CHAPTER 5: Excitatory and Inhibitory Amino Acids

• The major excitatory neurotransmitter in the brain is glutamate; the major inhibitory neurotransmitter is γ-aminobutyric acid (GABA).

• Glutamate receptors comprise two large families, ligand-gated ion channels called ionotropic receptors and G protein–coupled receptors called metabotropic receptors.

• Ionotropic glutamate receptors are divided into three classes—α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors, kainate receptors, and N-methyl-D-aspartate (NMDA) receptors—which are named after synthetic ligands that activate them. Individual excitatory synapses typically express several different subtypes of ionotropic glutamate receptors as well as metabotropic receptors.

• In contrast to AMPA receptors and kainate receptors, the NMDA receptor has two important biophysical properties. Because it is highly permeable to calcium and is voltage dependent, it allows calcium entry only if the cell is depolarized.

• AMPA receptors mediate the vast majority of excitatory synaptic transmission in the brain, whereas NMDA receptors play an important role in triggering synaptic plasticity and, when overactivated, in triggering excitotoxicity.

• There are eight subtypes of metabotropic glutamate receptors. When localized to the presynaptic terminal, they inhibit neurotransmitter release. When localized to the postsynaptic membrane, they exert complex modulatory effects through specific signal transduction cascades, which can lead to excitatory or inhibitory effects.

• The most extensively studied form of synaptic plasticity is long-term potentiation (LTP) in the hippocampus, which is triggered by strong activation of NMDA receptors and the consequent large rise in postsynaptic calcium concentration. Several other types of LTP are widespread in the nervous system and are mediated via distinct mechanisms.

• Long-term depression (LTD), a long-lasting decrease in synaptic strength, also occurs at most excitatory and some inhibitory synapses in the brain.

• The GABA_A receptor, a ligand-gated chloride channel, and the GABA_B receptor, a G protein–coupled receptor, are the two major classes of GABA receptors.

• GABA_A receptors, which are highly heterogeneous, mediate the bulk of inhibitory synaptic transmission in the brain. Many drugs, most notably benzodiazepines and barbiturates, bind to GABA_A receptors and enhance their function.

• GABA_B receptors are localized both presynaptically, where they inhibit neurotransmitter release, and postsynaptically, where they mediate a slow, inhibitory synaptic response.

• Glycine, like GABA, is an inhibitory neurotransmitter that activates receptors that are ligand-gated chloride channels. It is critical in inhibitory neurotransmission in the spinal cord and brainstem.

Amino acids are the building blocks of proteins involved in normal intermediary metabolism, but they can also function as neurotransmitters. The amino acids glutamate and, to a much lesser extent, aspartate mediate most of the fast excitatory synaptic transmission in the brain; likewise, the amino acids γ-aminobutyric acid (GABA) and, to a lesser extent, glycine mediate most fast inhibitory synaptic transmission. Excitatory amino acids are utilized by nearly every information-bearing circuit in the brain and have been implicated in such diverse pathologic processes as epilepsy, ischemic brain damage, and several psychiatric disorders. In addition, they are necessary for the development of normal synaptic connections. Consequently, amino acid neurotransmitters have been the subject of intensive research for many decades.

The Major Excitatory Neurotransmitter

Long before glutamate’s role in neurotransmission was established, investigators observed that glutamate excites most neurons in the CNS. In fact, it is responsible for most neurotransmitter action at excitatory amino acid receptors. Glutamate is present in high concentrations in the adult CNS and is released in a Ca²⁺-dependent manner by electrical stimulation. Enzymes responsible for glutamate synthesis and degradation are located in both neurons and glial cells, as are high-affinity glutamate receptors and excitatory amino acid reuptake transporters (EAATs), which terminate the synaptic actions of glutamate. Many of these reuptake and receptor proteins also respond to aspartate, which may mediate transmission at a small number of central excitatory synapses.

Synthetic and Degradative Pathways

Glutamate and aspartate are charged amino acids and therefore do not cross the blood–brain barrier. As a result, they must be synthesized in the brain from glucose and a variety of other precursors. Glutamate, which is the reduced form of glutamic acid, is in a metabolic pool with α-oxoglutaric acid and glutamine. It is packaged into presynaptic vesicles via several steps 5–1. After action potential–driven release, glutamate is primarily taken up by glial cells, where it is converted into glutamine by glutamine synthetase. The resulting glutamine is transported out of glia by system N-1 (SN1), a Na⁺- and H⁺-dependent pump that is homologous to the vesicular GABA transporter (VGAT). Glutamine is subsequently taken up by neurons, by means of a transport process that remains poorly described, and is converted back into glutamate by glutaminase. As discussed later in this chapter, glutamine also replenishes the transmitter pool of GABA.

Release and Reuptake

Synaptic vesicles actively accumulate glutamate through a Mg²⁺- and adenosine triphosphate (ATP)-dependent uptake process that is driven by an electrical gradient across their membranes. During uptake, concentrations of glutamate in these vesicles likely exceed 20 mM. Substances that destroy the electrochemical gradient across vesicular membranes prevent the transporters from concentrating this amino acid. The three isoforms of the vesicular glutamate transporter, VGluT1 to 3, are highly selective with a high affinity for glutamate and a low affinity for aspartate (Chapter 3).

After release from presynaptic terminals, the reuptake of glutamate and aspartate serves to control the extracellular concentrations of these amino acids in the CNS. Na⁺-dependent glutamate plasma membrane transporters are coupled to electrochemical gradients for Na⁺, K⁺, and H⁺ 5–1. Because such coupling permits the transport of glutamate and aspartate against their concentration gradients, glutamate transporters are capable of decreasing extracellular glutamate concentrations to submicromolar levels. These transporters take up glutamate and aspartate with similar affinity and maximum velocity (V_max). Glutamate transport is electrogenic, resulting in the net inward movement of positive charge during each transport cycle.

Five principal members of the high-affinity Na⁺-dependent family of glutamate plasma membrane transporters have been cloned: EAAT1 (also known as glutamate–aspartate transporter [GLAST]), EAAT2 (glutamate transporter-1 [GLT-1]), EAAT3 (excitatory amino acid carrier-1 [EAAC1]), and EAAT4 and 5. Each of these transporters belongs to a large superfamily whose members transport a wide range of neurotransmitters and related substances (Chapter 3). EAAT1 and 2, which are expressed by glial cells, appear to be responsible for the majority of glutamate reuptake in the CNS (see 5–1).

Several pathologic conditions, such as ischemia, can lead to the accumulation of glutamate or aspartate in extracellular spaces, which results in the excessive activation of glutamate receptors (Chapters 18 and 20). Such activation can in turn cause a variety of pathologic changes and can, in its extreme form, result in cell death. Glutamate transporters limit the concentrations of free glutamate and aspartate in extracellular spaces, and thereby prevent the excessive stimulation of glutamate receptors. Consequently, agents capable of facilitating transporter function might limit the damage caused by ischemia and other neurologic insults that cause extracellular glutamate to increase. However, most progress to date has involved developing inhibitors of glutamate transporters. Two such inhibitors are D,L-threo-3-hydroxyaspartate (THA), a broad-spectrum antagonist of glutamate and aspartate transport, and dihydrokainate (DHK), an inhibitor selective for the glial EAAT2 transporter. Although these drugs are useful experimental tools, they have no apparent clinical use.

Glutamate Receptors

Glutamate receptors comprise two large families: the ionotropic and the metabotropic receptors. Ionotropic glutamate receptors contain associated ion channels that are gated by agonist binding. Three classes of ionotropic glutamate receptors, N-methyl-D-aspartate (NMDA), α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA), and kainate receptors, were originally named based on the ability of these drugs to serve as selective agonists 5–1.

Metabotropic glutamate receptors (mGluRs) belong to the large superfamily of G protein–coupled receptors. These receptors, which are characterized by seven transmembrane domains (TMs), couple to G proteins and in turn mediate the biologic effects of receptor activation (Chapter 4). The term metabotropic was used to indicate that these receptors affect cellular biochemical (“metabolic”) processes, and do not form ion channels. However, mGluRs, like other G protein–coupled receptors, can exert profound effects on neuronal function through the regulation of other ion channels, second messenger cascades, and protein phosphorylation.

Mediation of fast excitatory transmission

Synaptically released glutamate interacts with postsynaptic receptors located on dendrites or nerve terminals of immediately adjacent cells or on the nerve terminals from which glutamate is released. The binding of glutamate to AMPA and NMDA receptors opens postsynaptic cation channels and initiates a two-component excitatory postsynaptic current (EPSC) at most central synapses 5–2. Considerable evidence suggests that AMPA and NMDA receptors colocalize at most functional excitatory synapses. However, the ratio of AMPA to NMDA receptors at individual synapses can vary greatly; indeed, some synapses may contain only NMDA or AMPA receptors. In contrast, smaller numbers of kainate receptors are present in most CNS regions.

The activation of an AMPA receptor mediates a synaptic current that has a rapid onset and decay, whereas the current mediated by an NMDA receptor has a slower onset and a decay that lasts as long as several hundred milliseconds (see 5–2). The decay time of the NMDA receptor–mediated current is approximately 100 times longer than the mean open time of its channel. Such prolonged activation is believed to be caused by glutamate’s high affinity (K_d = 3–8 μM) for and consequent slow dissociation from these receptors. In contrast, glutamate has a much lower affinity for AMPA receptors (K_d = 200 μM), from which it rapidly dissociates.

AMPA receptors respond to single vesicles of glutamate. NMDA receptors, however, because of their special voltage dependence (described later in this chapter), commonly require coordinated input from many synapses for substantial activation. This allows NMDA receptors to act as coincidence detectors that can sense the activity of many independent synaptic inputs converging on the same cell 5–3.

AMPA Receptors

Current is carried through AMPA receptors primarily by the movement of Na⁺ from the extracellular space into the intracellular compartment. However, because the reversal potential of current (the membrane potential at which net current flow is zero; see Chapter 2) through AMPA channels is close to 0 mV, an outward current carried by K⁺ must counterbalance the inward flow of Na⁺ ions. The resulting current–voltage relationship for these AMPA receptors is roughly linear (see 5–2).

Some AMPA receptors, on both neurons and astrocytes, are also permeable to Ca²⁺. The translocation of Ca²⁺ from the extracellular space to the intracellular compartment plays a key role in the regulation of several second messenger systems. Thus, the permeability of some AMPA receptors to Ca²⁺ may have great functional importance, particularly in cells that do not contain NMDA receptors (which can always flux Ca²⁺). AMPA receptor channels that are permeable to Ca²⁺ lack the GluA2 subunit (see below) and exhibit an inwardly rectifying type of current–voltage relationship; they pass current readily in the inward direction but not in the outward direction. This occurs because intracellular polyamines such as spermine and spermidine, which associate with the channel, prohibit outward current from passing through. A spermine analogue, N(1)-acetylspermine (NASPM), a selective antagonist of GluA2-lacking AMPA receptors, is a valuable tool to study the influence of such receptors in brain function and plasticity.

All AMPA receptors can be blocked selectively by certain quinoxaline diones, the most notable of which is 6-nitro-7-sulfamobenzo-quinoxaline 2,3-dione (NBQX; see 5–1). NBQX is a potent competitive antagonist of AMPA receptors, which is reasonably selective for AMPA receptors over kainate receptors. Drugs in the 2,3-benzodiazepine class, such as GYKI 53655, are noncompetitive, selective antagonists of AMPA receptors and are being explored as neuroprotective drugs for the treatment of stroke. Because it has minimal effects on kainate or NMDA receptors, this class of drug permits unequivocal separation of the AMPA receptors from other categories of glutamate receptors.

AMPA and kainite receptors desensitize within milliseconds of exposure to agonist and can be reliably distinguished from one another by their response to two drugs, cyclothiazide and the lectin concanavalin A. Cyclothiazide relieves AMPA receptor desensitization but does not affect kainate receptors. In contrast, concanavalin A relieves the desensitization of kainate receptors, most likely by interacting with surface sugar chains, but does not have significant effects on AMPA receptors.

The possibility that enhancement of AMPA receptor function may be beneficial therapeutically has received increased attention with the development of positive allosteric modulators (PAMs) of AMPA receptors, also termed AMPAkines or AMPA potentiators. These drugs enhance AMPA receptor–mediated excitatory synaptic transmission by slowing deactivation or desensitization of the receptors, with the former mechanism appearing to be more functionally important. Clinical trials are currently being conducted in a wide range of neuropsychiatric disorders (eg, Alzheimer disease, depression, and attention deficit hyperactivity disorder [ADHD]) to determine whether this class of compound may be clinically useful.

Molecular composition of AMPA receptors

Three families of ionotropic glutamate receptor subunits, encoded by at least 16 genes, assemble to form functional AMPA, NMDA, or kainate receptors (see 5–1). Within each family there is greater than 80% identity at the amino acid level over membrane-spanning domains. Between families, a lower degree of identity exists (~50%). Ionotropic glutamate receptors are believed to exist as tetramers. Different subunit combinations produce functionally different glutamate receptors. Moreover, there are striking regional differences in the expression of genes that encode these subunits.

Four subunits, termed GluA1 to GluA4 (also known as GluR1–GluR4), coassemble to form AMPA receptors. Each of these subunits is encoded by a distinct gene, and each exists in two forms, termed flip and flop. The flip and flop forms, which are generated by alternative splicing, exhibit region-specific patterns of expression in the brain, and give rise to receptors that differ in desensitization rates. The best characterized AMPA receptor subunit is GluA1, which is predicted to be 889 amino acids long; other ionotropic receptor subunits, such as those of nicotinic, GABA_A, or glycine receptors, are approximately 420 amino acids long. The extra length of GluRs results from unusually large N-terminal extracellular domains.

AMPA and kainate glutamate receptor families have a topology that differs from that of certain other ligand-gated ion channels, such as GABA_A and glycine receptors 5–4. Glutamate receptor subunits possess only three transmembrane-spanning domains. What appears to be an abridged TM, located between TM2 and TM3 in glutamate subunits, is a reentrant loop whose both ends face the cytoplasm. Homology mapping of glutamate receptors suggests that agonist binding requires portions of both the large N terminus and the short region between TM1 and TM2 (see 5–4).

As stated earlier, AMPA receptors lacking the GluA2 subunit are highly permeable to Ca²⁺ ions and are commonly found on GABAergic inhibitory interneurons throughout the brain. This Ca²⁺ permeability has been traced to a single amino acid in the reentrant loop of GluA2, which is known as the Q/R site (see 5–4). In GluA1, GluA3, and GluA4, a glutamine (Q) resides at this position, but in GluA2 an arginine (R) is present. When heteromeric receptors contain GluA2, they are relatively impermeable to Ca²⁺, most likely because the positive charge of the arginine residue repels Ca²⁺ from the channel pore. The replacement of glutamine with arginine in GluA2 occurs because of a process called RNA editing: the gene for GluA2 encodes a glutamine, and the unedited form of GluA2 is expressed in various regions of the brain during development. However, virtually all of the GluA2 mRNA present in adult mammalian brain is edited to a codon that encodes arginine at this site. This editing, which is catalyzed by adenosine deaminases acting on RNA (ADARs), changes a single base within GluA2 transcripts. In addition to Ca²⁺ permeability, the Q/R site of GluA2 influences the single channel conductance properties of associated AMPA receptors and the sensitivity of the receptor complex to blockage by polyamine spider toxins and polyamines (eg, NASPM). AMPA receptors that do not contain edited GluA2 show greater overall conductance. In addition, there is voltage-dependent inhibition of outward current flow by endogenous intracellular polyamines at positive membrane potentials giving rise to the inward rectification that is characteristic of GluA2-lacking AMPA receptors. A deficiency in GluA2 editing has been implicated in the pathophysiology of amyotrophic lateral sclerosis (ALS), acting perhaps by increasing Ca²⁺ flux into motor neurons and promoting excitotoxicity (Chapter 18).

The flip and flop splice variants of receptor subunits GluA1 to GluA4, together with the products of Q/R editing, yield a wide range of AMPA receptor subunit proteins. This diversity provides neurons with an extraordinary degree of flexibility in the construction of AMPA receptors. The degree of AMPA receptor heterogeneity employed by neurons in vivo remains unknown. Ca²⁺-permeable AMPA receptors lacking the GluA2 subunit are consistently found at excitatory synapses on GABAergic inhibitory interneurons, whereas most principal cells (ie, projection neurons) express Ca²⁺-impermeable AMPA receptors that contain the edited form of GluA2. There is growing evidence that the number of GluA2-lacking AMPA receptors at particular synapses is subject to dynamic regulation and may be an important mechanism of neural and behavioral plasticity.

Kainate Receptors

Kainate receptors, like AMPA receptors, are cation-selective ligand-gated ion channels that strongly depolarize neurons when activated. They are found on presynaptic terminals, of both excitatory and inhibitory synapses, where their activation can modify neurotransmitter release, both because of their depolarizing actions and because, depending on their molecular composition, they are permeable to Ca²⁺. Depending on the synapse and the degree of activation, presynaptic kainate receptors can either facilitate or depress transmitter release. Kainate receptors are also found postsynaptically on certain neurons, where they normally generate slow, small, but functionally important, postsynaptic potentials. In the hippocampus and cortex, they may be particularly important in the early development of neural circuits. Recent data suggest that kainate receptors can also be metabotropic, initiating a G protein signaling cascade independent of their ionotropic signaling.

Kainate receptors are composed of a distinct array of subunits. The subunits GluK1 to GluK3 coassemble with GluK4 or GluK5 subunits to form functional kainate receptors. Homomeric GluK1, GluK2, and GluK3 receptors expressed in mammalian cell lines bind kainate with an affinity of approximately 80 to 100 nM. These homomeric receptors may correspond to low-affinity kainate binding sites previously identified in membrane fractions of the brain. In contrast, homomeric GluK4 receptors bind kainate with an affinity of 4 nM and may correspond to high-affinity kainate binding sites in the brain. However, when they are expressed alone, GluK4 and GluK5 are virtually inactive because they lack functional channels. Consequently, it is believed that they serve as modulatory subunits that confer high-affinity kainate binding on channels formed by GluK1 to GluK3. GluK1 and GluK2 exist in several splice variants, and also undergo Q/R editing, meaning that kainate channels can vary in their Ca²⁺ permeability.

Selective kainate receptor antagonists are under development; those currently available, such as LY382884 5–1, are primarily selective for the GluK1 subunit. LU97175 generally antagonizes kainate receptors, although it is most potent for GluK3-containing receptors. SYM 2081 is potent agonist for GluK1- and GluK2-containing kainate receptors, and rapidly desensitizes the receptors; it can therefore be used as an antagonist. However, GYKI 53655 remains the most valuable agent for determining whether a given synaptic current is mediated by AMPA or kainate receptors, as stated earlier. Domoic acid, derived from red algae and implicated as a cause of shellfish poisoning, acts as a neurotoxin via activation of kainate receptors.

Kainate receptors are believed to contribute to the development of temporal lobe epilepsy, and genetic studies have linked alleles of GluK2 to Huntington disease (in which it might be a disease modifier). However, the role of kainate receptors in normal neuronal function, let alone in the development of complex neuropathologies, remains uncertain.

NMDA Receptors

NMDA receptors have several properties that set them apart from other ligand-gated receptors. At membrane potentials more negative than approximately −50 mV, the Mg²⁺ in the extracellular fluid of the brain virtually abolishes ion flux through NMDA receptor channels, even in the presence of glutamate. Thus, at the resting membrane potentials that are typical of most neurons, approximately −60 to −70 mV, the activation of these receptors results in little current flow. This is because the entry of Mg²⁺ into the channel pore blocks the movement of monovalent ions across the channel.

In the presence of Mg²⁺ ions, the receptor’s current–voltage relationship has a region of slope negativity that produces a characteristic J shape when plotted (see 5–2). As the receptor’s membrane potential becomes less negative (more depolarized), the affinity of Mg²⁺ for its binding site decreases, and the blocking action of Mg²⁺ becomes ineffective; consequently, ionic current can pass through the channel as the cell membrane depolarizes.

As previously mentioned, NMDA receptors can be thought of as coincidence detectors capable of sensing simultaneous activity at a number of adjacent synapses (see 5–3). They possess this capability because they function (ie, pass current) only when they are stimulated by presynaptically released glutamate at a time when the postsynaptic cell is depolarized. Repetitive stimulation is often required because the depolarization produced by single inputs is not sufficient to relieve the blockage of the NMDA receptor channel by Mg²⁺.

The activation of NMDA receptors, like that of AMPA receptors, produces a nonspecific increase in permeability to the monovalent cations Na⁺ and K⁺. However, unlike most AMPA and kainate receptors in the adult CNS, NMDA receptors are highly permeable to Ca²⁺. Although their activation results in appreciable current and tends to depolarize the cell membrane toward the threshold for action potential firing, such activity is unlikely to represent the primary function of these receptors. Instead, NMDA receptors likely provide one of the most significant mechanisms by which synaptic activity can increase the level of intracellular Ca²⁺ at individual synapses. The major endogenous agonist for NMDA receptors is glutamate itself, although there is some evidence that aspartate can also activate this receptor. Numerous competitive antagonists of the agonist recognition site are available, notably D-2-amino-5-phosphonopentanoic acid (AP-5 or APV) and 3-(2-carboxypiperazin-4-yl)-1-propenyl-1-phosphonic acid (CPP). These compounds are not useful clinically because they are polar and penetrate the blood–brain barrier poorly.

The NMDA receptor is unique among all neurotransmitter receptors in that its activation requires the simultaneous binding of two different agonists. In addition to the binding of glutamate at the conventional agonist-binding site, the binding of glycine appears to be required for receptor activation 5–5. Because neither of these agonists alone can open this ion channel, glutamate and glycine are referred to as coagonists of the NMDA receptor. The physiologic significance of the glycine-binding site is unclear; however, there is some evidence to suggest that D-serine may be the endogenous agonist for this site. D-Serine, made through the conversion of L-serine by serine racemase, is subject to regulated release and specific reuptake primarily from glial cells. Thus, the glial environment of neurons may have a critical influence on NMDA receptor synaptic function.

Regardless of its physiologic role, the glycine site on the NMDA receptor may prove to be an important drug target. Cycloserine, originally developed as an antitubercular drug, is a weak partial agonist at this site and thus can modulate NMDA receptor function in vitro and in vivo. Cycloserine is reported to enhance the effects of antipsychotic drugs in patients with schizophrenia. More specific agonists, such as HA966, have been used in laboratory animals but are not yet available for clinical investigation. Derivatives of kynurenic acid, such as 5,7-dichlorokynurenic acid (5,7-DCK) and quinolinecarboxylic acid, are also competitive antagonists at the glycine site. It is important to note that the glycine site on NMDA receptors is distinct from the strychnine-sensitive glycine receptor, which mediates the independent neurotransmitter functions of glycine (see below).

Another important site on the NMDA glutamate receptor binds phencyclidine (PCP) and related drugs such as MK801 and ketamine 5–5. These drugs, which bind at or near the Mg²⁺-binding site, occlude the NMDA receptor channel. Thus, they act as noncompetitive receptor antagonists, and their actions, like those of Mg²⁺, are somewhat voltage dependent. These drugs exert potent effects on the brain. At higher concentrations they are psychotomimetic and produce effects, such as cognitive impairment, hallucinations, and delusions that are similar to some of the symptoms of schizophrenia (Chapter 17). These effects have led to the hypothesis that schizophrenia may involve hypofunction of NMDA receptors, which has in turn prompted the development of drugs that block the reuptake of glycine for the treatment of schizophrenia. The idea is that glycine transporter-1 (GlyT1) inhibitors would increase levels of extracellular glycine in forebrain and thereby facilitate activation of NMDA receptors. At still higher doses, drugs such as PCP and ketamine are dissociative anesthetics, used primarily in veterinary and pediatric practice. Interestingly, doses of ethanol associated with the upper range of intoxication in humans exert effects on NMDA receptors that are similar to those produced by PCP, ketamine, and related drugs. More recent research has shown that lower doses of ketamine, devoid of most psychotomimetic effects, exert rapidly acting antidepressant effects in a subset of individuals with treatment-resistant depression (Chapter 15).

NMDA receptors also have one or more modulatory sites that bind polyamines. The occupancy of one of these sites relieves tonic proton block and thereby potentiates NMDA receptor activation. At higher concentrations, however, polyamines act on an extracellular site to produce a voltage-dependent block of the ion channel and consequently inhibit receptor activation. Many additional drugs, such as amantadine, dextromethorphan, and memantine (Chapter 18), are weak antagonists of NMDA receptors, although it remains uncertain as to whether such actions contribute to the clinical utility of these medications.

Because they mediate a large number of important physiologic functions, and also contribute to cell damage and death when they are overactivated (Chapters 18 and 20), NMDA receptors have been a prominent target for therapeutic drugs. Many drugs that target these receptors have been developed and tested in clinical trials for conditions such as stroke or head trauma in which excitotoxicity may play a significant role. However, compounds with clear efficacy and tolerable side effects have yet to be identified. The large number of modulatory sites on the NMDA receptor increases the likelihood that clinically useful compounds will be discovered.

Molecular composition of NMDA receptors

Two families of NMDA receptor subunits have been identified. One family is represented by a single gene (GluN1) that encodes proteins composed of approximately 900 amino acids; the other is represented by four genes (GluN2A–2D; see 5–1) that encode proteins composed of approximately 1450 amino acids. Although homomeric GluN1 receptors appear to possess all of the pharmacologic features characteristic of bona fide NMDA receptors, the physiologically relevant glutamate-binding site is probably located on the GluN2 subunit. Moreover, the very small currents supported by homomeric GluN1 receptors increase by more than 100-fold when such receptors are coexpressed with GluN2 subunits. It is therefore believed that NMDA receptors in the brain exist as GluN1–N2 heteromeric complexes. Some neurons also express the GluN3A or GluN3B subunit, which modulates NMDA receptor function, decreasing both the conductance and the Ca²⁺ permeability of the channel. The physiologic significance of the GluN3 subunit is not understood.

More than nine splice variants of GluN1 have been cloned. These variants differ with regard to their regional patterns of expression in the CNS, their regulation (eg, by phosphorylation, polyamines, and Zn²⁺), the electrophysiologic properties of channels they form, and their affinity for elements of the neuronal cytoskeleton. Such distinguishing features suggest that these variants may be involved in different components of the synapse.

The subtype of GluN2 subunit that combines with GluN1 subunits influences the biophysical and pharmacologic properties of endogenous NMDA receptors. During early postnatal development, many synaptic NMDA receptors are composed of GluN1 and GluN2B. This subunit combination yields a receptor that produces very long-lasting synaptic responses and one that is strongly inhibited by the NMDA receptor antagonist ifenprodil. Gradually during the first few weeks of brain development, it is thought that the GluN2B subunit is largely replaced by GluN2A (and perhaps GluN2C), yielding a receptor that produces shorter synaptic currents and that is no longer sensitive to ifenprodil. There is also evidence that different synapses in the adult brain preferentially express GluN1–N2A versus GluN1–N2B NMDA receptors, which makes the development of selective agonists and antagonists a high priority. Indeed, selective antagonists of GluN2B-containing NMDA receptors are being evaluated for the treatment of several neuropsychiatric disorders.

Metabotropic Glutamate Receptors

Eight mGluRs, termed mGlu1 to mGlu8, have been cloned (see 5–1). mGluRs are considerably larger than other G protein–coupled receptors, and a comparison of their amino acid sequences with those of other receptors reveals little homology or common features. mGluRs are therefore considered to constitute a separate family of receptors. Like other G protein–coupled receptors, mGluRs contain seven membrane-spanning domains; however, like the ionotropic receptors, they also possess an unusually large N-terminal extracellular domain that precedes the membrane-spanning segments 5–6.

There are three functional groups of mGluRs based on amino acid sequence homology, agonist pharmacology, and the signal transduction pathways to which they are coupled. Group I includes mGlu1 and mGlu5, which are generally found on postsynaptic neurons adjacent to excitatory synapses. Group II is composed of mGlu2 and mGlu3, while group III comprises mGlu4, mGlu6, mGlu7, and mGlu8. Group II and III mGluRs are often found on presynaptic terminals where they modulate transmitter release. Members of each group share approximately 70% sequence homology, with approximately 45% homology exhibited between groups. Alternatively spliced variants also have been described for mGlu1, mGlu4, mGlu5, and mGlu7.

The classification of mGluRs into three groups is supported by their signal transduction mechanisms. Group I mGluRs stimulate phospholipase C activity by means of the G protein G_q, and thereby release Ca²⁺ from cytoplasmic stores through IP₃ (Chapter 4). Yet group I mGluRs vary in their ability to increase intracellular Ca²⁺ levels, most likely because each receptor has a different affinity for G_q. Activation of phospholipase C leads to the formation of not only IP₃ but also diacylglycerol, which, in conjunction with increases in intracellular Ca²⁺, activates protein kinase C. In contrast to group I mGluRs, group II and III mGluRs inhibit adenylyl cyclase and regulate specific K⁺ and Ca²⁺ ion channels, actions believed to be mediated via coupling to G_i/o.

Glutamate activates all of the mGluRs, with potencies that range from 2 nM for mGlu8 to 1 μM for mGlu7. Highly selective agonists for each of the three groups also have been identified 5–1. 3,5-Dihydroxyphenylglycine (DHPG) appears to be a selective group I agonist; LY354740 is a highly selective agonist for group II mGluRs, and L-amino-4-phosphonobutyrate (L-AP4) is a selective agonist of the group III mGluRs. mGluR antagonists have been developed for the various subtypes, but few antagonize a whole group. LY341495 at low concentrations will inhibit group II mGluRs, but at high concentrations will inhibit all mGluRs. 2-Methyl-6-(phenylethynyl)-pyridine (MPEP) is a somewhat selective group I antagonist. Positive and negative allosteric modulators of specific mGluRs (termed mGluR PAMs or NAMs) are also in development.

Modulation of ion channel activity

mGluRs located on the postsynaptic membrane modulate a variety of ligand- and voltage-gated ion channels expressed on central neurons 5–7. The activation of each of the three groups of mGluRs has been found to inhibit L-type voltage-gated Ca²⁺ channels, and groups I and II are capable of inhibiting N-type Ca²⁺ channels. Additionally, mGluR activation can close voltage-gated K⁺ channels, resulting in a slow depolarization and neuronal excitation. The exact mechanism by which mGluRs inhibit K⁺ currents is not yet clear. In cerebellar granule cells, mGluRs increase the activity of Ca²⁺-dependent K⁺ channels, termed BK channels, and thereby reduce cell excitability. In some cells, mGluRs also activate G protein–coupled, inwardly rectifying K⁺ (GIRK) channels. Thus, postsynaptic mGluRs can have a wide range of effects, depending on the cell type involved and the mGluR subtype that is activated.

mGluR-mediated presynaptic inhibition at excitatory synapses

Several types of mGluRs are located on the presynaptic terminals of central neurons, and the activation of presynaptic mGluRs blocks both excitatory glutamatergic and inhibitory GABAergic synaptic transmission in many CNS regions 5–7. In the hippocampus, mossy fiber-evoked EPSPs on CA3 pyramidal neurons are blocked by DCG-IV, which acts on presynaptic group II mGluRs located on granule cell terminals. Similar effects have been observed throughout the cerebral neocortex. In contrast, synaptic transmission between Schaeffer collaterals and CA1 pyramidal cells is resistant to DCG-IV but is reduced by L-AP4, a group III mGluRs agonist. One mechanism by which activation of mGluRs decreases neurotransmitter release may involve the inhibition of voltage-gated Ca²⁺ channels on the presynaptic nerve terminal membrane. Such activity has been observed directly at certain glutamatergic synapses, where mGluR agonists suppress voltage-gated P/Q-type Ca²⁺ channels and thereby inhibit transmitter release. Accordingly, mGluRs most likely function as inhibitory autoreceptors at many glutamatergic nerve terminals.

The numerous effects that mGluRs can have on both postsynaptic cells and presynaptic terminals are confusing, especially because the conditions under which these receptors are physiologically activated are not always clear. Yet the diversity that characterizes mGluRs and their actions promises to further the development of subtype-specific drugs for the treatment of neuropsychiatric disorders. Several drugs that target specific mGluRs are under evaluation for the treatment of a variety of brain disorders; however, none have yet proved clearly efficacious in large clinical trials.

Synaptic Clustering of Glutamate Receptors

Effective synaptic communication requires the precise localization of high concentrations of appropriate presynaptic and postsynaptic proteins at the synapse. On the postsynaptic side of the synapse, neurons must be able to target glutamate receptors to excitatory but not inhibitory synapses and must ensure that the receptors are appropriately clustered opposite presynaptic terminals that release glutamate. Several members of the protein family involved in this clustering contain single or multiple copies of an amino acid sequence termed the PDZ domain, which is important for mediating many important protein–protein interactions 5–1. An evolving hypothesis is that each subtype of glutamate receptor interacts with distinct proteins at the synapse. Accordingly, the subcellular localization of each type of receptor may be independently regulated under physiologic conditions. Although these mechanisms remain poorly understood, they appear to be important for certain forms of neural plasticity.

Role of Glutamate in Neural Plasticity

The high permeability of NMDA receptor channels to divalent cations, especially to Ca²⁺, has many implications for cell function. The concentration of Ca²⁺ in the cell interior typically is heavily buffered to approximately 100 nM and is tightly regulated by several mechanisms, including storage in intracellular organelles. Ca²⁺ entry through NMDA receptor channels can lead to a transient increase in intracellular Ca²⁺ concentrations to the micromolar range. Such an increase can in turn result in the activation of many Ca²⁺-dependent enzymes, including Ca²⁺/calmodulin-dependent protein kinases (CaM-kinases), calcineurin (protein phosphatase 2B or PPP3), protein kinase C, phospholipase A₂, phospholipase C, nitric oxide synthase, and several proteases (Chapter 4).

One of the most important consequences of NMDA receptor activation is the generation of often long-lasting changes in synaptic function termed synaptic plasticity. Many forms of synaptic plasticity have been discovered in the mammalian CNS, but long-term potentiation (LTP) and long-term depression (LTD) of excitatory synaptic responses in CA1 pyramidal cells in the hippocampus have been the most extensively characterized. LTP and LTD are activity-dependent alterations in synaptic efficacy that can last weeks or months in vivo. Changes in receptor number (eg, via altered receptor trafficking 5–1 or transcription) or changes in receptor function (eg, due to phosphorylation 5–2) contribute to these changes in synaptic strength. The bidirectional control of synaptic strength by LTP and LTD is believed to be important for many forms of experience-dependent plasticity including some forms of learning and memory in the mammalian brain.

The best understood form of long-lasting synaptic plasticity is NMDA receptor–dependent LTP, which is found in many regions of the mammalian brain but has been most extensively studied in the hippocampus because of its key role in learning and memory (Chapter 14). The triggering of this form of LTP requires activation of NMDA receptors by synaptically released glutamate when the postsynaptic membrane is already strongly depolarized. This depolarization relieves the voltage-dependent block of the NMDA receptor channel by Mg²⁺, and allows Ca²⁺ to enter the postsynaptic dendritic spine when the receptor is activated by glutamate. The rise in postsynaptic Ca²⁺ concentration, the critical trigger for LTP, activates complex intracellular signaling cascades that include several protein kinases, most notably CaM-kinase II 5–8A. The primary mechanism underlying the increase in synaptic strength during LTP is a change in AMPA receptor trafficking that results in an increased number of AMPA receptors in the postsynaptic plasma membrane. Within a few hours, the maintenance of LTP requires protein synthesis and there is good evidence that LTP is accompanied by observable enlargements of dendritic spines. Such structural changes may be essential to cement the information storage process initiated at synapses on LTP induction.

At many excitatory synapses throughout the brain, weaker activation of NMDA receptors can elicit the opposite phenomenon, NMDA receptor–dependent LTD, which is thought to result from a smaller rise in postsynaptic Ca²⁺ than is required for LTP 5–8B. This more modest change in postsynaptic Ca²⁺ activates serine/threonine phosphatases, which dephosphorylate critical synaptic substrates including AMPA receptors themselves. The depression of synaptic strength during NMDA receptor–dependent LTD is due to the removal of synaptic AMPA receptors via dynamin- and clathrin-dependent endocytosis. An intriguing feature of NMDA receptor–dependent LTD is that NMDA receptor–mediated synaptic responses are also depressed via mechanisms distinct from those responsible for the LTD of AMPA receptor–mediated responses.

Surprisingly, activation of postsynaptic mGluRs can also lead to a form of LTD that was first described in the cerebellum but also occurs in the hippocampus and neocortex. At the cerebellar parallel fiber synapse, LTD requires both postsynaptic Ca²⁺ influx through voltage-gated Ca²⁺ channels and postsynaptic group I mGluR activation, while at other synapses activation of postsynaptic mGluRs alone appears to be sufficient. In most cases, however, this form of LTD is mediated by clathrin-dependent endocytosis of synaptic AMPA receptors 5–8C. At certain developmental stages, rapid protein synthesis is required for both mGluR-triggered AMPA receptor endocytosis and LTD. There is evidence from mouse models that this form of LTD may be relevant to fragile X mental retardation syndrome. Fragile X mental retardation protein (FMRP), which is deficient in fragile X syndrome due to abnormal methylation of the FMRP gene promoter, normally opposes group I mGluR-mediated induction of local protein synthesis in dendrites and the resulting LTD. Accordingly, it is proposed that fragile X symptoms arise in part through excessive dendritic protein synthesis and LTD. Consistent with this hypothesis are the recent findings that many of the fragile X–like symptoms exhibited in mouse models can be reversed by MPEP, a group I mGluR antagonist. This raises the possibility of novel treatments for fragile X syndrome in humans.

The aforementioned forms of synaptic plasticity are all triggered by activation of postsynaptic glutamate receptors and involve changes in the number of AMPA receptors at synapses. In contrast, there are also forms of LTP and LTD that involve long-lasting presynaptic changes in the release of glutamate. A presynaptic form of LTP was first described in the CA3 region of the hippocampus and is now known to occur in the striatum and cerebellum as well. This form of LTP does not require NMDA receptors and appears to be initiated by an activity-dependent rise in intracellular Ca²⁺ within the presynaptic terminals 5–8D. The Ca²⁺ rise may activate particular isoforms of adenylyl cyclases to produce cAMP, with subsequent activation of cAMP-dependent protein kinase (PKA) (Chapter 4). Via unknown mechanisms, this in turn leads to a persistent increase in the amount of glutamate released each time an action potential enters the nerve terminal. A potentially important form of presynaptic LTD may also occur at certain glutamatergic as well as at some inhibitory GABAergic synapses. This is commonly due to postsynaptic activation of mGluRs or voltage-gated Ca²⁺ channels, which triggers the synthesis of endocannabinoids, lipophilic molecules that are released by postsynaptic cells and can travel retrogradely across the synapse to bind to presynaptic cannabinoid receptors (discussed further in Chapter 8). Depending on the specific synapse, these endocannabinoids can either transiently depress neurotransmitter release for a period of many seconds or cause an LTD mediated by a long-lasting depression of transmitter release 5–8E. Why endocannabinoid release produces only a transient synaptic depression at some synapses, but more persistent LTD at others, is not fully understood.

The Major Inhibitory Neurotransmitter in the Brain

GABA is present in highly diverse inhibitory interneurons and projection neurons throughout the brain. In the past, the action of GABA on its receptors was considered to be solely inhibitory, based on the observation that GABA receptor activation moves the membrane potential of a cell away from action potential threshold. However, recent evidence suggests that the role of GABA is more complex. During brain development, GABA also may function as an excitatory neurotransmitter because in certain circumstances GABA transmission can depolarize neurons. In addition, GABA released from local-circuit inhibitory interneurons assists in generating membrane oscillations such as the theta rhythm (Chapter 13).

The wide-ranging and ubiquitous role of GABA as an inhibitory transmitter means that enhancement of GABAergic function is a highly effective approach for the treatment of some forms of anxiety, insomnia, and epilepsy. GABA may play a role in diverse neuropsychiatric disorders, including epilepsy, Huntington disease, tardive dyskinesia, alcoholism and other addictions, and sleep disorders. Animal models of epilepsy have been generated by use of agents that compromise GABAergic transmission; such interference results in an imbalance of excitation and inhibition and leads to hyperexcitability and various forms of epileptiform activity.

Synthetic and Degradative Pathways

The portion of cellular GABA that functions as a neurotransmitter is formed by a metabolic pathway commonly referred to as the GABA shunt 5–9. As with glutamate synthesis, the most common precursor of GABA is glucose; pyruvate also can act as a precursor. The first step in the GABA shunt is the conversion of α-ketoglutarate into glutamate by the action of α-oxoglutarate transaminase (GABA transaminase [GABA-T]). Glutamic acid decarboxylase (GAD) then catalyzes the decarboxylation of glutamic acid to produce GABA. Autoantibodies directed against GAD are the cause of stiff person syndrome, an example of an autoimmune disorder that targets the CNS (Chapter 12).

Like most neurotransmitters, GABA is packaged into small synaptic vesicles in presynaptic terminals for release into the synaptic cleft. Like the vesicular glutamate transporter, the VGAT is highly dependent on the electrical potential across the vesicle membrane (Chapter 3). The VGAT and vesicular glutamate transporter differ from the vesicular transporters for monoamines and acetylcholine in terms of this bioenergetic dependence. Specific inhibitors of vesicular GABA transport have not yet been identified.

After it is released, GABA is removed from the synaptic cleft by the actions of several types of plasma membrane GABA transporters (GATs; Chapter 3). Through this action GABA can be returned to GABAergic nerve terminals where it is repackaged for release; it is also taken up by glial cells. In glia, GABA is metabolized into succinic semialdehyde by the action of GABA-T. To conserve the available supply of GABA, this transamination step occurs only when the precursor α-ketoglutarate is also present to accept the amino group removed from GABA. GABA is then converted back into glutamic acid, which is transferred back to the neuron. GABA-T inhibitors are being developed for use as anticonvulsant agents. One such drug, vigabatrin (γ-vinyl-GABA), is now used for the treatment of several forms of epilepsy. A model GABAergic synapse is depicted in 5–10.

Release and Reuptake

Once released, free synaptic GABA is taken back into the presynaptic terminal or into glial cells by plasma membrane GATs. This is the major mechanism for terminating the synaptic actions of GABA. Such transport requires extracellular Na⁺ and Cl⁻; two Na⁺ ions and one Cl⁻ ion are transported for each GABA molecule. Molecular cloning has revealed the genes for four distinct GATs (GAT-1, GAT-2, GAT-3, and the low-affinity transporter BGT-1) (Chapter 3). GATs are expressed on nerve terminals and glial cell membranes throughout the nervous system.

Hydroxynipecotic acid, a rigid GABA analogue, has been used as an inhibitor of GABA transport. Its selectivity for glial transporters is approximately 20-fold greater than its selectivity for GATs on nerve terminals. Related inhibitors nipecotic acid and guvacine block nerve terminal and glial transporters with similar potency. The cloning of GATs led to a clearer understanding of their pharmacology. Guvacine, nipecotic acid, and hydroxynipecotic acid are equipotent blockers of the GAT-1 transporter, with IC₅₀s between 20 and 40 μM. L-2,4-Diaminobutyric acid (L-DABA) and ACHC are approximately four times less potent, and β-alanine, hypotaurine, and taurine all have a low potency at GAT-1. One of the most potent compounds at this transporter is NNC-711, a lipophilic derivative of guvacine with an IC₅₀ of less than 400 nM. NNC-711 is 40,000 times more selective for GAT-1 than for GAT-3. The anticonvulsant drug tiagabine and the lipophilic inhibitors Cl-966 and SKF89976 also act at the GAT-1 transporter. By inhibiting GABA transport, these drugs act to potentiate the inhibitory effects of GABA on CNS function.

GABA Receptors

Like glutamate receptors, GABA receptors are divided into two main functional groups. The ionotropic GABA_A receptor is a heterooligomeric protein complex that consists of a GABA binding site coupled to an integral Cl⁻ channel 5–11. This receptor is the site of action for anxiolytic benzodiazepines and other sedative–hypnotics. GABA_A receptors are blocked by the convulsant bicuculline, a competitive antagonist at the GABA binding site.

GABA_B receptors belong to the superfamily of G protein–coupled receptors. Thus, they are considered to be metabotropic: their ligand-binding domain is not directly associated with an ion channel effector. These receptors, which are resistant to bicuculline, are activated by baclofen, a competitive agonist, and inhibited by phaclofen, a competitive antagonist.

Recent evidence supports the existence of an ionotropic GABA_C receptor that is distinct from GABA_A and GABA_B receptors and resistant to the actions of both bicuculline and baclofen. These receptors are found in many CNS locations, primarily the retina, but also including the cerebellum, hippocampus, optic tectum, and spinal cord. It has been hypothesized that the more complex GABA_A receptor evolved from the simpler, possibly homomeric, GABA_C receptor. However, little is known about the functional significance of the GABA_C receptor, and it is likely just a subclass of GABA_A receptors.

Molecular composition of GABA_A receptors

The cloning of the first GABA_A receptor subunits occurred in 1987. Since then, 19 different subunits have been identified and placed into 8 functionally distinct families. The α subfamily comprises six known subunits, the β subfamily three, the ρ subfamily three, and the γ subfamily three; the δ, ε, θ, and π subfamilies each have one member 5–2. GABA_A subunits are approximately 50 kDa in size. All possess a similar putative membrane topology, comprising a long N-terminal extracellular domain, four α-helical transmembrane-spanning segments (M1–M4), a long intracellular sequence between M3 and M4, and a short extracellular C-terminal loop (see 5–4). A high degree of conservation exists among members of a subfamily, with 70% to 80% sequence homology at the amino acid level. Although GABA_A subunits exhibit a low sequence homology (approximately 10%–25%) with glycine, 5HT₃, and nicotinic acetylcholine receptors, they are distantly related and are placed within the same superfamily of ionotropic receptors.

The heterogeneity of GABA_A receptors is partly due to the existence of multiple splice variants. Alternative exon splicing generates, for example, two versions of the γ₂ and of the β₂ subunits. These isoforms differ in terms of the presence of a short peptide sequence between TMs M3 and M4, which carries a protein kinase C phosphorylation site (see 5–2).

Heteromeric subunit assembly can occur among members of the α, β, γ, δ, and ε subfamilies; members of the ρ subfamily, which are predominantly expressed in the retina, are believed to form the GABA_C receptors. The π subfamily is expressed primarily in reproductive tissues and also is believed to exist only in homomeric form.

Like the homologous glycine, 5HT₃, and nicotinic acetylcholine ionotropic receptors, GABA_A receptors probably exist as a pentameric complex positioned around a water-filled ion-conducting pore. Within most receptor complexes, members of α, β, and γ subunit families are believed to be identical; that is, a given receptor is not expected to contain two different forms of an α (or β or γ) subunit. However, a minority of receptors may contain more than one form. Based on the large number of receptor subunits and their potential combinations, calculations indicate that more than 2000 distinct GABA_A receptors may be formed. However, evidence suggests that less than 20 different GABA_A receptors are widely expressed, and a relatively few subtypes predominate.

The composition of GABA_A receptors in vivo is ultimately influenced by the relative levels of expression of subunits in a given cell type. The most prevalent subunit in the brain is α₁, with the major GABA_A receptor subtype in brain having a stoichiometry of α₁β₂γ₂. Receptors containing the α₂ subunit are most abundant in regions where the α₁ subunit is absent or expressed at low levels, such as the hippocampus, striatum, and olfactory bulb. Similarly, the α₃ subunit is expressed in regions complementary to the α₁ subunit, including the lateral septum, reticular nucleus of the thalamus, and brainstem nuclei. Notably, the α₆ subunit is expressed almost exclusively in cerebellum.

Surprisingly, the molecular composition of GABA_A receptors at different subcellular locations within a single cell may differ. The mechanisms responsible for this targeting remain unknown but likely involve anchoring proteins, based on analogy with mechanisms responsible for glutamate receptor clustering at synapses (see 5–1). Gephyrin is one protein that has been established to be important for the trafficking and clustering of GABA_A receptors as well as glycine receptors.

GABA_A receptor subunit expression changes during brain development, and it is likely that such plasticity is maintained throughout adult life. Indeed GABA_A receptor expression is altered in syndromes such as epilepsy and alcoholism, and such changes may contribute to their pathophysiology and also offer clues for the development of novel treatments.

GABA_A receptor pharmacology

GABA_A receptor pharmacology originated in the 1970s, when researchers discovered that the convulsant alkaloid bicuculline antagonizes certain inhibitory actions of GABA. This observation paved the way for the discovery of GABA’s role as the principal inhibitory neurotransmitter in the brain. Other antagonists of GABA_A receptors include pitrazepin, the convulsant securinine, the aminidine steroid analogue RU5135, and the pyridazinyl derivative SR95531 (gabazine). In addition to GABA, selective agonists at this site include muscimol, isoguvacine, 4,5,6,7-tetrahydroisoxazolo[5,4-c]-pyridone (THIP), and piperidine-4-sulfonic acid (see 5–2). Competitive agonists and competitive antagonists of GABA_A receptors interact with the GABA binding site (see 5–11), which for most receptor complexes is located on the β subunit. However, the affinity of GABA for the receptor can be modulated by other subunits in the heteromeric complex. For example, GABA receptors whose subunit composition is α₁β₁γ₂ have an EC₅₀ for GABA of 41 μM, whereas α₃β₁γ₂ receptors have an EC₅₀ of approximately 100 μM. In addition to the previously mentioned competitive GABA_A receptor antagonists are antagonists that do not bind to the GABA binding site. The best known are the potent convulsants pentylenetetrazol and picrotoxin (and the related picrotoxinin), which appear to bind at or near the Cl⁻ channel pore and occlude ion flow.

In the early 1970s, benzodiazepines were proven to potentiate the effects of GABA. Shortly thereafter, investigators discovered that benzodiazepines potentiate GABAergic inhibitory synaptic transmission by binding directly to the GABA_A receptor. The benzodiazepine site, which is clearly distinct from the GABA-binding site, is an allosteric modulatory site on the GABA_A receptor pentamer (see 5–11). Recent evidence indicates that the site is formed by the apposition of α and β subunits but also requires the presence of a γ subunit, which itself is not part of the benzodiazepine binding site.

Drugs targeted at the benzodiazepine site exist on a continuum that ranges from full agonist to full inverse agonist. Indeed, GABA_A receptor pharmacology played an important role in the development of the concept of inverse agonists (Chapter 1). Agonists at the benzodiazepine site, including anxiolytic benzodiazepines such as diazepam, chlordiazepoxide, lorazepam, alprazolam, and clonazepam, to name a few, act to increase the receptor’s affinity for GABA at its own binding site and consequently increase the frequency of channel openings. Such drugs, in addition to being anxiolytic, are anticonvulsant and sedative and sometimes function as muscle relaxants. Partial agonists increase the frequency with which channels open but exert a milder effect on GABA_A receptor functioning. Conversely, inverse agonists such as β-carboline, or β-CCE, decrease both the frequency of channel openings and the efficacy of GABA binding, and thereby antagonize GABA_A receptor function. Such inverse agonists tend to be convulsant and anxiogenic, although a weak partial inverse agonist may exert less of an activational effect. Antagonists such as flumazenil typically bind to and occlude the benzodiazepine-binding site, but do not affect channel function; they can be used to reverse the actions of an agonist or inverse agonist, as in the treatment of an overdose with a benzodiazepine agonist.

The responsiveness of GABA_A receptors to drugs targeted at the benzodiazepine site varies considerably. Although the γ₂ subunit appears necessary for benzodiazepine binding, the presence of a γ₁ subunit may exert the opposite effect. Of particular interest are the findings that numerous benzodiazepine site agonists and inverse agonists exert α subunit–selective effects on GABA_A receptors. For example, α₄ and α₆ subunits have a low affinity for classical benzodiazepines, particularly when combined with the δ subunit. These discoveries have important clinical ramifications, as discussed later in this chapter.

GABA_A receptors also are the site of action for a large number of sedative–hypnotic and anesthetic agents, including barbiturates and related drugs, ethanol and other alcohols, volatile anesthetics, and certain neurosteroids. Barbiturates and other sedative–hypnotics, such as methaqualone, chloral hydrate, and propofol, among many others, exert a modulatory effect on the GABA_A receptor similar to the effect produced by benzodiazepine agonists, but do so by binding to a distinctly separate location, most likely near the Cl⁻ channel pore (see 5–11). Barbiturates increase the duration of channel openings without affecting the frequency with which channels open. They can exert this effect even in the absence of GABA, which makes barbiturates more potent and potentially dangerous than benzodiazepines. Pentobarbital and phenobarbital are the most commonly used barbiturates; phenobarbital, because of its long duration of action, has been in use as an anticonvulsant since the early part of the 20th century. Variation in the subunit composition of GABA_A receptors does not seem to alter their sensitivity to barbiturates.

The effects that ethanol exerts on the GABA_A receptor complex are similar to those produced by benzodiazepines and barbiturates: it facilitates GABA’s ability to activate the receptor and prolongs the time that the Cl⁻ channel remains open. Interestingly, the alternatively spliced forms of the γ₂ subunit include a short form (γ_2S) and a long form (γ_2L), and ethanol potentiates GABA_A responses only in receptors containing γ_2L. This splice variant also contains a consensus sequence for protein kinase C phosphorylation; when this sequence is mutated, the receptor’s sensitivity to ethanol is lost.

Volatile anesthetics such as enflurane represent another class of drugs that prolong channel opening of GABA_A receptors (Chapter 13). Molecular mutagenesis has been used to identify sites in most GABA_A receptor subunits that are responsible for the potentiating effects of both ethanol and volatile anesthetics. Given their location, typically near the extracellular regions of TMs M2 and M3, these sites may be part of the protein pocket that binds both ethanol and the volatile anesthetics, although this hypothesis remains unproven.

Neurosteroids, such as 3α,5α-tetrahydroprogesterone (3α,5α-THPROG), similarly promote channel opening of the GABA_A receptor. Receptors containing the δ subunit are particularly sensitive, although most subtypes have some sensitivity to neurosteroids. GABA acts as only a partial agonist at δ subunit–containing receptors, and neurosteroids can serve to promote it into a full agonist. α₁ and α₃ subunits are more susceptible than other α subunits, while γ₁-containing receptors are relatively resistant compared with receptors expressing γ₂ or γ₃. Neurosteroids are produced by neurons from progesterone, and steroid levels are altered by stress, aging, and pregnancy, and also by ethanol, γ-hydroxybutyrate (GHB; often referred to as a date rape drug; see Chapter 13), and some antidepressants such as fluoxetine. Neurosteroids can produce anxiolytic, analgesic, sedative, hypotonic, anticonvulsant, and anesthetic effects. Curiously, the neurosteroid antagonist 17PA (1717-phenylandrost-16-en-3-0l) blocks the effects of 3α,5α-THPROG and related steroids, but does not inhibit the action of 3α,5β-THPROG and other 5β steroids, suggesting that there might be alternate binding sites for these steroids. See Chapter 10 for further discussion of neurosteroids.

Because barbiturates, benzodiazepines, and ethanol have related actions on a shared receptor substrate, the use of these agents can result in clinical complications. Their pharmacologic synergy increases the dangers associated with overdose and can lead to the development of cross-dependence. In fact, such cross-dependence often is exploited in alcohol detoxification: long-acting benzodiazepines (eg, chlordiazepoxide) are the treatment of choice and are used to prevent the emergence of alcohol withdrawal symptoms such as hallucinosis, delirium tremens, and seizures, which are potentially fatal.

Prolonged exposure to benzodiazepines results in tolerance to some of their pharmacologic actions, particularly to their sedative and anticonvulsant effects. Their effectiveness in the prolonged treatment of chronic insomnia or long-term treatment of epilepsy is therefore limited. Surprisingly, tolerance to the anxiolytic effects of benzodiazepines may not develop; patients often benefit from these effects for many years. In addition to tolerance, benzodiazepines (particularly at higher doses) and related sedative–hypnotics can produce profound dependence, which is manifested by rebound withdrawal symptoms that appear when drug administration ceases.

Both tolerance and dependence are characterized by a downregulation of GABAergic transmission and by a reduction in the benzodiazepine modulation of GABA_A currents. Despite considerable research, little consensus exists as to the mechanisms that underlie tolerance caused by prolonged exposure to benzodiazepines. Several reports have suggested that tolerance is caused by the downregulation of the benzodiazepine recognition site on GABA_A receptor complexes, indicating that distinct receptor subunits that lack such sites may be expressed in response to continued exposure to benzodiazepines.

The extensive heterogeneity of GABA_A receptors, and the broad array of pharmacologic agents that are currently known to modulate these receptors, offers hope for more specific therapeutic agents. Such optimism has been bolstered by the hypothesis that GABA_A receptors responsible for mediating the disparate effects of benzodiazepine agonists are located in distinct brain regions and are composed of distinct α subunits. Receptors containing the α₁ subunit are most important for the sedative actions of benzodiazepines, while α₂- or α₃-containing receptors primarily mediate the anxiolytic effects and α₅-containing receptors may be particularly important for the cognitive effects of these drugs. For example, zolpidem, eszopiclone, and zaleplon, nonbenzodiazepine sedative–hypnotics that bind at the benzodiazepine site of the GABA_A receptor, are somewhat selective for GABA_A receptors that possess the α₁ subunit. Another endorsement of this hypothesis was the development of RO 15-4513, a benzodiazepine site antagonist that acts selectively on GABA_A receptors that contain the α₆ subunit, which is concentrated in the cerebellum. This drug reduces the ataxia produced by ethanol without altering ethanol’s other actions. Thus, researchers have been encouraged in their attempts to develop drugs that selectively inhibit the anxiolytic, sedative, or anticonvulsant actions of nonselective agents such as diazepam.

Drug research is also aimed at developing agents that target GABA_A receptors without producing tolerance or dependence, or triggering abuse. One strategy may involve the development of partial agonists that are sufficiently efficacious but less likely than full agonists to elicit compensatory adaptations. A related approach involves the development of partial inverse agonists that are selective for GABA_A receptors associated with cognitive function and attentional states. Although a full inverse agonist is convulsant and anxiogenic, a weak partial inverse agonist with the right α subunit selectivity may effectively enhance cognition. Indeed, it is conceivable that a single agent might reduce anxiety and enhance cognitive function simultaneously by acting as a partial agonist at anxiolytic GABA_A receptors, and as a weak inverse partial agonist at cognitive GABA_A receptors. However, drugs with these various mechanisms of action have not yet been developed.

Early studies of benzodiazepines led to the discovery of a second target for some of these drugs named the peripheral benzodiazepine receptor. This protein, now known as translocator protein (TSPO), is expressed ubiquitously in the outer mitochondrial membrane, where it regulates steroidogenesis. TSPO is unrelated to the psychotropic effects of benzodiazepines.

GABA_B receptors

GABA_B receptors were initially recognized because of their insensitivity to bicuculline and other GABA_A ligands. Subsequently, the GABA analogue and muscle relaxant known as baclofen proved to be a potent agonist at GABA_B receptors. The physiologic roles of GABA_B receptors were further elucidated after the development and analysis of specific antagonists such as phaclofen, CGP 35348, and 2-OH-s-(2)-saclofen.

Molecular cloning techniques revealed that GABA_B receptors are members of the G protein–coupled receptor superfamily and contain seven TMs (see 5–6). GABA_B receptors, which are larger than most G protein–coupled receptors, are composed of 850 to 960 amino acids and are structurally similar to mGluRs. The receptors are G_i/o coupled and are a heterodimer of two major GABA_B subunits, GABA_BR₁ and GABA_BR₂. GABA_BR₁ binds ligand and has two isoforms: GABA_BR_1a and R_1b, while GABA_BR₂ subunit couples to G_i/o. The GABA_B receptors were the first G protein–coupled receptors known to function as multimeric complexes, although many more have been identified (eg, opioid receptors; see Chapter 7).

GABA_B receptors are expressed on both presynaptic and postsynaptic membranes, with their localization regulated by a series of accessory proteins. Like other G_i/o-linked receptors, GABA_B receptors open K⁺ channels, decrease Ca²⁺ conductances, and inhibit adenylyl cyclase. Compared with GABA_A receptors, postsynaptic GABA_B receptors produce a slower but longer-lasting form of inhibition, an effect largely ascribed to the opening of GIRK channels. In the hippocampus, thalamus, and cortex, low-intensity stimulation of inhibitory interneurons evokes inhibitory postsynaptic potentials (IPSPs) mediated entirely by GABA_A receptors; IPSPs mediated by GABA_B receptors require much stronger stimulation or stimuli of longer duration and higher frequency. These findings have led to the conclusion that the synaptic location of these receptor subtypes may differ. GABA_A receptors are believed to localize at the synapse proper, directly opposite the site of GABA release, and can therefore be activated by the release of a single quantum of neurotransmitter. In contrast, GABA_B receptors may be extrasynaptic. An extrasynaptic location would explain their need for high-intensity stimulation or for prolonged, high-frequency stimuli, either of which would increase GABA release and make it possible for GABA to diffuse out of the synaptic cleft.

On the presynaptic side, GABA_B receptors can function as autoreceptors and inhibit further GABA release from GABAergic nerve terminals. Other GABA_B receptors are located on excitatory presynaptic terminals, where their activation inhibits the release of glutamate. Receptors at this location are activated by synaptically released GABA that has, in a paracrine fashion, spilled out from an inhibitory synapse and diffused to adjacent excitatory synapses 5–12. The activation of presynaptic GABA_B receptors can also inhibit the release of several other neurotransmitters, including norepinephrine, dopamine, serotonin, and substance P. GABA_B-mediated inhibition of transmitter release occurs, at least in part, through the inhibition of Ca²⁺ channels and the consequent decreases in Ca²⁺ influx in response to action potentials. Such inhibition also may involve K⁺ conductances, or direct modulation of the release machinery in the terminal via regulation of cAMP signaling cascades.

The only major drug in clinical use that interacts with GABA_B receptors is baclofen, which acts as a muscle relaxant and is used to decrease spasticity in a variety of neurologic disorders. It is used less frequently for the treatment of trigeminal neuralgia. Inhibition of glutamate release is believed to underlie its clinical efficacy. GHB, mentioned earlier, is a weak agonist at GABA_B receptors. Other applications for GABA_B receptor agonists, or PAMs, are currently under investigation and may include their use in the treatment of seizure disorders, anxiety, and depression. GABA_B antagonists may also be used in the future to enhance cognition.

Synthetic and Degradative Pathways

Most of the glycine in the mammalian CNS is synthesized de novo from glucose through serine. Serine is converted to glycine by serine hydroxymethyltransferase (SHMT), a pyridoxal phosphate-dependent enzyme. Pyridoxal phosphate is a derivative of vitamin B₆. It is believed that this conversion occurs in the mitochondrial compartment because the distribution of mitochondrial SHMT mirrors the distribution of glycine 5–13.

Glycine is an amino acid primarily used for the synthesis of proteins. Only a small fraction of the cellular pool of glycine in a small subset of neurons is packaged into small synaptic vesicles for release as neurotransmitter. This packaging is mediated by a vesicular transport system identical to that previously described for GABA (VGAT). Glycine acts as a neurotransmitter predominantly in the brainstem and spinal cord, where, like GABA, it is a principal mediator of inhibitory neurotransmission.

The degradation of glycine occurs mainly by means of the glycine cleavage system (GCS), which has four protein components (see 5–13). A defect in the GCS, which is located in the inner mitochondrial membrane, causes a group of metabolic disorders termed nonketotic hyperglycinemias, which are characterized by high concentrations of glycine in the cerebrospinal fluid 5–3.

Glycine Release and Reuptake

As with the release of other neurotransmitters, the arrival of an action potential in the presynaptic terminal of a glycinergic neuron initiates a cascade involving vesicular fusion and the release of glycine into the synaptic cleft. Glycine thus released is free to diffuse and bind with its receptors clustered on the postsynaptic face of adjacent cells.

Glycine is removed from the synaptic cleft by reuptake transporters located on the plasma membranes of glial cells and of presynaptic nerve terminals (see 5–13). The electrochemical gradients of Na⁺ and Cl⁻ assist in transporting glycine against its concentration gradient. This uptake mechanism is electrogenic and results in a net movement of positive charge. Two glycine transporters have been cloned thus far: GLYT1 and GLYT2. GLYT1 exists in three isoforms, which most likely are generated by alternative splicing; these exhibit no known variation in their uptake properties but possess distinct patterns of expression in the CNS.

GLYT1 and GLYT2 can be differentiated pharmacologically; for example, only GLYT1 isoforms are sensitive to sarcosine (N-methylglycine). Both GLYT1 and GLYT2 are expressed in caudal regions of the CNS, a location consistent with their role in terminating glycinergic neurotransmission. In addition, GLYT1 is expressed in several forebrain regions that are devoid of glycinergic transmission. This forebrain GLYT1 might regulate NMDA glutamate receptor function in these areas by controlling levels of glycine in the extracellular fluid available to allosterically modulate these receptors (see earlier sections of this chapter). This would suggest the GLYT1 inhibitors may be useful clinically in promoting NMDA receptor function. GLYT1 is expressed in both astrocytes and neurons, whereas GLYT2 is localized on axons and terminal boutons of neurons that contain vesicular glycine.

Glycine Receptors

Glycine receptors are primarily restricted to the brainstem and spinal cord. Like GABA_A receptors, the glycine receptor is a receptor ionophore that contains a Cl⁻ channel. It is also similar in size to the GABA_A receptor and is believed to possess a quasisymmetrical pentameric structure that surrounds a water-filled ion conduction pore. Glycine receptors, which are unrelated to the glycine-binding sites present on NMDA glutamate receptors, are defined pharmacologically by strychnine, a selective antagonist and a potent convulsant. Agonists at these receptors, which are competitive and limited to a few ligands, are ranked according to potency as follows: glycine > β-alanine > taurine > L- and D-alanine > L-serine >> D-serine. The potent convulsants picrotoxin and picrotoxinin are noncompetitive inhibitors of some of these receptors and are believed to interact directly with the receptor’s ion channel to block Cl⁻ permeation.

Glycine receptors are composed of two types of glycosylated integral transmembrane proteins: 48-kDa α subunits and 58-kDa β subunits, which are closely associated with gephyrin, a large (93-kDa) cytoplasmic protein. The α and β subunits possess similar sequence homologies (see 5–4), which in turn are similar to those of other receptor ionophores; thus, glycine receptors belong to the same superfamily as nicotinic cholinergic, GABA_A, and 5HT₃ receptors. Like other members of this superfamily, glycine receptors possess a large N-terminal extracellular domain and four transmembrane-spanning regions. The glycine receptor pentamer is formed by α subunits either alone or in association with a β subunit; only the α subunit contains the functional glycine-binding site. Four α subunits and a single β subunit have been cloned, and each of these is known to have several splice variants. Gephyrin associates with the intracellular region of the β subunit and links the receptor to cytoplasmic tubulin; it thereby functions as an anchoring protein for glycine receptors and controls receptor clustering.

The composition of the glycine receptor is developmentally regulated. In rat embryonic tissue, these receptors are exclusively composed of α₂ subunits; however, in the adult they typically are composed of 3α₁ and 2β subunits 5–14. The transient expression of α₂ subunits during development occurs in most regions of the CNS. In mature animals, only sparse neurons in the spinal cord exhibit persistent expression of α₂. In adults, the distribution of α₁ corresponds closely to the distribution of strychnine-binding sites throughout the spinal cord, in brainstem nuclei, and in the reticular, auditory, and vestibular systems. An increase in α₃ transcription during development is restricted to the infralimbic system; in the adult such transcription is restricted to the hippocampus and cerebellar granule cells. Interestingly, β mRNA transcripts are found throughout the mammalian adult CNS, although β subunits alone are incapable of forming functional glycine receptors.

The pharmacology of glycine receptors remains limited. Among the known antagonists of these receptors, strychnine is the most commonly used experimentally because it blocks receptors composed of all α subunits. The binding pocket of the glycine receptor binds both strychnine and glycine agonists and is composed of several discontinuous domains of the α subunit. Interestingly, the residues important for ligand binding are homologous to amino acid positions that determine the ligand-binding affinities of nicotinic cholinergic and GABA_A receptors; thus, it appears that the ligand-binding pockets of all members of this receptor superfamily may share a common architecture. Other glycine receptor antagonists include cyanotriphenylborate (CTB), which preferentially binds to receptors containing the α₁ subunit; picrotoxin, which most effectively blocks recombinant α₂ homomeric receptors; and picrotoxinin, which most effectively blocks α₁ homomeric receptors. The presence of the β subunit enables glycine receptors to resist antagonism by picrotoxinin.