Discovery of a novel fluorescent chemical probe suitable for evaluation of neuropilin‐1 binding of small molecules

Abstract Neuropilin‐1 (NRP1) is emerging as an important molecule in immune signaling where it has been shown to modulate the actions of TGF‐β1 in macrophages and regulatory T cells. The development of cost‐effective and reliable assays for NRP1 binding is therefore important. We synthesized three new NRP1 small molecule fluorophores and examined their performance as fluorescent polarization probes. One molecule DS108 exhibited favorable binding and fluorescent characteristics and allowed us to establish a simple assay suitable for medium to high throughput screening of small molecules.


| INTRODUCTION
The ability of neuropilin-1 (NRP1) to bind and augment the action of growth factors such as vascular endothelial growth factor (VEGF), transforming growth factor-β1 (TGF-β1), placental growth factor (PLGF), HGF (scatter factor) and Semaphorins 3A, 4F (Pellet-Many, Frankel, Jia, & Zachary, 2008) are consistent with its emerging role as a tumor promoting receptor acting by a number of mechanisms. These mechanisms can be via modulation of the immune response to tumors through affecting the function of macrophages (Nissen, Selwood, & Tsirka, 2013) and regulatory T cells (T regs ) (Delgoffe et al., 2013), via angiogenesis by promotion of NRP1/VEGF-A signaling (Pan et al., 2007); through prevention of tumor cell migration by binding to NRP1 ; or a direct effect on the tumor cells (Grun, Adhikary, & Eckert, 2016). The binding site on NRP1 b1 is formed by loops in the protein structure forming a "receptor" for a C-terminal arginine residue (Jarvis et al., 2010). This is variously termed in the literature as the tuftsin site, arginine receptor, or aromatic box (Y297, W301, and Y353). The peptide and small molecule antagonists reported dating all bind to this site (Peng, Bai, Zhu, Hu, & Xu, 2019).
Many assay systems have been developed to detect NRP1 binding ranging from classical radiolabelled formats (Jia et al., 2006), surface plasmon resonance (SPR), luminescence (Powell et al., 2018) to bead based systems (Huang et al., 2019) and homogeneous time resolved fluorescence (HTRF) (Auriau et al., 2018). These assays may be expensive and complicated to implement. The development of simple and reliable assays for NRP1 binding and function is therefore of importance.
Carboxyfluorescein (Flu)-labeled RPARPAR peptide is a known The new probes were based on the structure of EG01377 2, a compound identified in our group and known to be selective for NRP1 over the closely related NRP2 receptor and possess submicromolar NRP1 antagonistic properties (Powell et al., 2018). The design incorporated several combinations of polyethylene glycol (PEG) and triazole click linker units to examine whether the fluorophore could be attached without compromising the intrinsic binding of EG01377 (Figure 1).
Three structurally distinct probes (3, 4, and DS108, Figure 1) were synthesized and evaluated against the literature RPARPAR probe. We used SPR as an orthogonal biophysical technique to assess the binding properties of these probes and to determine their respective K D values. From this, we were able to quickly identify conditions in which to assess the FP dynamic range of each probe, by optimizing probe and NRP1-b1 concentrations. Subsequent NRP1-b1 protein titration experiments and competition assays with known NRP-b1 ligands (EG01377 and EG03286) confirmed the utility of this assay for the medium-to-high throughput discovery of novel ligands for NRP1-b1 binding.

| SPR
All SPR analysis was performed on a BIAcore T200 system using series S CM5 sensor chips. The Biotage SPR is effectively a stop-flow instrument, and dissociation from the immobilized protein is initiated by the absence of analyte (ligand) when buffer alone is perfused, this setup allows the determination of on and off rates and equilibrium binding constants. Relatively high DMSO concentrations are normal for SPR experiments to limit solubility problems and minimize nonspecific aggregation. Extensive DMSO controls are included and automatically subtracted from the sensorgrams. Sensorgrams were double referenced by subtracting the response on a reference flow cell and a F I G U R E 1 Peptide fluorescein probe and chemical structures of novel NRP1-B1 fluorescent probes used in this study blank sample. Ligands were evaluated against the NRP1-b1 domain. NRP1-b1 was covalently attached to a CM5 chip via amine coupling (Powell et al., 2018) with a surface density of 2,000-3,000 RU. Binding of novel fluorescent probes (0.4-100 μM) were analyzed by multicycle sequential injections (30-120 s association time) followed by undisturbed dissociation (30-60 s). A regeneration step was not used. Peptide stocks were dissolved in dimethyl sulfoxide (DMSO), and the final sample solutions for kinetic affinity experiments contained 3% DMSO in 1× phosphate-buffered saline P20 buffer (PBS-P, Cat no 28995084, GE Healthcare Ltd.). DMSO solvent effects were corrected for with eight calibration solutions (0.5-1.8% DMSO in PBS-P). Equilibrium constants (K D ) were calculated using either kinetic or affinity models, assuming simple 1:1 (Langmuir) binding.
Data processing and analysis were performed using BIAevaluation and OriginPro software. The theoretical R max (the maximal feasible signal between a ligand-analyte pair) for each compound/protein pair was calculated using Scheme 1 (Marquart, 2017).
The experimentally observed R max was then calculated as a percentage of the theoretical R max as a quality control measure.

| FP
Initial experiments were performed using PBS-P buffer with 3% DMSO in a final volume of 80 μl. The reaction plates or tubes were kept on ice during pipetting.
Saturation binding: dynamic range and protein titration experiments were performed at probe concentrations selected with guidance from in-house SPR analysis, literature K D of the untagged compound, and an FP technical resource guide (Invitrogen, 2006). The NRP-b1 concentrations ranged from 10 nM to 30 μM. Samples were prepared in the following order-NRP1-b1 protein (serial dilution as shown in Figure X, 40 μl) in PBS-P buffer, and FP probes (concentrations shown in Tables 2, 40 μl) in PBS-P buffer.
Background FP was blanked using a PBS-P buffer only control.

| Final optimized protocol for FP
The final FP assays were realized at 50 nM of DS108, 300 nM of NRP1 b1b2 All solutions contained a buffer constituted of 10 mM HEPES and 0.5% DMSO, reflecting the low initial concentration of the molecule under investigation needed for the assay. EG00229 and EG01377 starting concentrations were 5 μM only (stock solution consisted of 10 mM of the compounds diluted in 100% DMSO). DMSO derived from DS108 (stock solution 10 mM in 100% DMSO) is at a very low concentration; 100 times lower compared to the 0.5% derived from the compounds and did not interfere with the acquisition of the competition assay curves. The final sample volume in each well was 30 μl. Data were analyzed using OriginPro software. The assays were performed using PHERAstar ® . From the general settings of the PHERAstar ® fluorescence polarization software the Settling time was changed from 0.3 to 0.5 s, which is important in order to enhance the assay's accuracy at lower probe concentrations.
Data processing: raw data were processed using OriginPro curve fitting software to obtain the IC 50 s and error values shown. A doseresponse curve fitting model was utilized for the competition experiments. A web based IC 50 -to-K i converter that computes K i values from experimentally determined IC 50 values was employed (Cer, Mudunuri, Stephens, & Lebeda, 2009) (https://bioinfo-abcc.ncifcrf.gov/IC50_Ki_ Converter/index.php) Each experiment was conducted in triplicate and repeated two times. The statistical reproducibility of this assay was evaluated using the z factor equation (Scheme 2).

| CHEMISTRY
All starting materials were from commercial sources or synthesized by literature procedures as indicated. FAM-RPARPAR was purchased from Peptide Protein Research Ltd, Hampshire, UK.
The reaction was stirred overnight and then water (0.

| Chemistry
The synthesis of the EG01377 derived probes was carried out using the unprotected EG01377 as the starting material. This minimized the number of synthetic steps required. Reaction mixtures could be purified directly by reverse phase chromatography. Fluorescent probe 3 was prepared from EG1377 in two steps ( Figure 2). First the Fmoc-PEG-acid was preactivated with PyBOP in DMF and DIPEA then the EG01377 added. The Fmoc group was removed in the work-up then the resulting amine coupled with 5,6-carboxyfluorescein. Probe 4 was synthesized using a similar coupling of alkyne-PEG-acid to EG01377 to provide alkyne 5 (Figure 3). This was followed by a copper mediated Huisgen click cycloaddition reaction (Rostovtsev, Green, Fokin, & Sharpless, 2002) with the fluorescein alkyl azide to give the product.
For DS108 we wanted to preserve the basic amine group on the EG01377 structure which was known to be important for binding.

| SPR and FP
We have previously validated the SPR system for NRP1 and it provides similar data to other assay systems such as biotinylated-VEGFluciferase or radiolabelled VEGF. Fluorescently tagged probes RPARPAR, 3, 4, and DS108 exhibited SPR K D values of 69.03 ± 22, 24.6 ± 6.3, 3.3 ± 0.2, and 2.13 ± 0.81 μM, respectively ( Figure 5, Table 1). This was compared with the positive control EG01377, which possessed a K D = 1.32 ± 0.08 μM (Powell et al., 2018). All probes demonstrated reasonable equilibrium binding characteristics as evidenced by their sensorgrams and calculated R max values as shown in Table 1.
With steady-state equilibrium binding dissociation constant K D values from SPR analysis in hand, fluorescently tagged RPARPAR, 3, 4, and DS108 were also evaluated for their utility as probes in FP experiments (Table 1). Based on this information and FP guidelines (Moerke, 2009) we assessed the dynamic FP window for the probes via a protein titration experiment ( Figure 6, Table 2).
Despite probe 4 demonstrating reasonable binding properties by SPR (K D = 3.3 ± 0.2 μM), only a very small dynamic window could be obtained (Table 2; Figures 5a, and 6a). In contrast, probe 3 and DS108 showed low μM SPR binding properties, and this correlated with improved FP dynamic window readouts (Table 1). DS108 in particular appeared to possess a superior readout window, and this was confirmed by running full NRP1-b1 titration curves for each probe ( Figure 6a). Pleasingly, DS108 not only showed the greatest FP window but this probe was also able to elicit robust FP dynamic windows at significantly lower NRP1-b1 protein concentrations than the literature FP probe, 5-Flu-RPARPAR, which is an important practical consideration for high throughput drug screening. As a check on the specificity of the probe we evaluated it in a protein titration experiment with NRP1 b1b2 and the NRP1 b1 Y297A mutated protein  of the probe was observed (Figure 6c). We also wanted to discover whether a structurally diverse competitor would displace this probe.
We also employed the bicyclic disulfide bonded peptide, EG03287, which is derived from the C-terminal domain of VEGF-A165, as a competitor compound in this assay. Gratifyingly, concentrationdependent competition of DS108 was also observed with EG03287 ( Figure 6c)  were in agreement (within 1 log concentration unit) with competition and/or binding assays in the literature (Table 3) (Jarvis et al., 2010;Jia et al., 2006;Powell et al., 2018). In addition, the z-factor for this assay was determined to be 0.90, suggesting its statistical robustness.
Final optimization of the assay was conducted by evaluating different buffers, and lower probe and DMSO concentrations. Then, 50 nM of probe concentration was chosen as it was still detectable by the plate reader without compromising the assay's accuracy. First, different concentrations of NRP1 b1 and b1b2 were titrated with 50 nM of probe in order to find out whether this low concentration was able to create a satisfactory assay window. As shown in Figure 6d NRP1 b1b2 domains created a good assay window. From the titration curve 300 nM of NRP1 b1b2 was chosen as it increased the initial signal by 65%. The optimal increase in signal is considered to lie between 50 and 80% as below 50% the assay window is poor and above 80% competition is visualized and the competition assay curve is expected to be a straight line. Secondly, a competition assay was performed using either EG00229 or EG01377. At 0.5% DMSO and using 10 mM of HEPES buffer the assay was able to generate a competition assay curve for both EG00229 and EG01377 (Figure 6e). HEPES buffer was found to be superior to PBS or PBS-P. Miniaturization of the initial FP assay can be exploited for high throughput screening for NRP1 b1b2 inhibitors with expected Ki at the micromolar range. FP assay instruments can read a 384-well plate at around 10 min and at the same time the miniaturized nature of the optimized protocol constitutes a cost effective alternative to the existing screening methods. Finally, the initial protocol was unable to generate a VEGF165 (natural ligand of NRP1 b1b2) competition assay curve. However, the optimized protocol revealed that the reason for this was nonspecific binding of DS108 to VEGF165, suggesting once more the sensitivity of the optimized protocol (Figure 6f).

| CONCLUSION
We have identified DS108, a novel and convenient EG01377-based fluorescent probe for evaluation of NRP1 binding, an important growth factor receptor that is implicated in the progression of various cancers. The probe was utilized as the ligand for a competitive FP binding assay. This probe could serve as a useful addition to the technical tools available for NRP1 study.