Portal de Periódicos da CAPES

PDZs and Receptor/Channel Clustering: Rounding Up the Latest Suspects

1996; Cell Press; Volume: 17; Issue: 4 Linguagem: Inglês

10.1016/s0896-6273(00)80190-7

1097-4199

Morgan Sheng,

Neurobiology and Insect Physiology Research

Electrical signaling in the nervous system is dependent on the function of a wide variety of ion channels in the surface membrane of neurons. These neuronal ion channels do not diffuse freely in the membrane, but are typically localized at specific subcellular locations, such as axon terminals, postsynaptic sites, and nodes of Ranvier. The regulated distribution of ion channels on the cell surface is clearly critical for signaling within and between neurons. An obvious example is the postsynaptic concentration of a specific class of neurotransmitter-gated ion channels (ionotropic receptors) underneath the appropriate presynaptic terminal. The molecular mechanisms underlying targeted ion channel clustering in neurons have been the focus of considerable interest, not only because this is a fundamental cell biological problem but also because an understanding of these mechanisms should throw light on the structural and functional organization of synapses. Studies of site-specific ion channel clustering have taken their cue from the neuromuscular junction (NMJ), where the mechanisms of postsynaptic clustering of nicotinic acetylcholine receptors (nAChRs) have been most thoroughly investigated. 43K/rapsyn is believed to be the molecule most directly responsible for nAChR clustering, though how it functions at the molecular level is still a mystery. Rapsyn is biochemically associated with nAChRs in the NMJ, it can cluster nAChRs in heterologous cells, and targeted disruption of the rapsyn gene abolishes nAChR clustering at the NMJ. A distant cousin of nAChRs is the glycine receptor (GlyR). Postsynaptic clustering of GlyRs, however, is dependent on gephyrin, a 93 kDa protein that is totally unrelated to rapsyn. Whether rapsyn and gephyrin (or relatives of these proteins) play a role in the synaptic organization of other kinds of receptors or ion channels remains an important open question. The clustering molecules rapsyn and gephyrin were initially recognized as channel-associated proteins during biochemical purification of nAChRs and GlyRs. No such clues existed, however, for the ionotropic glutamate receptors (NMDA, AMPA, and kainate receptors), which were cloned by functional expression or by molecular genetic means. Nevertheless, glutamate receptors are clearly concentrated at a subset of postsynaptic sites in the brain based on both electrophysiological and immunohistochemical evidence. Recently, the yeast two-hybrid system has been used to screen for potential clustering proteins that bind to the intracellular C-terminal tails of NMDA receptor subunits. The NR2 subunits of the NMDA receptor were found to interact specifically with a family of membrane-associated synaptic proteins (8Kim E. Cho K.-O. Rothschild A. Sheng M. Neuron. 1996; 17: 103-113Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar, 10Kornau H.-C. Schenker L.T. Kennedy M.B. Seeburg P.H. Science. 1995; 269: 1737-1740Crossref PubMed Scopus (1574) Google Scholar, 15Niethammer M. Kim E. Sheng M. J. Neurosci. 1996; 16: 2157-2163Crossref PubMed Google Scholar) homologous to the product of the Drosophila gene discs large (dlg; 18Woods D.F. Bryant P.J. Cell. 1991; 66: 451-464Abstract Full Text PDF PubMed Scopus (750) Google Scholar). In mammals, this family to date includes at least four closely related members: PSD-95/SAP90 (2Cho K.-O. Hunt C.A. Kennedy M.B. Neuron. 1992; 9: 929-942Abstract Full Text PDF PubMed Scopus (969) Google Scholar, 9Kistner U. Wenzel B.M. Veh R.W. Cases-Langhoff C. Garner A.M. Appeltauer U. Voss B. Gundelfinger E.D. Garner C.C. J. Biol. Chem. 1993; 268: 4580-4583Abstract Full Text PDF PubMed Google Scholar), SAP97/hdlg (12Lue R.A. Marfatia S.M. Branton D. Chishti A.H. Proc. Natl. Acad. Sci. USA. 1994; 91: 9818-9822Crossref PubMed Scopus (336) Google Scholar, 13Müller B.M. Kistner U. Veh R.W. Cases-Langhoff C. Becker B. Gundelfinger E.D. Garner C.C. J. Neurosci. 1995; 15: 2354-2366Crossref PubMed Google Scholar), chapsyn-110/PSD-93 (1Brenman J.E. Chao D.S. Gee S.H. McGee A.W. Craven S.E. Santillano D.R. Wu Z. Huang F. Xia H. Peters M.F. Froehner S.C. Bredt D.S. Cell. 1996; 84: 757-767Abstract Full Text Full Text PDF PubMed Scopus (1394) Google Scholar, 8Kim E. Cho K.-O. Rothschild A. Sheng M. Neuron. 1996; 17: 103-113Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar), and SAP102 (14Müller B.M. Kistner U. Kindler S. Chung W.J. Kuhlendahl S. Lau L.-F. Veh R.W. Huganir R.L. Gundelfinger E.D. Garner C.C. Neuron. 1996; 17: 255-265Abstract Full Text Full Text PDF PubMed Scopus (353) Google Scholar). In their N-terminal half, this family of proteins is characterized by the presence of three domains with a length of approximately 90 amino acids (Figure 1). (These domains are now termed PDZ domains because they were initially recognized as sequence repeats in PSD-95, Dlg, and a tight junction protein ZO-1; they are also known as DHRs or discs large homology regions.) In addition, these proteins contain an SH3 domain and a guanylate kinase-like (GK) domain in their C-terminal region (Figure 1), defining them as a subclass of the MAGUK (membrane-associated guanylate kinase) superfamily of proteins. Remarkably, a contemporaneous yeast two-hybrid screen using Kv1.4 as bait led to the realization that this same subfamily of Dlg-related proteins also binds to the cytoplasmic C-terminal tail of Shaker-type voltage-gated K+ channels (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar), a class of ion channel that is also concentrated in synapses as well as other subcellular locations. Most importantly from a functional point of view, coexpression of PSD-95 (or its relative chapsyn-110) results in the striking macroscopic clustering of Shaker K+ channels and NMDA receptors in heterologous cells (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar, 8Kim E. Cho K.-O. Rothschild A. Sheng M. Neuron. 1996; 17: 103-113Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar). But how is it that two groups of apparently unrelated (NMDA receptor and Shaker K+ channel) ion channel subunits interact with the same set of intracellular proteins? As with any whodunnit, the truth is only revealed at the end. At the very C-terminus of both Shaker and NR2 proteins are four highly conserved amino acids (consensus sequence -E-S/T-D-V). This C-terminal -E-S/T-D-V motif in the cytoplasmic tails of the ion channels is specifically recognized by the PDZ domains of the Dlg/PSD-95 family of proteins (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar, 10Kornau H.-C. Schenker L.T. Kennedy M.B. Seeburg P.H. Science. 1995; 269: 1737-1740Crossref PubMed Scopus (1574) Google Scholar, 15Niethammer M. Kim E. Sheng M. J. Neurosci. 1996; 16: 2157-2163Crossref PubMed Google Scholar). Domain analysis of PSD-95 in the above studies defined the PDZ repeats as modular protein–binding sites that recognize a short consensus peptide sequence, analogous to the better known SH2 or SH3 domains. And like SH2 and SH3 domains, PDZ domains show specificity of binding. For instance, the Shaker/NR2 C-terminal E-S/T-D-V sequence motif does not bind to PDZ domains from syntrophins, and furthermore, it shows different affinities for the three PDZ domains of PSD-95 family proteins: the second PDZ domain (PDZ2) has the highest affinity (KD < 10 nM), whereas PDZ1 has an intermediate (KD ∼30 nM), and PDZ3 has relatively low affinity (KD > 1 μm) (14Müller B.M. Kistner U. Kindler S. Chung W.J. Kuhlendahl S. Lau L.-F. Veh R.W. Huganir R.L. Gundelfinger E.D. Garner C.C. Neuron. 1996; 17: 255-265Abstract Full Text Full Text PDF PubMed Scopus (353) Google Scholar). The basis for specific peptide recognition by modular PDZs has been beautifully clarified by the X-ray crystallographic structure of a PDZ domain from PSD-95, complexed with its cognate peptide ligand (4Doyle D.A. Lee A. Lewis J. Kim E. Sheng M. MacKinnon R. Cell. 1996; 85: 1067-1076Abstract Full Text Full Text PDF PubMed Scopus (929) Google Scholar). A loop formed by the conserved -G-L-G-F- sequence in the PDZ domain binds to the carboxylate group of the terminal valine, explaining the specificity for a C-terminal peptide sequence. A prominent hydrophobic pocket on the surface of the PDZ is filled by the side chain of the terminal valine, accounting for the requirement for this hydrophobic amino acid at the very C-terminus of the peptide. Further side chain interactions explain the specific recognition of serine or threonine at the −2, and glutamine at the −3 positions. The penultimate (−1) residue of the peptide, however, makes only a backbone contact with the PDZ. This finding may account for the previously surprising result in which the −1 residue of the E-T-D-V sequence of the K+ channel C-terminus could be substituted without impairing PSD-95 binding (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar), despite the conservation of aspartate (-D-) at the −1 position in Shaker and NR2 proteins. An important conclusion from the three-dimensional X-ray structure is that at least four residues at the peptide C-terminus are clearly involved in specific PDZ binding. Thus, although the protein databases contain an extensive list of polypeptides that terminate with the sequence -S/T-X-V (presciently noted by Kornau et al. [1995]), where X is any amino acid, it cannot be presumed that each will bind a PDZ domain. Moreover, C-terminal hydrophobic amino acids other than valine may be accommodated by the PDZ domain, since the inward rectifying K+ channel subunit Kir2.3 (which ends with -E-S-A-I, where I is isoleucine) has also been shown to bind PSD-95 (Cohen et al., 1996 [this issue of Neuron]). Even more significantly, several of the critical contact residues defined in the crystallized PDZ domain from PSD-95 (4Doyle D.A. Lee A. Lewis J. Kim E. Sheng M. MacKinnon R. Cell. 1996; 85: 1067-1076Abstract Full Text Full Text PDF PubMed Scopus (929) Google Scholar) are different in other more distantly related PDZs, leading to the prediction that the various subclasses of PDZ domains will recognize different C-terminal peptide sequences. Indeed, PDZs of different flavors are found in a wide diversity of membrane-associated proteins (see Figure 1; for a more comprehensive list, see4Doyle D.A. Lee A. Lewis J. Kim E. Sheng M. MacKinnon R. Cell. 1996; 85: 1067-1076Abstract Full Text Full Text PDF PubMed Scopus (929) Google Scholar), and some appear to have binding specificities for C-terminal sequences that are distinct from -E-S/T-X-V. For instance, the PDZ domain of LIN-2/CASK binds to the C-terminal tail of neurexins, which ends in -E-Y-Y-V (6Hata Y. Butz S. Südhof T.C. J. Neurosci. 1996; 16: 2488-2494Crossref PubMed Google Scholar). The prevalence of a critical hydroxyl-containing residue at the −2 position (S/T and perhaps Y) has provoked interest in the potential regulation of PDZ interactions by phosphorylation. Indeed, phosphorylation of the −2 serine residue of Kir2.3 by cAMP-dependent protein kinase inhibits Kir2.3 binding to PSD-95 (3Cohen N.A. Brenman J.E. Snyder S.H. Bredt D.S. Neuron. 1996; 17 (this issue)Abstract Full Text Full Text PDF PubMed Scopus (225) Google Scholar), suggesting a novel mechanism for regulating ion channel interactions with intracellular proteins. When expressed alone in heterologous cells, PSD-95 or its close relative chapsyn-110 are diffusely distributed throughout the cell; this is in contrast with rapsyn or gephyrin, which form macroscopic aggregates by themselves in transfected cells. Shaker and NR2 proteins do not cluster in the absence of PSD-95. Coexpression of Shaker-type K+ channels (or NR2 subunits) with either PSD-95 or chapsyn-110, however, results in the coclustering of both binding partners (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar, 8Kim E. Cho K.-O. Rothschild A. Sheng M. Neuron. 1996; 17: 103-113Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar). This mutual clustering phenomenon has led to a simple cross-linking model for the mechanism of clustering. Since ion channels are multimeric and present several cytoplasmic C-termini for PDZ binding (four in the case of tetrameric Shaker-type channels), they can be linked together like a raft in the membrane by the concatenated PDZ domains in each PSD-95 polypeptide (see diagrams of models in5Gomperts S.N. Cell. 1996; 84: 659-662Abstract Full Text Full Text PDF PubMed Scopus (214) Google Scholar). Alternatively, PSD-95 and its relatives can form multimers, in which case only a single functional PDZ domain per PSD-95 monomer is required for multivalent cross-linking and clustering. PSD-95 does indeed form a stable complex with chapsyn-110, and both proteins are recruited into the same ion channel clusters (8Kim E. Cho K.-O. Rothschild A. Sheng M. Neuron. 1996; 17: 103-113Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar), suggesting that hetero- or homomultimerization of PSD-95 and its relatives could contribute to the mechanism of aggregation of their ion channel–binding partners. Rafting of ion channels by multimerization of PSD-95 (rather than by concatenation of PDZ domains) is a more attractive mechanism, since the three different PDZ domains found within each PSD-95 monomer appear to show significant differences in binding specificity. In this case, even homomultimerization of PSD-95 would generate a submembrane lattice containing three functionally distinct PDZ-binding sites. This would provide a simple mechanism for clustering a specific mixture of ion channels (possibly with other kinds of proteins such as cell adhesion molecules) in a spatially and stoichiometrically defined manner. Indeed, there are many candidate integral membrane protein ligands (see10Kornau H.-C. Schenker L.T. Kennedy M.B. Seeburg P.H. Science. 1995; 269: 1737-1740Crossref PubMed Scopus (1574) Google Scholar) that have C-terminal sequences potentially compatible with binding to PDZs. Thus, it remains critically important to detail the binding specificities of different PDZ domains and to work out how PDZ-containing proteins like PSD-95 function to cluster their ligands. It should also be pointed out that although these heterologous expression studies reveal the propensity of PSD-95-like proteins to form macroscopic aggregates with their ion channel–binding partners, they do not address the question of how these protein clusters localize in vivo to specific subcellular domains such as neuronal synapses. What is the evidence that PDZ-containing proteins like PSD-95 are important for ion channel clustering in vivo? For one, these proteins are found at the crime scene, in close association with their binding partners (Table 1). Dlg and PSD-95 family members are highly concentrated at the inner surface of synaptic membranes, and in rat neurons they have been shown to colocalize with Shaker-type K+ channels and with NMDA receptors, at both presynaptic and postsynaptic sites (7Kim E. Niethammer M. Rothschild A. Jan Y.N. Sheng M. Nature. 1995; 378: 85-88Crossref PubMed Scopus (877) Google Scholar, 10Kornau H.-C. Schenker L.T. Kennedy M.B. Seeburg P.H. Science. 1995; 269: 1737-1740Crossref PubMed Scopus (1574) Google Scholar, 14Müller B.M. Kistner U. Kindler S. Chung W.J. Kuhlendahl S. Lau L.-F. Veh R.W. Huganir R.L. Gundelfinger E.D. Garner C.C. Neuron. 1996; 17: 255-265Abstract Full Text Full Text PDF PubMed Scopus (353) Google Scholar). Different members of the PSD-95 family, however, are differentially distributed at the cellular and subcellular level in mammalian brain: SAP97 is axonal and presynaptic, while PSD-95, chapsyn-110, and SAP102 are largely postsynaptic. The physiological significance of this differential targeting, and whether they interact with distinct proteins at different subcellular sites, is unknown.Table 1Examples of PDZ Interactions with Ion Channels and ReceptorsPDZ-Containing ProteinBinding PartnersC-Terminal SequenceOther Candidate Binding PartnersPSD-95/SAP90, SAP97/hdlg, chapsyn-110/PSD-93, and SAP102Shaker K+ channels NMDAR2 subunits Kir2.3(ETDV) (ESDV) (ESAI)Voltage-gated Na+ channel (ESIV), GIRK2 channel (ESKV), and β1 adrenergic receptor (ESKV)LIN-2/CASKNeurexins(EYYV)LIN-7LET-23/EGFR(ETCL)INADTRP Ca2+ channelaINAD has been reported to bind to an internal sequence in the cytoplasmic tail of the TRP Ca2+ channel (Shieh and Zhu 1996).The C-terminal sequence requirements for binding to PDZ domains of LIN-2/CASK or LIN-7 are not well characterized; thus, no speculations are made regarding other potential binding partners for these proteins.a INAD has been reported to bind to an internal sequence in the cytoplasmic tail of the TRP Ca2+ channel (16Shieh B.-H. Zhu M.-Y. Neuron. 1996; 16: 991-998Abstract Full Text Full Text PDF PubMed Scopus (231) Google Scholar). Open table in a new tab The C-terminal sequence requirements for binding to PDZ domains of LIN-2/CASK or LIN-7 are not well characterized; thus, no speculations are made regarding other potential binding partners for these proteins. Genetic evidence for PDZ involvement in receptor localization at specific subcellular sites has been obtained in Caenorhabditis elegans (17Simske J.S. Kaech S.M. Harp S.A. Kim S.K. Cell. 1996; 85: 195-204Abstract Full Text Full Text PDF PubMed Scopus (228) Google Scholar). LIN-2 and LIN-7 are cell junction–associated proteins that act cell autonomously in epithelial vulval precursor cells. They both contain a single PDZ domain (see Figure 1), but are otherwise unrelated at the primary sequence level. LIN-7 binds to the intracellular region of LET-23, an EGF receptor (EGFR) homolog required in the vulval induction signaling pathway. Mutation of either lin-2 or lin-7 results in a vulva-less phenotype, apparently caused by the mislocalization of LET-23/EGFR. Thus, in this paradigmatic developmental pathway, the localization of a transmembrane tyrosine kinase receptor at a cell junction is dependent on proteins with PDZs. But are PDZ-containing proteins involved in anything other than the clustering of membrane ion channels and receptors? Although recent attention has focused on PDZ interactions with the C-termini of transmembrane proteins, PDZ domains have also been shown to bind to intracellular proteins. Intriguingly, PSD-95 and chapsyn/PSD-93 interact with neuronal nitric oxide synthase, apparently by a homotypic PDZ–PDZ association (1Brenman J.E. Chao D.S. Gee S.H. McGee A.W. Craven S.E. Santillano D.R. Wu Z. Huang F. Xia H. Peters M.F. Froehner S.C. Bredt D.S. Cell. 1996; 84: 757-767Abstract Full Text Full Text PDF PubMed Scopus (1394) Google Scholar). How PDZ–PDZ interactions are mediated is unclear, but in vivo this binding would bring into close proximity of the NMDA receptor channel an enzyme that is activated by Ca2+ influx through NMDA receptors. From such findings, the idea is emerging that PSD-95 and related proteins function not just as clustering molecules for ion channels, but as multimodular scaffolds that nucleate a complex of integral membrane proteins with their downstream signaling molecules. The synaptic junction is an example par excellence of a membrane specialization at which transmembrane and intracellular signaling proteins are congregated. The synaptic concentration of PSD-95-like proteins and the abnormal synaptic morphology of Drosophila dlg mutants (11Lahey T. Gorczyca M. Jia X.-X. Budnik V. Neuron. 1994; 13: 823-835Abstract Full Text PDF PubMed Scopus (253) Google Scholar) support the notion that this family of proteins plays a central role in the molecular organization of synapses. Recent advances in understanding the function of PDZs came largely from studies of the PSD-95 protein family, which are predominantly neuronally expressed and synaptically localized. PDZ domains, however, are present either singly or multiply in a variety of distantly related (e.g., the MAGUKs ZO-1, p55, and LIN-2/CASK) or in otherwise quite unrelated proteins (e.g., the tyrosine phosphatase FAP-1, dishevelled, syntrophins, etc). Where characterized, many of these proteins are associated with cell junctions: for instance, ZO-1 at tight junctions, and syntrophins at the sarcolemma and NMJ. It therefore seems likely that these PDZ domains are also involved in the binding and organization of specific membrane proteins at sites of cell surface specialization. LIN-2 and LIN-7 are compelling examples of this functional theme, and have been discussed above. The recently identified interactions between neurexin and the PDZ domain of LIN-2/CASK (6Hata Y. Butz S. Südhof T.C. J. Neurosci. 1996; 16: 2488-2494Crossref PubMed Google Scholar), and between the Drosophila TRP photoreceptor Ca2+ channel and the PDZ domain of INAD (16Shieh B.-H. Zhu M.-Y. Neuron. 1996; 16: 991-998Abstract Full Text Full Text PDF PubMed Scopus (231) Google Scholar), hint at the wealth and diversity of PDZ interactions still waiting to be uncovered in the brain. With their PDZ domains in the limelight, it is easy to overlook the fact that the Dlg/PSD-95 family of clustering proteins contain in their C-terminal half an SH3 domain and a GK domain, the function of which remain quite mysterious. Which proteins bind specifically to the SH3 domain? And given that no enzymatic activity has been detected, what is the function of the GK domain, which is approximately 30% identical at the amino acid level to yeast guanylate kinase? Another significant question is how PSD-95 molecules link to the cytoskeleton, as they surely must if they are involved in anchoring of receptor/ion channel clusters at specific membrane sites. Finally, how are these PDZ-containing proteins specifically targeted to their final synaptic (or other subcellular) destinations? Further analysis of PSD-95 and its relatives in vitro and in vivo has the potential to uncover the domains of the protein to which clustering, anchoring, targeting, and signaling functions are segregated. The continuing study of PSD-95-like proteins promises to illuminate our understanding of the structure, function, and plasticity of neuronal synapses.