Kinetic profiling and functional characterization of 8-phenylxanthine derivatives as A2B adenosine receptor antagonists

A2B adenosine receptor (A2BAR) antagonists have therapeutic potential in inflammation-related diseases such as asthma, chronic obstructive pulmonary disease and cancer. However, no drug is currently clinically approved, creating a demand for research on novel antagonists. Over the last decade, the study of target binding kinetics, along with affinity and potency, has been proven valuable in early drug discovery stages, as it is associated with improved in vivo drug efficacy and safety. In this study, we report the synthesis and biological evaluation of a series of xanthine derivatives as A2BAR antagonists, including an isothiocyanate derivative designed to bind covalently to the receptor. All 28 final compounds were assessed in radioligand binding experiments, to evaluate their affinity and for those qualifying, kinetic binding parameters. Both structure-affinity and structure-kinetic relationships were derived, providing a clear relationship between affinity and dissociation rate constants. Two structurally similar compounds, 17 and 18, were further evaluated in a label-free assay due to their divergent kinetic profiles. An extended cellular response was associated with long A2BAR residence times. This link between a ligand’s A2BAR residence time and its functional effect highlights the importance of binding kinetics as a selection parameter in the early stages of drug discovery.


Introduction
The A 2B adenosine receptor (A 2B AR) belongs to the superfamily of rhodopsin-like G protein-coupled receptors (GPCRs), being a member of the adenosine receptor (AR) family. It has been mapped on chromosome 17p11. [2][3][4][5][6][7][8][9][10][11][12], and as all GPCRs the encoded protein consists of a seven transmembrane (7TM) α-helix architecture [1,2]. Adenosine, a ubiquitous purine nucleoside, is the endogenous ligand for all ARs, i.e. A 1 , A 2A , A 2B and A 3 . These AR subtypes are coupled to different effectors and modulate different physiological and pathophysiological conditions. A 2B AR is the least well characterized of the four AR subtypes, possibly due to its low affinity for adenosine [3]. Under physiological conditions A 2B AR is considered to remain silent, as the extracellular concentration of adenosine ranges from 20 to 300 nM, much lower than the reported half maximal effective concentration (EC 50 ) of adenosine for A 2B AR (EC 50 = 24 μM). In contrast, under pathophysiological conditions extracellular concentrations of adenosine could rise up to 30 μM, therefore resulting in A 2B AR activation and signaling [4,5].
A 2B ARs are present in numerous tissues and organs, including bowel, bladder, lung, brain, as well as on hematopoietic and mast cells [2,6]. Interestingly, A 2B AR expression levels are often (up)regulated during disease. The high expression of the receptor in conjunction with the increased extracellular adenosine concentration under pathophysiological conditions render A 2B AR antagonists interesting pharmacological and therapeutic tools for a broad spectrum of diseases, such as asthma, chronic obstructive pulmonary disease [7,8], colon inflammation [9,10], diabetes [11] and cancer [12]. Adenosine production is upregulated in the tumor microenvironment and acts at both A 2A AR and A 2B AR to facilitate tumor progression in vivo [13]. In cancer models A 2B AR antagonists impede adenosine-induced tumor cell proliferation, angiogenesis and metastasis, and remove immune suppression [14].
Over the past years, various xanthine and non-xanthine derivatives have been synthesized and evaluated for their A 2B AR affinity and selectivity [15][16][17][18]. However, no A 2B ARselective antagonist has reached the market yet for therapeutic use. Only CVT-6883 has completed Phase I clinical trials with no adverse events reported, while another clinical trial for PBF-1129, as drug candidate for locally advanced or metastatic non-small cell lung carcinoma, is under recruitment [19,20].
The 3D structure of A 2B AR has not been elucidated yet, hence, the design of new potential drug candidates is mainly based on more classical structure-affinity relationships or on molecular modeling based on homology to the A 2A AR [21]. Although affinity is a key parameter in pharmacology, it does not necessarily predict in vivo efficacy. During the last decade an increasing number of studies suggested that the study of ligand binding kinetics, quantified by association (k on ) and dissociation (k off ) rate constants, is highly relevant in the early stages of drug discovery, as in vivo efficacy is linked to optimized kinetic characteristics in many cases [22]. A typical example is the neurokinin 1 (NK 1 ) receptor antagonist aprepitant, an antiemetic. Aprepitant has been found to have higher in vivo efficacy than other NK 1 receptor antagonists with similar thermodynamic affinities, due to its long residence time (RT: 1/k off ) at the NK 1 receptor [23].
Here, we report on the synthesis of a number of xanthine-based A 2B AR antagonists, on the affinities of these and a number of previously reported xanthines, and on their kinetic target binding parameters obtained in radioligand binding assays. Although not very selective most xanthine derivatives present high affinity for A 2B AR, while displaying a variety in association and dissociation rate constants. On the basis of these results we also synthesized and pharmacologically profiled compound 29, a xanthine antagonist that presumably binds covalently to A 2B AR. Additionally, we developed a label-free impedance-based assay using intact cells expressing A 2B AR for the further characterization of compounds with diverse kinetic profiles. Compounds 17 and 18 with a long and short RT on the receptor, respectively, were profiled in this assay. Compound 17, with the longer RT, had a more sustained effect than compound 18, suggesting this assay has translational relevance.

Chemistry
Synthetic reagents and solvents were purchased from Sigma-Aldrich (St. Louis, MO, USA), or prepared as reported. 1 H NMR spectra were obtained with a Bruker 400 spectrometer using CDCl 3 , DMSO-d 6 , acetone-d 6 or CDCl 3 as a solvent. The chemical shifts are expressed as ppm, and the coupling constants (J) are given in Hz. High resolution mass (HRMS) measurements were performed on a proteomics optimized Micromass Q-TOF-2 (Waters, Milford, MA, USA). All chemicals not further specified were from standard commercial sources. Compounds 2, 8, 11-19, 26 and 27 were reported in Kim et al. [24].

2.2.3.
Membrane preparation-CHO-spap-hA 2B AR cells were cultured as a monolayer in 15 cm ø plates to about 90% confluency. Cells were removed from the plates by scraping into 5 mL of phosphate-buffered saline (PBS) and centrifuged for 5 min at 1500 rpm. The resulting pellets were resuspended in ice-cold Tris-HCl buffer (50 mM Tris-HCl, pH 7.4) and homogenized using an Ultra Turrax homogenizer (IKA Werke GmbH & Co.KG, Staufen, Germany). Centrifugation at 31,000 rpm in an Optima LE-80 K ultracentrifuge with Ti-70 rotor (Beckman Coulter, Fullerton, CA) at 4 °C for 20 min, resulted in separation of membranes and cytosolic fraction. Subsequently, pellet was resuspended in 10 mL Tris-HCl buffer, homogenized and centrifuged once again. The final pellet was suspended in assay buffer (50 mM Tris-HCl buffer, 0.1% (w/v) CHAPS, pH 7.4), ADA was added to break down endogenous adenosine, and the homogenization step was repeated. Aliquots were stored at −80°C and the membrane protein concentration was determined by a BCA protein determination assay [26]. The BCA results were measured in a Wallac EnVision 2104 Multilabel Reader (Perkin Elmer, Groningen, The Netherlands).

Radioligand binding assay-
In all radioligand binding experiments, CHOspap-hA 2B AR membranes were thawed and homogenized using an Ultra Turrax homogenizer at 24,000 rpm (IKA-Werke GmbH & Co.KG, Staufen, Germany), diluted in assay buffer to the desired concentration (10-30 μg per well or Eppendorf tube). All materials were brought to 25 °C, 30 min prior to the experiment. ZM241385 (10 μM) was used to determine nonspecific binding (NSB). DMSO concentrations were 2% for all compounds except for 8, 12-18 and 20-22, where the concentration was 0.25%. The two different DMSO concentrations had negligible effects on the radioligand binding results. Finally, total radioligand binding (TB) did not exceed 10% of the [ 3 H]PSB-603 present in the assay in order to prevent ligand depletion.

Displacement experiments.:
Were performed using 1.5 nM [ 3 H] PSB-603 and a competing unlabeled ligand at multiple concentrations diluted in assay buffer (50 mM Tris-HCl, 0.1% (w/v) CHAPS, pH 7.4). Binding was initiated by addition of CHO-spap-hA 2B AR membrane aliquots to reach a total volume of 100 μL. Samples were incubated at 25 °C for 2 h to reach equilibrium. The incubation was terminated by rapid vacuum filtration over 96-well Whatman GF/C filter plates using a PerkinElmer Filtermate harvester (PerkinElmer, Groningen, Netherlands). Filters were subsequently washed ten times using ice-cold wash buffer (50 mM Tris-HCl, 0.1% (w/v) BSA, pH 7.4). Filter plates were dried at 55 °C for about 45 min and afterwards 25 μL Microscint (PerkinElmer) was added per well. Filter-bound radioactivity was determined by liquid scintillation spectrometry using a 2450 Microbeta 2 scintillation counter (PerkinElmer).

Saturation binding experiments.:
Were carried out by incubating increasing concentrations of [ 3 H]PSB-603 (from 0.05 to 5 nM) with membrane aliquots for 2 hr at 25 °C. Non-specific binding was assessed by three concentrations of the radioligand (0.05 nM, 1 nM and 5 nM) and analyzed by linear regression. Incubation was terminated by filtration through GF/C filters using a Brandel-harvester (Brandel Harvester 24w, Gaithersburg, MD, USA). Filters were washed three times using ice-cold wash buffer and collected in tubes. 3.5 mL Emulsifier-Safe scintillation fluid (Perkin Elmer, Groningen, the Netherlands) was Vlachodimou  added and the filter-bound radioactivity was determined in a Tri-Carb 2900TR liquid scintillation analyzer (PerkinElmer).

Association experiments.:
Were performed by incubation of [ 3 H] PSB-603 (1.5 nM) with membrane aliquots at 25 °C. The amount of receptor-bound radioligand was determined after filtration at different time intervals for a total incubation time of 45 min and samples were obtained as described under "Displacement experiments".

Dissociation experiments.:
Were carried out after a 45 min pre-incubation of 1.5 nM [ 3 H]PSB-603 and membrane aliquots. Subsequently, dissociation of the radioligand at different time points up to 150 min was initiated by addition of 5 μL ZM241385 (assay concentration 10 μM). Dissociation experiments were performed at 25 °C. The amount of receptor-bound radioligand was determined after filtration and samples were obtained as described under "Displacement experiments".

Competition association experiments.:
Were performed at 25 °C and in a total volume of 100 μL, by incubation of 1.5 nM [ 3 H]PSB-603 and a competing ligand diluted in assay buffer to reach IC 50 concentration. Addition of CHO-spap-hA 2B AR membrane aliquots initiated the association. The amount of receptor-bound radioligand was determined at different time points up to 3 hr. After 3 hr the incubation was terminated and samples were obtained as described under "Displacement experiments".

Washout experiments.:
Were performed at 25 °C and in a total volume of 400 μL. CHO-spap-hA 2B AR membranes were incubated for 2 hr with the unlabeled compounds (at a final concentration of 10 × IC 50 ) while shaking at 1,000 RPM in an Eppendorf thermomixer comfort (Eppendorf AG, Hamburg, Germany). Subsequently, the samples were centrifuged at 13200 rpm (16 100g) at 4 °C for 5 min and the supernatant containing unbound ligand was removed. Pellets were resuspended in 1 mL of assay buffer, and samples were incubated for 10 min at 25 °C in the thermomixer. After four centrifugation and washing cycles in total, supernatant was discarded and the membranes were resuspended in a total volume of 400 μL containing 1.5 nM [ 3 H]PSB-603. After 2 hr at 25 °C incubations were terminated by rapid filtration through GF/C filters using a Brandel harvester and the samples were obtained as described under "Saturation binding experiments".

Detection principle.:
Label-free assays were performed using the xCELLigence real-time cell analyser (RTCA) system [27,28], as described previously [29]. In short, cells attached to the gold-coated electrodes embedded on the bottom of E-plates are generating electrical impedance which is monitored by the RTCA system. Variations in cell number, adhesion, viability and morphology result in impedance changes (Z) which are constantly recorded at 10 kHz. Z is displayed in the unitless parameter called Cell Index (CI) [28,30], which is defined as (Z i -Z 0 ) Ω /15 Ω. Z i is the impedance at a given time point and Z 0 represents the baseline impedance in the absence of cells, which is measured prior to the start of the experiment.  50 values were obtained by non-linear regression curve fitting to a sigmoidal concentration-response curve using the "log(inhibitor) vs. response" GraphPad analysis equation. pK i values were converted from pIC 50 values and the saturation K D using the Cheng-Prusoff equation [31]: The k off value was obtained using a one-phase exponential decay analysis of data resulting from a radioligand dissociation assay. The value of k on was determined using the equation: k on = (k obs -k off )/ radioligand in which k obs was determined using a one phase association analysis of data from a radioligand association assay. The association and dissociation rate constants were used to calculate the kinetic K D value using: K D = k off / k on .
Association and dissociation rate constants for unlabelled A 2B AR inhibitors were determined by nonlinear regression analysis of competition association data as described by Motulsky and Mahan [32]. The data were fitted into the GraphPad "kinetics of competitive binding" analysis, where k 1 and k 2 are the k on (M −1 min −1 ) and k off (min −1 ) of [ 3 H]PSB-603 obtained from radioligand association and dissociation assays, respectively, L is the radioligand concentration (nM), I is the concentration of unlabeled competitor (nM), X is the time (min) and Y is the specific binding of the radioligand (DPM). Fixing these parameters resulted in the calculation of the following parameters: k 3 , which is the k on value (M −1 min −1 ) of the unlabeled ligand; k 4 , which is the k off value (min −1 ) of the unlabeled ligand and B max , that equals the total binding (DPM). All competition association data were globally fitted. The residence time (RT) was calculated using RT = 1 / k off [33].

Chemistry
2-(4-(2,6-Dioxo-1,3-dipropyl-2,3,6,7-tetrahydro-1H-purin-8-yl)phenoxy)acetic acid (xanthine carboxylic congener, XCC, 1) was synthesized as reported [24]. Its amide derivatives 3-7, 9, 10, 20-25 were prepared by reaction with the desired amine in the presence of EDAC and DMAP as carboxyl group activating agents (Scheme 1). XCC was also used as the starting reagent for the synthesis of isothiocyanate-containing 29, aimed to bind to the A 2B AR covalently (Scheme 2).  Table 1). When evaluated in a homologous displacement assay, unlabeled PSB-603 showed similar affinity, yielding a pK i value of 8.90 (Fig. 1D, Table 1). Subsequently, [ 3 H]PSB-603 was evaluated in kinetic binding assays in order to determine its kinetic binding parameters k on and k off (Fig. 1B To obtain kinetic binding parameters for unlabeled A 2B AR antagonists, a radioligand competition association assay was developed. The specific binding of [ 3 H]PSB-603 was measured in the absence and presence of unlabeled PSB-603 over a time course of 45 min (Fig. 1C) and k on , k off and kinetic K D values of unlabeled PSB-603 were calculated to be 0.109 nM −1 min −1 , 0.084 min −1 and 0.77 nM, respectively ( Table 1). As the values of the competition association assay were in excellent agreement with the ones from the association and dissociation assay (Table 1), the first was deemed validated for determining an unlabeled ligand's binding kinetics.

Validation of [ 3 H]PSB
In order to increase the throughput of the assay, a single concentration of PSB-603 (1.0-fold its IC 50 ) was tested. Association and dissociation rate constants were found to be similar to the aforementioned ones, i.e. 0.111 ± 0.014 nM −1 min −1 and 0.086 ± 0.007 min −1 for k on and k off respectively (data not shown). Consequently, all other compounds were tested only at one concentration equal to 1.0-fold their IC 50 determined from displacement experiments.

Determination of equilibrium binding affinity (K i values) of A 2B AR
antagonists-Once the necessary assays were developed and validated, various xanthinebased A 2B AR antagonists were examined. The affinities of all compounds were evaluated in an equilibrium radioligand displacement study using [ 3 H]PSB-603 as the radiolabeled competitor. All compounds fully displaced the radioligand from the hA 2B receptor in a concentration-dependent manner. The data were fitted in a one-phase competition model showing mono-phasic displacement. A wide spread of affinities was noticed, ranging from 61.4 μM for compound 4 to 1.78 nM for compound 13; all affinities are listed in Tables 2-4.

Evaluation of ligand binding recovery with a washout assay-To
validate the results of the competition association assay and distinguish between ligands with distinct kinetic binding parameters, a [ 3 H]PSB-603 washout assay was developed (Fig.  2, Table 5). Compounds 18 and 17 were selected as they presented a short (8 min) and long (87 min) RT compound, respectively, while they have similar structures and affinities. Compound 29, designed as a putative covalently binding antagonist was also tested in this assay.
Both the washed and the unwashed conditions were assessed. For the washed condition, the unlabeled compounds were incubated with the target for 2 hr, followed by four wash and centrifugation cycles. Subsequently, [ 3 H]PSB-603 was co-incubated which led to competition of the radioligand with the unlabeled ligand still bound after the washing procedure. For the unwashed condition no washing was performed before the determination of radioligand displacement. Based on the experimental set-up, the long RT compound and the covalent ligand would be predicted to remain bound to A 2B ARs, as they would not be easily removed during the washing steps, and thus to result in lower radioligand binding.
In the unwashed condition, all A 2B ARs were almost fully occupied by each of the compounds as there was little specific binding of [ 3 H]PSB-603 observed ( Fig. 2; control/ unwashed). When the washed and unwashed conditions for both compounds 17 and 18 ( Fig.  2; washed and unwashed) were compared, a significant increase in radioligand binding after washing was monitored, indicating that they had (partially) dissociated from the target and been washed away. This was hardly the case for compound 29. The short RT compound 18 did not show any significant difference in specific [ 3 H]PSB-603 binding compared to control (TB). Apparently, 18 was almost completely (85%) removed during the washing procedure. On the other hand, the long RT compound 17 was washed away for 56%, indicating that 44% of A 2B ARs were still occupied by this ligand after the applied washing cycles, showing a significant (P < 0.0001, Fig. 2, washed) decrease of [ 3 H]PSB-603 specific binding compared to control (TB). This was even more apparent with compound 29, with only 18% of material being washed off the receptor.

Structure-Affinity Relationships (SAR) and Structure-Kinetic
Relationships (SKR)-We started with the study of the prototypic A 2B AR antagonist MRS1754 (13). In the displacement assay an affinity of 1.78 nM was determined. The kinetic characterization of 13 resulted in a determined RT of 69 min, which made us increase the duration of the competition association assay from 45 min to 3 h for all compounds in order to allow for the longer RT compounds to reach equilibrium.
We initiated the investigation on the xanthine scaffold with compound 1 [24]. Its affinity was found to be higher than 100 nM, the limit set as threshold for the kinetic studies. By substitution of the acid moiety for an acetamide (2) the affinity increased approx. 7-fold. Therefore, this acetamide was incorporated in all other compounds synthesized and tested. For compound 2, it was not possible to determine its kinetic characteristics, most probably because they were outside the detection range of our method.
Further functionalization of the acetamide to incorporate a cyclopropyl group (3) decreased the affinity, an effect observed for every non-aromatic ring tested (3, 4, 5, 7). When pyrazole was incorporated (6) the affinity increased compared to compound 5, but it remained in the micromolar range. To the contrary, introduction of pyridine (9) and pyrazine (10) decreased affinity even further when compared to the cyclohexyl substitution (7). Only incorporation of a phenyl ring (8) resulted in a low nanomolar affinity (1.93 nM) and a RT of 46 min. In addition to the cyclic substituents, a linear one (11) was incorporated, leading to a decrease in affinity compared to compound 2. However, the affinity did not exceed the 100 nM threshold.
Taking these results into consideration we continued with para substitution of the phenyl ring and determined the influence of those substituents on affinity and kinetic binding parameters. When we substituted compound 8 with a p-methyl group (12) a slight decrease in affinity and a 10 min increase in RT were observed. Introduction of a pcyano group (13) increased RT further, while the association rate constant increased about 4 times, yielding a compound with (sub)nanomolar affinity. The introduction of other electron withdrawing groups at the p-position (14, 15, 16, 17, 18) also yielded high affinity values for the receptor. Introduction of a nitro (14) or methyl ketone (16) substituent resulted in a moderate RT of 58 and 54 min, respectively, while the k on value was largely varying, with 14 presenting a slow association to the receptor. The trifluoromethyl substituent (15) resulted in a high affinity for the receptor, while its kinetic characteristics could not be monitored due to the detection range of the assay. Introduction of a carboxylic acid (17) was responsible for the longest RT measured in this study, while a methylcarboxamide (18) resulted in a similar affinity but significantly shorter RT.
Subsequently, the introduction of a spacer between the acetamide and the phenyl ring was investigated (Table 4). By introducing a carbon linker (19) to compound 8, the affinity dropped to a value in the micromolar range. A similar trend was observed for compounds 12 and 20.  (26), the affinity slightly dropped compared to 24 and 25, which after kinetic analysis appeared due to a decrease in association rate constant. The benzyl substitution of the amido group (27) resulted in an increase in affinity compared to the unsubstituted 19, indicating that this benzyl moiety is well accommodated in the binding pocket of A 2B AR. However, when compared to 26, 27 showed a lower affinity, suggesting that 27 did not optimally fit. Finally, 29, bearing a reactive warhead, did not allow the determination of true equilibrium affinity values; its apparent affinity under the conditions tested was approx. 9 nM.

Correlation plots-
To obtain a better comparison of kinetic and affinity parameters and understand their relationship, correlation plots for all compounds with measurable rate constants were constructed (Fig. 3). The affinity obtained from traditional radioligand displacement assay (pK i ) and the kinetic affinity (pK D ) derived from the radioligand competition association assay were found to be significantly and strongly correlated (r = 0.95, P < 0.0001; Fig. 3A), validating the use of the competition association assay. When the association rate constants (log k on ) of all kinetically characterized compounds were plotted against the kinetic affinity (Fig. 3B), a low, non-significant correlation was observed (r = 0.36, P = 0.202). However, the kinetic affinity was found to be significantly correlated with the dissociation rate constants (pk off ) (r = 0.76, P = 0.0015) (Fig. 3C). 3.2.7. Functional characterization of compounds 17 and 18-Next to binding parameters we studied compounds 17 and 18 in a functional set-up, in order to investigate the link between binding kinetics and a possibly prolonged functional effect. For this purpose a label-free assay was developed as described in Materials and Methods and below.
Initial experiments with NECA (5 ′ -N-ethylcarboxamidoadenosine), a non-selective AR agonist, were performed on control CHO-spap and CHO-spap-hA 2B ARs cells (Fig. 4). No response was found on "empty" CHO-spap cells upon treatment with NECA, whereas a concentration-dependent response was measured on CHO-spap-hA 2B ARs cells, yielding a pEC 50 value of 8.95 (Table 5). In order to validate that the response measured on CHO-spap-hA 2B ARs was only A 2B AR mediated, we pre-incubated cells with selective antagonists for each AR subtype prior to NECA stimulation. Only PSB-603, the A 2B AR-selective antagonist, inhibited the NECA response, confirming the A 2B AR specific hypothesis (Fig.  5A-D). As a result, further experiments for the study of A 2B AR were performed on CHOspap-hA 2B AR cells.
After this assay development, chemically similar compounds 17 and 18 but with an 11-fold difference in RT, were selected for further experiments. First, their inhibitory potency in the presence of an EC 80 concentration of NECA was determined resulting in pIC 50 values of 7.12 ± 0.13 and 6.44 ± 0.21, respectively ( Fig. 6A-D, Table 5). These potencies were found to be approximately 1.5 log unit lower than their affinities determined in the radioligand binding studies (8.64 ± 0.05 and 7.70 ± 0.06 for 17 and 18, respectively; Table 3), in line with the presence of a high agonist concentration (i.e. EC 80 ) in this assay set-up.
Subsequently, a washout assay was performed and the cell response to the compounds was monitored and evaluated (Fig. 7). In short, cells were pre-treated with a concentration of 30 × IC 50 of the compounds, and 4 hr later cells were washed, and fresh"serum-free medium" was added (Fig. 7A). For the evaluation of the unwashed condition, the medium was not refreshed but was pipetted up and down in order to mimic the possible mechanical stress induced to the washed cells. Cells were then stimulated with an EC 80 concentration of NECA, enabling us to monitor the response exerted by only those receptors that were not bound to compounds 17 (Fig. 7B) and 18 (Fig. 7C). Based on the experimental set-up, it was hypothesized that the short RT compound was removed more readily during washing, resulting in an increased number of receptors available for NECA to bind and cause a cellular response.
Cells pre-treated with long RT compound 17 showed no significant increase (p > 0.05) in NECA signaling after washing (9.8% and 12% for unwashed and washed cells, respectively) (Fig. 7D, Table 5). On the contrary, the increase in NECA signaling between unwashed and washed cells was significantly higher (P < 0.001) with the short RT compound 18 (19% and 68% for unwashed and washed cells, respectively) (Fig. 7D, Table 5), verifying our hypothesis. This assay simulates a non-equilibrium condition, not unlike human physiology.

Binding affinity of selected compounds at other adenosine receptors
-Selected xanthine derivatives were compared in binding assays [24] at four adenosine receptors (human and rat) as shown in Table 6. The A 2B receptor selectivity was generally low, especially at the rat A 2B receptor. However, several derivatives, e.g., 20 and 22, displayed mixed higher affinity at the human A 2B and A 2B receptors compared to the other subtypes.

Discussion
Intrigued by earlier studies [34] we combined previously synthesized xanthines with a series of newly synthesized derivatives. In this way we obtained a total of 28 final compounds that were subsequently tested in a variety of assays. The aim was to analyze both their structureaffinity and structure-kinetics relationships, and to test the translational properties of two selected derivatives in a number of label-free assays. We had performed similar studies on other GPCRs before, and learned that the sole determination of (equilibrium) affinity values provides an incomplete profile of the pharmacological characteristics of receptor ligands. Adding kinetic information informed us better, and made clear that target binding kinetics can dictate the duration of action and the sustainability of a pharmacological effect, thus having an impact on the ligand's pharmacokinetic and pharmacodynamic behavior in vivo [35].
We first tested the behavior of [ 3 H]PSB-603, the radioligand used in the present study, and learned that the K D value we derived was comparable to the value previously reported by Bormann et al. [36] A similar K D value was derived from the kinetic association and dissociation experiments, providing a reliable framework for our further experiments. We then used the radioligand to determine the affinity of the compounds for the hA 2B AR. Some of the previously synthesized xanthines had been tested on a number of other human and rat ARs (Table 6), showing a limited selectivity at best. It is important to stress that selectivity was not pursued in the current study.
We then determined the target binding kinetics of the full series of xanthine derivatives following an established protocol by Motulsky and Mahan [32]. This so-called competition association assay has been used to investigate the binding kinetics of ligands for various targets, such as GPCRs [37], kinases [38], transporters [39,40] and other proteins [41]. As the values from the competition association assay were in excellent agreement with the ones from the association and dissociation assay (Table 1), the first was deemed validated for determining an unlabeled ligand's binding kinetics. Evaluation and incorporation of target binding kinetics has been found to be a crucial parameter in drug optimization. The association rate constant is crucial for high target occupancy, due to the resulting rebinding effect [42], as well as for drug selectivity over different targets and ultimately for increased drug safety [43]. Last but not least, a fast association is crucial for an immediate drug response in case of an acute pathological event [44,45]. As far as the dissociation rate constant is concerned, a slow rate, hence a long RT is required for a longer and/or a more durable, sustained effect [46]. If RT exceeds the pharmacokinetic half-life, the drug could maintain its effect even past plasma clearance, resulting in potential advantages like a decreased frequency of drug dosing and a reduction in off-target toxic effects [43,47]. The compounds showed largely varying equilibrium affinity values as well as kinetic rate constants and corresponding residence times (RTs). Among the compounds was the prototypic A 2B AR antagonist MRS1754 (13). In the displacement assay an affinity of 1.78 nM was determined for 13, which was in excellent agreement with data reported by Ji et. al. [48]. The kinetic characterization of 13, yielding an RT of 69 min, made us increase the duration of the competition association assay from an initial 45 min to 3 h in order to allow for the longer RT compounds to reach equilibrium. As a result the 3 hr incubation assay was used for the determination of kinetic binding parameters for all qualifying A 2B AR antagonists. Compound 3 had the shortest RT (0.9 min), while compound 17 was endowed with the longest (87 min). Overall, there was a significant correlation between affinity and dissociation rate constants of all compounds for which we were able to obtain the kinetic parameters (Fig. 3C).
Due to its high chemical similarity we chose compound 18 (RT = 8.0 min) as a benchmark comparator for compound 17. We performed radioligand washout experiments with both compounds and compared their behavior with that of a putatively covalent antagonist (29). It appeared that 18 was rapidly dissociating from the receptor under washing conditions, while 17 was much more resistant, although less so than 29. Currently, A 2B AR antagonists are in clinical stages of testing for their immuno-oncological behavior (vide supra and ref [14]).
The tumor micro-environment where the compounds supposedly act is under the influence of high adenosine levels. Hence antagonists with longer RT, counterbalancing this adenosine "pressure", may be very useful.
Next to binding parameters we studied compounds 17 and 18 in a functional set-up, in order to investigate the link between binding kinetics and a possibly prolonged functional effect. For this purpose a so-called label-free assay was developed with xCELLigence technology, measuring changes in cellular impedance that are expressed as a unitless parameter named Cell Index [28,30]. As mentioned before we found a relatively high potency [49,50] for the reference, non-selective, AR agonist NECA, although it complies with other data in literature [51][52][53][54]. Hence, we made sure that the effects seen were entirely due to interactions with the A 2B AR. The same assay was used to study wash-out of the two compounds, simulating a non-equilibrium condition not unlike human physiology. As a result, findings from this assay, with 17 largely maintaining its effect, constitute a possible translational step towards in vivo experiments [27,55].
In conclusion we reported the synthesis and pharmacological evaluation of a series of xanthine-based analogues designed as hA 2B AR antagonists. A radioligand competition association assay was developed to evaluate kinetic binding parameters next to affinity. Structure-affinity and structure-kinetic relationships (SAR and SKR) were examined and a great spread in target residence time (RT) was observed, from 0.9 min (3) to 87 min (17). Based on correlation plots, the dissociation rate constant appeared the driving force for affinity unlike the association rate constants. Subsequently two compounds (17 and 18) with long and short RT, respectively, were selected and tested in a label-free impedancebased assay. These experiments confirmed the link between long RT and an extended pharmacological effect under non-equilibrium conditions. To our knowledge, this is the first SKR study performed on hA 2B AR antagonists, which could pave the way to the development of clinically meaningful antagonists, e.g., in immune-oncology, with a high affinity and long residence time at the A 2B AR.   Correlation plots between affinity and kinetic parameters. Kinetic affinity (pK D ) is plotted against (A) affinity determined from typical displacement assays (pK i ); (B) association rate constant (logk on ); (C) dissociation rate constant (pk off ).      Table 1 Comparison of the affinity, dissociation constants and kinetic parameters of PSB-603, obtained with different radioligand binding assays performed on CHO-spap-hA 2B AR membranes.  Table 5 Potency values (pEC 50 and pIC 50 ) of compounds in a label-free assay, and NECA signaling therein before and after washing, all on CHO-spap-hA 2B AR cells.