VEGF 165 b-TMR added at x = 0. Kinetic data were fitted to a global association model without a constrained k on from the 90-min time course. For clarity, the 10-nM data set has not been included in the figure. (b) The derived rate constant, k obs , was obtained from exponential association curves fitted for each of the four fluorescent ligand concentrations. These were plotted against VEGF 165 b-TMR concentration and fitted against a linear regression (HiBiT complex y = 0.0023x + 0.034, R 2 = 0.46; SmBiT complex y = 0.0026x + 0.01, R 2 = 0.65). (c) Cells were stimulated with four concentrations of VEGF 165 a-TMR. Kinetic data were fitted to a global association model without a constrained k on from the initial 20 min. For clarity, the 5-nM data set has not been included in the figure. (d) The derived k obs for each fluorescent ligand at all four concentrations were plotted against each VEGF 165 a-TMR concentration and fit with a linear regression (HiBiT complex y = 0.0024x + 0.10, R 2 = 0.72; SmBiT complex y = 0.0025x + 0.09, R 2 = 0.62). Emissions were simultaneously measured on the PHERAstar FS platereader every 30 s for 90 min at 37 C. BRET ratios were baseline corrected to vehicle at each time point per replicate. Data represent mean ± SEM from five independent experiments with duplicate wells in each independent experiment. Derived k on , k off , and kinetic pK d parameters are shown in Table 1 NanoLuc-VEGFR2 when it was expressed alone . VEGF 165 b is, however, selective for VEGFR2 and unable to interact with NRP1 . The real-time BRET signal for VEGF 165 b-TMR remained elevated in intact cells at the NanoBiT complex over the full 90-min time course. This resembled observations made with NanoLuc-VEGFR2 in membrane preparations and was quite different to the decline in BRET signal normally observed in intact HEK293T cells (Peach et al., 2019). In contrast, the profile for VEGF 165 a-TMR at the HiBiT complex had a small decrease at latter time points, albeit to a lesser extent than at NanoLuc-VEGFR2 in intact cells (Peach et al., 2019).
This reduction in BRET signal for NanoLuc-VEGFR2 following F I G U R E 9 Ligand binding of VEGF 165 a-TMR at a NanoBiT complex with a binding-dead NRP1 mutant. (a) VEGF 165 a-TMR ligand binding was monitored at a defined NanoBiT complex between LgBiT-VEGFR2 and a HiBiT-NRP1 VEGF-A binding-dead mutant in the b1 domain (Y297A). (b) HEK293T cells were transiently transfected in 96-well plates with LgBiT-VEGFR2, HiBiT-NRP1 Y297A, or SmBiT-NRP1 Y297A (50-ng cDNA per well). Dual expression cells expressed a complemented NanoBiT complex with the HiBiT or SmBiT tag. Cells also expressed single constructs (empty bars) were transfected with 50-ng cDNA per well of empty pcDNA3.1/Zeo vector. Cells were incubated with 10-μM furimazine in HBSS/0.1% BSA for 10 min (37 C). Luminescence emissions (475-505 nm) were measured by the PHERAstar FS platereader. Data were normalised to untransfected cells (0%) and HiBiT-NRP1 Y297A/LgBiT-VEGFR2 (100%) per experiment. Data are expressed as mean ± SEM from five independent experiments, each with triplicate wells. (c,d) HEK293T cells were transfected in six-well plates with equal amounts of LgBiT-VEGFR2 (750-ng cDNA per well) and HiBiT-NRP1 Y297A (750-ng cDNA per well). Following 24 h, transfected cells were seeded in 96-well plates. (c) On the day of experimentation, cells were incubated with increasing concentrations of VEGF 165 a-TMR, with or without 100-nM VEGF 165 a to determine non-specific binding. Following 60 min at 37 C, 10-μM furimazine was added for 10 min (37 C). Emissions were measured on the PHERAstar platereader (550-LP/460-480 nm). BRET ratios are expressed as mean ± SEM from three independent experiments with duplicate wells. Derived equilibrium dissociation constants (K d ) are in the text. (d) Cells were pretreated with furimazine (10 μM) and left to equilibrate at 37 C for 10 min. Cells were incubated with four concentrations of VEGF 165 a-TMR. Kinetic data were fitted to a global association model without a constrained k on from the initial 20 min. Emissions were simultaneously measured on the PHERAstar FS platereader every 30 s for 90 min at 37 C. BRET ratios were baseline corrected to vehicle at each time point per experimental replicate. Data represent mean ± SEM from five independent experiments with duplicate wells. Derived k on , k off , and kinetic pK d parameters are noted in the text 20 min has been linked to VEGF-A/VEGFR2 endocytosis leading to a change in localisation and local pH, as this decline was absent in membrane preparations and not observed for binding to NanoLuc-NRP1 (Peach et al., 2019). These data suggest that the presence of NRP1 in VEGFR2 heteromeric complexes may reduce the extent of VEGFR2 endocytosis normally seen when VEGFR2 is expressed alone.
Imaging studies exploited the compatibility of HaloTag and SnapTag technologies to label distinct receptors co-expressed by the same cell to monitor co-localisation at 37 C. Unlike immunofluorescent antibody labelling, these experiments can be performed in living cells and do not require cell fixation or cell permeabilisation to access internalised receptors. These distinct tags confirmed that VEGFR2 was largely intracellular whereas NRP1 was highly localised around the plasma membrane when they were both co-expressed in the same cell. NRP1 was also localised in filopodia-like projections in HEK293T cells that resembled the filopodia of endothelial tip cells (Fantin et al., 2013(Fantin et al., , 2015. Although co-localisation studies were limited by the axial resolution limit of basic confocal microscopy, experiments monitoring receptor-receptor BRET confirmed that VEGFR2 and NRP1 were in proximity (<10 nm). Live cell confocal imaging and bioluminescence imaging data both suggested that VEGFR2 and NRP1 were co-localised in both intracellular compartments and at the plasma membrane. VEGFR2 is subject to macropinocytosis in the absence or presence of ligand . This bulk transport mechanism could therefore non-selectively engulf surrounding NRP1 in living cells. There is evidence in HUVECs for co-localisation between VEGFR2 and NRP1 both at the plasma membrane in the absence of stimulation (Lee-Montiel et al., 2015) or within intracellular sites following 20-min VEGF 165 a stimulation (Muhl et al., 2017). As the NanoLuc/NanoBiT substrate furimazine is membrane permeable, luminescence could be emitted from complexes anywhere in the cell regardless of subcellular localisation.
NanoBiT technologies take advantage of NanoLuc, a small enzyme engineered from a deep sea shrimp with bright, ATPindependent luminescence emissions (Hall et al., 2012). The small, 11-amino-acid NanoBiT fragment also has mutations that confer differing intrinsic affinities for the LgBiT fragment. For example, HiBiT has a much higher intrinsic affinity for LgBiT than SmBiT (Dixon et al., 2016). Luminescence emissions from HiBiT-containing complexes were higher than for the corresponding SmBiT-containing complex, as observed previously for NanoBiT-tagged GPCRs (Botta et al., 2019). This is likely to be due to differences in the affinity of LgBiT for SmBiT or HiBiT. Dixon et al. (2016) confirmed the rapid kinetics of NanoBiT tag complementation (seconds) and, therefore, this is unlikely to affect the ligand binding kinetics monitored at the VEGFR2/NRP1 complex (minutes). The intrinsic affinity between HiBiT and LgBiT can vary according to the expression system and protein conformation, as observed for chemokine GPCRs using the purified exogenous tag in different assay set-ups (White et al., 2020).
While the intrinsic affinity between NanoBiT tags must be taken into consideration, luminescence emissions from both HiBiT and SmBiT complexes were displaceable by increasing amounts of competing NRP1 (Figure 3d). The kinetic parameters derived from HiBiT and SmBiT complexes were also comparable, suggesting that VEGFR2-NRP1 complex formation was not being driven by the affinity of the HiBiT tag for LgBiT.
Despite its ability to up-regulate VEGF-A/VEGFR2 signalling in physiological and pathophysiology, the mechanism by which NRP1 up-regulates VEGFR2 signalling remains largely unknown. NRP1 can interact with a number of other growth factors (Banerjee et al., 2006;Rizzolio et al., 2012;West et al., 2005). Therefore, understanding how NRP1 co-expression influences RTK function has implications for other receptors contributing to cancer drug resistance. Our NanoBiT approach allowed us to isolate VEGF-A ligand binding at a defined complex of VEGFR2 and NRP1 and suggested that NRP1 did not increase the affinity or association binding kinetics of VEGF 165 a at VEGFR2. While NRP1 appeared to have no direct effect on ligand binding to a VEGFR2/NRP1 complex expressed within the same cell, NRP1 (which is quite often expressed endogenously at higher levels than VEGFR2) could still act as a reservoir for growth factors and create a localised gradient due to its interactions with the extracellular matrix (Shintani et al., 2006;Windwarder et al., 2016).
In summary, we have described here an approach using NanoBiT technology and NanoBRET to monitor in real time the binding of VEGF-A isoforms to defined heteromeric complexes containing both VEGFR2 and NRP1. Understanding the ligand binding properties of a specific heteromeric receptor oligomer is important for studying the molecular pharmacology of individual VEGF-A isoforms in primary cells, such as endothelial and tumour cells, which often endogenously co-express both receptors. This specific technique allowed us to determine for the first time the ligand binding kinetics of VEGF 165 a-TMR and VEGF 165 b-TMR to a defined VEGFR2-NRP1 complex. We were able to use bioluminescence imaging and confocal microscopy to determine that VEGFR2-NRP1 complexes are localised in both intracellular compartments and at the plasma membrane. At the plasma membrane, the presence of NRP1 within the heteromeric complex appeared to reduce the extent of agonist-induced VEGFR2 endocytosis normally observed when it is expressed alone. The presence of NRP1 within the VEGFR2-NRP1 heteromeric complexes did not enhance VEGF 165 a-TMR binding, and a NRP1 binding-dead mutant (Y297A) had no effect on the binding of VEGF 165 a-TMR, or the formation of VEGFR2-NRP1 complexes, suggesting that the high affinity binding site for VEGF 165 a on NRP1 might be masked within the heteromeric complexes. In keeping with this conclusion, VEGF 165 b-TMR, which does not bind to NRP1, had a very similar binding profile to the heteromeric complex to that observed with VEGF 165 a-TMR. This approach to monitor the binding profile of defined oligomeric complexes should be applicable to a wide range of receptor systems and facilitate drug discovery aimed a heteromeric complex. This approach could also be developed further to observe ligand interactions with a specified oligomer in vivo due to the small size of the HiBiT/SmBiT tag and bright luminescence emissions, in a similar way to that we have recently reported for a GPCR tagged with the full-length NanoLuc (Alcobia et al., 2018). Given the clinical importance of therapeutic agents targeting VEGF or VEGFR2 in cancer and other pathologies, understanding the mechanism by which NRP1 up-regulates VEGF-A/VEGFR2 signalling is a priority for identifying new targets to improve the long-term efficacy and adverse effects of VEGF-targeted therapeutics in cancer.