Sticking together.

R signals at the neuromuscular junction and at synapses between neurons are carried by small molecules, called neurotransmitters, that are released from the presynaptic terminal and bind to ligand-gated ion channels in the postsynaptic membrane. When transmitter binds, a small pore opens through which ions flow, resulting in a transient depolarization or hyperpolarization of the membrane and translating the chemical signal into an electrical one. The transmitters at the neuromuscular junction, and at excitatory synapses in the central nervous system, are acetylcholine and glutamate, respectively. Inhibitory signals are carried by two major transmitters: glycine, predominantly in the spinal cord, and g-aminobutyric acid, or GABA, predominantly in the brain. Receptors for these transmitters are important targets for drugs used to treat mental disorders, or to modulate sleep and mood. In particular, benzodiazepine-related drugs, such as Valium, Halcion, and Xanax, which are widely used for the treatment of anxiety and insomnia, appear to act by binding directly to a specific site on the GABA type A (GABAA) receptor, the principal GABA-gated ion channel (1, 2). Molecular biologists have spent much productive effort over the last decade working out the molecular structures of receptors for each of the major transmitters, while biophysicists have unraveled the detailed kinetics of transmitter binding and gating (1, 3, 4). The pharmaceutical industry concentrates enormous resources on determining the specificity and physiological consequences of binding of pharmacological agents to these receptors. In recent years, many researchers have turned to a new issue governing receptor function: What is the immediate protein environment of the receptors, and how does that environment influence receptor function at different synapses? Receptors for the rapid transmitters are not distributed randomly over the surface of the membrane. Rather, they are ‘‘targeted’’ to the postsynaptic membrane and are thus concentrated adjacent to the sites of transmitter release. The acetylcholine receptor is an extreme example of this process, clustering into a near crystalline array at the neuromuscular junction (5). In addition, the receptors are tightly associated with a cytoplasmic meshwork of proteins comprising regulatory enzymes that may modify the receptor itself to alter its kinetics or location, or transmit biochemical signals deeper into the cytoplasm (6–9). The N-methyl-D-aspartate (NMDA) subtype of the glutamate receptors is an extreme example of this situation. It is equipped with unusually long carboxylterminal ‘‘tails’’ that extend into the cytoplasm and associate with a variety of scaffold and signaling molecules (10). Several recent papers, including two in PNAS, by Kneussel et al. (11) and by Chen et al. (12), have begun to clarify the mechanism of clustering of the inhibitory receptors for glycine and GABA, and the nature of the matrix of intracellular proteins associated with these receptors. A possible glycine receptor clustering protein of '93 kDa, termed gephyrin (from the Greek word for bridge), originally was identified based on its copurification with the glycine receptor (13). Recent genetic experiments show that gephyrin is required for synaptic clustering of both glycine and GABAA receptors (8, 14). These experiments are puzzling, however, because gephyrin itself shows no significant affinity for GABAA receptors in vitro (8). A partial solution to the puzzle was offered in a set of experiments demonstrating that a small protein of '14 kDa, termed GABARAP (GABAA receptorassociated protein) can bind to a subunit of the GABAA receptor (15) and to gephyrin (11), perhaps forming the missing physical link between the two (Fig. 1). GABARAP has an N-terminal tubulinbinding domain and is closely related to the previously described ‘‘late-acting intra-Golgi transport factor,’’ termed p16 (11). These findings suggest the hypothesis that GABARAP may link gephyrin and the GABAA receptor together as they are carried along microtubules and targeted to the proper membrane site (16). The propensity of gephyrin to form multimeric assemblies in vitro may be important for later stabilization of receptor complexes at the postsynaptic site. Kneussel et al. (11) have found a f ly in the hypothetical ointment, however. In cortical neurons in culture, where GABARAP previously had been shown to colocalize with GABAA receptors (15), they found that GABARAP is primarily in intracellular vesicles and does not colocalize with gephyrin. Furthermore, in the retina, GABAA receptors and gephyrin are tightly colocalized, but GABARAP shows no significant colocalization with either of them. Thus, in neurons GABARAP is not usually clustered at inhibitory synaptic sites; thus, its interaction with the GABAA receptor and gephyrin may be important for cellular functions other than receptor an-