LPS2336, a New TREK-1 Channel Activator Identified by High Throughput Screening

Abstract

TWIK-related K+ (TREK-1) channels are involved in pain perception and their pharmacological activation has potential for pain relief. The development of new pharmacological tools to study these channels and enrich our knowledge of structure-activity relationships is therefore important. We optimized a high throughput screening method based on thallium flux monitoring for the detection of TREK-1 activators in chemical libraries. We screened 1040 compounds from the French National Essential Chemical Library and identified LPS2336 as a potent TREK-1 activator with an EC₅₀ of 11.76 µM. Thirty-three LPS2336 analogs were subsequently tested but none of them retained activity on TREK-1. In vivo, LPS2336 produces antinociceptive activity when administered systemically and, to a lesser extent, intracerebroventricularly, but not intrathecally, showing that targeting peripheral TREK-1 channels may be important to produce pain relief, with the interest of reducing potential central adverse effects. LPS2336 was shown to produce sedation and hypothermia with a narrow therapeutic window. As these adverse effects are also observed in TREK-1 knock-out mice, they are likely mediated by off-targets. Our work provides key optimization steps for thallium-based assays and a new pharmacological tool for the study of TREK-1 channels. It also raises the importance of investigating adverse effects in vivo at early stages of drug discovery.

Keywords: TREK-1; analgesia; high throughput screening; pain; structure–activity relationship.

Figure 1

Optimization of the thallium assay. (A–D) Optimization of thallium concentration in naive and HEK-hTREK-1 cells treated with vehicle or 10 µM BL-1249. n = 42 wells over 3 plates. (A) Average normalized fluorescence traces of BL-1249-treated HEK-hTREK-1 cells. Dots represent the time-point used for subsequent ΔFmax/F0 extraction. (B) Thallium dose–effect curves constructed with ΔFmax/F0 from naive and HEK-hTREK-1 cells treated with BL-1249 or vehicle. Optimal thallium concentration, determined in D, is shown in green. (C) Normalization from 0 to 100% of the HEK-hTREK-1 curves from B. (D) In HEK-hTREK-1 cells, ΔFmax/F0 of BL-1249-treated samples was divided by the average ΔFmax/F0 of the vehicle group to evaluate the influence of thallium concentration on the separation between treated and non-treated cells. The purple zone defines the range of thallium concentrations recommended by the manufacturer. In red is the optimal thallium concentration used for subsequent experiments and high throughput screening. (E,F) Optimization of DMSO concentration in naive and HEK-hTREK-1 cells treated with vehicle or 10 µM BL-1249. n = 48 wells over 3 plates. Normalized DMSO dose–effect curves are shown before (E) and after (F) normalization from 0 to 100%. The final DMSO concentration used in high throughput screening is indicated in green. (G,H) Optimization of the time-point used for Fmax extraction in HEK-hTREK-1 cells treated with 10 µM BL-1249 or vehicle. n = 180 wells over 4 plates. (G) ΔFmax/F0 of BL-1249-treated samples were divided by the average ΔFmax/F0 of the vehicle group to determine which maximum time-point allows the best separation between treated and control wells. The arrow indicates the time-point retained for subsequent analyses and high throughput screening as the best compromise between separation and variability. (H) Z scores calculated in BL-1249 treated wells. In all panels, means ± standard deviations are shown.

Figure 2

Dose–response curves of different known TREK-1 modulators. Dose–response curves were generated for 9 compounds in HEK-hTREK-1 cells using the previously optimized thallium assay conditions (0.5 mM thallium, 0.5% DMSO, Fmax measured up to 13 s after thallium injection). Some compounds were salified (ML335.Na2, ML402.K, GI-503159.HCl) to improve solubility, and the results obtained with salts are compared with those obtained with native molecules. For Bl-1249, the 100 µM datapoints (shown in gray) were excluded from the logistic regression (least square fit, 4 parameters) as visible toxicity occurred. Results are normalized to vehicle-treated wells. n numbers (wells) are indicated on each panel. Means ± standard deviations are shown. **** p < 0.0001 vs. salified compound, 2way ANOVA followed by Šídák’s multiple comparisons test.

Figure 3

Screening for TREK-1 activators in a collection of 1040 compounds. Compounds from the Essential Chemical Library (ChemBioFrance) were screened at 25 µM on HEK-hTREK-1 cells, on four 384-well plates replicated 3 times. Previously determined optimal conditions were used (0.5 mM thallium, 0.5% DMSO, Fmax measured up to 13 s after thallium injection). (A) Z-score of 180 vehicle-treated wells (HEK-hTREK-1 cells) spread over the 12 screening plates. A positivity threshold of 3 standard deviations was determined for high throughput screening to minimize false positive detection. (B) High throughput screening results expressed as Z scores. 1040 triplicates are shown. One compound, LPS2336, was over the positivity threshold 3/3 times (triplicates in red). One compound, RN-1-025, was over the positivity threshold 1/3 time (triplicates in purple). (C) Structures of LPS2336 and RN-1-025. (D) In silico analysis of LPS2336. (Left) Oral bioavailability radar of piperidine LPS2336 (shown as red line). The pink area depicts a suitable physicochemical space for bioavailable drugs. INSATU: insaturation; POLAR: polarity; INSOLU: insolubility; LIPO: lipophilicity; FLEX: flexibility; SIZE: molecular weight. (Right) Summary table of LPS2336’s predicted physicochemical properties, pharmacokinetic profile, drug-likeness and toxicity by molinspiration, SwissADME and preADMET.

Figure 4

TREK-1 activation by LPS2336 and RN-1-025. (A,B) Current densities recorded in HEK-hTREK-1 cells treated with 50 µM riluzole (n = 18 cells), GI-530159 (n = 9), ML335 (n = 13), LPS2336 (n = 17), RN-1-025 (n = 17) and vehicle (0.01% DMSO, n = 41) (A), and LPS2336 (n = 15), salified LPS2336.HCl (n = 13) and vehicle (0.01% DMSO, n = 14) (B). Left panels show current-voltage relationships from −100 to +30 mV. Currents are normalized by cell capacitances. Right panels show average current densities ± SEM recorded at 0 mV. In (A) are also shown currents recordings made in untransfected HEK293 cells treated with LPS2336 or RN-1-025 (n = 3 cells). (C) Dose–response curves of LPS2336.HCl and ML335 generated by thallium assay on HEK-hTREK-1 and untransfected HEK293 cells (0.5 mM thallium, 0.5% DMSO, Fmax measured up to 13 s after thallium injection). n = 5 wells. Mean ΔFmax/F0 ± SEM are shown. (D) Average current densities ± SEM recorded at 0 mV in HEK-hTREK2 cells. ** p < 0.01, **** p < 0.0001 vs. vehicle, Kruskal–Wallis test followed by Dunn’s post hoc test (A), one way ANOVA followed by Dunnett’s post hoc test (B) and Mann–Whitney test (D).

Figure 5

Evaluation of LPS2336 analogs. (A) 33 analogs of LPS2336 were tested on HEK-hTREK-1 cells using the thallium assay (2 mM thallium, 0.5% DMSO, Fmax measured up to 13 s after thallium injection). All compounds are used at 100 µM final. ML335 (n = 15), ML67-33 (n = 3) and LPS2336.HCl (n = 12 wells) were used as positive controls. n = 3 wells for all LPS2336 analogs but 2c (n = 6). Results are shown normalized to vehicle-treated wells (n = 240). Gray and green dashed lines indicate the mean of vehicle and LPS2336.HCl groups, respectively. Means ± standard deviations are shown. Structures of the 12 analogs from Contreras et al., (2001) [44] and of the 21 analogs synthetized in the frame of this study are shown in (B) and (C), respectively.

Figure 6

Effect of systemic and central treatment with LPS2336 on pain thresholds in a mouse model of inflammatory pain. Subacute paw inflammation was induced in WT CD1 mice by intraplantar injection of carrageenan (20 µL, 2%) 3.5 h before administration of different doses of LPS2336.HCl, morphine or vehicle (saline) given intraperitoneally (A), intrathecally (B) or intracerebroventricularly (C). Thermal pain thresholds were evaluated using the Hargreaves test before carrageenan injection (−210 min) and before and for 2 h after treatment (0–120 min). Doses given account for LPS2336 after removal of the HCl salt. Left panels show time course evaluation of pain thresholds and right panels show the area under curve measured from 0 to 120 min. Mean ± SEM are shown. n = 8 animals in all groups. * p < 0.05, ** p < 0.01 vs. vehicle, one way ANOVA followed by Dunnett’s post hoc test.

Figure 7

Adverse effects of systemic treatment with LPS2336. (A) Evaluation of motor coordination, balance and fatigue using the rotarod test in WT CD1 mice 15 min after intraperitoneal injection of LPS2336 at 0, 10, 20, 30, 40 and 50 mg/kg (n = 8 animals). (B) Evaluation of motor coordination and grip strength using the horizontal bar test in WT CD1 mice 15 min after intraperitoneal injection of LPS2336 at 0, 10, 20, 30, 40 and 50 mg/kg (n = 8). (C) The same horizontal bar test was performed in WT and TREK-1 KO littermates of C57Bl6/J background injected with 0 or 50 mg/kg LPS2336 (n = 10). (D) Evaluation of spontaneous locomotor activity of WT CD1 mice in an open field 15 min after intraperitoneal injection of LPS2336 at 0, 10, 20, 30, 40 and 50 mg/kg (n = 6–7). (E) The same open field test was performed in WT and TREK-1 KO littermates of C57Bl6/J background injected with 0 or 50 mg/kg LPS2336 (n = 10). (F) Rectal temperature measured in WT and TREK-1 KO littermates of C57Bl6/J background injected intraperitoneally with 0 or 40 mg/kg LPS2336, before and 15, 30 and 45 min after injection. (G) For each animal in (F), the body temperature before injection was subtracted from the body temperature 15 min after injection. * p < 0.05, ** p < 0.01, *** p < 0.001 vs. vehicle. Kruskall-Wallis followed by Dunn’s multiple comparisons test (A–E) and one way ANOVA followed by Dunett’s multiple comparisons test (G). Doses given account for LPS2336 after removal of the HCl salt.