Screening of a natural product library Set IV for modulators of human P2X7

We acquired a natural product library (Set IV) containing 419 compounds from NCI DTP and tested these on P2X7 responses using a YO-PRO-1 dye uptake in HEK293 cells stably expressing human P2X7 (HEK-hP2X7). This screen used a two-injection procedure on a Flexstation 3 plate reader, whereby the compounds (at a fixed concentration of 10 µM) were injected first, and 60 s later, the agonist ATP was injected, similar to the method employed for screening crude spider venoms [27]. This allowed us to assess for any immediate effect of the injected compound, and secondly, it allowed for any potential negative modulators to be pre-incubated with P2X7. Our criteria for identifying potential positive modulators were compounds increasing the ATP response by > 25%, and such compounds are listed in Supplementary Table 1. Conversely, any potential negative modulators were defined as any compounds that reduced the YO-PRO-1 uptake by > 50% (Supplementary Table 2). We excluded compounds that had shown an immediate increase in YO-PRO-1 fluorescence upon injection as this indicated immediate cell toxicity.

We first investigated potential PAMs by re-testing several selected compounds in a co-injection experiment (Fig. 1A). The compound showing the largest potentiating effect was scopafungin, and we investigated whether it would cause a leftward shift in the ATP concentration–response curve. Figure 1B shows that scopafungin could increase the maximum response to ATP and increase the sensitivity to ATP; the EC50 to ATP (DMSO) was 131.3 µM (95% confidence interval 95.8 to 180 µM), whereas in the presence of scopafungin, this was reduced to 66.8 µM (95% confidence interval 2.9 to 151.6 µM). We considered that scopafungin could be a non-specific compound similar to solanine and saikosaponin A from our previous study [28]. Scopafungin did cause an increase in YO-PRO-1 uptake in the absence of ATP; however, the response induced by ATP plus scopafungin could be prevented by pre-treatment with the P2X7 selective antagonist, AZ10606120 (Fig. 1C), suggesting that some of its activity could be P2X7-dependent. This compound could be further studied as a potential new PAM at P2X7; however, its large molecular weight and complex structure make this an unlikely lead molecule for future development.

Fig. 1

Purified Natural Product Set IV chemicals with action at hP2X7. A HEK-hP2X7 cells were pre-treated with diversity set IV compounds or ginsenoside CK (10 µM) for 10 min. ATP-dependent YO-PRO-1 iodide dye uptake was measured over 300 s following administration of ATP (1 mM). B Concentration–response curve for ATP in the presence of ginsenoside CK (10 µM) or scopafungin (10 µM) using the YO-PRO-1 dye uptake assay. C Inhibitory effect of AZ10606120 (10 µM) on responses elicited by ATP in the presence of either CK or scopafungin. D HEK-hP2X7 cells were pre-treated with diversity set IV compounds or AZ10606120 (10 µM) for 10 min. ATP-dependent YO-PRO-1 iodide dye uptake was measured over 300 s following administration of ATP (1 mM). E Concentration–response curve for ATP in the presence of digallic acid, confertifolin, or acronine (10 µM) using the YO-PRO-1 dye uptake assay. F Dose inhibition curve for confertifolin and digallic acid at hP2X7. The symbol “*” represents statistical significance (P < 0.05) from one-way ANOVA with Sidak’s multiple comparison post hoc test

Investigations into potential negative allosteric modulators revealed 19 compounds (Supplementary Table 2) meeting the screening criteria. We investigated several of these compounds in further experiments using a longer pre-incubation time (10 min) in the YO-PRO-1 dye uptake assay. Acronine, confertifolin, digallic acid, and tetrahydroberberine (10 µM) were all able to reduce ATP-mediated YO-PRO-1 uptake (Fig. 1D). Other compounds with potential NAM activity (Supplementary Fig. 2) were not confirmed in this study. We also tested several compounds on the ATP concentration–response curve (Fig. 1E) revealing a potential non-competitive mechanism of action for digallic acid and confertifolin. Dose inhibition experiments calculated the IC50 values for confertifolin and digallic acid to be 3.86 µM and 4.05 µM respectively. These may be interesting lead molecules for future antagonist development.

Screening of a traditional Chinese medicine (TCM) plant extract library revealed both positive and negative modulators of P2X7

In the second part of this study, we screened organic and aqueous plant extracts that are commonly used in traditional Chinese medicine practices to identify novel compounds that could allosterically modulate hP2X7. We screened almost 400 extracts, the identity of which was unknown to us at the time of testing allowing for an unbiased approach to screening. Again, we used a two-injection procedure in a Flexstation 3 plate reader, whereby the crude extracts were injected first (to a final concentration of 30 µg/mL), and 60 s later, the agonist ATP was injected. Our criteria for identifying potential negative modulators were extracts reducing the YO-PRO-1 uptake by > 75% (Fig. 2A), and 71 extracts from 39 identified plant species meeting these criteria are listed in Supplementary Table 3. Conversely, positive modulators were identified as extracts increasing the ATP response to > 150% of control (Fig. 2B), and 29 extracts from 17 plant species meeting these criteria are detailed in Supplementary Table 4. We next confirmed the activity of these plant extracts by performing further blinded tests using alternative assays for hP2X7 activity, namely an IL-1β secretion assay on LPS-primed THP-1 monocytes and a cell viability assay on HEK-hP2X7 cells. We treated THP-1 monocytes with LPS (100 ng/mL) together with the crude plant extracts for 4 h, prior to stimulation of P2X7 responses with 3 mM ATP for the final 30 min. The data indicated that all but two of the extracts (Sparganium stoloniferum and Prunus mume) also had inhibitory effects on the secretion of IL-1β following ATP stimulation (Supplementary Fig. 1A). Another physiological function linked to the activation of P2X7 is the induction of cell death pathways [20]. P2X7 activation has been linked to cell death via numerous pathways including apoptosis, pyroptosis, and necrosis [29]. Using HEK-hP2X7 cells, we pre-treated the cells with 30 µg/mL of the inhibitory plant extracts (1 h) and added a lethal concentration of ATP (3 mM). Cells were assessed for viability using the AlamarBlue metabolic assay after 24 h (Supplementary Fig. 1B). Treatment with AZ10606120, a selective P2X7 antagonist, served as the control for this experiment. Of the 71 extracts tested, 51 were observed to rescue ATP-induced cell death by ≥ 50% (Supplementary Fig. 1) and 20 plant extracts exhibited inhibition comparable to AZ10606120. All results were then collated and submitted to the NCI DTP office to release the identities of the plant extracts we considered to be hits. These identified inhibitory hit extracts were further tested for specificity against hP2X7 by counter-screening against HEK-293 cells expressing human P2X4 using the YO-PRO-1 dye uptake assay. Only seven extracts from plants Polygonum multiflorum, Paeonia suffruticosa, Rosa rugosa, Rheum palmatum, Paeonia lactiflora, Paeonia veitchii, and Andrographis paniculata also showed an inhibitory effect against this receptor (Supplementary Fig. 2). This data could be useful for identifying compounds with inhibitory activity against hP2X4.

Fig. 2

Extracts exhibiting inhibition or potentiation of hP2X7. HEK-hP2X7 cells were pre-treated with 30 µg/mL of TCM extract for 60 s prior to the administration of 1 mM ATP. YO-PRO-1 uptake was measured for an additional 210 s, and the YO-PRO-1 uptake between 200 and 300 s was quantified as a percentage of the ATP response in the absence of extracts. The addition of 10 µM AZ10606120 was added as a control for inhibition, 10 µM CK was used as a control for potentiation, and DMSO was used as a vehicle control. A Extracts exhibiting more than 75% inhibition were considered inhibitory. B Extracts enhancing ATP responses to > 150% of the original ATP response were considered potentiators. Data are representative of two technical replicates but one independent experiment due to the availability of the extracts. Red line indicates the control response in each case

Compounds contained in the inhibitory plant extracts are predicted to bind to the negative allosteric site on P2X7

Plant extracts that showed inhibitory activity in all three assays (P2X7-dependent YO-PRO-1 uptake, IL-1β secretion, and cell death) (Supplementary Table 3) were taken to be true hits and were taken forward for further investigation. A basic (but not exhaustive) search of plant catalogues identified 157 distinct compounds that could be found within these inhibitory extracts. The SMILES strings for each of these compounds were obtained from PubChem and converted into a ligand library using CACTUS prior to computational docking. A single negative allosteric modulator site on P2X7 has been well characterised and is able to accommodate many diverse P2X7 antagonists [29]. Preliminary docking using AutoDock Vina software highlighted 59 compounds to be capable of interacting with this region, many of which were chemically distinct (Supplementary Table 5). These compounds included the known P2X7 inhibitors emodin (known to be found in P. multiflorum, R. palmatum, and A. tatarinowii) and berberine (known to be found in C. yanhusuo and C. chinensis), but also several other compounds such as flavonoids kaempferol, quercetin, genistein, epicatechin gallate (ECG), epigallocatechin gallate (EGCG), and bergapten, methyl syringate, scopoletin, and palmitic acid. We purchased these nine commercially available purified compounds to further investigate their ability to inhibit P2X7-dependent YO-PRO-1 uptake in HEK-hP2X7 cells. Berberine, quercetin, ECG, and EGCG showed inhibition of ATP-induced YO-PRO-1 uptake with IC50 values less than 10 µM (Table 1). Several of the purified compounds (scopoletin, palmitic acid, bergapten, and genistein) showed a reduction of YO-PRO-1 uptake in hP2X7 cells at high concentrations (> 30 µM) (Fig. 3D, E). Induced fit docking of these compounds was then performed on a human P2X7 homology model confirming potential interactions with the hP2X7 NAM site (Fig. 3).

Table 1 IC50 values for potential negative allosteric modulators at hP2X7 from YO-PRO-1 dye uptake assay; nc, not calculatedFig. 3

Inhibitory effect of purified chemicals on hP2X7 responses. HEK-hP2X7 cells were pre-treated with purified compounds for 10 min, and ATP-dependent YO-PRO-1 iodide dye uptake was measured over 300 s following administration of ATP (1 mM). Concentration–response curve for A berberine and emodin; B quercetin and kaempferol; C ECG and EGCG; D bergapten and scopoletin; E palmitic acid, methyl syringate, and genistein; and F AZ11645373. Induced fit docking for G berberine (green), H quercetin (cyan), and I EGCG (yellow) at hP2X7 NAM site using a homology model built on zfP2X4. AZ10606120 is shown in red as a comparison

We were concerned that the identified flavonoid compounds could be interfering with the YO-PRO-1 fluorescence assay [30, 31], and therefore, we designed a cell-free interference assay to determine if this was the case. The entry of the YO-PRO-1 dye into the cytoplasmic space of cells enables binding to nucleic acids such as RNA and DNA which causes a measurable increase in dye fluorescence. Therefore, any interference of the ability of YO-PRO-1 dye to interact with cellular DNA/RNA would reduce the measured fluorescence. In the cellular assay, the compounds are pre-treated with cells in a buffer containing YO-PRO-1; therefore, compounds could interact with YO-PRO-1 dye directly or could enter cells and interact with cellular nucleic acids to act as interfering agents. The cell-free assay was designed to inject a known concentration of DNA into the YO-PRO-1 buffer and measure the resulting fluorescence (Fig. 4A). AZ11645373 has no interfering effect on YO-PRO-1 fluorescence (Fig. 4B); however, some of the flavonoids can be seen to inhibit YO-PRO-1 fluorescence, suggesting that these could be interacting with DNA or quenching dye fluorescence (Fig. 4). Quercetin, kaempferol, ECG, EGCG, emodin, and genistein all displayed interference with the assay (Fig. 4). Bergapten, berberine, scopoletin, and methyl syringate demonstrated no interference. Quercetin was tested on ATP-induced calcium responses using fura-2-loaded HEK-hP2X7 cells, and this data further supports a lack of real P2X7 antagonist activity (Supplementary Fig. 3), whereas EGCG and berberine both demonstrated P2X7 antagonist activity in this assay.

Fig. 4

Cell-free YO-PRO-1 assay to measure compounds with interference. Compounds were diluted in a low divalent cation buffer containing 2 µM YO-PRO-1 iodide and 90 µL plated per well. Calf thymus DNA (1 µg/mL) was injected at 30 s using a Flexstation 3 and fluorescence measured over a further 60 s. A A typical response to DNA under control conditions. B Lack of interference by AZ11645373. Dose–response curves show the effects of C bergapten, berberine, and emodin; D genistein, palmitic acid, scopoletin, and methyl syringate; E quercetin and kaempferol; and F ECG and EGCG over the concentration range 100 nM–100 µM

Plant extracts identified with potentiator activity can enhance P2X7-dependent IL-1β release and enhance P2X7-mediated cell death

We next focused our investigation on the crude TCM plant extracts that could potentiate P2X7-mediated responses. Those extracts that stimulated YO-PRO-1 responses themselves, or gave variable effects, were removed from further experiments. The extracts were ranked from greatest potentiating effects to the least potentiating effects (Fig. 2B). To date, the most effective PAM of P2X7 we have identified is the ginsenoside compound K (CK) [12, 18, 28], which we used here as a positive control. From the screen, we identified crude extracts from four other plants that could increase the hP2X7 response more than ginsenoside CK (Fig. 2B), from plants Kochia scoparia, Dioscorea nipponica, Platycodon grandiflorum, and Polygala tenuifolia. Some of the identified plant extracts that potentiated P2X7 responses were from Panax notoginseng and Panax ginseng, which contain compounds we already know act as PAMs on P2X7 [12], and this was an important unbiased internal control. We first confirmed the effects of the PAMs on the potentiation of P2X7-dependent IL-1β secretion. THP-1 cells were stimulated for 4 h with LPS, prior to stimulation with 500 µM ATP and the respective plant extract for the final 30 min of the incubation. This resulted in some variability, and unexpectedly, the greatest potentiators identified in the YO-PRO-1 assay did not stimulate the greatest IL-1β release. Instead, Isatis indigotica elicited the greatest IL-1β response, followed by P. tenuifolia, P. notoginseng, and S. marianum (Fig. 5A). Extracts of Kochia scoparia and Dioscorea nipponica were capable of potentiating ATP-induced IL-1β release (Fig. 5A). We then tested whether the PAMs could potentiate the cell death of HEK-hP2X7 cells exposed to a non-lethal concentration of ATP (500 µM) after a 24-h incubation [28]. Under these conditions, extracts from K. scoparia, D. nipponica, P. tenuifolia, T. farfara, P. aviculare, and L. heterophyllus enhanced the killing of HEK-hP2X7 cells to a level comparable to the positive control, ginsenoside CK (Fig. 5B). Extracts from C. morifolium and P. communis were the only hit extracts to not enhance cell death in these experiments (Fig. 5B). However, it must be kept in mind that there was high variability therefore meaningful conclusions due to the data being from one experiment due to a limited sample material (Fig. 5B). Using our counter-screen against the related receptor hP2X4, none of the extracts that had a positive modulatory effect at P2X7 had a positive modulator effect at P2X4 (Supplementary Fig. 2).

Fig. 5

Plant extracts that potentiate hP2X7 dye uptake responses can enhance the secretion of IL-1β and cell death responses. A THP-1 cells were treated with LPS (100 ng/mL) for 4 h followed by the addition of TCM plant extract (30 µg/mL) and ATP (500 µM) for the final 30 min of incubation. IL-1β release into supernatants was quantified and compared to the secretion of IL-1β in the absence of plant extract. Ginsenoside CK (10 µM) was used as the positive control and 3 mM ATP was used as a control to demonstrate P2X7-induced IL-1β secretion. B HEK-hP2X7 cells were treated with ATP (500 µM) in the presence or absence of selected TCM plant extracts (30 µg/mL) for 24 h prior to quantification of viable cells by using AlamarBlue assay. Ginsenoside CK (10 µM) was used as a positive control for the enhancement of cell death. Data are representative of two technical replicates from one independent experiment due to the limited availability of the TCM plant extracts

Dioscin, a compound found in D. nipponica is a potential PAM for P2X7

From the screening results, we researched the crude plant extracts to identify compounds as active PAMs at P2X7. We chose to investigate the D. nipponica (organic) extracts further since this plant contained several glycosides with structural similarity to the ginsenosides (dioscin, protodioscin, and gracillin), and we tested whether a fixed concentration of extract (10 µg/mL) could potentiate ATP responses in a P2X7-dependent manner. D. nipponica extract could potentiate all concentrations of ATP tested, and this response could be inhibited by the selective P2X7 antagonist AZ10606120 (Fig. 6A). We previously identified that protopanaxadiol glycosides from P. ginseng are PAMs for P2X7 [12, 28], and for this reason, glycosides dioscin, protodioscin, and gracillin and the diosgenin aglycone (control) were docked against P2X7. Dioscin was predicted to bind to a site previously described to be involved in positive allosteric modulation (Supplementary Fig. 4). Diosgenin and protodioscin were not predicted to interact with this site at all, and gracillin was not predicted to make polar interactions at this site. We tested the ability of these purified glycosides to potentiate P2X7-dependent dye uptake responses in HEK-hP2X7 cells. In the presence of 200 µM ATP (approx. EC50 value), dioscin was able to enhance YO-PRO-1 dye uptake and this could be reduced with AZ10606120, indicating the involvement of P2X7 in the response (Fig. 6B). Whole-cell patch-clamp electrophysiology confirmed that dioscin could potentiate hP2X7 ion channel opening and that dioscin alone was not inducing ion channel opening itself during the short 5-s application (Fig. 6C). The aglycone diosgenin was not capable of potentiating P2X7 channel opening (Fig. 6C, D). Exploring the type of PAM activity at hP2X7, we performed a dose response to ATP in the absence/presence of 10 µM dioscin using the YO-PRO-1 assay and observed an increase in the maximum response with no leftward shift and no major effect on the EC50 value for ATP (110 µM vs 160 µM in the presence of dioscin) (Fig. 6E). However, dioscin could also increase YO-PRO-1 dye uptake in non-transfected HEK-293 cells in the presence of ATP (Fig. 6F) suggesting a non-specific effect. We investigated whether this was due to cytotoxicity. Dioscin could induce cell death in non-transfected HEK-293 cells over 24 h (Fig. 6G), and dioscin was able to induce propidium iodide staining of HEK293 cells within 10 min of addition (Fig. 6H) suggesting a non-specific membrane permeabilisation and cytotoxic action.

Fig. 6

Dioscin enhances P2X7-dependent pore formation and potentiates channel opening. A Dioscorea nipponica (organic) plant extract (30 µg/mL) increased ATP-dependent YO-PRO-1 iodide uptake. AZ10606120 (10 µM, grey bars) was added to block hP2X7. B YO-PRO-1 uptake following a co-stimulation of HEK-hP2X7 cells with 200 µM ATP ± the named compounds in the presence or absence of AZ10606120. YO-PRO-1 uptake was quantified between 50 and 300 s (n = 5). C Whole-cell patch-clamp recordings were performed at room temperature. Cells were voltage-clamped at − 60 mV, and responses to a 5-s stimulation of ATP (100 µM) followed by a 5-s stimulation of ATP (100 µM) + dioscin/diosgenin (10 µM) were measured from HEK-hP2X7 cells. Representative traces are shown. D Summary of normalised current amplitudes represented as a fold change compared to ATP responses in the absence of dioscin/diosgenin (n = 10–18 cells). E Concentration response to ATP in the presence of dioscin (10 µM, green) or vehicle (DMSO, black) using YO-PRO-1 uptake assay in HEK-hP2X7 and F HEK-293 cells. G Cell viability assay using AlamarBlue shows dioscin (10 µM) is toxic over 24 h in HEK-293 cells. 1 mM ATP does not induce cell death. H HEK-293 cells were treated with dioscin or CK (10 µM) for 10 min and stained with propidium iodide to reveal permeabilised cells. Images taken with EVOS microscope