Visual Overview
Abstract
Members of the α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionic acid (AMPA) subtype of ionotropic glutamate receptors mediate the majority of fast synaptic transmission within the mammalian brain and spinal cord, representing attractive targets for therapeutic intervention. Here, we describe novel AMPA receptor modulators that require the presence of the accessory protein CACNG8, also known as transmembrane AMPA receptor regulatory protein γ8 (TARP-γ8). Using calcium flux, radioligand binding, and electrophysiological assays of wild-type and mutant forms of TARP-γ8, we demonstrate that these compounds possess a novel mechanism of action consistent with a partial disruption of the interaction between the TARP and the pore-forming subunit of the channel. One of the molecules, 5-[2-chloro-6-(trifluoromethoxy)phenyl]-1,3-dihydrobenzimidazol-2-one (JNJ-55511118), had excellent pharmacokinetic properties and achieved high receptor occupancy following oral administration. This molecule showed strong, dose-dependent inhibition of neurotransmission within the hippocampus, and a strong anticonvulsant effect. At high levels of receptor occupancy in rodent in vivo models, JNJ-55511118 showed a strong reduction in certain bands on electroencephalogram, transient hyperlocomotion, no motor impairment on rotarod, and a mild impairment in learning and memory. JNJ-55511118 is a novel tool for reversible AMPA receptor inhibition, particularly within the hippocampus, with potential therapeutic utility as an anticonvulsant or neuroprotectant. The existence of a molecule with this mechanism of action demonstrates the possibility of pharmacological targeting of accessory proteins, increasing the potential number of druggable targets.
Introduction
Glutamate is the primary excitatory neurotransmitter in mammalian brain. The α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionic acid (AMPA) subtype of glutamate receptors are ligand-gated ion channels expressed primarily on postsynaptic membranes of excitatory synapses in the central nervous system. AMPA receptors (AMPARs) mediate the majority of fast synaptic transmission within the central nervous system (CNS). Thus, inhibition or negative modulation of AMPARs is an attractive strategy for therapeutic intervention in CNS disorders characterized by excessive neuronal activity. With the notable exception of pore blockers (which are selective for calcium-permeable AMPA receptors; see Stromgaard and Mellor, 2004), no AMPAR inhibitors have been found to have selectivity among the AMPAR subtypes, or to exhibit regional specificity. Since AMPAR activity is ubiquitous within the CNS, general antagonism results in undesired effects, such as ataxia, sedation, and/or dizziness. In clinical use, AMPAR antagonists have very narrow therapeutic dosing windows: the doses needed to obtain anticonvulsant activity are close to or overlap with doses at which undesired effects are observed (Rogawski, 2011).
Over the past two decades, investigations into the quaternary structure of native AMPA receptors have revealed a remarkably large set of interaction partners. Heterologous expression of individual members of the AMPA subtype of ionotropic glutamate receptor (GluA) is sufficient to form functional AMPA receptors. However, full recapitulation of the trafficking, localization, gating characteristics, and pharmacology of native AMPA receptors requires coassembly with a large and diverse set of accessory proteins (Jackson and Nicoll, 2011; Schwenk et al., 2012; Straub and Tomita, 2012). These auxiliary subunits include cytoskeletal and anchoring proteins, other signaling proteins, and several intracellular and transmembrane proteins with largely unknown functions. The wide variety of proteins which can participate in AMPA receptor complexes vastly increases the ability of a neuron to tune the response characteristics of its synapses. Here, we demonstrate that these accessory proteins can be used as novel pharmacological targets.
Members of the transmembrane AMPA receptor regulatory protein (TARP) family (CACNG2, 3, 4, 5, 7, and 8) are associated with most, if not all, AMPARs in the brain. These proteins were originally discovered and named due to their homology to the gamma subunit of voltage-gated calcium channels (Letts et al., 1998; Burgess et al., 1999; Klugbauer et al., 2000). TARPs were subsequently found to associate with and to modulate the activity of AMPA receptors (Hashimoto et al., 1999; Tomita et al., 2003). Several TARPs have distinct region-specific expression in the brain, leading to physiologic differentiation of the AMPA receptor activity. It has been theorized that targeting individual TARPs may enable selective modulation of specific brain circuits without globally affecting synaptic transmission (Gill and Bredt, 2011). The expression pattern of TARP-γ8 is particularly attractive in this respect. Based upon in situ hybridization studies, TARP-γ8 is the predominant TARP throughout the hippocampus, and is expressed within essentially all neurons within the stratum pyramidale and stratum granulosum. In addition, it is expressed in a substantial proportion of neurons in the amygdala, olfactory bulb, and olfactory nucleus, and in certain layers within the frontal cortex. In contrast, TARP-γ8 shows very little expression within the hindbrain, midbrain, or thalamus (Tomita et al., 2003; Lein et al., 2007; http://mouse.brain-map.org/experiment/show/72108823).
Negative modulation of AMPA receptors with a molecule selective for TARP-γ8 offers the possibility of selectively reducing excitatory transmission within brain circuits associated with neuropsychiatric or neurologic disorders. Such an agent could be a useful therapeutic in pathologic conditions characterized by hyperactivity within the hippocampus—for example, temporal lobe epilepsy. This approach should mitigate the side-effect profile attributed to nonselective AMPAR antagonists (Ko et al., 2015).
Here, we describe the in vitro and in vivo characterization of 5-[2-chloro-6-(trifluoromethoxy)phenyl]-1,3-dihydrobenzimidazol-2-one (JNJ-55511118) and 2-(3-chloro-2-(2-oxo-2,3-dihydro-1H-benzo[d]imidazol-5-yl)phenyl)acetonitrile (JNJ-56022486). These compounds are potent negative modulators of AMPA receptors containing TARP-γ8. They show exquisite selectivity, with no measurable effects upon AMPARs containing other TARPs, or upon TARP-less receptors. Using chimeric proteins comprising various segments of TARP-γ8 and -γ4 followed by site-directed mutagenesis, we identified the specific amino acids responsible for this remarkable selectivity. We demonstrate in vivo target occupancy using ex vivo autoradiography, and provide a preliminary investigation of the in vivo pharmacological effects of TARP-γ8–selective AMPA receptor inhibition.
Materials and Methods
3-(2-Chlorophenyl)-2-[2-[6-[(diethylamino)methyl]-2-pyridinyl]ethenyl]-6-fluoro-4(3H)-quinazolinone hydrochloride (CP-465022; Menniti et al., 2000), 1-(4-Aminophenyl)-3-methylcarbamyl-4-methyl-3,4-dihydro-7,8-methylenedioxy-5H-2,3-benzodiazepine hydrochloride (GYKI-53655; Bleakman et al., 1996), and (S)-N-[7-[(4-Aminobutyl)amino]heptyl]-4-hydroxy-α-[(1-oxobutyl)amino]benzenepropanamide dihydrochloride (Philanthotoxin-74; Kromann et al., 2002) were purchased from Tocris (Bristol, UK). N-[4-[1-(propan-2-ylsulfonylamino)propan-2-yl]phenyl]benzamide (LY-395153; Linden et al., 2001) was purchased from Diverchim (Roissy-en-France, France). Perampanel (Hanada et al., 2011) was purchased from Alsachim (Illkirch-Graffenstaden, France). Unless otherwise noted, all data analyses, statistics, and data plots were performed using Origin 2015 or OriginPro 2015 (OriginLab, Northampton, MA). Grubbs’ test was performed prior to statistical analysis; if identified, a single extreme outlier was excluded from further analysis. Unless otherwise noted, averages are expressed as the mean ± S.E.M. Significance levels in figures are denoted as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Unless otherwise noted, parameters from linear and nonlinear least-squares fitting procedures are expressed as the value ± standard error.
Animal studies described in this article that were performed in the United States were in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council, 2011). Studies performed in Europe were in accordance with the European Communities Council Directive 2010/63/EU (European Union, 2010) and local legislation on animal experimentation. Facilities were accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. Animals were allowed to acclimate for 7 days after receipt. They were housed in accordance with institutional standards, received food and water ad libitum, and were maintained on a 12-hour light/dark cycle.
Chemical Synthesis
General Synthetic Methods.
All reagents were purchased from Sigma-Aldrich (St. Louis, MO), Strem Chemicals (Newburyport, MA), or Combi-Blocks (San Diego, CA) and used without further purification, except where noted. Solvents were purchased from EMD Millipore (Cincinnati, OH) and dried by passing through activated alumina columns maintained under argon. All reactions were conducted under a nitrogen atmosphere unless otherwise noted. Flash chromatography was performed on Teledyne Isco CombiFlash systems using commercially available RediSep silica gel cartridges (Teledyne Isco, Lincoln, NE). Reverse-phase high-performance liquid chromatography purifications were performed on an Agilent 1100 Series system (Agilent Technologies, Santa Clara, CA) with a Waters XBridge C18 OBD 5 μM preparative column (Waters Corporation, Milford, MA) unless otherwise noted. NMR spectra were recorded on a Bruker UltraShield-400, Bruker UltraShield-500, or Bruker UltraShield-600 spectrometer (Bruker AG, Fallanden, Switzerland) and were referenced to trimethylsilane. Chemical shifts were recorded in parts per million relative to trimethylsilane, and indirect dipole-dipole coupling constants (J values) are reported in Hertz. Combustion analysis was performed at Intertek Pharmaceutical Services (Whitehouse, NJ). Tritium labeling was conducted at Moravek Biochemicals (Brea, CA). The reaction scheme for the synthesis of JNJ-55511118 is shown in Supplemental Fig. 1. The reaction scheme for the synthesis and tritiation of JNJ-56022486 is shown in Supplemental Fig. 2.
Molecular Biology
Molecular Cloning of GluA Receptors and Their Accessory Proteins from Different Species.
cDNAs for human GluA1-FLIP; GluA1-FLOP; GluA2-FLOP; GluA2-FLIP; GluA3-FLOP; GluA4-FLOP; and and monkey, dog, mouse, and rat GluA1-FLOP, as well as their accessory proteins, including human CACNG2, CACNG3, CACNG4, CACNG7, CACNG8, CNIH2, monkey CACNG8, mouse CACNG8, and CACNG8, were polymerase chain reaction (PCR) amplified from brain cDNAs from respective species. A point mutation was introduced into the GluA2 constructs at the Q/R editing site to allow calcium permeability in the expressed protein (Burnashev et al., 1992). Dog CACNG8 was synthesized with codon optimization based on the published sequence (GenBank accession no. KT749896). The sequences for PCR primers are listed in Supplemental Table 1. The PCR products were cloned into mammalian expression vectors as indicated: pCIneo (Promega, Madison, WI), pcDNA3.1(+) (Life Technologies, Carlsbad, CA), or pcDNA4/TO (Life Technologies). Cloning sites (highlighted in shaded letters) were introduced into primers to facilitate the cloning process. The insert regions were sequenced to confirm the sequence identities. FLOP and FLIP splice variants are designated with o and i suffixes, respectively (e.g., the FLIP variant of GluA1 is designated GluA1i).
Generation of GluA1o-CACNG8 Fusion Protein Expression Constructs.
To ensure a 1:1 stoichiometry of GluA1o and γ8 in the expressed channel, a fusion of the cDNAs for GRIA1o and CACNG8 was used. Following Shi et al. (2009), we fused the cDNA encoding the C terminus of GluA1o to the cDNA encoding the N terminus of γ8. We inserted a linker sequence encoding QQQQQQQQQQEFAT between the two full-length cDNAs. The channels expressed with this construct appear to have identical properties to channels formed by coexpression of GRIA1o with an excess of CACNG8 (Shi et al., 2009). Human, mouse, and rat GluA1o-CACNG8 fusion protein expression constructs were generated by overlapping PCR followed by cloning into mammalian expression vectors. The human GluA1o-CACNG8 fusion protein expression DNA was cloned into pCIneo between EcoR1 and Not1 sites, whereas the mouse and rat GluA1o-CACNG8 expression constructs were cloned into pcDNA4/TO between HindIII and Not1 sites. The primers and templates for the overlapping PCRs are listed in Supplemental Table 2. All clones were sequenced, and the identities were confirmed. DNA coding and predicted amino acid sequences for the fusion constructs are listed in Supplemental Table 3.
Construction of Chimeric Proteins Using TARPs γ8, γ4, and γ2.
The sequences for the human variants of each protein were aligned using the UniProt alignment tool, which also predicted the transmembrane segments of the proteins. The protein sequences were divided into nine regions separated near the borders of the predicted transmembrane sections; these nine regions were the N and C termini (CT), the four transmembrane domains (TM1–TM4), the two extracellular domains (EX1, EX2), and the intracellular domain. The predicted topology of the TARP is shown in Fig. 3A, and the splice points between the TARPs are shown in Supplemental Table 4. The chimeras were designated by a nine-digit number; each digit indicates the TARP used for that section of the protein, starting from the N terminus. Graphical representations of the chimeric TARPs are shown in Supplemental Fig. 3. The chimeric expression constructs were generated using overlapping PCRs, except those indicated otherwise. First, two separate PCR reactions (5′ end PCR and 3′ end PCR) that generated overlapping PCR products were performed. Next, the 5′ end and 3′ end PCR products were mixed to serve as the template for the PCR reactions that generated the full-length PCR product for molecular clonings. The primers and templates used for PCR reactions are listed in Supplemental Table 5. DNA coding and predicted amino acid sequences for the chimeric constructs are listed in Supplemental Table 6.
Generation of Point Mutations.
All mutant expression constructs were generated by overlapping PCR using the human wild-type CACNG8 or CACNG4 cDNA as the template. The primers used for generation of the mutants are listed in Supplemental Table 7. DNA coding and predicted amino acid sequences for the chimeric constructs are listed in Supplemental Table 8.
Calcium Flux Assay
A clonal cell line stably expressing the human GluA1o-γ8 fusion construct under geneticin selection in human embryonic kidney 293 (HEK-293) cells was established for the primary calcium flux assay. All other combinations of GluA subunits and TARPs were performed using cotransfections of the respective plasmids into HEK-293-F cells. AMPA receptors formed by cotransfections are designated with the plus symbol (e.g., GluA1i cotransfected with TARP-γ8 is referred to as GluA1i+γ8).
For assays with transiently transfected cells, the cells were generated by bulk transfection. Prior to transfection, 293-F cells were cultured in FreeStyle-293 Expression Medium (Gibco, Grand Island, NY) at 0.5–2 million cells/ml in shaker flasks at 37°C and 8% CO2 at 120 rpm. At the time of transfection, cells were diluted to 1 million/ml with FreeStyle-293 medium. Cell viability was above 90% for transfections to be considered successful. Transfection was performed by combining equal amounts of pAdvantage vector (Promega) and target DNA. Total DNA was 50 µg per 40-ml transfection. The DNA ratio of AMPA receptor to TARPs was 4:1. The transfection reagent was FreeStyle MAX (Invitrogen, Carlsbad, CA). Cells were seeded into 384-well polylysine-coated plates at 15,000 cells/well at 16–24 hours after transfection, and used for assays 24–48 hours after transfection.
The calcium flux assays were performed as follows. Cell plates were washed with assay buffer (135 mM NaCl, 4 mM KCl, 3 mM CaCl2, 1 mM MgCl2, 5 mM glucose, and 10 mM HEPES, pH 7.4, 300 mOsm) using a Biotek EL405 plate washer (Biotek, Winooski, VT). The cells were then loaded with a calcium-sensitive dye according to the manufacturers’ instructions (Calcium-5 or Calcium-6; Molecular Devices, Sunnyvale, CA) combined with the test compounds at a range of concentrations. Calcium flux following the addition of 15 µM glutamate was monitored using a FLIPR Tetra (Molecular Devices).
The fluorescent response in each well was normalized to the response of negative and positive control wells. The negative control wells had no added compounds, and the positive control wells had been incubated with 50 µM CP-465022 (a non–subtype-selective AMPAR antagonist; Lazzaro et al., 2002). The responses (R) to glutamate as functions of the test compound concentrations (x) were fitted to a four-parameter logistic function (eq.1):
(1)The fitted parameter corresponding to the midpoint (ϰ0) was taken to be the potency of inhibition of the compound (IC50; 50% inhibitory concentration). Potency is expressed as Equation 2:
(2)
where pIC50 is the negative log of the 50% inhibitory concentration.
Knockout Animals
The TARP-γ8 knockout mouse line Cacng8tm1Ran was originally described by Rouach et al. (2005). This mouse line, generated by homologous recombination in embryonic stem cells to replace exons 2 and 3 with a neomycin resistance gene, was rederived by back-crossing into a C57BL/6J mouse line at Jackson Laboratory (Bar Harbor, ME).
Patch-Clamp Electrophysiology
Heterologous Cells.
Studies with GluA1i were performed with transiently transfected HEK-293 cells. Human GluA1i with or without human TARP-γ8 was transfected into HEK-293 cells using Lipofectamine 2000 (Life Technologies) following the manufacturer’s instructions. Eight to 24 hours after transfection, cells were plated onto 12-mm glass coverslips in Dulbecco’s modified Eagle’s medium with high glucose, without glutamine (Sigma-Aldrich), supplemented with 10% fetal bovine serum, and incubated at 37°C in a humidified 5% CO2 incubator. To increase surface expression, cells were transferred to a humidified 5% CO2 incubator at 30°C for 6–24 hours immediately prior to use. Recordings were performed 48–72 hours post-transfection.
For patch-clamp electrophysiology on heterologously expressed GluA1o-γ8 and human GluA1o-γ2, we established single-cell clones stably expressing these constructs in Chinese hamster ovary cells (T-Rex-CHO; Invitrogen) using a tetracycline-inducible expression vector. Cells were cultured in Ham’s F-12 supplemented with 10% fetal bovine serum, 100 µg/ml zeocin, and 5 µg/ml blasticidin. To induce expression, 1 µg/ml tetracycline was added to the culture medium 1–4 days prior to use. Cells were plated onto 12-mm plain glass coverslips and incubated at 37°C in a humidified 5% CO2 incubator. To increase surface expression, cells were transferred to a humidified 5% CO2 incubator at 30°C for 6–24 hours immediately prior to use.
Hippocampal Neurons.
Acute hippocampal neurons were obtained from 8- to 12-week old C57BL/6J male mice, following the protocol described by Brewer (1997) with the following modifications. Medium was prepared by supplementing HibernateA with 2% B27 and 0.5 mM Glutamax (HABG medium; all reagents from Life Technologies). Mice were asphyxiated with CO2 and then decapitated in accordance with National Institutes of Health (NIH) animal and use guidelines. The brain was rapidly removed, then placed into ice-cold HABG medium. Sagittal slices, 300 µm thick, were obtained using a VT1200S microtome (Leica Biosystems, Buffalo Grove, IL). Slices were cut in ice-cold solution composed of 150 mM sucrose, 50 mM NaCl, 25 mM NaHCO3, 10 mM glucose, 7 mM MgSO4, 2.5 mM KCl, 1.25 mM Na3PO4, and 0.5 mM CaCl2 equilibrated with 95% O2 and 5% CO2. The hippocampus was isolated from the rest of the slices and transferred to a calcium-free HibernateA Minus Calcium solution (BrainBits, Springfield, IL) containing 20 mg of papain (Worthington Biochemical, Lakewood, NJ) and 0.5 mM Glutamax (Life Technologies) and digested at 30°C under gentle shaking for 30 minutes. Then, the papain solution was aspirated and replaced with HABG. Slices were gently triturated with fire-polished Pasteur pipettes. The supernatant containing dissociated neurons was collected, and then centrifuged for 2 minutes at 200g. The cell pellet was collected and then resuspended in HABG. The cell suspension was then plated over coverslips, and isolated neurons were picked under visual inspection for whole-cell patch-clamp recordings. The extracellular and intracellular solutions were the same as described earlier for transfected HEK-293 cells.
Cerebellar Granule Cells.
Cerebellar granule cell cultures were prepared following Brewer (1997) with the following modifications. Cerebella were harvested from newborn Sprague-Dawley rat pups (1–4 days old, males and females were mixed). Tissue was minced manually prior to trypsin digestion. Dissociated cells were plated onto glass coverslips coated with poly-d-lysine and fibronectin, and then cultured for 2–4 weeks prior to use.
Electrophysiology.
Whole-cell and outside-out patch electrophysiology (Hamill et al., 1981) was performed using 1.5-mm-diameter glass capillary tubes (TW150-4; World Precision Instruments, Sarasota, FL) pulled to a fine tip with a Sutter P-97 micropipette puller (Sutter Instruments, Novato, CA). The intracellular buffer was 90 mM potassium fluoride, 30 mM KCl, 10 mM HEPES, and 5 mM EGTA (pH 7.4, 290 mOsm). The extracellular buffer was 135 mM NaCl, 4 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 5 mM glucose, and 10 mM HEPES (pH 7.4, 300 mOsm). The open-tip resistances of the micropipettes using these solutions were 2–4 MΩ. Recordings were performed in voltage-clamp mode using an Axopatch 200B amplifier and Digidata 1440A digitizer (Axon Instruments, Sunnyvale, CA). Recordings were controlled and measured using pClamp 9.2 software (Axon Instruments). Current was measured by holding the interior of the cell at −60mV, using a 5-kHz low-pass filter. The cells were continuously perfused through 7-mm square glass barrels using a solenoid-controlled solution switching device (PF-77B; Warner Instruments, Hamden, CT). The peak current in response to a 500-ms exposure to 10 mM glutamate every 5 seconds was measured before and after exposure to test compound; 10 mM glutamate was chosen as a saturating concentration for the peak responses (Robert and Howe, 2003). Steady-state currents were measured during the last 50 ms of the glutamate application. Upon establishing stable glutamate-evoked responses, JNJ-55511118 was applied before and during glutamate application until a steady-state inhibition was observed (typically 50–60 seconds). For analysis, the mean peak current of five traces in the presence of test compound was divided by the mean peak current of five traces prior to the addition of test compound.
For ultra-fast glutamate perfusion, a piezo-driven perfusion system was used (Siskiyou, Grants Pass, OR). Recordings on outside-out patches were performed using an AxoPatch 200B amplifier (Axon Instruments), and signals were filtered at 10 kHz and digitized at 50 kHz. Data acquisition and online analysis were performed using pClamp 9 (Axon Instruments). Current decay kinetics were fitted with a double exponential function using Origin (OriginLab, Northampton, MA) and expressed as a weighted decay time constant. For recovery from desensitization, an initial desensitizing pulse of glutamate was followed by a second pulse of glutamate at varying time intervals. The recovery from desensitization was expressed as the current peak amplitude fraction of the second pulse to the first pulse at a given time interval, and was fitted using a single exponential function (for TARP-γ8–containing AMPA receptors) or a double exponential function (for TARP-less AMPA receptors) using Origin (OriginLab).
Brain Slice Whole-Cell Patch Clamp Electrophysiology
Male mice (2–3 weeks) were anesthetized with isoflurane and then decapitated in accordance with NIH animal care and use guidelines. Transverse hippocampal slices (300 µm thick) were cut in ice-cold high sucrose buffer containing 87 mM NaCl, 2.5 mM KCl, 0.5 mM CaCl2, 7 mM MgSO4, 1.25 mM NaH2PO4, 25 mM NaHCO3, 25 mM glucose, and 75 mM sucrose equilibrated with 95% O2 and 5% CO2. Slices were then placed in artificial cerebrospinal fluid (ACSF) at 35°C for 30 minutes, and then allowed to recover for at least 1 hour in ACSF at room temperature. ACSF for electrophysiological recordings contained 119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl2, 1.3 mM MgSO4, 1 mM NaH2PO4, 26.2 mM NaHCO3, and 11 mM glucose equilibrated with 95% O2 and 5% CO2. The intracellular recording solution contained 130 mM CsMeSO4, 10 mM HEPES, 8 mM KCl, 4 mM MgATP, 0.4 mM NaGTP, 10 mM sodium creatine, and 10 mM 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid.
Excitatory postsynaptic currents (EPSCs) intracellularly recorded from a neuron in the CA1 pyramidal cell layer were evoked by electrical stimulation of the Schaffer collateral/commissural pathway using a monopolar glass stimulating electrode (filled with ACSF) placed in the stratum radiatum of CA1 (0.1-Hz stimulation frequency). In cases where trains of five stimulations at 50 Hz were done, the pulse train was alternated with one stimulation (0.1 or 0.05 Hz between trains). Test compounds were bath-applied. To evoke AMPA EPSCs, neurons were held at −70 mV. N-methyl-d-aspartate (NMDA) EPSCs were recorded at +40 mV 50 ms after the AMPA EPSC at −70 mV. NMDA EPSCs were recorded prior to and after establishing the steady-state inhibition of peak AMPA EPSCs. Recordings were performed using a Multiclamp 700B patch-clamp amplifier (Axon Instruments); signals were filtered at 4 kHz, digitized at 10 kHz, and displayed and analyzed online using pClamp 9.2 (Axon Instruments).
In some experiments, trains of 50-Hz stimulation were used to determine the paired-pulse ratio. In this case, the EPSC magnitude for each pulse was measured from the baseline current immediately after the pulse to the peak current. The paired-pulse ratio was calculated as this EPSC magnitude, divided by the immediately preceding EPSC magnitude.
Brain Slice Field Excitatory Postsynaptic Potentials Electrophysiology
Male mice (7–9 weeks) were anesthetized with isoflurane and then decapitated in accordance with NIH animal care and use guidelines. Horizontal hippocampal slices (300 µm thick) were cut in ice-cold high-sucrose ACSF (composition described earlier). Slices were then placed in ACSF at 35°C for 30 minutes, and then allowed to recover for at least 1 hour in ACSF at room temperature. Slices were then placed on a perforated multielectrode array chip (Multichannel Systems, Reutlingen, Germany) and perfused with ACSF heated to 35°C. Excitatory postsynaptic potentials were evoked with electrodes placed in the CA1 radiatum once every minute, sampling at 50 kHz. Multielectrode array responses across multiple electrodes were chosen based on stability of the response and amplitude (>0.4 mV), which were then averaged to generate an N = 1 for each slice. Test compounds were bath-applied.
Radioligand Binding
Male Sprague-Dawley rats (6–14 weeks) were anesthetized with isoflurane and then decapitated in accordance with NIH animal care and use guidelines. The brain was removed, and hippocampi were rapidly dissected and then frozen at −80°C until use. For each assay, two hippocampi per five 96-well plates were used. On the day of the experiment, the hippocampal tissue was thawed, and then homogenized in assay buffer (50 mM Tris, pH 7.4) for 30 seconds at high speed. The homogenate was centrifuged at 1500 rpm for 5 minutes followed by careful decanting of the supernatant, which was centrifuged at 39,000g for 30 minutes. Ice-cold assay buffer was added to the cell pellet. The protein concentration within the pellet was determined by colorimetry using a Pierce bicinchoninic acid protein assay kit (Thermo Fisher Scientific, Rockford, IL), then diluted with assay buffer to obtain a concentration of 200–400 µg protein per milliliter (10–20 µg protein per well).
Binding assays were performed in Whatman GF/B 96-well filter plates (GE Healthcare, Little Chalfont, United Kingdom) presoaked with 0.3% polyethylenimine. When manufactured, the stock solution of tracer was 34.5 µM single-labeled [3H]JNJ-56022486 and 10.3 µM unlabeled JNJ-56022486 in ethanol. The actual stock concentration was calculated at the time of use based upon the decay rate of tritium. Ten microliters of 10× test compound, 40 µl of 2.5× tracer, and 50 µl of membrane homogenate were placed into each well. The reaction was incubated for 2 hours at 4°C on a shaker, then terminated by filtration followed by washing with ice-cold assay buffer four times. After drying for 30 minutes at 50°C, 60 µl/well MicroScint-O (PerkinElmer, Waltham, MA) was added to the plate. Radioactivity retained on the filters was measured using a TopCount liquid scintillation counter (PerkinElmer, Waltham, MA). The scintillation counter was calibrated with a linear least-squares fit to the radioactivity counts from known quantities of [3H]JNJ-56022486. All analyses were performed in Origin 2015 (OriginLab).
For saturation binding experiments, a 2× serial dilution of [3H]JNJ-56022486 in quadruplicate wells was used for total binding (TB), with 50 µM JNJ-55511118 in quadruplicate wells for determination of nonspecific binding (NSB). Ligand depletion was determined by comparing the radioactivity counts of total binding to the counts in a separate plate spiked with an equivalent amount of radioligand; ligand depletion was <15% at each concentration. Specific binding (SB) at each radioligand concentration was calculated as SB = TB – NSB at each radioligand concentration. NSB was fitted to a linear function using linear least-squares analysis. SB was converted to pmol/mg protein, and then fitted to a single-site binding model to determine the dissociation coefficient (KD) and total receptor concentration (Bmax):
For competition binding experiments, a serial dilution of the test compound was prepared in assay buffer and combined with a final concentration of 20 nM radioligand. NSB was determined using eight wells containing a blocking concentration of 50 µM JNJ-55511118 along with radioligand and tissue membranes, and total binding was determined using eight wells containing only radioligand and tissue membranes. Four replicates for each test compound concentration were used. After incubation and washing as described earlier, the radioactive counts (SB) in each well were normalized to the total and nonspecific binding counts and then fitted to a single-site logistic function. The equilibrium dissociation constant (Ki) was calculated from the midpoint parameter of the fit, adjusted for radioligand concentration using the Cheng-Prusoff correction (Cheng and Prusoff, 1973).
Plasma Protein Binding
Plasma protein binding was determined by equilibrium dialysis using the Rapid Equilibrium Dialysis (RED) device (Thermo Fisher Scientific), consisting of a Teflon base plate, and RED Device inserts comprising two (sample and buffer) side-by-side chambers separated by a dialysis membrane (molecular weight cut-off ≈ 8000). Compounds were prepared as 100 µM dimethylsulfoxide (DMSO) stocks and spiked into 1 ml of mouse, rat, and human plasma (BioreclamationIVT, Westbury, NY) to make a final concentration of 1 µM. Plasma (300 µl) was dispensed into the sample well, and dialysis buffer (100 mM potassium phosphate, pH 7.4, 500 µl) was dispensed into the buffer well. Each compound was tested in triplicate. The RED device was sealed, and equilibrium was permitted for 6 hours in a 37°C incubator with gentle agitation at 100 rpm. After incubation, plasma samples were prepared by transferring 10 µl from plasma wells to 90 µl of fresh dialysis buffer, and buffer samples were prepared by transferring 90 µl from buffer wells to 10 µl of naïve plasma. In addition, a reference sample without equilibration was prepared in triplicate by mixing 10 µl of plasma containing 1 µM compound with 90 µl of buffer to determine compound recovery from the assay. Two volumes of 1:1 acetonitrile:methanol spiked with the internal standard phenytoin (0.2 µg/ml) was added to the reference and samples. Precipitation of plasma protein binding was allowed for 15 minutes before the reference and samples were centrifuge clarified. Supernatant (10 µl) was used for liquid chromatography–tandem mass spectrometry (LC-MS/MS) analyses.
Brain Tissue Binding
Brain tissue binding was assessed by an equilibrium dialysis technique similar to the procedure described for plasma protein binding. Rat brain tissue homogenate prepared in phosphate-buffered saline buffer [pH 7.4, 1:10 (w/v)] was spiked with compound DMSO stock solution to yield a final concentration of 5 μM. The dialysis was carried out in a shaking incubator at 37°C for 5 hours. After incubation, 25 μl of homogenate or 50 μl of buffer was extracted with 50 μl of DMSO and 300 μl of acetonitrile and analyzed by LC-MS/MS using the calibration curves across an appropriate concentration range and quality control samples. All determinations were conducted in triplicate. The apparent unbound fraction (f u,app) was determined as the ratio of the concentration measured in the homogenate to the concentration measured in the buffer. The unbound fraction in undiluted brain was calculated as
where D is a dilution factor of 10. Subsequently, the percentage of compound bound to brain tissue (%BTB) was determined as
Pharmacokinetic Studies
Single-dose pharmacokinetic studies of JNJ-55511118 in male Sprague-Dawley rats were conducted by BioDuro, LLC (Beijing, China) following i.v. (1 mg/kg) and per os (p.o.; 5 mg/kg) administration as a solution in 20% hydroxypropyl-β-cyclodextrin with three equivalents of sodium hydroxide. Blood was sampled at predose and at 0.033 (i.v.), 0.083 (i.v.), 0.25, 0.5, 1, 2, 4, 8, and 24 hours postdose. Plasma concentrations were quantitated by LC-MS/MS. Pharmacokinetic parameters were derived from noncompartmental analysis of the plasma concentration versus time data using WinNonlin software (Pharsight, Palo Alto, CA).
Single-dose pharmacokinetic studies of JNJ-55511118 in male C57/BL6 mice were conducted by BioDuro, LLC following p.o. (10 mg/kg) administration as a suspension in 0.5% hydroxypropyl methylcellulose (HPMC). Blood was sampled at predose and at 0.5, 1, 2, 4, 8, and 24 hours postdose.
Blood-Brain Barrier.
Adult male animals were dosed by oral administration of a suspension in HPMC. The animals were euthanized using carbon dioxide and decapitated at specified time points after drug administration. Brains were rapidly frozen on powdered dry ice and stored at −80°C before sectioning for receptor occupancy studies or for compound concentration determination by LC-MS/MS. The blood-brain barrier ratio was calculated as the compound concentration in the brain divided by the concentration in the plasma for each animal.
LC-MS/MS.
JNJ-55511118 was quantified on an API4000 MS/MS System (Applied Biosystems, Concord, Ontario, Canada) interfaced with an Agilent 1100 Series high-performance liquid chromatographer. Samples were loaded onto a 2.1 × 30-mm ACE 5µm C4 100A column (Advanced Chromatography Technologies Ltd., Aberdeen, Scotland) under a flow rate of 0.9 ml/min using 5 mM ammonium acetate (0.1% formic acid) as mobile phase A and acetonitrile (0.1% formic acid) as mobile phase B. Starting with 87% mobile phase A for 0.4 minute, mobile phase B was increased from 13 to 90% using a linear gradient for 0.8 minute, held at 90% B for 0.3 minute, and equilibrated at 13% B for 1.0 minute for an overall run time of 2.5 minutes. JNJ-55511118 was quantified by MS/MS in the positive ion mode by monitoring the transition of 328.95 to 208.10 m/z.
Ex Vivo Receptor Occupancy
Receptor occupancy was assessed by ex vivo autoradiography using the TARP-γ8 receptor antagonist [3H]JNJ-56022486. Coronal and sagittal tissue sections of 20-μm thickness were prepared for autoradiography as previously described (Langlois et al., 2001). Tissue sections were incubated for 10 minutes in 50 mM Tris HCl containing 0.1% bovine serum albumin (pH 7.4) with 5nM [3H]JNJ-56022486 at room temperature. Nonspecific binding was characterized with a structurally distinct TARP-γ8 receptor antagonist. Sections were rinsed in 50 mM Tris HCl containing 0.1% bovine serum albumin on ice four times for 10 minutes per rinse, followed by two dips in ice-cold deionized water, then dried under a stream of cold air. Digitized images were acquired with β-Imager DFine or TRacer (Biospacelab, Paris, France).
In Vivo Electrophysiology
Male Sprague-Dawley rats (Charles River Laboratories, San Diego, CA) weighing approximately 300–450 g were used for these experiments. The jugular vein was precannulated by the vendor to facilitate intravenous administration of compound. Animals were singly housed, given food and water ad libitum, and maintained on a 12-hour light and dark cycle.
Evoked population spikes from the CA1 region of the hippocampus were recorded following established procedures (Jeggo et al., 2014) with the following modifications. Animals were anesthetized with isoflurane for the duration of the surgical preparation and recording periods while body temperature was maintained with a homeothermic heating pad. A small piece of skull overlaying the hippocampus was removed using a hand-held drill before a concentric bipolar stimulating electrode (FHC, Bowdoin, ME) and tungsten recording microelectrode (World Precision Instruments) were inserted into CA1 using the following stereotaxic coordinates (from the bregma):
Stimulating electrode: anterior-posterior = 3.4 mm, medial-lateral = 2.75 mm
Recording electrode: anterior-posterior = 4.4 mm, medial-lateral = 2.25 mm.
Electrodes were typically inserted to a depth of 2–2.5 mm below the pial surface before test stimuli were used to help optimize the evoked signal and determine the final recording depth. Stimulation intensities evoking a 30–60% maximal response were used. Signals from the recording electrode were amplified and filtered (1 Hz to 10 kHz, DAM80 bio-amplifier; World Precision Instruments), then digitized and collected (40-kHz sampling) using a PowerLab 16/35 data acquisition unit controlled by LabChart Pro software (ADInstruments, Colorado Springs, CO). Brief stimulus pulses were continuously delivered to the hippocampus at a rate of 0.33 Hz, and the evoked neural responses were recorded. A stable baseline period of 10 minutes was obtained before administration of the compound, and evoked responses were obtained for an additional 60 minutes thereafter. JNJ-55511118 was formulated in 5% N-methyl-2-pyrrolidone plus 20% Cremophor (BASF, Ludwigshafen, Germany) plus 75% water and dosed intravenously. At the end of the recording session, the brain was electrically lesioned to determine the final positions of the stimulating and recording electrodes. Additionally, the brain was removed and plasma samples collected to determine in vivo concentrations of compound via LC-MS/MS. The population spike amplitude for each stimulus was extracted from the recording following the procedure outlined by Jeggo et al. (2014), normalized to the mean baseline amplitude, then averaged for each dose group according to time relative to the injection of drug.
Electroencephalogram Recording and Locomotor Activity Studies in Rats
Experiments were conducted in male Sprague-Dawley rats (350–450 g; Harlan Laboratories, Livermore, CA). Animals were chronically implanted with telemetric devices (PhysioTel F40-ETT; Data Sciences International, St. Paul, MN) for the recording of electroencephalogram (EEG) with two epidural electrodes placed in the frontal and parietal cortex, electromyogram (EMG), and locomotor activity as described previously (Dugovic et al., 2009). EEG and EMG signals were digitized at a sampling rate of 100 Hz. High- and low-pass filters were set at 1 and 30 Hz for the EEG signal. Polysomnographic wave forms were analyzed per 10-second epoch and classified as wake, non-rapid eye movement (NREM), or rapid eye movement (REM) sleep using the computer software program SleepSign (Kissei Comtec, Nagano, Japan). EEG activity within specific vigilance states was determined by power spectral analysis (fast Fourier transform) within a frequency range of 1–30 Hz. Values for power spectra were divided into four frequency bands: delta (1–4 Hz), theta (4–10 Hz), alpha/sigma (10–15 Hz), and beta (15–30 Hz). Locomotor activity counts were analyzed into 1-minute bins and averaged into 5-minute intervals for each animal. All results were averaged and expressed as the mean ± S.E.M. in defined time intervals for each animal. To determine whether differences were significant at a given interval, either a one-way analysis of variance (ANOVA) or two-way repeated-measures ANOVA followed by Dunnett’s multiple comparison test was performed.
Anticonvulsant Studies
Anticonvulsant studies were performed by NeuroAdjuvants, Inc. (Salt Lake City, UT). Unless otherwise noted, male albino CF1 mice (Charles River Laboratories, Portage, MI) were used as experimental animals. All animals were allowed free access to both food and water except when they were removed from their cages for the experimental procedure. All mice were housed, fed, and handled in a manner consistent with the recommendations of the National Research Council (2011). No insecticides capable of altering hepatic drug-metabolizing enzymes were used in the animal facilities. All animals were euthanized in accordance with the Institute of Laboratory Resources policies on the humane care of laboratory animals. All test substances were administered orally in 0.5% methylcellulose in a volume of 10 ml/kg body weight. For the maximal electroshock (MES), 6 Hz, and corneal kindling assays, a drop of anesthetic/electrolyte solution (0.5% tetracaine hydrochloride in 0.9% saline) was applied to the eyes of each animal prior to placement of the corneal electrodes.
Mouse 6 Hz Psychomotor Seizure and MES Tests.
Details of the procedures have been described previously (Barton et al., 2001; Rowley and White, 2010). In brief, an acute seizure was induced via electrical stimulation through electrodes applied to the corneas of test animals. For the 6 Hz tests, the stimulus was 32 or 44 mA at 6 Hz for 3 seconds, and the induced seizure was characterized by jaw chomping, vibrissae twitching, forelimb clonus, and Straub tail. For the MES test, the stimulus was 50 mA at 50 Hz for 0.2 second, and the induced seizures were characterized by a tonic hindlimb extension. Cohorts of eight mice for each test concentration and stimulation intensity were treated with a single oral dose 2 hours prior to challenge with the electrical stimulation. Mice not displaying the described seizure phenotypes were considered protected.
Mouse Corneal Kindling.
The corneal kindling assay followed established procedures (Matagne and Klitgaard, 1998; Rowley and White, 2010). Kindling was achieved by twice-daily corneal stimulation (3 mA, 3 seconds, 60 Hz) until all mice reached an established criteria of five consecutive secondarily generalized seizures (Racine stage 5). For compound testing, a cohort of eight fully kindled mice were administered a single oral dose of the test compound 2 hours prior to challenge with the kindling stimulus. The Racine seizure score (0–5; Racine, 1972) was recorded for each mouse and averaged. Animals with seizure scores of 3 or lower were considered protected.
Timed Intravenous Infusion of Metrazol Test.
A single dose of each test compound or vehicle was administered p.o. to cohorts of 10 mice 2 hours prior to the test. Mice were challenged with 0.5% heparinized Metrazol solution [5 mg/ml; pentylenetetrazol (PTZ), Sigma-Aldrich, St. Louis, MO], infused at a constant rate of 0.34 ml/min into a lateral tail vein of an unrestrained mouse (Orloff et al., 1949; White et al., 1997). The time in seconds from the start of the infusion to the appearance of the “first twitch,” and then to the onset of sustained clonus, was recorded. The times to each endpoint were converted to mg/kg of PTZ for each mouse, taking into account the rate of infusion, concentration of PTZ, and weight of the animal.
Amygdala Kindling.
Male Sprague-Dawley rats (Charles River Laboratories) were surgically implanted with stimulation/recording electrodes unilaterally into the amygdala according to the procedure described by McNamara (1995). Rats received daily subthreshold stimulation followed by behavioral and electrographic monitoring for a period of 2–3 weeks, during which time a majority of rats were considered fully kindled (five generalized seizures, Racine scale 4–5, over a period of 8 days). Fully kindled rats received either vehicle (0.5% HPMC) or JNJ-55511118 (suspension in 0.5% HPMC, oral gavage, 10 ml/kg). They were then challenged 2 hours later with the same kindling stimulation, and their behavioral seizure score and electrographic after-discharge duration were recorded.
Rotarod
Immediately prior to seizure testing, all mice were subjected to a rotarod test to assess motor coordination. Animals were placed on a 1-inch knurled rod that rotates at a speed of 6 rpm. The animal was considered motor-impaired if it fell off this rotating rod three times during a 1-minute period.
Morris Water Maze
The procedure followed the water maze task described by Atcha et al. (2009), with the following modifications. Video tracking software (EthoVision XT 9.0; Noldus, Wageningen, The Netherlands) was used to measure the path, time taken, and swim speed for each animal to reach the platform. Male Long-Evans rats (Janvier, Le Genest-Saint-Isle, France; N = 12 per dose group) were trained for 4 days in three daily trials with random starting positions to find the hidden platform (days 1–4). The location of the hidden platform was maintained throughout the study. When an animal failed to find the platform within 60 seconds, it was guided to the platform and allowed to stay there for another 5–10 seconds. Directly after the last acquisition trial on day 4, the animals were subjected to a probe trial for 60 seconds, during which the platform was removed.
Statistical analyses were performed for the averages per day and per trial. For analysis of “latency to platform,” a Cox proportional hazards model was used. For the other parameters, a repeated-measures ANOVA model was used. For probe trials and percent-per-quadrant measures in acquisition, one-way ANOVA statistics were run using InVivoStat software (Clark et al., 2012; http://invivostat.co.uk/), with dose used as the treatment factor.
V-maze
Male Long-Evans rats (body weight 250–300 g; Janvier Laboratories) were individually housed for 7 days before testing and habituated to the experimental procedures. The apparatus consisted of two enclosed arms with walls of different visual contexts positioned at a 90° angle to each other, and connected to a center zone (Embrechts and Ver Donck, 2014). A top-mounted video camera recorded the movements of the animals, and images were analyzed for distance made and time spent in each arm using EthoVision XT 9.0 (Noldus).
Two hours after dosing with vehicle or test compound, animals were subjected to a habituation trial (T1), during which one arm was closed and the other arm was free to be explored for 5 minutes [defined as the familiar arm (FAM)]. Then during the retention trial (T2), the novel arm was opened (NEW arm), and the animal explored both arms of the maze for 5 minutes. The discrimination index indicating preference for the novel arm was calculated from the time spent in each arm of the maze during the retention trial: discrimination index = (NEW − FAM)/(NEW + FAM).
Statistical analyses were performed using “InVivoStat” software (Clark et al., 2012; http://invivostat.co.uk/). The discrimination index was analyzed using single measures parametric analysis, and the overall effects were determined using ANOVA followed by all-to-one comparisons without adjustment for multiplicity (Fisher's least-significant differences tests).
Delayed Nonmatch to Position
Standard operant chambers (Med Associates, St. Albans, VT) were used. One wall was equipped with two retractable response levers and stimulus lights. The opposite wall contained the reward magazine, equipped with a reward light and an infrared sensor. In addition, each box was equipped with a small “house” light, a small speaker, and a metal grid floor. Forty-five-milligram dustless precision pellets (standard chow; Bio-Serv, Flemington, NJ) were used as reward.
Animals (male Lister-Hooded rats, aged approximately 18 months and weighing approximately 350–450 g at the time of testing; Harlan Laboratories, Horst, The Netherlands) received treatment within a counter-balanced design (0, 1, 3, and 10 mg/kg JNJ-55511118, suspended in 0.5% HPMC, 10 ml/kg p.o., 120 minutes prior to testing). During testing, the rats experienced 15 trials at four different delays (1, 10, 20, and 30 seconds) per session. A session ended once animals completed all 60 trials or after 45 minutes. A trial started with the illumination of the magazine, followed by a nose poke at the magazine, turning off the receptacle light, and starting the sample phase. One of two levers was extended in a pseudorandom fashion, and the rat had 20 seconds to respond. This caused the lever to be retracted and started the delay phase (variable 1–30 seconds). Once the delay had passed, the rat had 10 seconds to respond at the receptacle, causing the magazine light to go off and the two levers to be extended (choice phase). A response at the lever opposite the sample lever within a limited hold of 10 seconds was counted as a correct response, and led to the retraction of the levers, magazine illumination, a short tone, and delivery of a food pellet. Pellet collection turned the magazine light off and started a 5-second intertrial interval. A response to the same lever that was presented during the sample phase was counted as an incorrect response and resulted in the retraction of the response levers, a short “time out” (10 seconds) in darkness. A lack of response during the sample or choice phases is counted as an omitted trial.
Percentage correct served as a measure of working memory. The relative number of omitted trials and various response latencies served as measures of responsivity. Delayed nonmatch to position (DNMTP) percentage correct data were fitted to repeated-measures logistic regression mixed-effect models, latencies were analyzed using mixed-effect models, and percentage of errors of omission was determined by a logistic regression model. Main treatment effects were then analyzed by ANOVA followed by post-hoc contrasts.
Results
JNJ-55511118 and JNJ-56022486 were discovered through directed medicinal chemistry following a high-throughput screening campaign targeting AMPA receptors containing TARP-γ8. The structures of these molecules are shown in Fig. 1.
Calcium Flux.
JNJ-55511118 and JNJ-56022486, along with the nonselective AMPAR inhibitors CP-465022 and GYKI-53655, were evaluated for their ability to inhibit glutamate-evoked calcium flux in HEK-293 cells heterologously expressing GluA subunits with and without TARPs. Inhibition as a function of concentration for these compounds in assays using various combinations of human GluA and TARP subunits is shown in Fig. 2. The fitted values for the potency of inhibition at each target, averaged over multiple experiments, are summarized in Table 1. JNJ-55511118 and JNJ-56022486 potently inhibited every tested GluAx subunit, provided TARP-γ8 was present. Both compounds showed little inhibitory activity up to the highest concentrations tested at TARP-less AMPARs and at AMPARs coexpressed with other TARPs or with cornichon family AMPA receptor auxiliary protein 2 (CNIH2). In contrast, the noncompetitive inhibitors CP-465022 and GYKI-53655 showed no selectivity among AMPA receptor subtypes, or among the TARP-containing AMPARs.
To explore the molecular basis of the selectivity of these compounds, we constructed chimeric proteins of TARP-γ8 and -γ4. Functional interaction with the AMPA receptor was established by determining the potency of inhibition of GluA1o coexpressed with the chimeric TARPs. Figure 3A shows the topology and nomenclature of the constructs used in these experiments. Figure 3C shows the change in potency of inhibition for each compound as compared with the potency at GluA1o-γ8. In the first group of chimeras, we interchanged EX1 and the CT of TARP-γ8 and -γ4. Neither exchange altered the potency or efficacy of the compounds: chimeras 448444444 and 444444448 (γ4 with only EX1 or CT replaced) were insensitive, whereas 884888888 and 888888884 (γ8 with only EX1 or CT replaced) were potently inhibited.
In the next set of chimeras (888888884–844444444), we progressively replaced sections of γ8 with the corresponding regions of γ4, starting from the C terminus. JNJ-55511118 and JNJ-56022486 both lost their ability to inhibit the response beginning with chimera 888888844, implicating TM4 in the functional activity of the compounds. Indeed, chimera 888888848 was completely insensitive to inhibition. The inverse chimera 444444484 was sensitive to these compounds, with a reduced potency.
In chimeras 488888888–444444488, we progressively replaced domains of TARP-γ8, starting from the N terminus. The potency of JNJ-55511118 and JNJ-56022486 was unchanged in chimeras 488888888–444448888. However, when TM3 was replaced (444444888), the compounds lost approximately 10-fold in potency. This suggests that TM3 is also involved in the functional activity of the compounds, although somewhat lower in magnitude than TM4.
We aligned the sequences for the human TARPs to identify candidate residues that could account for the pharmacology of these compounds (Fig. 3B). TM4 contains four amino acids unique to TARP-γ8, and TM3 contains three. We generated point mutations of γ8, in each case mutating individual residues to the corresponding one from γ4. Figure 3D shows the change in potency of inhibition for GluA1o coexpressed with each of these constructs. Two of these point mutations showed altered potency of the TARP-selective compounds: G210A and V177I. G210A completely abolished activity of the compounds, whereas V177I caused a 10-fold loss of potency. Double mutations of γ4 and γ2 in the corresponding locations to the γ8 residues conferred sensitivity of those TARPs to inhibition by JNJ-55511118 and JNJ-56022486.
To identify additional residues which may be involved with the functional activity, we scanned TARP-γ8 in the vicinity G210 and V177. We made single-point alanine mutations from N173 through G176 in TM3, and G209 through I214 in TM4. Most of these point mutations, when coexpressed with GluA1o, did not alter the potency of the γ8-selective inhibitors. N173A caused a complete loss of inhibitory efficacy, whereas the compounds were 10- to 100-fold less potent with G209A and F213A.
Selectivity.
The selectivity of JNJ-55511118 and JNJ-56022486 was evaluated at 1 µM against a panel of 52 receptors, ion channels, and transporters using radioligand displacement assays (Cerep S.A., Poitiers, France). Data are summarized in Supplemental Table 9. The compounds showed less than 50% binding at all tested targets, except for activity of JNJ-55511118 at the serotonin receptor 2B (78% effect) and JNJ-56022486 at the melatonin receptor (57% effect). JNJ-55511118 was further evaluated in a cell-based functional assay using the recombinant human serotonin receptor 2B, and was determined to be an antagonist at this receptor, with an IC50 of 6 µM (data not shown). Thus, both compounds were a minimum of 100-fold selective against all tested targets.
Electrophysiology.
The AMPAR modulation of JNJ-55511118 was explored in further detail using electrophysiological recordings. Figure 4A shows representative traces of the glutamate-evoked responses of cells expressing AMPA receptors. In outside-out patches of Chinese hamster ovary (CHO) cells expressing GluA1o-γ8, a saturating concentration of JNJ-55511118 produced a partial inhibition. Peak currents with 1 µM JNJ-55511118 were reduced to 57.2 ± 2.0% (mean ± S.E.M.; N = 7) relative to currents in the same patches prior to addition of the compound. In contrast, peak currents in patches of CHO cells expressing GluA1o-γ2 were virtually unaffected by the presence of 1 µM JNJ-55511118 (97.6 ± 2.0%, N = 8).
We used cultured cerebellar granule cells from neonatal mouse as our model system for native AMPA receptors expressing TARP-γ2, and acutely dissociated hippocampal neurons from adult mouse for native AMPA receptors expressing TARP-γ8. Figure 4A shows representative traces of the glutamate-evoked responses of outside-out patches from cultured cerebellar neurons and whole-cell currents from acute hippocampal neurons. Analogous to the results with heterologously expressed AMPA receptors, 1 µM JNJ-55511118 reduced peak glutamate-evoked currents in hippocampal neurons to 60.7 ± 2.6% (N = 6) relative to control, with virtually no effect on cerebellar currents (98.5 ± 1.7%, N = 7). Supplemental Fig. 4 shows a summary of the effect of the glutamate-evoked peak current by JNJ-55511118 in a variety of cell types; no inhibition was seen in cells expressing GluA1o, GluA1o-γ2, GluA1i+γ2, cultured cerebellar granule cells, or hippocampal neurons from TARP-γ8 knockout mice.
Figure 4B shows the glutamate-evoked peak current as a function of concentration of JNJ-55511118 in outside-out patches from CHO cells expressing human GluA1o-γ8 and from acutely dissociated mouse hippocampal neurons. In both cell types, the peak currents were partially inhibited at a saturating concentration. By nonlinear least-squares fitting to a logistic function with slope fixed to unity, the maximal inhibition was 55.8 ± 1.9%, and the midpoint was 3.8 ± 1.0 nM for GluA1o-γ8. For hippocampal neurons, the maximal inhibition was 60.1 ± 3.4%, and the midpoint was 15.0 ± 4.6 nM.
To assess the activity and selectivity of JNJ-55511118 during synaptic transmission, we recorded field excitatory postsynaptic potentials (fEPSPs) from the CA1 region of hippocampal slices from wild-type and TARP-γ8 knockout animals (Fig. 4C). We did not observe an effect of JNJ-55511118 (1 µM) on the fEPSPs in CA1 of the hippocampus from TARP-γ8 knockout animals (105.3 ± 6.9% of baseline; N = 5), whereas the fEPSPs were inhibited in slices from wild-type littermates (63.4 ± 11.0% of baseline; N = 6). We used whole-cell patch clamp in pyramidal CA1 neurons from hippocampal slices to test the activity of JNJ-55511118 (1 µM) on synaptic AMPAR currents. Electrical stimulation of Schaffer collaterals evoked AMPAR-mediated EPSCs which were inhibited by JNJ-55511118 to 73.0 ± 4.0% of control (N = 10; P = 0.001 by paired-sample Student’s t test) (Supplemental Fig. 5A). This effect appeared to be specific to postsynaptic AMPA receptors since we did not observe changes in the paired-pulse ratio or NMDA EPSCs (123 ± 16% of control; N = 10; P = 0.59 by paired-sample Student’s t test) (Supplemental Fig. 5, C and D). We observed reduced synaptic summation when stimulated at 50 kHz in the presence of JNJ-55511118 (Supplemental Fig. 5B).
To investigate the effect of JNJ-55511118 on desensitization of γ8-containing AMPA receptors, we used 500-ms applications of 10 mM l-glutamate on outside-out patches of HEK-293 cells expressing heterologous GluA1i+γ8 (Fig. 4D). JNJ-55511118 (1 µM) reduced peak currents to 63.8 ± 4.1% of control and steady-state currents to 25.4 ± 3.7% of control (N = 6). In the presence of cyclothiazide, which removes desensitization, 1 µM JNJ-55511118 inhibited glutamate-evoked currents to 59.5 ± 1.7% of control (N = 6; Supplemental Fig. 6). Inhibition in the presence of cyclothiazide was not significantly different from inhibition of peak currents (P = 0.52), whereas it was significantly reduced compared with inhibition of the steady-state current (P < 0.001; one-way repeat measures ANOVA followed by Tukey’s test).
In addition to reducing peak and steady-state currents, the presence of 1 µM JNJ-55511118 accelerated the rate of desensitization, which can be observed by normalizing to the peak currents (Fig. 4D, middle panel). After application of 10 mM l-glutamate, GluA1i+γ8 desensitized with a time constant of 7.9 ± 0.8 ms (N = 6). In the presence of 1 µM JNJ-55511118, the time constant decreased to 5.7 ± 0.5 ms (P < 0.001 by paired-sample Student’s t test). In the absence of TARP, GluA1i desensitized considerably faster, with a time constant of 3.5 ± 0.3 ms (N = 5). We determined the recovery from desensitization by measuring the peak current in response to paired applications of 10 mM l-glutamate, separated by a variable delay (Fig. 4E). Recovery time constants were similar for GluA1i+γ8 in the absence or presence of JNJ-55511118 (control τr,desens = 93 ± 29 ms, N = 7; JNJ-55511118 τr,desens 102 ± 33 ms, N = 6; P = 1). Receptors in patches from cells expressing TARP-less GluA1i recovered more slowly than GluA1i+γ8 (GluA1i τr,desens = 163 ± 10 ms, N = 8; P = 0.02 by Student’s t test).
We investigated the effects of JNJ-55511118 on the deactivation of TARP-γ8–containing AMPA receptors. Fast application (1 ms) of glutamate to excised outside-out patches generated rapidly deactivating currents. Figure 4F (top panel) shows that JNJ-55511118 partially inhibited the glutamate-evoked peak currents to 67 ± 2% of control (N = 9). Normalization of peak currents shows that JNJ-55511118 increased the deactivation rate (Fig. 4F, middle panel). After application of 10 mM l-glutamate, GluA1i+γ8 deactivated with a time constant of 4.6 ± 0.9 ms (N = 9). In the presence of 1 µM JNJ-55511118, the time constant decreased to 3.1 ± 0.6 ms (P = 0.001 by paired-sample Student’s t test). In the absence of TARP, GluA1i deactivated considerably faster, with a time constant of 1.2 ± 0.3 ms (N = 6).
To test for the possibility that JNJ-55511118 may disrupt the interaction between TARP γ-8 and AMPA receptors, we looked at the effects of JNJ-55511118 on kainate efficacy. TARPs enhance kainate efficacy for AMPA receptors (Tomita et al., 2005; Turetsky et al., 2005); the ratio of response to kainate versus glutamate is a sensitive assay for TARP/AMPAR stoichiometry (Shi et al., 2009). As previously reported, TARP-less AMPA receptors had a considerably lower kainate/glutamate ratio of steady-state currents when compared with γ8-containing AMPARs (Fig. 4G, top panels). In the absence of TARP-γ8, the kainate/glutamate ratio for GluA1i was 0.21 ± 0.04 (N = 4), whereas the kainate/glutamate ratio for GluA1i-γ8 was 5.1 ± 1.1 (N = 9). In the presence of JNJ-55511118 (Fig. 4G, bottom-left panel), the ratio was 4.2 ± 0.7 (N = 9). JNJ-55511118 failed to significantly change the kainate/glutamate ratio of the steady-state currents (P = 0.58 by paired-sample Student’s t test); thus, the inhibition observed with JNJ-55511118 does not involve the dissociation of the TARP from the receptor complex.
Binding Assays.
We incorporated tritium into JNJ-55511118 and JNJ-56022486 to explore their utility in radioligand binding assays. [3H]JNJ-56022486 proved to have lower nonspecific binding than [3H]JNJ-55511118, likely due to its lower lipophilicity and higher free fraction in brain tissue (see Supplemental Table 10). Therefore, we focused on [3H]JNJ-56022486 for additional binding studies.
Saturation binding in membranes from rat hippocampus using [3H]JNJ-56022486 is shown in Fig. 5A. The fitted value of the binding affinity was 27 ± 3 nM, with Bmax = 3.8 ± 0.3 pmol/mg protein. In competition binding experiments (Fig. 5B), JNJ-55511118 and JNJ-56022486 fully displaced the radioligand (20 nM) with Ki = 26 ± 7 and 19 ± 6 nM, respectively (N = 3). Neither Philanthotoxin-74 nor glutamate showed appreciable displacement of [3H]JNJ-56022486, whereas LY-395153 [a positive allosteric modulator (PAM)] and perampanel partially displaced the radioligand (Fig. 5B). Displacement by LY-395153 required the presence of 500 µM glutamate; the PAM did not displace the radioligand in the absence of glutamate (data not shown).
Autoradiograms showing total binding of [3H]JNJ-56022486 in brain slices from mouse, rat, and monkey are shown in Fig. 5, D–H. An image from the Allen Brain Atlas of the expression of CACNG8 in a corresponding coronal brain slice from an adult mouse (Lein et al., 2007) is shown in Fig. 5C. These images indicate a high concentration of specific binding that corresponds well to the expression pattern of CACNG8.
Pharmacokinetics.
We determined the pharmacokinetic and in vivo target occupancy profiles for JNJ-55511118 and JNJ-56022486. JNJ-55511118 achieved high plasma concentrations upon oral (p.o.) and i.v. dosing (Fig. 6A). The compound was orally bioavailable in both species. In vivo clearance and volume of distribution in rats were 4.8 ml/min/kg and 1.8 l/kg, respectively. Target occupancy was determined by ex vivo autoradiography of brain slices of animals dosed with test compound, using [3H]JNJ-56022486 to probe for unoccupied receptors. JNJ-55511118 was highly brain-penetrant, and showed high target occupancy upon oral dosing in both rat and mouse (Fig. 6, B and C). This compound also displayed linear exposure as a function of dose (Fig. 6D). Although JNJ-56022486 also showed good oral bioavailability, the brain penetration and target occupancy at 10 mg/kg were substantially lower compared with JNJ-55511118 (Supplemental Fig. 7). Therefore, we focused on JNJ-55511118 for additional in vivo studies. In vitro measurements of tissue binding of JNJ-55511118 showed 1.48 and 0.88% free fraction in rat and mouse plasma, respectively, and 0.24% free fraction in rat brain tissue. Supplemental Table 10 shows a summary of the pharmacokinetic parameters derived for these two compounds.
In Vivo Electrophysiology.
For a more direct measurement of the functional consequences of target engagement, we performed in vivo electrophysiological recordings in the pyramidal cell layer of CA1 in the rat hippocampus, while stimulating the Schaffer collateral projections from CA3. When dosed intravenously, JNJ-55511118 showed a rapid, dose-dependent inhibition of the evoked population spike amplitude, with a fitted half-maximal effective dose (ED50) of 0.4 mg/kg (Fig. 7, A and B). One hour after dosing, recordings were terminated; plasma and brain tissue were harvested for bioanalysis to measure the concentration of JNJ-55511118. Fig. 7C shows the population spike inhibition at 1 hour postdosing as a function of the measured brain concentration for each animal. Superimposed in this graph is the receptor occupancy as a function of brain concentration for rat from the experiments shown in Fig. 6E. The calculated half-maximal effective concentration (EC50) values for these two assays were quite similar (591 ng/ml, in vivo electrophysiology; 545 ng/ml, autoradiography).
EEG.
To investigate the effect of functional γ8-selective inhibition on the cortical oscillations in freely behaving rats, sleep-wake architecture and EEG power spectral density were evaluated after oral administration of JNJ-55511118 in a dose-response experiment. A group of seven animals were orally dosed at 2 hours into the light phase with vehicle (0.5% HPMC) or JNJ-55511118 (1, 3, and 10 mg/kg; 0.5% HPMC suspension) in a randomized crossover design. EEG/EMG signals and locomotor activity were recorded for up to 20 hours after each pharmacological treatment; here, we show the effects during the first 8 hours after administration.
Oral administration of JNJ-55511118 elicited a clear dose-dependent decrease in EEG activity during the wake state in all frequency bands above 4 Hz (Fig. 7G). Specifically, a dose-related reduction of the power spectral density in the theta (4–10 Hz) [F(3, 18) = 26.73, P < 0.001], alpha (10–15 Hz) [F(3, 18) = 35.80, P < 0.001], and beta (15–30 Hz) [F(3, 18) = 168.10, P < 0.001] oscillations was observed from the lowest dose tested. The power spectral density in the delta oscillations was minimally affected, and only a small but significant decrease was revealed at a dose of 1 mg/kg. Consequently, the contribution of the EEG delta activity over the total power (relative delta power) was enhanced, since the absolute values for power spectra in all of the higher frequency bands were decreased.
The 10-mg/kg dose, which should have achieved approximately 90% target occupancy (see Fig. 6F), represents the near-saturating drug effect. At this dose, power in the wake state in the theta, alpha, and beta bands was reduced to 81.2 ± 1.0, 70.6 ± 1.3, and 63.8 ± 1.2%, respectively, relative to control. This substantial reduction in EEG power was accompanied by a significant increase in locomotor activity for approximately 1 hour postdose (Fig. 7H). In addition, there was a dose-dependent increase in the latency to REM and NREM sleep (Supplemental Fig. 8).
JNJ-55511118 also produced dose-dependent decreases in absolute EEG power during REM and NREM sleep states (Supplemental Fig. 9). During NREM sleep, EEG oscillations were significantly reduced from the 1-mg/kg dose onward in all measured frequency bands (1–30 Hz). During REM sleep, delta (1–4 Hz), theta (4–10 Hz), and beta (15–30 Hz) EEG oscillations were significantly reduced from the dose of 1 mg/kg onward. In contrast, EEG oscillations in the sigma frequency range (10–15 Hz) were minimally affected.
Anticonvulsant Assays.
We explored the anticonvulsant profile of JNJ-55511118 using several in vivo models. Figure 7, D and E shows the dose-response relationship of the activity in the corneal kindling and 6 Hz models, as well as the rotarod test. Based on the target occupancy in mouse shown in Fig. 6F, target occupancy should be well above 90% at doses above 40 mg/kg. Any additional pharmacology above that dose would likely be due to off-target effects. In the corneal kindling model, JNJ-55511118 provided near-complete seizure protection at and above 5 mg/kg (p.o., 1 hour postdosing). Twenty-five out of 32 animals had Racine scores of zero at doses of 5–40 mg/kg, indicating the complete absence of observable effects of the stimulus. In comparison, 8/8 animals in the vehicle control group had Racine scores of 5 (rearing and falling with forelimb clonus). The curve fits to the corneal kindling data in Fig. 7, D and E indicate that ED50 = 3.7 mg/kg, with a brain concentration EC50 of 938 ng/ml.
JNJ-55511118 showed partial protection in the 6 Hz models. At doses of 10 mg/kg and above, projected to give target occupancies of 80% or greater, 12/32 (37.5%) animals were protected in the 6 Hz 32 mA model, and 11/40 (27.5%) were protected in the 6 Hz 44 mA model. This compound was also tested in the MES model; at 40 mg/kg p.o., 50% of the animals showed seizure protection (N = 8; data not shown). The curve fits to the 6 Hz 32 mA data indicate that ED50 = 18.3 mg/kg, with a plasma concentration EC50 of 4644 ng/ml and 76% maximum protection. For the 6 Hz 44 mA test, ED50 = 6.5 mg/kg, with a plasma concentration EC50 of 2533 ng/ml and 22% maximum protection.
The Metrazol (PTZ) test was performed at t = 0.5 hour after dosing with 40 mg/kg JNJ-55511118. The compound showed strong protection (Fig. 7F); the mean threshold for clonus increased from 33.3 ± 4.9 to 49.5 ± 10.0 mg/kg PTZ (P < 0.001, two-sample t test), and for twitch, from 30.7 ± 4.6 to 42.4 ± 7.5 mg/kg (P < 0.001, two-sample t test).
JNJ-55511118 was also tested in the amygdala kindling model in rats. This compound protected 11/12 rats from seizure (2 hours postdose, 10 mg/kg p.o., 0.5% HPMC suspension), compared with 1/12 rats protected with vehicle alone. The after-discharge duration, shown in Supplemental Fig. 10, was significantly reduced by JNJ-55511118 to 35 ± 7 seconds, compared with 92 ± 12 seconds with vehicle alone (mean ± S.E.M., N = 12; P = 4 × 10−4, two-sample t test).
Effects in Models of Learning/Memory/Cognition.
Inhibition of AMPA receptors, particularly in the hippocampus and prefrontal cortical areas, might be expected to impact hippocampal plasticity and learning/memory mechanisms important for performance on spatial working memory tasks.
The Morris water maze (MWM) provides a key method to investigate spatial learning and memory in rodents (D'Hooge and De Deyn, 2001). Figure 8, A and B shows that vehicle-treated animals obtained a typical learning curve during the 4 training days evidenced by shortening of latency times (Fig. 8A) and path length (Supplemental Fig. 11) to reach the platform. Based upon the occupancy as a function of dose shown in Fig. 6E, the dose range of 0.63, 2.5, and 10 mg/kg p.o. should have given target occupancies of 50–90%. Surprisingly, the performance was not severely impacted; at the 0.63- and 2.5-mg/kg doses during the training days, there was no statistically significant deviation from performance of vehicle-treated animals, apart from a small transient but statistically significant improved latency time to platform on day 2 (which was not confirmed in an independent repeat study; data not shown). The animals in the 10-mg/kg cohort showed a modest deficit in learning the platform location (increased time and path length to locate the platform) on the first (P = 0.05, Cox proportional hazards model) and second (P = 0.04) training days; however, on the third and fourth training days, their performance was indistinguishable from vehicle-treated animals.
In the DNMTP test, there was an overall effect of dose (P < 0.01), with JNJ-55511118 decreasing the percentage of correct responses at the two highest doses tested (3 and 10 mg/kg; P < 0.05 and P < 0.01, respectively), although the effect was rather moderate (approximately 5% reduction; Fig. 8C). This was accompanied by a small, but significant, reduction in responsivity, as indicated by an increase in errors of omission and response latencies (Supplemental Fig. 12). Although there was no significant drug × delay interaction (P > 0.05), post-hoc contrasts were computed against the vehicle group for the percentage of correct responses, showing significant effects of the two higher doses at the shortest delay (1 second), but not at longer delays (Fig. 8D). All tested doses increased the percentage of trials omitted, and the two highest doses tested induced a significant increase on latency measures (see Supplemental Fig. 12). The magnitudes of these effects were generally modest (less than 0.5 second for most measures).
The V-maze was used to study the effects of JNJ-55511118 on working memory in rodents. This paradigm exploits the natural tendency of rodents to explore a novel, nonthreatening environment. The V-maze avoids the use of aversive test conditions, such as electric shocks or deprivation, that may have nonspecific influences on the responses, and does not require prior learning of a rule.
Vehicle-treated animals show a clear preference for the new arm over the familiar arm they explored just before (discrimination index = 0.40 ± 0.06; Fig. 8E). A partial impairment of this parameter is found in the 10-mg/kg treatment group (P = 0.0055, ANOVA). Concomitantly, a small increase in distance moved is seen in a dose-dependent manner in the habituation phase (P < 0.0001) and at a dose of 10 mg/kg only in the test phase (P = 0.0075; Fig. 8F).
Discussion
AMPA receptor signaling is a critical component of normal and pathophysiological excitatory neuronal function, and has been an attractive pharmacological target for both positive and negative modulation. To date, all negative modulators of AMPA receptors target the pore-forming GluA subunits. Although the presence of TARPs can modify potency and efficacy of modulators (Cokic and Stein, 2008), no published reports have shown any selectivity of inhibitors among the TARPs. Here, we report a novel pharmacological approach to negatively modulate AMPA receptor signaling, with molecules selective for TARP-γ8.
JNJ-55511118 and JNJ-56022486 are highly potent, and are exquisitely selective for AMPA receptors containing TARP-γ8, with no detectable functional effect on TARP-less AMPA receptors or those containing TARP-γ2, -γ3, -γ4, or -γ7. The data presented here suggest that these compounds exert their effects through a novel mechanism of action, via partial disruption of a protein-protein interaction. JNJ-55511118 exhibits excellent pharmacokinetics and brain penetration, and achieves high-target occupancy upon oral and intravenous dosing. Tritiated JNJ-56022486 is useful as a radioligand for competitive binding studies, and for ex vivo autoradiography to determine target occupancy.
TARP Selectivity.
The γ8–γ4 chimeras were designed to locate the site of selectivity. Previous studies have demonstrated that the TARP C terminus and EX1 domains have strong effects on trafficking and functional interaction with the GluA subunits (Tomita et al., 2005; Turetsky et al., 2005; Cais et al., 2014). Thus, a drug interacting at one of these two regions could differentially impact the function of the AMPA receptor complex. Chimera pairs (448444444, 884888888) and (444444448, 888888884) directly probed this question. As shown in Fig. 3C, JNJ-55511118 and JNJ-56022486 inhibited 888888884 (γ8 with the C terminus from γ4) and failed to inhibit 444444448 (γ4 with the C terminus from γ8). This pair of results suggests that the C terminus is not involved with the pharmacological activity of these two compounds. The chimera pair (448444444, 884888888) showed a similar failure to alter the pharmacology of the backbone TARP, again suggesting that domain EX1 is not involved with the pharmacological activity. Instead, domain scanning and direct substitutions indicated that TM3 and TM4 govern the selectivity.
The point-mutation studies allowed us to determine that selectivity is entirely determined by two amino acids predicted to lie within TM3 and TM4. These two amino acids (G210 and V177) are unique to TARP-γ8. Altering these two amino acids to their corresponding components of TARP-γ4 completely abolished the potency of the compounds, whereas the TARP-γ8 versions of these same two amino acids added to both TARP-γ2 and TARP-γ4 confer identical sensitivity of these TARPs to the compounds. According to the predicted topology of the TARP (Fig. 3A), G210 and V177 are both 2–3 residues deep within the outer membrane surface.
Data and alignments were retrieved from the UniProt database (UniProt Consortium, 2015) for genes identified as CACNG8, from a selection of species. Sequence alignments of TM3 and TM4 for the TARP-γ8 in selected species are shown in Supplemental Fig. 13; these regions are highly conserved across species. Curiously, the two key amino acids that determine selectivity against TARP-γ2 and -γ4 (G210A and V177I) are also conserved across species (Supplemental Fig. 14). Indeed, we found no species differences in potency or selectivity in rat, mouse, dog, monkey, or human isoforms (Fig. 2, E and F; Table 1). These two key amino acid changes are relatively small; each involves adding a single methyl group in making the wild-type TARP-γ8 protein completely insensitive to JNJ-55511118. In addition to these two key residues, alanine scanning indicated that N173 in TM4 and G209 and F213 in TM3 influence the potency of the compounds. These amino acids are adjacent to, or one alpha helix turn away from, the key selectivity locations. These five amino acids may form or contribute to the binding site for these TARP modulators.
Mechanism of Action.
The radioligand binding assays showed that JNJ-55511118 and JNJ-56022486 fully displaced [3H]JNJ-56022486, with affinities comparable to the potencies as measured in the calcium flux assays. In contrast, ligands that bind to the agonist site (glutamate) and the pore (Philanthotoxin-74) showed no displacement. Ligands that bind to the GluA PAM site (LY-395153) and the noncompetitive antagonist site (perampanel), which couple allosterically when TARPs are associated with the AMPA receptor (Schober et al., 2011), partially displaced [3H]JNJ-56022486. Partial displacement implies allosteric coupling between these sites and the binding site for the TARP-γ8 modulators; further studies are required to investigate this phenomenon.
The electrophysiology experiments allowed detailed characterization of the effects of JNJ-55511118 upon the AMPA receptor currents. The results are consistent with a partial disruption of the interaction between TARP-γ8 and the pore-forming GluA subunits. Type I TARPs provide a positive modulatory influence upon the AMPA receptor (Howe, 2015). The electrophysiological studies showed that JNJ-55511118 reverses some, but not all, of the effects of the TARP on the AMPA receptor complex. A saturating concentration of JNJ-55511118 reduces peak responses by 36–44%, and steady-state currents by ∼75%. This reduction is consistent with masking the increases in single-channel conductance observed in studies of single-channel kinetics of TARPed versus TARP-less AMPA receptors (Shelley et al., 2012; Zhang et al., 2014). This inhibition is accompanied by an increase in the desensitization and deactivation kinetic rates.
Other aspects of the inhibition by JNJ-55511118 indicate that TARP-γ8 remains associated with the receptor complex. First, the kainate/glutamate ratio does not change as would be expected with changes in TARP stoichiometry (Shi et al., 2009). TARPs enhance kainate efficacy for AMPA receptors (Tomita et al., 2005; Turetsky et al., 2005), and the kainate-to-glutamate response ratio is a sensitive assay for TARP/AMPAR stoichiometry (Shi et al., 2009). As previously reported, TARP-less AMPA receptors had a considerably lower kainate/glutamate ratio of steady-state currents when compared with AMPA receptors containing TARP-γ8. JNJ-55511118 failed to significantly change the kainate/glutamate ratio of the steady-state currents, strongly suggesting that the inhibition observed with JNJ-55511118 does not involve the dissociation of the TARP from the receptor complex. Second, the deactivation and desensitization kinetics observed are considerably faster in the TARP-less receptors when compared with the TARP-γ8–containing receptors in the presence of JNJ-55511118. Third, JNJ-55511118 does not affect the recovery from desensitization, whereas the TARP-containing AMPA receptors recover from desensitization faster than TARP-less AMPA receptors (Priel et al., 2005). JNJ-55511118 may partially affect the GluA-TARP interaction required for reduced desensitization and deactivation, leaving intact those interactions required for kainate efficacy. Alternatively, TARP-γ8 may induce a unique conformational state on AMPARs not observed with other TARPs, enabling a binding site on AMPARs for the inhibitors to reduce open-channel probability.
These features suggest that JNJ-55511118 does not cause the TARP to dissociate from the AMPA receptor complex, but instead modifies the protein-protein interaction between the TARP and GluA subunits. In a simplified gating model containing single open, resting, and desensitized states (Sun et al., 2002), these effects are consistent with destabilization of the open state by JNJ-55511118. Additional kinetic studies comparing TARP-less AMPA receptors to those containing TARP-γ8 with and without JNJ-55511118 should reveal more details of the mechanism of action. In addition, mutations of TARP-γ8 at the key amino acids described earlier may provide additional mechanistic insight and structure-activity relationships regarding the interaction between TARPs and the GluA subunits.
Electrophysiological measurements in acutely dissociated hippocampal neurons indicate that the modulation by JNJ-55511118 closely recapitulates the behavior in heterologous systems. This suggests that, consistent with previous findings, native hippocampal AMPA receptors are highly TARPed (i.e., contain a high level of TARP proteins), primarily with TARP-γ8. It also indicates that the presence of the additional accessory proteins in the native system does not substantially alter the functional impact of the compound or the binding pocket.
In addition to their modulatory effects on gating and channel conductance, TARPs have multiple additional effects upon AMPARs: they modify surface expression (Rouach et al., 2005; Tomita et al., 2005); alter the pharmacology of the PAM, agonist, and noncompetitive antagonist binding sites (Tomita et al., 2005; Turetsky et al., 2005; Schober et al., 2011); and participate in anchoring AMPARs to the synaptic scaffolding (Chen et al., 2000). Whether JNJ-55511118 impacts these additional interactions between TARPs and AMPARs remains an open question. Indeed, this molecule may prove useful in determining the structure-function relationships of these diverse phenomena.
Synaptic Transmission.
We tested the activity of JNJ-55511118 in hippocampal slices, a preparation in which the native synaptic receptors are found in association with components of the postsynaptic densities and associated auxiliary subunits. Here, we found that JNJ-55511118 inhibited the synaptic responses in the hippocampal CA1 region from wild-type mice but not from TARP-γ8 knockout littermates. Consistent with specificity of JNJ-55511118 on postsynaptic AMPA receptors, the compound only inhibited the AMPA EPSCs from CA1 pyramidal neurons, but not the NMDA EPSCs, and did not affect the paired-pulse ratio. Taken together, these data suggest that JNJ-55511118 does not affect the presynaptic glutamate release probability, but instead acts directly on postsynaptic AMPA receptors containing TARP-γ8. We also observed reduced synaptic summation when stimulated at 50 kHz in the presence of JNJ-55511118 (Supplemental Fig. 5), suggesting a potential indication of this compound in states of hyperactive hippocampal activity, such as seizures. These results confirm the activity of JNJ-55511118 on native postsynaptic AMPA receptors under basal and high-frequency synaptic stimulation, and set the stage for a physiologic role of this compound in hippocampal activity.
In anesthetized rats, JNJ-55511118 caused near-complete inhibition of population spikes in CA1 driven by Shaffer collateral stimulation. The concentration dependence of population spike inhibition closely matched the target occupancy. The magnitude of inhibition of the population spike amplitude is somewhat surprising, given the partial inhibition of AMPA receptor current in in vitro electrophysiology (Fig. 4, A and B) and of the fEPSP slope in ex vivo slices (Fig. 4C). Partial inhibition of AMPA receptors to the extent observed with JNJ-55511118 is apparently sufficient to reduce the synaptic drive below the spiking threshold of the postsynaptic neurons.
Effects In Vivo.
Table 2 summarizes the results of behavioral tests performed using a saturating dose of JNJ-55511118. Quantitative EEG analysis in freely behaving rats during the wake state revealed a dose-dependent decrease in the power bands of theta, alpha, and beta activity. These data indicate that JNJ-55511118 produced some degree of EEG slowing, starting at a dose corresponding to 60% receptor occupancy. This effect is consistent with the anticonvulsant properties of the compound. JNJ-55511118 gave strong protection in the corneal kindling and PTZ models in mice, and in the amygdala kindling model in rats. Efficacy in the corneal kindling model was dose-dependent, and roughly approximated the target occupancy as a function of plasma concentration. The compound also showed partial protection in the 6 Hz and MES tests in mice.
Modulation of AMPA receptor signaling has long been considered an attractive strategy for the treatment of epilepsy (Rogawski, 2011). AMPAR antagonists show strong anticonvulsant activity in preclinical models, and the noncompetitive AMPAR inhibitor perampanel was recently approved as an adjunctive treatment of partial-onset seizures (Ko et al., 2015). The efficacy of topiramate may be mediated in part by modulation of phosphorylation of neuronal AMPA receptors. As with all known anticonvulsant medications, both of these drugs exhibit dose-limiting side-effect profiles that limit their clinical utility (Perucca and Gilliam, 2012). Modulation of AMPAR signaling using a TARP-γ8–selective mechanism has two distinct advantages that may result in an improved therapeutic margin. First, the expression of TARP-γ8 within the brain indicates that the drug will have its largest effect within the hippocampus, while avoiding direct inhibitory effects on brain regions involved with motor coordination and wakefulness. Second, the compounds negatively modulate but do not completely inhibit AMPAR signaling. This inherently limits the maximal effect of the drug.
AMPA receptors have been associated with regulation of hippocampal plasticity and short-term memory mechanisms important for performance on spatial and working memory tasks. GluA1 knockout mice show impaired short-term habituation to recently experienced stimuli, which has been suggested to affect performance in hippocampus-dependent spatial working memory tasks (Sanderson et al., 2009; Sanderson and Bannerman, 2012). Hippocampal lesions affect learning and memory in both human (Manns et al., 2003) and nonhuman species (Morris et al., 1982). Similarly, knockout of TARP-γ8 reduces hippocampal AMPA receptor function and synaptic plasticity (Rouach et al., 2005; Fukaya et al., 2006), although the impact of learning, memory, and cognition in these transgenic animals has not yet been reported.
We addressed this issue by testing JNJ-55511118 in rats using the delayed nonmatch to position, V-maze, and Morris water maze tests. Performance in MWM is severely attenuated by manipulations that negatively impact hippocampal signaling (Morris et al., 1982; Riedel et al., 1999). DNMTP is generally thought to be dependent upon hippocampal function (Steckler et al., 1998a), with degradation in performance following hippocampal lesions (Chudasama and Muir, 1997; Winters and Dunnett, 2004) and direct intrahippocampal infusion of drugs affecting cholinergic, GABAergic, or glutamatergic function (Robinson and Mao, 1997; Mao and Robinson, 1998). Moreover, numerous studies have found that DNMTP is sensitive to systemic glutamatergic manipulations (Steckler et al., 1998b; Smith et al., 2011), although effects of AMPA receptor blockade in the delayed match to position test, a closely related paradigm to DNMTP, has been reported to be more subtle than effects of NMDA receptor blockade (Stephens and Cole, 1996).
In the MWM, animals in the highest-dose cohort showed attenuated learning within training days 1 and 2 (as evidenced by nearly flat learning curves), but it is remarkable that the latency time during the first trial on the second and third training day is close to that of the vehicle group on the last trial of the day before, suggesting that the animals consolidate memories during the nights between the training sessions. This is consistent with absence of treatment effects on performance during the probe trial: all treatment groups showed indistinguishable preference for the area in the pool where the platform was located during the training days. JNJ-55511118 showed impairment in V-maze, MWM, and DNMTP. However, compared with effects following systemic administration of the NMDA receptor antagonists dizocilpine or phencyclidine (Willmore et al., 2001), or the muscarinic antagonist scopolamine (Chudasama and Muir, 1997), the impairment induced by JNJ-55511118 was relatively small. Our data are in line with those reported for GluA1 knockout mice, which also show relatively mild learning and memory impairments compared with animals with hippocampal lesions (Sanderson and Bannerman, 2012).
Even at saturating doses, JNJ-55511118 showed a benign side-effect profile, with no loss of motor coordination or sedation (Fig. 7; Supplemental Fig. 8). Overall behavior of animals dosed with JNJ-55511118 appeared largely normal, with only transient hyperlocomotion immediately after dosing and upon transfer into a novel environment. Considering the robust anticonvulsant profile, the strong inhibition of EEG signals, and the expression of the target within the hippocampus and cortex, the relatively mild impact upon learning and memory in the Morris water maze, DNMTP, and V-maze assays is quite surprising. Thus, TARP-γ8 inhibition with molecules such as JNJ-55511118 shows strong potential clinical utility as an anticonvulsant, particularly for those forms of epilepsy with a strong hippocampal component, such as temporal lobe epilepsy.
In summary, TARP-γ8 modulators represent a novel pharmacological class of molecules which possess an unprecedented mechanism of action: partial disruption of a protein-protein interaction between the pore-forming GluA subunit of AMPA receptors and the TARP-γ8 accessory protein. These compounds provide important tools at several levels: 1) the molecular pharmacology of this interaction, 2) dissection of the structure-function relationship of the GluA-TARP interaction, and 3) the in vivo behavioral and therapeutic potential of partial inhibition of hippocampal excitability. Neuropsychiatric disorders can be viewed as pathologic disruption of the excitation/inhibition balance of specific structures, circuits, or sets of neurons. TARP-γ8 modulators have the potential to be transformational anticonvulsants, particularly for medial temporal lobe epilepsy. By avoiding the midbrain and hindbrain, AMPA receptor modulators selective for TARP-γ8 may attenuate seizures without side effects such as ataxia and sedation that are often seen with less-selective anticonvulsants. Other clinical applications for this mechanism include schizophrenia, particularly in the early stages where excessive limbic activity has been observed (Schobel et al., 2013), and anxiety disorders, given the reciprocal relationship between hippocampal function and the effects of anxiolytic drugs (Gray and McNaughton, 2003).
The existence of this mechanism of action has the potential to greatly expand the number of druggable targets. There are at least 30 additional proteins in the AMPAR proteome (Schwenk et al., 2012), each with unique expression profiles and functional impact upon the AMPA receptor complex. Thus, just as TARP-γ8 shows tissue specificity for the hippocampus, it may be possible to tune the effects of a drug targeting other subunits for specific brain regions or neuronal subtypes. Beyond AMPA receptors, most ion channels, and indeed drug targets from other protein classes, associate with accessory proteins, some of which show expression patterns more specific than the pore-forming subunit.
Acknowledgments
The authors acknowledge the contributions of the following individuals associated with Janssen Research and Development for their contributions to the previously described experiments. Ning Qin cloned several GluA and TARP constructs. Raymond Rynberg developed formulations for the in vivo studies. Nancy Aerts and John Talpos performed the DNMTP experiments. Steven Sutton managed the back-crossing of the transgenic mice. Sofie Embrechts performed the water maze and V-maze experiments. Tom Van de Casteele performed the statistical analysis of the behavioral data from learning/memory assays. Caroline Lanigan advised on statistical analyses. Leslie Nguyen, Minerva Batugo, and Brian Scott performed the bioanalytical studies. The authors also thank the following individuals associated with NeuroAdjuvants, Inc. for performing the anticonvulsant experiments: H. Steve White, University of Utah (study oversight and consultation, study reporting); Cameron S. Metcalf, NeuroAdjuvants, Inc. (6 Hz, MES, corneal kindling, amygdala kindling, study design and management, data analysis and reporting); Misty D. Smith (study review/quality control, tissue collection); Timothy Pruess, University of Utah (i.v. Metrazol test); Fabiola Vanegas, University of Utah (novel object and open field assays); Jenny Huff, University of Utah (amygdala kindling); Carlos H. Rueda, University of Utah (surgical amygdala electrode implantation). Staff members at BioDuro, LLC (Beijing, China) performed the pharmacokinetic studies in rats and mice.
Authorship Contributions
Participated in research design: Maher, Ameriks, Savall, Liu, Dugovic, Wickenden, Carruthers, Lovenberg.
Conducted experiments: Maher, Wu, Ravula, Liu, Lord, Wyatt, Matta, Dugovic, Yun.
Contributed new reagents or analytic tools: Ravula.
Performed data analysis: Maher, Wu, Liu, Lord, Wyatt, Matta, Dugovic, Ver Donck, Steckler.
Wrote or contributed to the writing of the manuscript: Maher, Ameriks, Liu, Lord, Wyatt, Matta, Dugovic, Ver Donck, Steckler.
Footnotes
- Received December 23, 2015.
- Accepted March 11, 2016.
↵This article has supplemental material available at jpet.aspetjournals.org.
Abbreviations
- ACSF
- artificial cerebrospinal fluid
- AMPA
- α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionic acid
- AMPAR
- AMPA receptor
- ANOVA
- analysis of variance
- CHO
- Chinese hamster ovary
- CP-465022
- 3-(2-Chlorophenyl)-2-[2-[6-[(diethylamino)methyl]-2-pyridinyl]ethenyl]-6-fluoro-4(3H)-quinazolinone hydrochloride
- CNS
- central nervous system
- CT
- carboxyl terminus
- DMSO
- dimethylsulfoxide
- DNMTP
- delayed non-match to position
- EC50
- half-maximal effective concentration
- ED50
- half-maximal effective dose
- EEG
- electroencephalogram
- EMG
- electromyogram
- EPSC
- excitatory postsynaptic current
- EX
- extracellular domain
- FAM
- familiar arm
- fEPSP
- field excitatory postsynaptic potential
- fu
- unbound fraction
- GluA
- AMPA subtype of ionotropic glutamate receptor
- GYKI-53655
- 1-(4-Aminophenyl)-3-methylcarbamyl-4-methyl-3,4-dihydro-7,8-methylenedioxy-5H-2,3-benzodiazepine hydrochloride
- HABG
- HibernateA supplemented with B27 and Glutamax
- HEK-293
- human embryonic kidney 293
- HPMC
- hydroxypropyl methylcellulose
- J values
- indirect dipole-dipole coupling constants
- JNJ-55511118
- 5-[2-chloro-6-(trifluoromethoxy)phenyl]-1,3-dihydrobenzimidazol-2-one
- JNJ-56022486
- 2-(3-chloro-2-(2-oxo-2,3-dihydro-1H-benzo[d]imidazol-5-yl)phenyl)acetonitrile
- LC-MS/MS
- liquid chromatography–tandem mass spectrometry
- LY-395153
- N-[4-[1-(propan-2-ylsulfonylamino)propan-2-yl]phenyl]benzamide
- MES
- maximal electroshock
- MWM
- Morris water maze
- NEW
- novel arm
- NIH
- National Institutes of Health
- NMDA
- N-methyl-d-aspartate
- NREM
- non-rapid eye movement
- NSB
- nonspecific binding
- PAM
- positive allosteric modulator
- PCR
- polymerase chain reaction
- Philanthotoxin-74
- (S)-N-[7-[(4-Aminobutyl)amino]heptyl]-4-hydroxy-α-[(1-oxobutyl)amino]benzenepropanamide dihydrochloride
- p.o.
- per os
- PTZ
- pentylenetetrazole
- RED
- Rapid Equilibrium Dialysis
- REM
- rapid eye movement
- SB
- specific binding
- TARP
- transmembrane AMPA receptor regulatory protein
- TB
- total binding
- TM
- transmembrane domain
- Copyright © 2016 by The American Society for Pharmacology and Experimental Therapeutics