Preclinical characterization of AMPA receptor potentiator TAK‐137 as a therapeutic drug for schizophrenia

Abstract The downregulation of the glutamate system may be involved in positive, negative, and cognitive symptoms of schizophrenia. Through enhanced glutamate signaling, the activation of the α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazole‐propionic acid (AMPA) receptor, an ionotropic glutamate receptor, could be a new therapeutic strategy for schizophrenia. TAK‐137 is a novel AMPA receptor potentiator with minimal agonistic activity; in this study, we used rodents and nonhuman primates to assess its potential as a drug for schizophrenia. At 10 mg kg−1 p.o., TAK‐137 partially inhibited methamphetamine‐induced hyperlocomotion in rats, and at 3, 10, and 30 mg kg−1 p.o., TAK‐137 partially inhibited MK‐801‐induced hyperlocomotion in mice, suggesting weak effects on the positive symptoms of schizophrenia. At 0.1 and 0.3 mg kg−1 p.o., TAK‐137 significantly ameliorated MK‐801‐induced deficits in the social interaction of rats, demonstrating potential improvement of impaired social functioning, which is a negative symptom of schizophrenia. The effects of TAK‐137 were evaluated on multiple cognitive domains—attention, working memory, and cognitive flexibility. TAK‐137 enhanced attention in the five‐choice serial reaction time task in rats at 0.2 mg kg−1 p.o., and improved working memory both in rats and monkeys: 0.2 and 0.6 mg kg−1 p.o. ameliorated MK‐801‐induced deficits in the radial arm maze test in rats, and 0.1 mg kg−1 p.o. improved the performance of ketamine‐treated monkeys in the delayed matching‐to‐sample task. At 0.1 and 1 mg kg−1 p.o., TAK‐137 improved the cognitive flexibility of subchronic phencyclidine‐treated rats in the reversal learning test. Thus, TAK‐137‐type AMPA receptor potentiators with low intrinsic activity may offer new therapies for schizophrenia.


| INTRODUC TI ON
Schizophrenia consists of a spectrum of symptoms: positive symptoms (hallucinations and delusions), negative symptoms (blunted affect and deficits in social functioning), and cognitive symptoms (deficits in attention, working memory, and cognitive flexibility). 1,2 The hyperdopamine hypothesis postulates that excessive activation of dopaminergic neurons in the subcortical regions of the brain is deeply involved in the pathophysiology of the positive symptoms of schizophrenia. 3 Current antipsychotics, based on dopamine D2 receptor (D2R) antagonism, are effective against the positive symptoms of schizophrenia; however, their efficacy against the negative and cognitive symptoms is limited. 3,4 In addition, their side effects such as extrapyramidal symptoms and metabolic changes, limit their clinical application. 5 Owing to the limitations of the efficacy and safety of the current medications, there is a large unmet need for novel therapeutic strategies for schizophrenia.
As postulated by the hypoglutamate hypothesis, the downregulation of glutamate signaling, especially the dysfunction of the N-methyl-D-aspartate (NMDA) receptor in the medial prefrontal cortex plays an important role in schizophrenia. [6][7][8] antagonists, such as phencyclidine (PCP) and ketamine, induce not only psychotic symptoms, but also negative and cognitive symptoms in healthy subjects. [9][10][11] Thus, the activation of the NMDA receptor may offer a potential therapeutic approach against schizophrenia. Bitopertin (RG1678), a glycine transporter type 1 (GlyT1) inhibitor, which can activate the NMDA receptor by increasing the level of the NMDA receptor co-agonist (glycine), 12 improved the negative symptoms of schizophrenia. 13 Moreover, glycine or D-serine-agonists at the glycine site of the NMDA receptor, significantly improved the negative symptoms of schizophrenia. 14,15 However, strategies that directly activate NMDA function have not been successful, possibly owing to the excitotoxic side effects 16,17 or the desensitization of the NMDA receptors. 18 The α-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid (AMPA) receptor is an ionotropic glutamate receptor that plays a key role in learning and memory. 19 Glutamate receptor subunit 1-knockout mice exhibited cognitive impairment, hyperdopaminergia, and psychosis-like behaviors. 20 The AMPA receptor is involved in the regulation of NMDA receptor activation, which induces ion influx into cells, triggering the release of channelblocking magnesium ion from the NMDA receptor. This results in the activation of NMDA receptor signaling through an increase in NMDA receptor-mediated calcium influx. 21 Therefore, the potentiation of the AMPA receptor is expected to offer a new therapeutic strategy for schizophrenia through the enhancement of glutamate signaling.

Methamphetamine (METH) enhances dopamine release and
activates dopaminergic neurons in the subcortical regions of the brain, which causes psychotic symptoms that resemble the positive symptoms of schizophrenia. 24 Based on hyperdopamine and hypoglutamate hypotheses for psychosis, METH and MK-801 (an NMDA receptor antagonist) were used, respectively. As for cognitive functions, we evaluated the effects of TAK-137 on multiple cognitive domains associated with schizophrenia, such as attention, working memory, and cognitive flexibility, using behavioral test batteries in rats and monkeys. 2,25 The data suggest that TAK-137-type AMPA receptor potentiators with low intrinsic activity may be a promising treatment for the multiple symptoms of schizophrenia, especially for negative and cognitive symptoms.

| Animals
Animal species and strains were selected based on the previous publications for each experiment. Male Sprague-Dawley (SD) rats were purchased from Charles River Laboratories, Japan (Yokohama, Japan) for the measurements of locomotion (7-week-old rats). 26 Male ICR mice and Wistar rats were purchased from CLEA Japan Inc. (Tokyo, Japan) for the measurement of locomotion (6-week-old mice) 26 and the social interaction (SI) test (6-week-old rats), respectively. Male Long-Evans rats were purchased from Japan SLC, Inc. (Hamamatsu, Japan) for the five-choice serial reaction time task (5-CSRTT) (6-week-old rats) and the radial arm maze (RAM) test (8-week-old rats). 27 Female Lister Hooded rats (21 days postnatal) were supplied by Harlan Laboratories UK, Ltd. (Bicester, UK) for the reversal learning test. 28 Female rats were used for reversal learning because females are highly sensitive to PCP 29 and showed more robust performance in certain cognitive tasks compared with male rats. 30,31 Rats and mice were housed in groups of four or five per cage in a light-controlled room (12-hour light/dark cycle; lights on at 7:00 AM) with free access to food and water.
The room temperature and humidity were 20°C-25°C and 40%-60%, respectively, and the animals were given a minimum acclimation period of 1 week prior to the experiment. The animals were randomly assigned to the vehicle-or compound-treated groups.
The delayed matching-to-sample (DMTS) task was evaluated in 4-6-year-old male cynomolgus monkeys (Macaca fascicularis) weighing 4-6 kg (Keari Co. Ltd., Osaka, Japan). The monkeys were housed individually in cages stored at a room temperature of (24 ± 1)°C and a humidity of (55 ± 15) %, with a 12-hour light/dark cycle (lights on at 7:00 am). The monkeys were fed a complete, nutritionally balanced diet with fruit once daily (approximately 3:00-4:00 pm) and water was available ad libitum.
All monkeys were housed and handled in strict accordance with good animal practice under the supervision of veterinarians.

| Locomotion measurement in rats
Locomotion was automatically evaluated using an infrared sensor system (SuperMex, Muromachi Kikai, Tokyo, Japan). The infrared sensor was placed on the center of the cover on the cage with a hole to detect locomotion. Each rat was placed in a Plexiglas covered cage (38 × 25 × 32 cm) for more than 2 hours before the experiment was started to allow acclimation to the experimental environment. Food and water were available ad libitum. After acclimation, TAK-137 or the corresponding ve-

| Locomotion measurement in mice
Locomotion was automatically counted with an infrared sensor automated activity monitoring system. Each of the 36 test cages (30 × 40 × 20 cm) was equipped with a pyroelectric infrared sensor. Each mouse was allowed to acclimate to the test cage for more than 2 hours before the experiment was started. After acclimation, TAK-137 or the corresponding vehicle was administered p.o.

| SI test
The experiment was performed as described previously 32 with some modifications. One day before the experiment, the rats were placed in the experimental room for 2 hours and orally administered distilled water for acclimation. On the day of the experiment, the rats were again Two rats in the same drug treatment group, from the different home cages but with a body weight difference of less than 15 g were placed diagonally opposite in a test box (60 × 60 × 60 cm, 35-40 lux). The SI between two rats was determined from the total time spent participating in social behaviors such as sniffing, genital investigation, chasing, and fighting. The interaction time within a 10 minutes test period was measured by a researcher through the monitoring of the CCD camera viewing the test box under blind condition. Between each session, the test box was cleaned with 10% ethanol. In addition to the interaction time, the locomotor activity of each animal was measured by the tracking system in the same software (TopScan, CleverSys Inc., Reston, VA).
The data are presented as the mean ± SEM of the interaction time and distance moved during the test.

| 5-CSRTT
The experiment consisted of two sessions: training and testing. The training session started with 7-week-old male Long-Evans rats, and 13 rats were used for the test at 12-13 months of age. These rats were also used to assess the effects of drugs other than TAK-137 on 5-CSRTT performance between ages 6 and 12 months. They had a 2-week washout period prior to the start of the study with TAK-137. From the training session, food was restricted to 80%-85% of the animals' free-feeding body weight, throughout the experimental period. The training and testing were conducted by using four operant chambers enclosed in sound-attenuating boxes (Med Associates Inc., St Albans, VT). Each chamber contained a curved wall with five contiguous apertures. Food pellets were supplied automatically into a magazine located in the opposite wall of the five contiguous apertures in the chamber, and a photocell beam was used to detect head entries into the magazine. In the training session, a pellet was delivered into the magazine at the start of each session to initiate the first trial. After a 5 seconds intertrial interval (ITI), a light stimulus was presented in one of the 5 apertures, followed by a 5 seconds limited hold in the absence of light stimulus. The duration of the light stimuli was set to 5 seconds and gradually decreased during training to 2 seconds. The rats were required to nose-poke into the illumi-

| RAM test
The experiment was performed as previously described 33,34 with a minor modification. The dimensions of each arm were 50 cm × 10 cm × 40 cm (length × width × height); the maze was elevated 50 cm above the floor. After a 24-hour fast, the rats' food intake was restricted to 80%-85% of the free-feeding body weight on the first day of exposure to the maze and throughout the experimental period. The RAM test consists of two sessions: training and testing. In the training session, rats were acclimated to the maze and then trained. Reinforcement consisted of 3 45 mg food pellets in a food cup. On the first day of acclimation to the maze, reinforcements were placed near the entrance and at the mid-point of each arm. Three rats were placed on the maze at one time and allowed to explore and consume the pellets for 8 minutes. On the second day of acclimation, each rat was placed on the maze and allowed 5 minutes to consume the pellets in the food cups placed at the mid-point and at the end of each arm. From the third day, reinforcements were placed in a food cup at the end of each arm. The rats were well-trained to collect the pellets placed on the edge of each arm. The learning criterion for the testing session was defined as 2 errors or fewer for 2 consecutive days. From a group of 50 rats which were trained for 15 days, 36 rats that achieved the performance criterion were used for the experiment. On the previous day of the testing session, the baseline level of performance of the rats was assessed to select the rats that would complete the collection of all the pellets placed in the 8 arms with 2 errors or fewer. In the testing session, each rat was placed on the maze facing away from the researcher and facing the fixed arm at the start of the trial. The entry of rats into each arm was recorded in sequence. The rats were allowed to explore until all the pellets in the 8 arms were consumed, or 5 minutes had elapsed; entry into an

| DMTS task
The experiment was performed on 4-6-year-old male cynomolgus monkeys. The monkeys were maintained on 80% of their freefeeding body weight throughout the experiment. Four monkeys were trained to perform the DMTS task by using the Cambridge Neuropsychological Test Automated Battery system (CeNes, Cambridge, UK). 35 Briefly, a trial was initiated by the presentation

| Reversal learning test
The reversal learning test was conducted by b-neuro ™ (University of Manchester, UK). The training and testing methods have been previously described. 28 All rats were tested in operant chambers with two-lever Skinner boxes. The details of the apparatus were described in the previous report. 37 The boxes were controlled by Med-PC software (Version 2.0 for DOS or Med-PC for Windows, Med Associates, Inc. Lafayette, IN). Programs were written using Medstate notation. At 12 weeks post weaning, the female Lister Hooded rats (n = 60) were initially trained to respond to food (45-mg Noyes pellets, PJ Noyes Co Inc., Sandown Chemical Ltd) on a fixed ratio 1 (FR1) schedule of reinforcement in standard two-lever Skinner boxes for 30 minutes. In the FR1 schedule, one press of either lever resulted in the delivery of a food pellet. The rats were food-restricted to approximately 85% of free-feeding body weight, maintained throughout the training and testing by feeding them approximately 12 g of standard lab chow per rat per day. Following a stable level of response to the FR1 schedule, the rats were trained to respond to food in the presence or absence of a visual cue in the form of a light stimulus above the lever. At the start of each session, the house light was turned on; both levers were introduced into the chamber, and the activation of one lever resulted in the delivery of a food pellet. One half of the animals were trained to respond in the presence of visual cue and the other half in the absence of the cue. Following a response on one lever, the house light was turned off and both levers were retracted for a period of 3 seconds; subsequently, the cycle was repeated. The rats were tested until 128 responses were obtained, or until the experimental session was terminated 30 minutes after the initiation of the training.
The active lever was changed from session to session according to a pseudorandom Gellerman schedule. The rats participated in approximately 10-15 sessions over 2-3 weeks of training on the initial reward contingency and we ensured that they met the criterion on this initial task; that is, achieved at least 90% of correct responses with each lever active in at least two sessions. In the next step, they were trained on the opposite reward contingency, the reversal task.
Once the criterion (90% of correct responses) was achieved on the reversal task, PCP (2 mg kg −1 , as a salt) or vehicle was administered (i.p.) at a volume of 1 mL kg −1 twice per day for 7 days. 38 During the PCP treatment period, the animals did not receive training to avoid association between drug treatment and cognitive performance.
After 1-week drug-free period without training, 39 the reversal testing was carried out on the animals. One day before the reversal testing, the rats were trained to respond to food using a randomly assigned contingency (ie, in the presence or absence of a cue). The session was terminated after the consumption of 128 pellets. Both PCP-and vehicle-treated groups were required to achieve 90% of correct responses. If the animals failed to reach this criterion, they were subjected to further training until they were able to sustain 90% of the correct responses. The reversal testing consisted of two experimental phases: initial and reversal. In the initial phase, a consistent reward contingency like that of the previous training day was presented to the rats for 5 minutes or until the rats had earned 20 pellets. Following the initial phase, there was a time-out period of 2 minutes (the house light was turned off). The animal stayed in the Skinner box during this time-out period and the reversal phase was then initiated. In the reversal phase, the reward contingency was reversed so that the animals must respond in the opposite way from the initial test. The reversal test was also performed for 5 minutes.

| Statistical analysis
The data are presented as the mean ± SEM Statistical analysis was performed by EXSUS (CAC Croit Corporation). The F test, followed by Student's t test (for data with homoscedasticity with P ≥ 0.2 by F test) or Aspin-Welch test (for data with heteroscedasticity with P < 0.2 by F test) with multiplicity adjustment by Bonferroni correction were used for comparisons between two groups: vehicle/ vehicle and vehicle/METH ( Figure 1A

| Effect of TAK-137 on the positive symptoms of schizophrenia
The total count of locomotion over 120 minutes after METH administration in the METH-treated group (14470 ± 1136 counts) was significantly higher than that in the vehicle-treated group (2154 ± 588.7 counts) (P = 0.0394, F test followed by P < 0.01, Aspin-Welch test; Figure 1A test; Figure 1D).
Unlike reported antipsychotics with dopamine D2 antagonism, TAK-137 did not induce a significant cataleptic response and increase of prolactin and glucose levels in plasma ( Figure S1). with MK-801, respectively) ( Figure 2B).

| Effects of TAK-137 on the cognitive symptoms of schizophrenia
The plasma and brain concentrations of TAK-137 under fasted conditions were 67% and 71% of those under the fed conditions (Table   S1). Thus, 0.2 and/or 0.6 mg kg −1 of TAK-137 were used in the experiments requiring restricted food consumption.

| Attention
The  Figure 3B). The number of premature responses (a response prior to light stimulation) was 5.15 ± 1.57 and 6.92 ± 1.00 in the vehicle-treated group and the TAK-137-treated group, respectively ( Figure 3C).

| Working memory
In the testing session of the RAM test, treatment with MK-801  Figure 4A).

| Cognitive flexibility
We evaluated the effects of TAK-137 on subchronic PCP-induced deficits in a reversal learning task. In the initial phase of the test-

| Effects of TAK-137 under the combination with olanzapine
We studied the effects of the combination of olanzapine (3 mg kg −1 ), one of the antipsychotics, and TAK-137 (0.1, 1, and 10 mg kg −1 ) on METH-induced hyperlocomotion, cataleptic response, and plasma prolactin level ( Figure S2). The results showed that the inhibitory effect of olanzapine on METH-induced hyperlocomotion was not affected by TAK-137 ( Figure S2A). TAK-137 did not exacerbate the cataleptic response and plasma prolactin level ( Figure S2B and C, respectively). In addition, co-treatment of TAK-137 (1 mg kg −1 , p.o.) and olanzapine (3 mg/kg, p.o.) did not affect the effect on cognitive improvement in novel object recognition test in rats ( Figure S2D).

| D ISCUSS I ON
Schizophrenia is a chronic psychiatric disorder with a spectrum of symptoms: positive, negative, and cognitive symptoms. Several hypotheses for the pathophysiology of schizophrenia have been indicated based on clinical findings. 40 The hyperdopamine hypothesis is mans. In addition, we reported that LY451646 showed narrow safety margins between exposure at the effective dose in rat novel object recognition test (NORT) and at the maximum dose in the absence of seizures, that is, 3.1-and 7.5-fold in the AUC brain and brain C max , respectively. 23 Thus, it can be presumed that the doses of AMPA receptor potentiators in previous clinical studies could not be increased to the exposure level required for efficacy owing to their narrow safety margin against seizure. Therefore, compounds with improved bell-shaped responses and reduced seizure risk should be generated.
We revealed that the bell-shaped response was related to the agonistic property of compounds, which was detected only in primary cells and not in recombinant cells. 22 Thus, the agonistic property of each compound should be characterized using the respective physiological receptor, although such complex conditions may not be appropriate for a high throughput screening (HTS) assay. Therefore, we established an original screening strategy, which included a unique binding assay for HTS, and identified TAK-137. 23 In the DMTS test in naive monkeys, TAK-137 enhanced cognitive performance at 0.03, 0.1, and 1 mg kg −1 , whereas LY451656 enhanced performance only at 0.1 mg kg −1 when dosed at 0.03, 0.1, and 1 mg kg −1 . The safety margins between the exposure yielding cognitive enhancement in NORT and the maximum exposure in the absence of seizure were 116-and 43.7-fold in the AUC brain and brain C max , respectively.
Therefore, TAK-137 is superior to LY451646 in terms of the bellshaped responses and the safety margin.
In this study, we investigated effects of TAK-137, a compound with low agonistic properties, on animal models of schizophrenia. At 10 mg kg −1 , TAK-137 significantly inhibited METH-induced hyperlocomotion in rats to 33.6 ± 7.66% of that in the control ( Figure 1A and B); however, the percentage of inhibition was lower than that of antipsychotics, which exhibited greater than 50% inhibition under the 60% occupancy of D2R. 26 Another critical issue in the development of drugs for schizophrenia is the heterogeneity of the disease etiology and biology of schizophrenia. Indeed, the importance of patient segmentation by biophenotype has been suggested. 67 Patients in the hypoglutamate state may be determined, for example, by mismatch negativity or gamma oscillations. 68,69 The proof of concept should be investigated in the clinical studies of patients selected by such biomarkers.
In conclusion, TAK-137, an AMPA receptor potentiator with a low agonistic activity, broader effective dose range, and greater safety margin against seizure, was shown to be efficacious in various animal models of schizophrenia. Thus, TAK-137-type AMPA receptor potentiators may be promising therapeutic options in neuropsychiatry and neurological diseases.

ACK N OWLED G EM ENTS
We wish to express our sincere thanks to Yasukazu Tajima for supporting the locomotion and RAM studies, Kazuhiro Hamajo for supporting the monkey DMTS study of TAK-137, b-neuro ™ for conducting the reversal learning test, Tomohiro Kaku and Takanobu Kuroita for providing the chemical compound, and Youhei Kosugi and Yuu Moriya for providing the pharmacokinetic data of TAK-137.

D I SCLOS U R E S
This work was funded by Takeda Pharmaceutical Company Limited.