Preclinical development of an EP2 antagonist for post-seizure cognitive deficits.

Ten cohorts of 15-20 C57Bl/6CR mice were batch-randomized to receive saline (n=54) or pilocarpine HCl (n=121), typically in a ~1:2 ratio to account for expected mortality in SE. One mouse did not enter SE after pilocarpine. Out of 120 pilocarpine-treated mice that experienced SE, a total of 41 mice died within 18 hr of SE onset. Of these most died in SE or very shortly thereafter; two vehicle-treated mice and two treated with up to two doses of TG11-77 died between 9 and 18 hr of SE onset. For the 79 survivors we monitored delayed mortality and body weight daily for 8 days. Like TG6-10-1, our first generation EP2 antagonist ( Jiang et al., 2013 ), TG11-77 (8.8 mg/kg ip administered 4, 8 and 19 hr after SE onset) reduced delayed mortality compared to vehicle-treated mice ( Fig 1A ; p = 0.0185 by Gehan-Breslow-Wilcoxon test). In contrast to TG6-10-1, however, TG11-77 did not accelerate weight regain or neurological recovery after SE ( Figs 1B , C ), or improve the ability to build a nest on day 4 ( Fig 1D ). The small temporary loss of weight caused by TG11-77 in saline-treated control mice ( Fig 1B ) was also observed during oral exposure to much higher daily doses (55-57 mg/kg/day) of TG11-77 ( Banik et al., 2021 ; Rawat et al., 2022). The cause of this transient weight loss is not known, but it does not appear to be adverse. Mice had fully regained motor activity 8-10 days after SE. The batch randomization was successful in that mice were approximately evenly distributed across the four treatment groups in each cohort. They were then subjected to a single 8-min spontaneous alternation trial in a 3-arm Y-maze 8-17 days after SE. Successive arm choices during free exploration of a 3-arm Y-maze is considered a measure of spatial working memory, because the innate curiosity of a mouse leads to exploration of the arm less recently visited ( Hughes, 2004 ). However, habituation to novelty during prolonged time in the maze, and associated changes in motivation, can reduce the frequency of spontaneous alternation towards the end of the trial. Indeed, we have observed that for many mice their strategy appears to change from exploration to escape as the trial progresses. For this reason, in this study mice were allowed to explore the Y-maze freely for 8 min, but we limit the alternation analysis to the first 17 arm entries. The experiment ( Fig 2A ) was analyzed by one-way ANOVA. For the experiment, F(3,47)=9.599 (P<.0001). Control mice that had received saline followed by vehicle alternated arm choices 75.3% of the time ( Fig 2A , blue symbols), whereas pilocarpine-treated mice that had been subsequently injected with vehicle performed at near-random level (median = 50.8%, green symbols) (P<.0001). However, mice with fleeting exposure to TG11-77 (8.8 mg/kg as free base) after SE performed similar to the control mice (74.1%, red symbols). Pilocarpine-treated mice tend to be hyperactive, and Fig 2B shows that the improved performance of TG11-77 treated mice in the Y-maze after SE was not due to any difference in motor ability. For the pilocarpine-treated mice, there was no correlation between distance traveled and Y-maze performance (R 2 = 0.09 and 0.10, not shown). SE causes profound neuroinflammation in rats and mice that is alleviated by post-SE treatment with two other EP2 antagonists, TG6-10-1 and TG8-260 ( Jiang et al., 2013 ; Jiang et al., 2019 ; Rojas et al., 2021a ; Rojas et al., 2015 ; Varvel et al., 2021 ). Moreover, TG6-10-1, which has brain/plasma ratio of 1.6-1.7, was neuroprotective in three SE models in mice and rats( Jiang et al., 2013 ; Jiang et al., 2019 ; Rojas et al., 2015 ; Varvel et al., 2021 ), whereas TG8-260, which is restricted to the periphery, provides minimal to no neuroprotection ( Rojas et al., 2021a ). Based on these results we decided to examine gliosis and neurodegeneration after pilocarpine-induced SE in mice treated with different doses of TG11-77 as the HCl salt, ip at 4, 8 and 19 hr after SE onset, and euthanized 4 days after SE. In the first experiment, 41 mice (n=18 saline-treated and n=23 pilocarpine-treated) were processed through immunohistochemistry for Iba1 (microglia and monocyte-derived macrophages) and GFAP (astroglia). Images were acquired from four brain regions in each mouse. Fig 3A shows representative Iba1 images from the amygdala, with all acquisition parameters held constant for all mice and brain regions. A profound microgliosis is apparent after SE, which is attenuated by post-SE administration of TG11-77 at 8.8 mg/kg as free base. Figure 3B quantifies the % area covered by microglia in all four brain regions in the four groups of mice, from which we conclude that TG11-77 quells microgliosis consistently. The data were analyzed by 1-way ANOVA; for the amygdala, one outlier in the saline-vehicle group was identified by the Grubbs test, and F(3,36)=17.72 with p<.0001. For each region two preselected individual comparisons were made with Sidak correction, as indicated in Fig 3B . A similar analysis was done for GFAP-stained astrocytes and interestingly, no effect of TG11-77 was discernable on astrogliosis ( Fig 3C ). We then repeated this experiment at higher (22 mg/kg) and lower (4.4 mg/kg) doses of TG11-77, as free base, focusing on the amygdala region, which had shown the most consistent effects of TG11-77 at 8.8 mg/kg. The results are shown in Table 1 , alongside pharmacokinetic parameters acquired from separate groups of mice. From these data we conclude that the lowest effective dose of TG11-77 that caused statistically significant suppression of microgliosis in this assay was 8.8 mg/kg, which is associated with a plasma exposure (AUC inf ) of 1046 hr•ng/ml. We next examined neurodegeneration in the hippocampus 4 days after SE by staining sections from each mouse with FluoroJade B. As shown in Fig 4 , although TG11-77 caused a small increase in neurodegeneration at 8.8 mg/kg, this was not replicated in nearby brain regions nor enhanced at the higher dose, indicating that TG11-77 exerted no consistent changes in neuronal injury in hippocampus. In saline-treated mice that did not undergo SE, TG11-77 itself (8.8 mg/kg) caused no neuron death in the three hippocampal regions examined (n=10 mice). Pilocarpine can elevate plasma IL-6 level ( Holtman et al., 2013 ) and can transiently activate circulating T-lymphocytes and monocytes ( Marchi et al., 2009 ). The molecular targets of pilocarpine, M1 and M3 muscarinic receptors, are expressed by monocytes ( Koarai et al., 2012 ), and we have provided evidence that infiltrating activated monocytes contribute strongly to the neuroinflammation seen after pilocarpine-induced SE ( Varvel et al., 2021 ; Varvel et al., 2016 ). With these considerations in mind, we used traditional complete blood cell (CBC) counts to determine whether pilocarpine-induced SE is associated with evidence of peripheral inflammation 4 days after SE. ANOVA was used to analyze the results with F(11,162)=18.53 (P<.0001), with selected comparisons indicated in Fig. 5 and Sidak’s adjustment to p-values. The ratios of the number of circulating lymphocytes, monocytes and neutrophils to the total number of white blood cells revealed a modest reduction in the percentage of lymphocytes and modest elevation in monocytes and neutrophils ( Fig 5A ), none of which reached significance. The platelet to lymphocyte ratio, which can be a sensitive indicator of peripheral inflammation ( Gasparyan et al., 2019 ), was elevated 4 days after SE ( Fig 5B ), and the monocyte to lymphocyte ratio was modestly increased. Taken together these data are consistent with mild peripheral inflammation 4 days after pilocarpine-induced SE. None of the cellular indices of peripheral inflammation that we measured was, however, influenced by prior treatment with TG11-77 ( Fig 5 ). Encouraged by our findings of anti-neuroinflammatory and cognitive preservation effects in the SE mouse model, and the dose-related anti-inflammatory effect, we advanced TG11-77 for IND-enabling preclinical tests as required by the FDA for initiating clinical trials with an EP2 antagonist. In vitro hepatic intrinsic clearance of TG11-77 was moderate to very high in liver microsomes and hepatocytes from all five species ( Table 2 ). The predicted in vivo hepatic clearance (Cl H ) is, however, equivalent to or lower than hepatic blood flow in all species. The first pass hepatic extraction calculated from these data (E H in Table 2 ) is a substantial 50-80%. We verified this in the rat in separate experiments by sampling blood from both portal vein and cardiac puncture after oral administration of TG11-77; the measured hepatic extraction was 48-83% of the absorbed dose. Assuming 100% absorption, these data predict moderate oral bioavailability ( Table 2 ), which was confirmed in the rat and dog by the absolute oral bioavailability data in the in vivo pharmacokinetic studies ( Supplemental Table 1 ). By allometric scaling of the measured in vivo total clearance in the rat (81 ml/min/kg; Supplementary Table 1 ) and in the dog (10 ml/min/kg), the predicted human in vivo clearance is 21 ml/min/kg based on the rat data and 7 ml/min/kg based on data from dogs, which was similar to that predicted by in vitro studies (10-14 ml/min/kg). TG11-77 produces one principal metabolite (hydroxylation on indole) across species (Rat, Dog, Human) after a four hour incubation with mammalian hepatocytes under incubation conditions that allowed both Phase 1 and Phase 2 metabolic reactions to occur ( Fig 6 ). Importantly there were no unique human metabolites detected, and no metabolite exceeded 5% of the parent compound, indicating that from a metabolic perspective rat and dog are appropriate for IND-enabling toxicology studies. To identify the CYP isoforms that metabolized TG11-77, a reaction phenotyping assay was performed, indicating that CYP3A4 accounted for 93% of TG11-77 metabolism in human hepatocytes ( Table 3 ). TG11-77 showed low to modest reversible CYP inhibition as judged by IC50, except for CYP2C8, which was inhibited by TG11-77 with IC50 of 4 μM ( Table 3 ). Likewise CYP induction was modest and inconsistent over the three hepatocyte donors. TG11-77 appeared to be a weak time-dependent inhibitor of CYP3A4, which may require further investigation before phase 2 clinical trials. TG11-77 exhibited >500-fold selectivity for EP2 compared with other Gαs-coupled prostanoid receptors ( Amaradhi et al., 2020 ). To further explore selectivity, we used two off-target panels of binding assays performed by Eurofins that, together, probed 75 targets. The IC50 values were >10 μM for all targets except those bolded in Supplementary Table 4 . Of these, the only target of potential concern would be if TG11-77 acted as an agonist at 5HT2B ( Cavero and Guillon, 2014 ), but we found in a follow-on cell-based functional study that TG11-77 acts as a weak antagonist of 5HT2B, thus has no liability for valvulopathy. Toxicologically relevant functional effects of TG11-77 on the other bolded targets in Supplemental Table 4 would, if present, be revealed in animal toxicity studies. Additional ADME properties are presented in Supplementary Table 5 . Directional transport of a molecule across Caco-2 cells is used to predict oral bioavailability of test compounds. TG11-77 was highly permeable across membranes of Caco-2 cells, suggestive of high absorption across intestinal epithelia. Similarly, MDCK cells that express the multi-drug resistance protein 1 (MDR1) or breast cancer resistance protein (BCRP) transporters were used to evaluate whether compounds are substrates or inhibitors of MDR1 or BCRP, two major P-glycoprotein efflux transporters at the blood brain barrier. TG11-77 proved to be an inhibitor and a substrate for both transporters ( Supplemental Table 5 ), perhaps accounting for its moderate brain-to-plasma ratio of 0.4 to >1 ( Amaradhi et al., 2020 ; Banik et al., 2021 ). TG11-77 was also a weak inhibitor of the liver and kidney uptake transporters OATP1B1 and MATE1 ( Supplementary Table 5 ), the effects of which, if any, can be monitored in clinical trials. Finally, TG11-77 exhibited high plasma protein binding in all five species examined, and was evenly distributed between red blood cells and plasma, suggesting that the compound was not sequestered within erythrocytes. Plasma pharmacokinetics were compared in mouse, rat and dog after intravenous administration of TG11-77, and in rat and dog after oral administration. The purpose of these studies was to assist in the interpretation of in vivo efficacy data in the mouse as described above, and to prepare for dose range-finding toxicology studies in rat and dog. At low doses, plasma exposure in the dog was considerably greater than in the rat ( Figs 7A , B ). Plasma exposure after oral administration in 0.5% methylcellulose (4000 CP) suspension was linear with dose in the rat up to 1000 mg/kg but unexpectedly plateaued in the dog between 60 and 200 mg/kg ( Fig 7C ). Exposure in the dog at 60 mg/kg was not improved by two organic formulations, one of which (vehicle F in Supplementary Table 1 ) protected the compound in micelles. TG11-77 had a high volume of distribution in rat, mouse and dog, indicative of extensive distribution of the compound from plasma into tissues. Clearance was higher in the rat and mouse than in the dog ( Table 2 ), which is reflected in much larger plasma AUC values in the dog at low doses ( Fig 7C ). TG11-77 was orally bioavailable in all three species tested ( Fig 7 , Supplemental Table 1 ). TG11-77 was brain permeable with brain/plasma ratio from 0.4 to >1 in rats and mice ( Amaradhi et al., 2020 ; Banik et al., 2021 ). The compound was stable in both simulated gastric fluid pH 1.2 (t 1/2 >400 min) and simulated intestinal fluid pH 6.8 (t 1/2 = 197 min). Likewise, it was stable in blood (t 1/2 >400 min) and plasma (t1/2 >1000 min) in all five species tested (mouse, rat, dog, cynomolgus monkey and human). TG11-77 was tested in vitro in a mini-Ames (non-GLP) bacterial mutagenicity assay and as an inhibitor of the hERG cardiac potassium channel. TG11-77 was negative for cytotoxicity and mutagenicity in the mini-Ames test (5-100 μM tested in 4 strains ± S9; Eurofins) and had a 71X margin for hERG in a patch clamp assay based on the free plasma Cmax of a minimally effective dose in a mouse SE model. Moreover, TG11-77 inhibited [3H]-dofetilide binding to hERG protein by less than 25% at 10 μM (Eurofins). These results are consistent with lack of mutagenic and hERG liability. To establish dose-proportional pharmacokinetics, TG11-77 was dosed orally to rats and dogs in a 0.5% methylcellulose suspension. Plasma exposure in the rat was dose-proportional over a wide range (15 to 1000 mg/kg) but not, as mentioned above, in the dog ( Fig 7C ). For this reason, we have so far only carried out a dose-range-finding toxicity study in the rat. SRI Biosciences (Menlo Park, CA) performed this study. After oral administration in a methylcellulose suspension, TG11-77 continued to show dose related Cmax and AUC last from 150 through 1000 mg/kg/day x 7 days, with similar plasma exposure values on days 1 and 7 ( Fig 8 ). We conclude that TG11-77 does not accumulate in the plasma compartment after daily doses up to 1000 mg/kg in rats. Female rats had higher plasma concentrations than males (~2-FOLD) when treated with 500 mg/kg (the only dose tested), suggesting a possible sex-related difference in clearance. No adverse clinical signs were apparent at any dose for male or female rats. There were no notable differences in hematology or clinical chemistry in animals treated with 1000 mg/kg/day TG11-77 for 7 days compared to vehicle control animals. Importantly, coagulation parameters (prothrombin time, activated partial thromboplastin time, fibrinogen) were not changed ( supplemental Table 6 ), and serum bilirubin was below detectable level, which suggests that the observed in vitro low potency inhibition of liver transporters MATE1 and OATP1B1 ( Supplemental Table 5 ) did not have clinical consequence in the rat. No drug-associated changes were evident in absolute body weights, body weight changes, absolute organ weights, organ to body weight ratios, organ-to-brain ratios, macroscopic and histopathologic findings of 11 tissues. Based on these data, the dose of TG11-77 eliciting no observed adverse effects (NOAEL) was the highest dose studied for each sex – 1000 mg/kg in males and 500 mg/kg in females. The Day 7 mean AUC last values at these doses were 32,300 hr•ng/ml in males and 36,900 in female rats ( Supplemental Table 1 ). AMRI (Curia) carried out a salt and polymorph screen on TG11-77 to identify a stable salt form of TG11-77 to carry forward into GLP toxicology studies and the clinic. They evaluated salts of 13 different acids and identified 8 acids that formed crystals with TG11-77. Evaluation of organic solvent content, solid phase composition, physical stability at ~90% relative humidity as assessed by X-ray Powder Diffraction, approximate aqueous solubility, and behavior upon thermal analysis led to the selection of the mesylate A form, an anhydrous/unsolvated monomesylate, as the recommended salt for further development. The X-ray diffraction pattern of individual crystals showed sharp peaks that could be indexed, confirming stoichiometry. The mesylate salt of TG11-77 was physically stable upon exposure to 90% relative humidity at room temperature for at least 7 days with no visible signs of deliquescence. Its aqueous solubility was visually estimated as ~7 mg/mL, and we confirmed this by HPLC. Note that TG11-77 has an ionizable nitrogen with pKa=5.8 so its solubility is expected to be pH dependent. We have verified this.

Experimental procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The mouse efficacy protocol (201700490) was reviewed and approved by the Institutional Animal Care and Use Committee of Emory University. The experiments reported here are in accordance with the ARRIVE guidelines. The ADME studies, and toxicology studies in rats and dogs were conducted consistent with pertinent FDA and International Conference on Harmonization guidance. For pharmacokinetics studies, male Sprague-Dawley rats (8-12 weeks old, 200-300 g) were sourced from Charles River or Global, India. Male beagle dogs were sourced from Marshall BioResources, New York (approximately 7 months old, 6.4-6.9 kg) or Isoquimen, Hyderabad India (above 1 year old, 8-11kg). For toxicokinetics and toxicology studies, male and female rats (9-10 weeks, 263-339g males and 204-275g females) were sourced from Charles River, and male beagle dogs (11 months old, 9.0-10.5 kg) were sourced from SRI Biosciences. We predetermined the number of mice per group for behavioral studies to be 12-14, and 8-12 mice per group for cellular studies, based on previous experience ( Jiang et al., 2020b ; Jiang et al., 2013 ; Rojas et al., 2016 ; Varvel et al., 2021 ). Groups of 15-20 C57Bl/6 male mice (8-12 weeks old, from Charles River) were numbered and batch-randomized to receive saline (10 ml/kg ip) or pilocarpine HCl (freshly prepared 270 mg/kg as free base, ip, Sigma P6503) in saline. All mice first received terbutaline (4 mg/kg ip, Sigma T2528) and methylscopolamine (4 mg/kg ip, Sigma S8502) to alleviate respiratory and cardiovascular effects of pilocarpine, then 30 min later received saline or pilocarpine. Pilocarpine-treated mice experienced SE for 1 hr, after which all mice (including saline-treated) were administered diazepam (10 mg/kg ip) to interrupt SE. During this period, mice were scored for seizure severity every 5 min ( Varvel et al., 2021 ). Following SE, mice were fed moistened rodent chow, monitored daily, and injected with 5% (w/v) dextrose in lactated Ringer’s solution (0.5 ml i.p.) when necessary. Immediately after receiving diazepam, mice were matched into two groups to approximately equalize the number of stage 4 seizures experienced during SE, then a coin flip determined whether a group would receive TG11-77•HCl ip at 4, 8 and 19 hr after SE onset, or its vehicle (10% DMSO, 50% PEG400, 40% H2O, by volume), dosing at 8.8 mg/kg and 8 ml/kg. The experimenter was blinded as to whether each group received drug or vehicle and was not unblinded until the data had been analyzed. Mice were group housed in a warm (28°C) and humid environment for 2-3 days, after which they were separated into individual cages. Weight and overall health as judged by a modified Irwin test ( Varvel et al., 2021 ) were monitored daily, and the ability to build a nest from a supplied nestlet was evaluated ( Deacon, 2006 ) on day 4 after SE. The starting dose of TG11-77 for the efficacy experiments, 8.8 mg/kg, was based on earlier preliminary experiments and our assessment from pharmacokinetics data that the brain concentration should be in a therapeutic range. The rate of spontaneous alternation in arm choice was tested in a single-trial 3-arm Y-maze, in which each mouse was allowed to explore freely for 8 min. Three symmetrical arms each 36 cm in length were used, and the ambient light was dimmed to approximately 20-25 lux to minimize anxiety. As shown by the composite heat map inset of Fig 2A (compiled from a preliminary experiment with 4 saline treated mice), once a mouse committed to an arm it tended to travel all the way to the end, so there was rarely any question whether a mouse had entered an arm. The number of choices resulting in an alternation (e.g., arm entry sequence of ACB but not ACA) was expressed as a fraction of the total number of choices. The number of arm entries in different mice ranged between about 20 to over 70 during the 8 min period. To mitigate this difference, and to restrict the assay to the period of time in which the mouse focused on exploring the maze rather than escaping, we analyzed the first 17 arm choices of each mouse. These experiments were identical in design to those above except for the following. Pilocarpine HCl was administered at 280 mg/kg ip, and terbutaline and methylscopolamine were administered as 2 mg/kg ip each, 20 min before mice received saline or pilocarpine. Immediately after receiving diazepam, mice were randomized into two groups to receive TG11-77 or its vehicle ip at 4, 8 and 19 hr after SE onset. Four days after SE, mice were anesthetized deeply with isoflurane, perfused through the heart with ice-cold PBS solution, and their brains rapidly removed from the cranium. The brain was immediately bisected through the midline, and the left hemisphere was fixed in 4% (w/v) paraformaldehyde for 24 hours at 4°C. After fixation, the left hemisphere was cryoprotected in 30% (w/v) sucrose in PBS solution. The brains were then frozen in 2-methylbutane chilled with dry-ice and sectioned coronally at 25 μm using a freezing/sliding microtome. Every 12 th section for a total of three sections from one hemisphere of each mouse was used for FJB staining to label degenerating neurons. Brain sections were mounted on Superfrost Plus Microscope Slides (Thermo Fisher Scientific) and allowed to air-dry overnight. Sections were immersed in 0.06% potassium permanganate for 15 minutes with gentle agitation, rinsed for 1 min in distilled water, and then transferred to the FJB staining solution (0.0001% w/v FJB in distilled water with 0.1% acetic acid, Millipore AG310) for 30min in the dark with gentle agitation. Sections were rinsed with three 1-minute changes of distilled water and air-dried. The slides were immersed in xylene and then coverslipped with D.P.X. mounting medium (Electron Microscopy Sciences 13510). Sections between bregma −1.3 and −2.3 were examined with a fluorescent microscope. Following FJB staining, images were obtained from three hippocampal areas (hilus, CA1, CA3) in each section. A researcher unaware of experimental conditions counted the number of FJB-positive neurons in the hippocampus. Only positive neurons with a near-complete cell body shape and size were tabulated. Cell counts were expressed as the total number of FJB-positive cells per section for each region, and were averaged across the three sections counted from each mouse. For immunofluorescence staining, 30 μm thick free-floating sections were blocked for 45 minutes in 5% goat serum (Sigma, G9023) and incubated approximately 16 hours at 4°C with rabbit polyclonal anti-Iba1 for microglia/macrophages (1:1000, Wako AB839504 ) and rat anti-GFAP for astroglia (1:1000, Invitrogen AB2532994). Primary antibody incubation was followed by extensive washing with PBS solution, incubation with secondary antibodies Alexa Fluor 488 goat anti-rabbit (1:500, Invitrogen, AB143165 ) and Alexa Fluor 594 goat anti-rat (1:500, Invitrogen, AB141374 ) for 1 h at room temperature. After washing with PBS solution, stained brain sections were mounted on glass slides and mounting medium with DAPI and antifade (Vector H-1200) was applied. Fluorescent images of the amygdala, dorsal hippocampus and neocortex overlying the hippocampus were acquired from Iba-1 and GFAP-stained brain sections with a Carl Zeiss Axio Observer A1 epifluorescence microscope equipped with an AxioCam Mrc5 camera. A threshold level for each glial marker was determined to detect the maximum difference between saline control and pilocarpine-treated SE mice subject to vehicle. The camera settings for Iba1-stained sections were −0.11 (brightness), 9.24 (contrast), and 750 ms (exposure time) with the “Moments” threshold setting used in NIH ImageJ Fiji software. The camera settings for GFAP-stained sections were −0.27 (brightness), 3.00 (contrast), and 629 ms (exposure time) with the “Li” threshold setting used in NIH ImageJ Fiji software. The camera settings and threshold levels were employed for all the images stained with a given marker. Percentage area occupied by each glial marker indicates the area above the threshold level. The percentage areas of three sections were averaged to provide a single value for each animal. The data were analyzed by ANOVA with post hoc Sidak test. Two predetermined comparisons were made: between saline-vehicle and pilocarpine-vehicle cohorts, and between pilocarpine-vehicle and pilocarpine-TG11-77 cohorts. Whole blood was analyzed automatically using the animal blood counter VetScan HMS (ABAXIS). Healthy male or female C57BL/6 mice (22-27g) or Sprague Dawley rats (260-300 g) were administered TG11-77•HCl in groups of three by intravenous (iv), intraperitoneal (ip) or oral (po) route in formulation and dosing volumes, and dose, shown in Supplementary Table 1 . Blood samples were collected from the retro-orbital plexus (mice) or from an indwelling venous catheter (rats) at various times from 15 min to 24 hr after dosing. Samples were collected into labeled microtubes containing 20% K 2 EDTA solution as an anticoagulant. Plasma was immediately harvested from the blood by centrifugation at 4000 rpm for 10 min at 4 ± 2 °C and stored below −70 °C until bioanalysis. To determine the brain/plasma ratio in mice, immediately after collection of cardiac blood mice were perfused with ice cold PBS and brain samples were then collected and homogenized in 2:1 (vol:wt) ice-cold phosphate buffer saline (pH 7.4); homogenates were stored below −70±10 °C until analysis. Concentrations of TG11-77 in plasma and brain samples were determined by fit-for-purpose LC-MS/MS methods (Sai-life, India and Southern Research Institute, Birmingham, AL). Phoenix WinNonlin ® (version 8.0) was used to analyze the pharmacokinetics parameters presented in Supplementary Table 1 . The data were plotted using Origin 9.8 software. Three male Beagle dogs (6-7 kg, fasted overnight (16-18 hr), or fed 2-4 hr prior to dose administration) were administered an iv dose of 3 mg/kg as free base, then after washout (≥ 7 days later) an oral dose of 15 mg/kg as free base. Other groups of beagles were administered a single po dose of TG11-77 at 20 mg/kg as free base. Following a ≥ 7-day washout, the same 3 dogs were administered 60 mg/kg of TG11-77, po, as free base of TG11-77. Following another ≥ 7-day washout, 200 mg/kg po as free base was administered. The vehicle for each oral dose was 0.5% methylcellulose in sterile water (w/v) and the dosing volume was consistent at 10 ml/kg. Clinical observations were recorded predose, immediately postdose, 2-4 hr postdose, and prior to the last blood collection. The 3 dogs in all three dose groups appeared normal throughout the study. Blood was collected via an indwelling jugular catheter and processed to plasma pre-dose and at various times up to 24 hr post-dose. An LC-MS/MS bioanalytical method was developed to analyze the dog plasma samples with a lower limit of quantitation (LLOQ) of 2.00 ng/ml. These studies were done at Southern Research (Birmingham, AL), SRI Biosciences (Menlo Park, CA), and SAI Ltd (Hyderabad, India). Subsequent studies were done with different formulations as described in Supplementary Table 1 . In vitro ADME assays were performed at AMRI (now Curia), Absorption Systems or Eurofins laboratories, unless specified. Methods for the following assays were as described in ( Amaradhi et al., 2022 ): CYP inhibition in human liver microsomes, plasma protein binding across species, metabolic clearance in liver microsomes across species, P-glycoprotein (PgP) substrate and inhibitor assessments in Caco-2 and MDR1-MDCK monolayers. Additional assays are described here. A 10 μM solution of TG11-77 in pre-warmed, blood, plasma or simulated gastric fluid was prepared in a final DMSO concentration of 1%. The mixture was incubated at 37 °C in a shaking water bath. At various times 150 μL of the incubation mixture was removed into 300 μL of cold acetonitrile to stop the reaction. The samples were stored at −80°C or allowed to sit on ice for at least 15 minutes and then the plate was centrifuged at 3,600 rpm for 15 minutes. The supernatants were diluted 1:1 (v/v) with water containing 0.15 μM verapamil (internal standard) in a 96-well shallow injection plate, which was sealed for LC-MS analysis. All measurements were performed in duplicate. Stability half-life values were calculated from the slope of the plot natural log [ln](% remaining) vs time. Cryopreserved hepatocytes from humans (pooled from 10 male donors), male Sprague Dawley rats, beagle dogs, ICR/CD1 mouse, and cynomolgus monkey were obtained from Bioreclamation IVT (Baltimore MD). A prewarmed solution containing 1 μM TG11-77 and 1 million hepatocytes in InvitroGro HI medium was incubated at 37 C in 5% CO2 for 0, 15, 30, 60, 90 and 120 minutes; aliquots were removed into 1.5 vol of ice-cold acetonitrile, placed on ice for at least 15 min and centrifuged to precipitate protein. Supernatants were diluted 1:1 with H2O containing verapamil (0.15 μM) and ibuprofen (20 μM) as internal standards for LC-MS analysis done in duplicate. Metabolic half-life (t 1/2 ) and intrinsic clearance (CL int ) values were calculated from the slope of the plot of ln(% remaining) vs. time and the concentration of hepatocytes present in the incubation; CL int = slope / (mg protein/ml). In vivo hepatic clearance (CL H ) was estimated by (Q H x CL’ int ) / (Q H + CL’ int ), where Q H is total liver blood flow and CL’ int = CL int x (10 6 cells/g liver weight) x (g liver weight/kg body weight). These equations assume the fraction unbound (fu) = 1 (i.e., 100% of the compound is unbound); CL H is reduced if fu is <1. The predicted hepatic extraction ratio, E H = CL H /Q H . The species-specific parameters used in these equations are presented in Supplementary Table 2 . Hepatocytes were sourced from rat, dog and human and incubated as described above; the Trypan Blue exclusion method indicated 70-87% viability across the three species. Incubation reactions consisted of 500,000 active hepatocytes and 50 μM TG11-77 in InVitroGro HI medium, in a final DMSO concentration of 0.5%. After a 4 hour incubation in a CO 2 shaking incubator at 37 °C, 5% CO 2 and saturating humidity, a 2-fold volume excess of acetonitrile was added and samples were incubated on ice for 15 min then centrifuged (10,000 g, 15 min) to remove precipitated protein. Supernatants were passed through a 0.45 μm Whatman filter, dried at 37°C under nitrogen and redissolved in 20% acetonitrile for LC/MS-MS analysis. The method consisted of a TOF MS scan (m/z 150-1000) with information dependent acquisition (IDA) triggering thresholds set appropriately to collect MS/MS spectra of potential peaks of interest; UV data were acquired at 254 nm and 280 nm. Data analysis and structure assignment were performed using the MetabolitePilot 2.0 software. Eight concentrations (3-fold serial dilutions) of TG11-77 were incubated at 37°C with pooled (male and female) human liver microsomes (0.1 mg/ml) in a buffer containing a CYP substrate appropriate for each CYP tested, the concentration of the substrate being at or near the Km value for its cognate CYP isoform. NADPH was added to one set of reactions in a 30 min preincubation step at 37°C, then NADPH (1 mM) was added to both reaction sets and the incubation continued for 10 min (CYP isoforms 3A4 and 2D6) or 20 min (1A2, 2C19), times selected to ensure linear kinetics. Reactions were terminated by adding ice-cold acetonitrile containing the appropriate internal standard for each substrate. LC/MS/MS was used to follow disappearance of each substrate; measured IC50s in the presence and absence of NADPH preincubation were used to calculate a fold IC50 shift as the ratio of (IC50 without preincubation) / (IC50 with preincubation). For time-dependent inactivation kinetic studies, eight concentrations of TG11-77 were preincubated at 37°C in duplicate with human liver microsomes (1 mg protein/ml) in the presence of an NADPH regenerating system. At different times aliquots of the preincubation mixture were diluted 9-fold with CYP-specific substrates and the reaction terminated after 10 min with ice cold acetonitrile. Measurement of inhibition was carried out by mass spectrometric detection of the CYP-specific metabolite of the probe substrate using multiple reaction monitoring. Extent of inhibition was calculated by comparing metabolite formation in samples incubated with TG11-77 versus control wells incubated with the blank solvent. To determine kobs values, the decrease in natural logarithm of the activity over time was plotted for each TG11-77 concentration, and k obs values were described as the negative slopes of the lines. k inact and K I values were calculated using a non-linear regression of the data to the Michaelis-Menten steady state model fit with XLFit 5.3 from IDBS Software (Emeryville, CA). Cryopreserved hepatocytes from three donors were used, one female 71 yr old and two males, 50 and 33 yr old. Hepatocytes in culture were treated in triplicate with vehicle and 0.3, 3 and 30 μM TG11-77 for 48 hr and then known CYP-specific substrates were added for 1 hr. Induction was calculated as the ratio of metabolite level relative to vehicle-treated cells. Hepatocyte viability was tested with the CellTiterGlo assay and was above 96% relative to vehicle for each of the three TG11-77 concentrations. The objective of this assay was to determine contributions of different human cytochrome P450 isoforms (rhCYPs) to metabolism of TG11-77 by extrapolating experimental metabolic clearance data obtained with individual recombinant human CYPs to clearance in human liver microsomes (HLM). This approach addresses differences between rhCYP and HLM systems that result in different activities of CYP enzymes in recombinant systems versus HLM. The analysis also accounts for the relative abundance of CYP isoforms in microsomes. Extrapolation of rhCYP experimental data to the HLM system is achieved using intersystem extrapolation factors (ISEFs), which are derived from ( Chen et al., 2011 ; Palacharla et al., 2018 ). These values and the abundances of CYP isoforms in HLM are presented in Supplementary Table 3 . TG11-77 (1 μM) was incubated in duplicate separately with 80 nM of each recombinant CYP in a NADPH regenerating system containing 1% acetonitrile at 37°C for 30 min in a shaking water bath. Reactions were terminated by the addition of 2-fold volume of acetonitrile containing 40 nM verapamil. After placing incubation samples on ice for 15 min, samples were centrifuged (10,000 g , 15 min, 4 °C) to remove precipitated protein. The supernatants were filtered using a Teflon ® syringe filter (4 mm PTFE, 0.45 μm) from Whatman, Inc. (Clifton, NJ), and then transferred to HPLC vials for LC/MS/MS analysis of parent compound and metabolites. The binding of TG11-77 (10 μM) to each target in a panel of 56 receptors, ion channels, and monoamine transporters was estimated by % inhibition of binding of target-specific radiolabeled probes. A second profile of 27 receptor binding assays and functional assays of 10 enzymes was carried out with 10 μM TG11-77. Altogether 75 unique targets were examined. Inhibition ≥ 50% was considered presumptive positive. These studies were performed at Eurofins. This was evaluated by both patch clamp electrophysiology and inhibition of [ 3 H]dofetilide binding to hERG channels over a range of TG11-77 concentrations. TG11-77 was evaluated for agonist and antagonist effects on 5HT2B receptors by functional cell-based assays. Accumulation of IP1 in human CHO-K1 cells stimulated by 5HT or TG11-77 over a range of concentrations was measured by TR-FRET. HEK293 cells transfected with individual uptake transporters (OAT1, OAT3, OCT1, OCT2, OATP1B1, OATP1B3, MATE1, and MATE2K) were used to assess the substrate and inhibitor potential of TG11-77. To assess substrate potential, transfected and vector-control cells were cultured to confluence and then incubated with 0.5 or 5 μM TG11-77 for 2 and 10 min. The uptake incubation was stopped by two washes with ice cold HBSS, then the cells were lysed with 75% acetonitrile and aliquots of the lysate were analyzed for TG11-77 by LC-MS/MS. The influx rate (pmol/min/mg protein) was determined for each condition and the influx rate ratio to the vector control cells calculated for each condition. An influx rate ratio ≥ 2.0 is considered presumptive substrate positive. To assess inhibitor potential, 10 μM TG11-77 plus transporter-specific substrates were added to transfected and vector-control cells in duplicate. After 2-10 min incubation (time is transporter-specific), uptake was stopped by washing twice with ice cold HBSS then cells were lysed with acetonitrile containing transporter-specific internal standards. Substrate concentrations within the cells were measured by LC-MS/MS and percent inhibition of the influx rate was calculated. Percent inhibition ≥ 50% was considered presumptive positive. The goal was to prepare crystalline salts of TG11-77 with various acids to identify a crystalline salt form that exhibits improved physical properties over the free base. Thirteen acids were selected for testing based on industry precedence, molecular variety, and pKa values. TG11-77 free base was stirred at room temperature with 1-3 molar equivalents of acid in a slurry with solvent; precipitated solids were collected for analysis. The resulting solids were qualitatively evaluated for crystallinity by polarized light microscopy and/or X-ray powder diffraction (XRPD). XRPD indexing was attempted for new crystalline patterns. Solution proton NMR spectroscopy was used to confirm composition and stoichiometry, verify that chemical degradation did not occur, and evaluate the amount of organic solvent present. Ion chromatography was also utilized for further confirmation of counterion content for select salts with acids that cannot be detected by proton NMR. Solvated materials were dried under various conditions to remove the organic solvent. Salts with confirmed composition were further analyzed by differential scanning calorimetry and thermogravimetric analyses to assess volatiles on heating, as well as the likely melt onset. Unsolvated crystalline salts exhibiting a melt onset above 125 °C and confirmed to consist of a single crystalline phase were selected for further study. Aqueous solubility values were estimated visually, and samples were stressed at elevated relative humidity for qualitative assessment of hygroscopicity. Ideal candidates would exhibit aqueous solubility > 1 mg/mL and no changes upon stressing at 90% RH. The desired salt was subjected to single-crystal X-ray diffraction to confirm its stoichiometry and hydration state. The structure was solved by using the SHELXT algorithm. Solutions of the salt were also subjected to a variety of rapid and slow cooling, lyophilization, fast and slow evaporation, and milling, and the resulting solids evaluated by X-ray powder diffraction and indexing for the stable form. Cohorts of 5 male rats were administered TG11-77 by oral gavage (10 ml/kg) at 0, 150, 500 and 1000 mg/kg, and female rats at 500 mg/kg, daily for 7 consecutive days. All doses were as the free base. The vehicle was 0.5% methylcellulose in sterile water (w/v). Separate groups of 3 male and 3 female rats used to assess toxicokinetics (TK) were administered the same dose levels of TG11-77 and were subject to blood collection via a catheter in the jugular vein up to 24 hr post-dose on days 1 and 7 for analysis of TG11-77 in plasma samples. Blood was collected on day 8 for clinical chemistry, hematology, and coagulation assessment prior to necropsy. Main subset animals were observed cage-side pre-dose, immediately and 2-4 hr post-dose, and prior to gross necropsy on day 8. Cage-side observations were collected from the TK animals once daily. Upon necropsy, an examination of all cranial, thoracic and abdominal organs was performed, eight organs were weighed (brain, heart, kidneys, liver, spleen, ovaries, testes, without epididymides, thymus), and 15 tissues were preserved in formalin and processed through 5 μm paraffin sections for histopathology with H&E staining. Tissues processed for histopathology include forebrain, midbrain, hindbrain, eyes, heart, colon, duodenum, kidneys, liver, pancreas, spleen, stomach, testes, urinary bladder, optic nerve. In addition to these tissues, gross examination was made of the thymus and lungs with bronchi. This was a non-GLP study performed by SRI Biosciences, Menlo Park, CA.

Cognitive impairments from SE and epilepsy are primarily treated with rehabilitation and are usually refractory to available pharmaceutical treatments or rehabilitation therapy ( Brooks-Kayal et al., 2013 ). Our aim is to develop pharmacological strategies for seizure-related cognitive deficits by blocking EP2 receptors. In this study, we describe results in which our lead candidate EP2 antagonist TG11-77 attenuates both neuroinflammation and SE-induced cognitive deficits evaluated by Y-maze. Interestingly, delayed mortality after pilocarpine-induced SE is also attenuated by TG11-77. To develop the candidate EP2 antagonist for clinical use, FDA and other regulatory agencies mandate a number of preclinical tests to improve the chance that the clinical candidate is safe in humans. These include numerous in vitro ADMET studies, as well as in vivo safety pharmacology and toxicology evaluations that are conducted under non-GLP and GLP conditions. We have completed in vitro studies and a 7-day dose-range finding toxicology study in rats, all under non-GLP conditions. The results of these tests are generally positive but point to some potential limitations for use in humans that can be evaluated in clinical trials. Neuroinflammation is exacerbated by EP2 activation on myeloid cells, both microglia as shown here and elsewhere ( Chan et al., 2018 ; Ganesh et al., 2018 ; Jiang et al., 2020a ; Jiang et al., 2020b ; Jiang et al., 2013 ; Jiang et al., 2019 ; Quan et al., 2013 ; Rojas et al., 2021b ) and brain-infiltrating monocytes ( Varvel et al., 2021 ). Remarkably, the cellular and behavioral effects of the EP2 antagonists in SE models completely recapitulate the consequences of a conditional COX2 knockout limited to principal forebrain neurons ( Serrano et al., 2011 ; Levin et al., 2012 ), suggesting that EP2 activation is a major mediator of the deleterious effects of COX2 induction in the brain. Reduced microgliosis is a feature of all three EP2 antagonists tested (TG6-10-1, TG8-260 and TG11-77), across three rodent models of SE (kainate, pilocarpine and diisopropylfluorophosphate), two models of sepsis and one model of Alzheimer’s ( Jiang et al., 2020b ; Jiang et al., 2013 ; Jiang et al., 2019 ; Rojas et al., 2015 ; Rojas et al., 2016 ; Varvel et al., 2021 ). Our previous work ( Quan et al., 2013 ; Fu et al., 2015 ; Varvel et al., 2021 ), and our unpublished in situ hybridization work, indicate that microglia express EP2, and expression is increased in activated microglia. By contrast astrocytes appear to express little or no EP2. The cell-specific differences in EP2 expression among microglia, monocytes and astrocytes might underlie our observation that across a wide variety of animal models, EP2 inhibition appears to dampen microgliosis more than astrogliosis. The molecular mechanisms by which EP2 activation exacerbate inflammation in myeloid cells likely involves a switch from aerobic to anaerobic metabolism ( Minhas et al., 2021 ), coupled to upregulation of inflammatory mediators via NFkB transcriptional pathways ( Aoki et al., 2017 ; Lee et al., 2019 ; Thumkeo et al., 2022 ). It has been challenging to determine the cellular basis for cognitive deficits after SE, which have been variously attributed to neurodegeneration, neuroinflammation or aberrant neurogenesis, all of which invariably accompany SE ( Dingledine et al., 2014 ; Parent and Lowenstein, 2002 ; Parent and Murphy, 2008 ; Rojas et al., 2021b ; Scharfman and Gray, 2007 ; Vezzani et al., 2015 ). For several reasons we currently favor neuroinflammation playing a major role in the cognitive deficits observed after SE. First, TG11-77 provides no neuroprotection after SE but is anti-inflammatory in the brain and abolishes the Y-maze deficit. Second, prolonged febrile seizures can produce learning and memory deficits in the absence of neurodegeneration but with a pronounced neuroinflammatory component ( Brennan et al., 2021 ; Dube et al., 2009 ). Finally, EP2 antagonism in a peripheral inflammation model that lacks neurodegeneration nonetheless attenuates the ensuing brain inflammation and restores cognitive function in mice ( Jiang et al., 2020b ). These converging findings redirect attention away from neurodegeneration and towards the alleviation of EP2-dependent neuroinflammation as a possible mode of action to improve cognition and lessen delayed mortality after SE. We speculate that the reason TG6-10-1, but not after TG11-77 or TG8-260, is neuroprotective after SE may lie in higher brain penetration of TG6-10-1. Whether EP2 receptors regulate neurogenesis has yet to be evaluated. Inhibition of peripheral EP2 receptors is sufficient to restore cognition in aged mice ( Minhas et al., 2021 ). Moreover, preventing brain infiltration of circulating monocytes either by a systemic EP2 antagonist or CCR2 knockout replicates many of the consequences of EP2 antagonists after SE ( Varvel et al., 2021 ; Varvel et al., 2016 ). In contrast to microglia grown in culture ( Fu et al., 2015 ; Quan et al., 2013 ), microglia freshly isolated from the brain express a low level of EP2 mRNA, approximately 100-fold lower than that of monocytes ( Varvel et al., 2021 ). These results together implicate EP2 receptors expressed on brain-infiltrating monocytes, and surprisingly not brain-resident microglia, as a major driver of neuroinflammation and cognitive failure after SE. Whether these monocytes originate from the marrow of long bones or from restricted marrow sites in the skull that have privileged access to brain parenchyma ( Cugurra et al., 2021 ) is a question for future study. After seven daily oral doses of TG11-77 up to 1000 mg/kg/day in the male rat and 500 mg/kg/day in the female rat, we observed no toxicologically significant effects on behavior, weight gain, clinical pathology parameters, macroscopic organ appearance or histopathology. Thus, the NOAEL after 7 days of treatment was 1000 mg/kg/day in males and 500 mg/kg/day in females. The AUC, but not Cmax, progressively increases as the dose is increased ( Table 1 ). The ratio between the highest non-toxic AUC achieved in the rat DRF study (36,900 hr*ng/ml, supplemental Table 1 ), and the minimally effective AUC as an anti-inflammatory effect in mice after SE (1,046 hr*ng/ml, Table 1 ) can be used to estimate a therapeutic margin. Based on nearly identical plasma protein binding in mouse (98.6%) and rat (98.5%), the therapeutic margin based on free plasma drug levels is 35. In in vitro assays of metabolic enzymes, all but one of the seven CYPs tested showed low inhibition by TG11-77. Potential problems with the observed 4 μM IC50 for CYP2C8 inhibition should be mitigated by the fact that none of the anticonvulsant drugs commonly used in efforts to abort SE (midazolam, lorazepam, valproic acid, fosphenytoin, levetiracetam and lacosamide) is metabolized significantly by CYP2C8 so clinical drug-drug interactions are unlikely to be problematic for treatment of SE. From the measured in vitro intrinsic clearance in hepatic microsomes and hepatocytes across species, TG11-77 is predicted to have lower in vivo clearance in humans than in rat, mouse, dog or non-human primate (predicted clearance in humans = 10-14 ml/min/kg). Thus the plasma exposure in humans is predicted to be higher than that in other animals, for the same dose. Testing for time-dependent inhibition (TDI) is important in drug development because it can portend clinically relevant drug-drug interactions. TDI typically results from covalent or very tight reversible binding of a chemically reactive intermediate to the enzyme that metabolizes the drug ( Grimm et al., 2009 ), and is often flagged for additional study if the ratio of compound IC50’s in absence and presence of NADPH during a preincubation step is >1.5 to 2-fold. By this criterion TG11-77 is a potential time-dependent inhibitor of CYP2B6, 2C19 and 3A4 ( Table 3 ). CYP3A4 is almost exclusively responsible for metabolizing TG11-77 ( Table 3 ), as well as many other FDA approved drugs including midazolam, triazolam, lovastatin, sildenafil, nisoldipine and others (Hachad et al., 2010). CYP3A4 is thus the CYP of potential concern for clinical DDIs. A subsequent kinetic study that measured the apparent rate of inactivation at different inhibitor concentrations allowed estimation of K inact , the maximum rate of inactivation, and K i , the inhibitor concentration that causes half-maximal inactivation rate for TG11-77 interactions with CYP3A4. A conservative estimate of the natural degradation rate constant for CYP3A4 (kdeg = 0.000146/min ( Chan et al., 2018 ; Yang et al., 2008 ), together with K inact and K i and a projected upper limit for the TG11-77 free plasma Cmax of 50-200 nM at efficacious doses, allows prediction of the magnitude of a potential drug-drug interaction. Thus, the projected ratio of plasma exposure (AUC) of a drug metabolized exclusively by CYP3A4 in the presence and absence of TG11-77 would be 1.06-1.24 ( Mayhew et al., 2000 ). TG11-77 is therefore categorized as a potentially weak time-dependent inhibitor of CYP3A4 ( Grimm et al., 2009 ), and some caution should be exercised when TG11-77 is administered together with other CYP3A4 substrates, inhibitors and inducers. Pharmacokinetic-based drug-drug interactions with TG11-77 will be explored directly in clinical trials. For additional context it is useful to point out that many FDA-approved and marketed drugs are also time-dependent inhibitors of CYP3A4, including cannabidiol, fluoxetine, midazolam, nicardipine, tamoxifen, verapamil and others ( Grimm et al., 2009 ). TG11-77 is orally available (F%=12% in rats and 30% in dogs; supplemental Table 1 ) and brain permeable, with brain/plasma ratio in mice of 0.4 to ~4 depending on the timing of sample collection ( Banik et al., 2021 ). Three other structural classes of EP2-selective antagonist have been introduced. The candidate from Pfizer, PF0448948, is a potent, selective EP2 antagonist ( af Forselles et al., 2011 ) that is excluded from the brain ( Minhas et al., 2021 ) and was advanced to phase 1 clinical trials. This compound displayed excellent plasma exposure in humans and was well tolerated but caused a dose-dependent increase in plasma bilirubin levels at concentrations in the expected efficacious range, which was traced to inhibition of the liver and kidney transporters OATP1B1, OATP1B3, and MATE1. Likewise, the Amgen EP2 antagonist, “compound 52” ( Fox et al., 2015 ), inhibits the liver transporters with potency similar to that at EP2 (Michael Venuti, personal communication). We tested our candidate molecule TG11-77 against these transporters and found that it modestly inhibits OATP1B1 and MATE1 ( Supplemental Table 5 ). However, unlike the Pfizer and Amgen compounds, which were intended for chronic use, we intend to use TG11-77 subchronically over a period of 3-14 days, which could mitigate the potential consequences of inhibition of these transporters. Finally, Ono Pharmaceuticals introduced a series of EP2 antagonists (patent AU20203063887A1), but little is known about them. Our first generation “research lead” EP2 antagonist, TG6-10-1, is brain-permeant and exerts beneficial effects in mouse models of status epilepticus ( Jiang et al., 2013 , 2019 ; Varvel et al., 2021 ), sepsis ( C Jiang et al., 2020b ; unpublished), endometriosis ( Greaves et al., 2017 ), glioblastoma and neuroblastoma ( Qui et al., 2019 ). We developed TG11-77 to have more aqueous solubility and to avoid the acrylamide moiety of TG6-10-1, which has the potential to be carcinogenic. The remaining challenge for clinical development of TG11-77 for cognition preservation after SE is to achieve high enough plasma exposure in a non-rodent species to facilitate dose range-finding and pivotal IND-enabling toxicology studies. Currently, alternatives to cynomolgus monkeys (such as dogs) are preferred as the non-rodent species due to supply chain issues (Nonclinical Considerations for Mitigating Nonhuman Primate Supply Constraints Arising from the COVID-19 Pandemic. Guidance for Industry, February 2022), so efforts to improve plasma exposure in dogs are a priority. Possible solutions might be to dose more frequently (e.g., BID) or develop refined pharmaceutical formulations to alleviate this problem. We demonstrate that EP2 antagonism mitigates a prominent cognitive deficit in the pilocarpine mouse model of status epilepticus. These results confirm and extend preclinical findings targeting EP2 receptor inhibition after seizures or LPS-induced inflammation in order to reduce neuroinflammation and improve functional recovery. We further demonstrate that the novel EP2 antagonist TG11-77 reduces delayed mortality after SE, without side effects in rodents. Thus TG11-77 is a promising lead candidate for treatment of the cognitive comorbidities of SE. Successful completion of this project would lay the groundwork for a clinical test of the hypothesis that EP2 modulation after seizures can provide the first preventive treatment for one of the chief comorbidities of epilepsy. Success in these studies is also likely to underlie a more generalizable strategy for disease modification of acquired and genetic epilepsies, and perhaps other chronic CNS disorders driven substantially by COX-2 associated inflammation.

The different types of epilepsy, taken together, represent the fourth most prevalent neurologic disorder after stroke, Alzheimer’s Disease, and migraine. Among neurological diseases, epilepsy accounts for the highest disability-adjusted life years lost ( Beghi, 2016 ). Epilepsy is frequently associated with cognitive and psychiatric comorbidities that degrade quality of life ( Mula et al., 2022 ). In many pharmacoresistant patients these issues can be more debilitating than the seizures themselves . The past decade has witnessed several notable clinical advances in treating epilepsy, such as MRI-guided stereotactic laser surgery, neurostimulation and the introduction of new anticonvulsants. However, one third of patients cannot be managed adequately by existing drugs or surgery, and no strategy yet exists to modify either epileptogenesis or epilepsy comorbidities, two of the chief challenges in the epilepsy field today ( Galanopoulou et al., 2016 ; Goldman et al., 2016 ). In one study of comorbidities, 58% of 248 adults who experienced convulsive status epilepticus (SE) lasting a median of 85 min had poor cognitive outcome at 3 months, i.e., could not return to work, were unable to live independently, or worse ( Legriel et al., 2010 ). A following study of ICU patients with refractory SE reported a similarly poor long-term outcome ( Kantanen et al., 2017 ). The incidence of status epilepticus in the USA is 18-41 per 100,000 ( Lu et al., 2020 ) . Mounting an efficacy trial directed to comorbidities such as memory impairment ( Abrahams et al., 1999 ; Meador, 2019 ; Power et al., 2018a ) is likely to be more immediately tractable than tackling epileptogenesis itself, due to the shorter time required to demonstrate potential efficacy of therapeutic intervention and the availability of validated assessment tools such as the NIH Toolbox Cognition Battery ( Weintraub et al., 2013 ). Status epilepticus (SE) is well known to cause memory deficits in patients that appear within weeks and last a year or more ( Power et al., 2018a ; b ). For epilepsies not caused by SE, cognitive deficits are often present at the time of acquired epilepsy diagnosis ( Witt and Helmstaedter, 2015 ). Likewise, in rat pilocarpine models of SE, memory functions can be impaired before the onset of spontaneous recurrent seizures ( Hort et al., 1999 ; Mohammad et al., 2019 ). Cognitive comorbidities substantially reduce quality of life in people with epilepsy. Epidemiological and preclinical observations support the idea that temporal lobe epilepsy and its associated comorbidities including memory problems and anxiety disorders are driven by the same or similar pathological processes ( Bell et al., 2011 ; Goldman et al., 2016 ; Helmstaedter, 2007 ; Kanner, 2016 ; Kleen et al., 2012 ; Leeman-Markowski and Schachter, 2016 ). The mechanistic basis of cognitive comorbidities is suggested to involve sclerosis-associated neuronal death and/or neuroinflammation ( Goldman et al., 2016 ; Paudel et al., 2018 ). Inflammation is a component of all acute brain injuries such as stroke and SE as well as chronic diseases including epilepsy and is mediated by several broad signaling cascades including cyclooxygenase-2 (COX-2)-associated pathways. We have shown that activation of the EP2 receptor for prostaglandin E2 is responsible for blood-brain barrier leakage and much of the inflammatory reaction, neuronal injury and cognitive deficit that follows seizure-provoked COX-2 induction in rodent brain ( Jiang et al., 2012 ; Jiang et al., 2013 ; Jiang et al., 2015 ; Jiang et al., 2019 ; Levin et al., 2012 ; Rojas et al., 2021a ; Rojas et al., 2015 ; Rojas et al., 2016 ; Serrano et al., 2011 ; Varvel et al., 2021 ). These effects can be mostly accounted for by EP2-mediated immunomodulation of microglial activation and brain-infiltrating monocytes ( Fu et al., 2015 ; Quan et al., 2013 ; Varvel et al., 2021 ; Varvel et al., 2016 ). Following hit identification in a high throughput screen for EP2 antagonists, we have synthesized and tested >500 compounds as antagonists of the human EP2 receptor. These compounds displayed a competitive mode of inhibition of the human EP2 receptor and demonstrated in vivo efficacy in three animal models of status epilepticus. From these, we selected one for progression to clinical development. This compound, TG11-77 (2-((4,6-dimethylpyridin-2-yl)amino)-N-(2-(2-methyl-1H-indol-3-yl)ethyl)pyrimidine-5-carboxamide, shown in Figure 1A , and also known as BPN300343), is a potent EP2 antagonist (Schild K B 10 nM), >300-fold selective against the other eight prostanoid receptors, orally active with acceptable plasma half-life (2.4 hr) and brain-to plasma ratio (0.4-4) in mice depending on the timing of sample collection ( Amaradhi et al., 2020 ; Banik et al., 2021 ). The structure-activity relationships for potency and selectivity are well understood ( Amaradhi et al., 2020 ; Ganesh et al., 2014a ; Ganesh et al., 2014b ). We now present in vivo efficacy data for this lead candidate with the pilocarpine-induced mouse SE model and discuss the relationship between efficacy and plasma exposure. We present ADME results, in vitro toxicity, and in vivo toxicology results for this compound, all of which support the continued development of this preclinical lead candidate for advancement into clinical trials.