Evaluation of chemical constituents of Rooibos (Aspalathus linearis) and Honeybush (Cyclopia intermedia) as adenosine A1/A2A receptor ligands

Rooibos (Aspalathus linearis) and Honeybush (Cyclopia intermedia) are popular tisanes in South Africa and are of growing interest due to the wide variety of flavonoids and other phytochemicals they contain. Despite their history as herbal teas and traditional medicines, the chemical constituents of these tisanes have yet to be studied for their effects on adenosine receptors. A series of 30 commercially available chemical constituents of Rooibos and Honeybush were investigated via radioligand binding studies to determine their adenosine A1 and A2A receptor affinity at both rat and human subtypes in order to establish structure-activity relationships and identify novel adenosine receptor ligands. In addition, in silico evaluations of the 30 test compounds were also performed to predict their physiochemical and pharmacokinetic properties. The most promising chemical constituent was kaempferol (28) which showed sub-micromolar affinity towards the rat A1 subtype (rA1Ki = 0.7287 μM; hA1Ki = 9.88 µM) and acted as an antagonist toward adenosine rA1 receptors. Additionally, quercetin (2), chrysoeriol (8), luteolin (9), eriodyctiol (12), and naringenin (27) also showed adenosine A1 and/or A2A receptor affinity. It was observed that a flavonol scaffold is preferred to flavone and flavanone scaffolds, and within the flavonols, C4’-OH substitution on ring B is preferred to C3’,4’-diOH substitution. These phytochemicals, specifically kaempferol (28), may be considered lead-like and valuable in designing novel ligands, based on in vitro and in silico evaluation.


Introduction
Adenosine is involved in cellular activities, and thus, engaged in both physiologic and pathophysiologic processes.Four adenosine receptor (AR) subtypes, A 1 , A 2A , A 2B and A 3 [1,2] are members of the G protein-coupled receptor (GPCR) superfamily, which is also a therapeutic target for several pathologies [3].A 1 and A 3 ARs mainly couple to inhibitory G-proteins, while A 2A and A 2B ARs couple to stimulatory G-proteins [4].A 1 ARs are mostly found in the central nervous system (cerebral cortex, cerebellum, hippocampus, thalamus, brainstem and spinal cord) as well as some peripheral organs.A 2A ARs are concentrated in the striatum, nucleus accumbens, and olfactory tubercle [5][6][7].The abundant yet distinct distribution of the A 1 and A 2A ARs make them suitable drug targets for neurological conditions, including Parkinson's and Alzheimer's disease [8][9][10][11], and depression [12].
A number of herbs have traditionally been used as herbal infusions or "teas" in South Africa.Two Cape fynbos plants, Aspalathus linearis (Brum.f)Dahlg.(family Fabaceae; tribe Crotalarieae), better known as Rooibos, and various species of Cyclopia Vent.(family Fabaceae; tribe Podalyrieae), commonly known as Honeybush, have enjoyed commercial success as herbal teas [17].These teas are popular tisanes enjoyed for their taste and aroma and medicinal purposes [17,18].
Aspalathus linearis, known as Rooibos, is a leguminous bush indigenous to the mountain areas of the Western Cape Province of South Africa.It is gaining popularity as an alternative beverage worldwide, mainly due to the absence of caffeine and its low tannin content [19].Rooibos tea is widely known as a natural cognitive enhancer as it can maintain alertness and concentration in the short-term and may also be beneficial for long-term cognitive health [20,21].Traditional medicinal uses of Rooibos in South Africa include alleviating infantile colic, allergies, asthma, and dermatological problems [22].It is asserted that Rooibos has a low toxicity profile with several health benefits [19,23] which include antioxidant [24], antiaging [25], anticancer [26], anti-inflammatory [27], antimutagenic [28] and antidiabetic [29] properties.
The short, woody shrub Honeybush (Cyclopia intermedia) is grown in the mountain slopes of the Langkloof district between the Eastern and Western Cape regions of South Africa.Honeybush tea is usually prepared by infusion of fermented leaves, stems, and flowers of the plant.During the fermentation process, the plant material changes colour from green to dark brown as the phenolic compounds are oxidized [18].
In this pilot study, the chemical constituents of Rooibos and Honeybush (structures found in Appendix A, Table A1) were investigated for their potential A 1 and A 2A AR affinity and the possibility of identifying novel AR ligands.The results would give insight into structure-activity relationships toward gaining, or even losing, AR affinity, and previously unknown AR drug leads and scaffolds may be identified.

In silico prediction of physiochemical and pharmacokinetic properties
The physiochemical and pharmacokinetic properties of the test compounds 1-30 were computed via SwissADME, and the results are summarised in Appendix A, Tables A2-4 and Figs.A1-2.
For the purpose of this discussion, focus was placed on compounds 2, 8, 9, 12, 27 and 28 which showed promising in vitro rA 1 and/or rA 2A AR affinity.Table 1 summarises the results of the in silico predicted physiochemical properties, whereas Table 2 gives the drug-likeness and medicinal chemistry friendliness of selected chemical constituents found in Rooibos and/or Honeybush plants.
The drug-likeness properties of the compounds were assessed by rule-based filters from Lipinski [39], Ghose [40], Veber [41], Egan [42], and Muegge [43].The violated rules can be observed in Appendix A, Table A4.None of the test compounds that show adenosine A 1 and/or A 2A receptor affinity (2, 8, 9, 12, 27 & 28) violated any of Lipinski's five rules which outline the physiochemical properties that support the likelihood of absorption upon oral administration [39].
The WLOGP-versus-TPSA referential, also known as the BOILED-Egg model (Appendix A, Fig. A2), can be used to predict the passive human gastrointestinal absorption (HIA) and blood-brain-barrier (BBB) permeation [44].Since A 1 and A 2A ARs are greatly expressed in the central nervous system, a drug that targets these receptors should be able to cross the BBB.Unfortunately, none of the test compounds that show promising affinity are predicted to cross the BBB but are expected to have high GI absorption.Yet, compounds 15, 16, 17, 24 and 26 are expected to permeate the BBB.According to Daina & Zoete (2016) it is predicted that moderately polar (TPSA < 79 Å 2 ) and relatively lipophilic (MLogP 0.4-6.0)molecules have a high probability of entering the CNS, and thus, crossing the BBB.
It was also predicted that compounds 2, 8, 9, 12, 27 and 28 are soluble to moderately soluble in water.A soluble molecule makes drug development easier [45], and also influences GI absorption [46].
Two methods were employed to identify potentially problematic fragments in the present series.The first method identifies pan assay interference compounds (PAINS) and compounds having substructures that are likely to hinder biological assays and exhibit target promiscuity are identified [47].The Brenk method identifies fragments that may be potentially toxic, chemically reactive, metabolically unstable, or have poor pharmacokinetics [48].Only three out of the six test compounds with affinity toward rA 1 and/or A 2A subtypes produced alerts; quercetin (2), luteolin (9) and eriodyctiol (12), which all contain a catechol.It is seen as problematic since a catechol is rapidly oxidized to reactive quinones, which can damage biological macromolecules and interfere with membrane function.They can also produce toxic reactive oxygen species [49].

Competitive binding assays: degree of binding affinity
The degree of binding affinity that the test compounds 1-30 showed towards rat A 1 and A 2A ARs was determined via radioligand binding studies in either duplicate (specific binding (%)) or triplicate (inhibition constant (K i , μM)) and expressed as mean ± standard error of the mean (SEM).Test compounds were initially assessed at a high maximum concentration of 100 µM using rat whole brain membranes (rA 1 ) or rat striatal membranes (rA 2A ).Only compounds 2, 8, 9, 12, 27 and 28, which displayed specific binding values of ≤ 20% at a maximum tested concentration of 100 μM, underwent full biological assays for determination of inhibition constant values (K i , μM).These compounds were further investigated at human (h) A 1 and/or A 2A receptor subtypes.All other test compounds displayed specific binding values > 20%; thus, no K i values were determined.The A 1 and A 2A AR radioligand binding assays were validated with N 6 -cyclopentyladenosine (CPA, A 1 agonist), 8cyclopentyl-1,3-dipropylxanthine (DPCPX, A 1 antagonist) and istradefylline (IST, A 2A antagonist) as reference compounds.Results are summarized in Table 3.

Structure-activity relationships
Structure-activity relationships (SAR) based on K i values (μM) of test compounds 2, 8, 9, 12, 27 and 28 showed a coherent relationship between the constituents and rA 1 and rA 2A AR affinity (Fig. 9).The influence on AR affinity of polar (-OH) and/or non-polar (-OCH 3 ) groups, the number of these groups and their position on rings A, B and C of the flavones, flavonols and/or flavanones was explored.For the flavones, a comparison of chrysoeriol (8: C3'-OCH 3 ) to luteolin (9: C3'-OH) showed that C3'-OH substitution decreased rA 1 affinity.
Although both substitution patterns led to dual AR affinity, it may be said that a C3'-OCH 3 -C4'-OH substitution pattern ( 8) is preferred for rA 1 affinity and a catechol group (9) for rA 2A affinity.For the flavonols, a mono-hydroxy substituted compound, such as kaempferol (28), exhibited higher affinity than one which was 3',4'-dihydroxy substituted (for example quercetin ( 2)).Indicating that a catechol group (and potentially increased polarity), led to decreased rA 1 affinity.However, when the flavanones were compared, naringenin (27: C3'-OH) showed lower activity than its di-hydroxy substituted counterpart eriodyctiol (12).From the limited data at hand, it seems that not only the number and position of the -OH or -OCH 3 group(s) influence affinity, but also the type of flavonoid.It was observed that the flavonols have the best affinity for the A 1 /A 2A ARs, followed by the flavones and flavanones in decreasing order of affinity (flavonol>flavone>flavanones).

Functional binding assays: type of binding activity
A guanosine triphosphate (GTP) shift assay determined the type of binding affinity a test compound exhibits at the rat A 1 AR (agonist, antagonist, or partial agonist).The six compounds that showed promising rA 1 affinity (2, 8, 9, 12, 27 and 28) were selected to determine their functionality at the A 1 AR.These results are also summarized in Table 3. GTP works by uncoupling the A 1 AR from its G-protein leading to a shift from a high to a low affinity state for agonists [57].It is therefore possible to determine if a compound will act as an agonist, a partial agonist or an antagonist by comparing a test compound′s binding curves (or K i value) in the presence and absence of GTP [58].With an antagonist, the binding curve is unaffected in the presence of GTP (as is the K i value), and the antagonist test compound will have a GTP shift value of approximately 1, while an agonist test compound's binding curve will shift to the right in the presence of GTP (and the K i value will increase) and have a GTP shift value greater than 1 [59,60].The results obtained suggested that compounds 2, 8, 9, 12, 27 and 28 act as A 1 AR antagonists since there is no significant rightward shift of the binding curve (or change in the K i values) in the presence of GTP and compounds have calculated GTP shifts of ≤ 1 (Fig. 10).Interestingly, GTP and endogenous adenosine binding interactions are controlled similarly by human and rat proteins in A 1 and A 2A AR binding studies [57].As reference standards, CPA (A 1 agonist) and DPCPX (A 1 antagonist) were used to validate the GTP shift assays.

Correlation between species
The pK i values of selected reference standards and test compounds (2, 8, 9, 12, 27 and 28) which showed affinity toward rat and human A 1 and/or A 2A receptor subtypes were compared, and thus, correlation coefficients were calculated.The R-squared value of the linear regression analysis is equal to the correlation coefficient (Appendix A, Fig. A3, 4).
Site-directed mutagenesis studies on several members of the G-protein coupled receptor superfamily exposed that a single amino acid can modulate ligand affinity, and thus, explain species differences [61,62].Predictably, the rat and human A 1 receptor subtype (R squared 0.88) showed better correlation than the A 2A AR subtype (R squared 0.81); since a 95% sequence homology at the amino-acid level between rA 1 and hA 1 receptor subtypes exists, compared to only 82% between rA 2A and hA 2A [63].Prototypical xanthine derivatives DPCPX (selective A 1 receptor antagonist) and IST (selective A 2A receptor antagonist) used as reference standards were markedly less potent at the human than rat receptors (Table 3).DPCPX and IST were, respectively, eight and fourteen times more selective toward the rat than human A 1 and A 2A receptor subtype; indeed, the affinity of xanthine-based antagonists is species-dependent.Therefore, it was suggested that different structural determinants may be responsible for binding of xanthine derivatives (such as DPCPX and IST) compared to other scaffolds.

Conclusions
Despite their history as herbal teas and traditional medicines, the commercially available chemical constituents of Rooibos and Honeybush plants were not investigated for their effects on A 1 and/or A 2A ARs prior to this pilot study which established structure-activity relationships and identified test compounds with lead-like properties.Overall, the chemical constituents of Rooibos and Honeybush plants did not show mentionable affinity toward either A 1 or A 2A receptor subtypes, except for compounds 2, 8, 9, 12, 27 and 28.Of these compounds, the flavonol kaempferol (28) showed sub-micromolar AR affinity, making it the most promising test compound of the present evaluated natural products derived from either Rooibos or Honeybush.Additionally, it has drug-and lead-like properties and can be used as starting point for future studies.It was observed that a flavonol scaffold is preferred to flavone and flavanone scaffolds, and within the flavonol scaffold C4'-OH substitution on ring B is preferred to C3',4'-OH substitution.The observed structure-activity relationships may be used as reference to design a novel series of para-benzyl substituted flavonols which are structurally related to the parent compound kaempferol (28) in hopes of improving in vitro affinity and in silico physiochemical and pharmacokinetic properties (such as BBB permeation).
In vitro evaluation of adenosine A1 and A2A receptor affinity and activity

Membrane preparation
The North-West University Animal Care, Health and Safety Research Ethics Committee (NWU-AnimCare) approved the collection of tissue samples from adult male Sprague-Dawley rats for the radioligand binding studies (application number NWU-00035-10-A5).Rat brain membranes were prepared (whole brain and striatal membranes were pooled separately) and stored as described in literature [58].Chinese hamster ovary (CHO) cells expressing the human A 1 or A 2A AR were kindly donated by Prof. K.N.Klotz from the Fig. 9 Summary of structureactivity relationships of chemical constituents of Rooibos and Honeybush plants University of Würzburg, Germany.CHO cells were grown adherently and maintained in Dulbecco's Modified Eagles Medium with nutrient mixture F12 (DMEM/F12) without nucleosides, containing 10% foetal bovine serum, penicillin (100 U/ml), streptomycin (100 µg/ml), L-glutamine (2 mM) and Geneticin (G418, 0.2 mg/ml) at 37 °C in 5% CO 2 /95% air.Cells were split 2 or 3 times weekly at a ratio between 1:5 and 1:20, respectively.For radioligand binding studies, the culture medium was removed, cells were washed with phosphate-buffered saline (PBS) and frozen until preparation of membranes, upon which frozen cells in ice-cold hypotonic buffer (5 mM Tris/HCl, 2 mM EDTA, pH 7.4) was thawed again.The cell suspension was then homogenized on ice (Ultra-Turrax, 2 X× 15 s at full speed) and the homogenate was spun for 10 min (4 °C) at 1000 g.The supernatant was then centrifuged for 30 min at 100,000 g.The membrane pellet was resuspended in 50 mM Tris/HCl buffer pH 7.4, frozen in liquid nitrogen and stored at −80 °C.Protein content for rat and CHO cell membranes were determined using Bradford reagent and bovine serum albumin as reference standard [65].
Human (h): To determine hA 1 AR affinity, CHO cell membranes expressing A 1 ARs and [ 3 H]DPCPX was used, and for hA 2A affinity, CHO cell membranes expressing A 2A ARs and [ 3 H]NECA [68].Each incubation of the hA 1 or hA 2A assay consisted of: (i) test compound (2 µL), (ii) 0.1 nM [ 3 H]DPCPX or 30 nM [ 3 H]NECA (radioligand solution, 20 µL) and (iii) 20 µg CHO cell membranes expressing hA 1 AR or 65 µg CHO cell membranes expressing hA 2A AR and 0.2 units/mL adenosine deaminase (membrane suspension, 178 µL).Samples were incubated for 1 h at 25 °C, filtered and washed three times with 4 mL ice-cold binding and filtered individually as described for conventional radioligand binding studies.Non-specific binding of [ 3 H]DPCPX and [ 3 H]NECA was defined as binding in the presence of 100 µM CPA [58,66,67].Specific binding was defined as the total binding minus the non-specific binding [58].
GTP shift assay: The guanosine triphosphate (GTP) shift assay resembles the [ 3 H]DPCPX radioligand binding assay; however, it requires the addition of 100 µM GTP.Nonspecific binding was defined as binding in the presence of 10 µM DPCPX [69,70].

Data analysis
The competitive binding studies' data analysis was done using Microsoft Excel and GraphPad Prism Software.Sigmoidal dose-response curves, from which half maximal inhibitory concentration (IC 50 ) values were calculated, were obtained by plotting the specific binding against the logarithm of the test or reference compounds′ concentrations.Subsequently, the IC 50 values were used to calculate the inhibition constant (K i ) values for the competitive inhibition of [ 3 H]DPCPX (rK d = 0.36 nM [66]; hK d = 3.86 nM [68]) against A 1 ARs and [ 3 H]NECA (rK d = 15.3 nM [67]; hK d = 20.1 nM [68]) against A 2A ARs by the test compounds using the Cheng-Prusoff equation [71].Descriptive statistics were used to represent K i values (µM) as the mean ± standard error of the mean (SEM) of experiments performed in triplicate.Specific binding (%) values of the radioligand at a maximum tested concentration of 100 µM were represented as the mean of experiments performed in duplicate.Where applicable, selectivity index (SI) values were calculated for the A 1 isoform as a ratio of A 2A K i and A 1 K i values of test compounds.GTP shifts were calculated by dividing the K i values of compounds in the presence of Fig. 10 The binding curve of kaempferol ( 28) is an example of an A 1 AR antagonist which is determined via a GTP shift assay (with 100 μM GTP) in rat whole brain membranes expressing rA 1 ARs with [ 3 H]DPCPX as radioligand.28 has a GTP shift of 1

Table 2
In silico predicted drug-likeness and medicinal chemistry friendliness properties of selected chemical constituents found in Rooibos and/or Honeybush plants

Table 3
Competition of test compounds for [ 3

Table 3 (
i , µM) is presented as the mean ± standard error of the mean (SEM) of experiments performed in triplicate b Specific binding (%) at maximum tested concentration of 100 µM is presented as the mean of experiments performed in duplicate c Dissociation constant (K d ): 0.36 nM d K d : 3.86 nM e K d : 15.3 nM f K d : 20.1 nM; Values without SEM are taken from the literature; r: rat; h: human;