PDZ Domain Proteins: Plug and Play!

See allHide authors and affiliations

Science's STKE  22 Apr 2003:
Vol. 2003, Issue 179, pp. re7
DOI: 10.1126/stke.2003.179.re7


Protein-protein interactions are key elements in building functional protein complexes. Among the plethora of domains identified during the last 10 years, PDZ domains are one of the most commonly found protein-protein interaction domains in organisms from bacteria to humans. Although they may be the sole protein interaction domain within a cytoplasmic protein, they are most often found in combination with other protein interaction domains (for instance, SH3, PTB, WW) participating in complexes that facilitate signaling or determine the localization of receptors. Diversity of PDZ-containing protein function is provided by the large number of PDZ proteins that Mother Nature has distributed in the genome and implicates this protein family in the wiring of a huge number of molecules in molecular networks from the plasma membrane to the nucleus. Although at first sight their binding specificity appeared rather monotonous, involving only binding to the carboxyl-terminus of various proteins, it is now recognized that PDZ domains interact with greater versatility through PDZ-PDZ domain interaction; they bind to internal peptide sequences and even to lipids. Furthermore, PDZ domain-mediated interactions can sometimes be modulated in a dynamic way through target phosphorylation. In this review, we attempt to describe the structural basis of PDZ domain recognition and to give some functional insights into their role in the scaffolding of protein complexes implicated in normal and pathological biological processes.

Discovery of PDZ Domains

The discovery of PDZ domains was made on the basis of sequence repeats apparent in proteins (post-synaptic density 95, PSD-95; discs large, Dlg; zonula occludens-1, ZO-1) that contained these motifs. They were originally named GLGF (Gly-Leu-Gly-Phe) domains, because of a repetitive motif within the amino-terminal sequence (1, 2), or DHR domains, for discs large homology region (3). Shortly thereafter, the acronym PDZ (from the initial letters of the PSD-95, Dlg, and ZO-1 proteins) was proposed and adopted by the scientific community (4). With the sequences of several genomes now available, we know that PDZ domain proteins are widespread in metazoans (92 proteins in Caenorhabditis elegans; 131 proteins in Drosophila melanogaster; more than 400 proteins in Homo sapiens), plants (33 proteins), and bacteria (307 proteins), but are surprisingly rare in the Schizosaccharomyces pombe (three proteins) and Saccharomyces cerevisiae (two proteins) genomes (5).

PDZ Domain Families

PDZ domains can occur in one or multiple copies and are nearly always found in cytoplasmic proteins. The only exception is interleukin-16, a soluble cytokine that contains a PDZ domain-like structure (6). PDZ proteins can be classified into three principal families according to their modular organization (Fig. 1). The first family contains proteins consisting entirely of PDZ domains. The number of PDZ domains can vary from two [for example, in syntenin or NHERF (Na+/H+ exchanger regulatory factor), also called EBP50 (ezrin-radixin-moesin-binding phosphoprotein-50)] to more than 10 in certain proteins. No other protein interaction domains, except WD40 or LRR domains, are found in such high numbers in proteins; however, oligomerization of the named domains is required for ligand binding, whereas PDZ domains function independently. The MAGUKs (membrane-associated guanylate kinases, including PSD-95, Dlg, and ZO-1), which contain PDZ domains (one or three), one SH3 domain, and a guanylate kinase domain (GuK), comprise a second family. Lin-2, which is also known as CASK, is a MAGUK protein with a calmodulin kinase (CAMK) domain replacing the first two PDZ domains and two L27 domains. CAMK and GuK domains in MAGUKs have no catalytic activity, but behave as protein-protein interaction modules. The third family encompasses proteins that contain PDZ domains as well as other protein domains, such as ankyrin, LIM, L27, C2, PH, WW, DEP and LRR domains. SH2 domains are never associated with PDZ domains and only the X11 (also known as Mint) family displays a PTB domain along with two PDZ modules. InaD and MUPP1 have multiple PDZ domains (5 and 13, respectively) and a single L27 domain in the amino-terminus. The LAP (LRR and PDZ) protein family (Erbin, Scribble, Densin) contains 16 LRR and one or four PDZ domains (Fig. 1).

Fig. 1.

Examples of PDZ domain proteins classified according to their modular organization. Proteins are represented by black lines scaled to the length of the primary sequence of the protein. PDZ domains are shown in pink. SH3, Src-homology 3 domain; GUK, guanylate-kinase-like domain; CaMK, calmodulin kinase domain; L27, domain present in receptor targeting proteins Cask or Lin-2 and Veli or Lin-7; LIM, zinc binding domain present in Lin-11, Isl-1, Mec-3; PTB, phosphotyrosine binding domain; LRR, leucine-rich repeats; LAPSD, LAP-specific domain; CRIB, Cdc42/Rac Interactive Binding domain. More information on these protein-protein interaction domains can be found on the Web site:

All of these proteins act as adaptors that hold receptors and signaling molecules in large molecular complexes. Other PDZ proteins have intrinsic enzymatic activity and, as such, can directly participate in signaling events. Unlike SH2- or SH3-containing proteins, no PDZ proteins have intrinsic tyrosine kinase activity; however, some have serine-threonine kinase activity (such as LIM kinase and the microtubule-associated serine-threonine kinase family) or other enzymatic properties, including tyrosine phosphatase (PTP-Bas) activity, guanine nucleotide exchange factor (Tiam-1) activity, or guanosine triphosphatase (RhoGTPase-1) activity. An exhaustive list of PDZ proteins from several genomes can be searched on the SMART site (5).

Binding Specificities of PDZ Domains

Peptide Binding

It rapidly became clear that PDZ domains can bind the carboxyl-terminal sequences of proteins. The first demonstration of a PDZ domain-peptide interaction showed that the first two PDZ domains of PSD-95 could bind to the C-terminal T or S-X-V (T is threonine, S is serine, X is any amino acid, V is valine) motif of Shaker-type K+ channels (7) or of the NR2B NMDA (N-methyl-D-aspartate) receptor (8, 9). Simultaneously, Sato et al. demonstrated that the PDZ domain of the protein tyrosine phosphatase FAP interacted with the C-terminus of Fas (10). These studies shed light on which carboxyl-terminal residues were crucially required for protein interaction with a PDZ domain. Deletion of the C-terminal hydrophobic valine (Val has the position 0; subsequent residues toward the amino-terminus are numbered as residue -1, residue -2, and so forth), or mutation to alanine drastically reduced binding to the PDZ partner, as did any mutation of the residue in the -2 position. In contrast, any amino acid could substitute for the residue in position -1. Songyang et al. confirmed these findings using a degenerate peptide library and described two major classes of PDZ domains on the basis of binding specificity (11). One class of PDZ domains (class I, for example, PSD-95) prefers an S or TXV (Ser or Thr-any-Val) motif, similar to the T or S, D or E, V (Thr or Ser-Asp or Glu-Val) motif found in the NMDA receptor and Shaker-type K+ channels, whereas the second class (class II) interacts with a carboxyl-terminal ψXψ motif (ψ is a hydrophobic residue, such as Val, Tyr, Phe, Leu, Ile) (Table 1).

Diversity of PDZ domain binding is much greater than initially described because of the great variability in PDZ sequences. For instance, the neuronal nitric oxide synthase (nNOS) PDZ domain (sometimes called a class III PDZ domain) prefers a DXV (Asp-any-Val) motif, whereas one of the Mint-1 (X11) PDZ domains binds to a DXWC (Asp-any-Trp-Cys) sequence. Bezprozvanny and Maximov proposed classifying PDZ domains into 25 groups on the basis of the nature (hydrophobic, polar, aromatic, negative, and so forth) of residues in two positions: the αB1 position and the position immediately after the βB strand (Fig. 2). Accordingly, different binding specificities were attributed to each group thus specified (12).

Fig. 2.

Structure of the PSD-95 PDZ3 domain. (A) Ribbon model showing the overall structure of the PSD-95 PDZ3 (residues 306 to 394). β-Strands are shown in yellow, and α-helices are shown in red. (B) Structure of the PSD-95 PDZ3 domain complexed with a C-terminal peptide (KQTSV) from CRIPT (green and red). The peptide inserts itself between the αB helix and the βB strand (blue). Residues important for the interaction are highlighted. (C) Chemical interactions involved in peptide binding. Hydrogen bonds are depicted (dashed lines) between residues of the PDZ domain (blue) and the peptide ligand (green and red). [PDB accession number 1BE9].

In most cases, interaction between a PDZ domain and its target is constitutive, with a binding affinity of 1 to 10 μM. However, agonist-dependent activation of cell surface receptors is sometimes required to promote interaction with a PDZ protein. For example, interaction of AF-6 (ALL-1 fusion partner from chromosome 6), a Ras-binding protein, is increased after activation of the EphB3 tyrosine kinase receptor by its ligand (13). Similarly, NHERF interacts only with the activated β2-adrenergic receptor (14). Conversely, PDZ interactions are sometimes disrupted by phosphorylation of PDZ-binding sites by serine-threonine kinases. Protein kinase A (PKA)-dependent phosphorylation of the serine residue in position -2 of the inwardly rectifying K+ channel Kir 2.3 disrupts interaction with PSD-95 (15). Similarly, the serine-threonine kinase GRK-5 uncouples NHERF from the β2-adrenergic receptor by phosphorylating serine in the -2 position (16).

Solving the structural basis of peptide recognition of PDZ domains was critical to understanding the mechanisms by which they interact with their ligands and their binding specificity. PDZ domains consist of 80 to 90 amino acids comprising six β strands (βA to βF) and two α-helices, αA and αB, compactly arranged in a globular structure (Fig. 2). The solved three-dimensional structures of several PDZ domains, complexed with their respective ligands or ligand-free, have elucidated the mechanism of PDZ domain interactions (17-19). Peptide binding of the ligand takes place in an elongated surface groove as an antiparallel β strand interacts with the βB strand and the αB helix, as shown in Fig. 2 for the PSD-95 PDZ3 (class I PDZ domain) and the KQTSV (Lys-Glu-Thr-Ser-Val) peptide (17). The structure of PDZ domains allows binding to a free carboxylate group at the end of a peptide through a carboxylate-binding loop between the βA and βB strands that contains the R or KXXXGLGF signature. The carboxylate of the C-terminal valine is coordinated by hydrogen bonds with amides in the loop (GLGF residues) and by a water molecule coordinated with the R or K charged residues. The side chain of the C-terminal valine points into a large pocket formed by hydrophobic residues, whereas the threonine in position -2 points into a pocket containing a histidine of helix αB (position αB1). A hydrogen bond coordinates the N-3 nitrogen of the histidine and a hydroxyl group of the threonine or the serine in the -2 position. Phosphorylation of serine or threonine adds a large charged phosphate group that disrupts the interaction, explaining how PKA modulates the interaction between Kir 2.3 and PSD-95 (15). Structures of CASK and Grip PDZ6 domains have shown that class II PDZ domains differ from class I domains through formation of a second hydrophobic binding pocket composed of hydrophobic residues within the βB strand and the αB helix (18, 20). In class II, a hydrophobic residue (for example, Leu, Val, Tyr, Phe) at the -2 position in the peptide interacts with a corresponding hydrophobic residue in helix αB (position αB1). In the nNOS PDZ domain, a tyrosine residue in the αB1 position recognizes aspartic acid at position -2 in the DXV peptide sequence (19, 21). Whereas the side chain of residues in the -1 position usually points away from the binding pocket, in a few cases, specific interactions take place between the interaction surface and this residue. For example, the guanido group of arginine (position -1) found in the C-terminal motif (DTRL) of the cystic fibrosis transmembrane regulator (CFTR) forms two salt bridges and hydrogen bonds with the PDZ1 residues of NHERF (22). High-affinity binding between the EFCA sequence of NorpA (phospholipase Cβ isoenzyme) and the first PDZ domain of InaD requires formation of an intermolecular disulfide bond between cysteine in position -1 of the NorpA peptide and Cys31 of the PDZ domain ("dock-and-lock" interaction) (23). Amino acids closer to the amino-terminal (positions -3, -4, and so on) diverge in their PDZ domain-binding sites and confer greater binding specificity by interacting with nonconserved residues of the PDZ domain that lie outside of the groove.

The structure of bacterial PDZ-like domains has also begun to be appreciated. The Scenedemus obliquus photosystem II D1 protease has a PDZ-like fold (24). The regions that correspond to the amino-terminal βA strand in metazoan PDZ domains lie at the carboxyl terminus of the domain. Nevertheless, this strand is positioned in the correct orientation and location in the structure as in the case of the corresponding strand in the canonical PDZ domains. The entire protease recognizes its substrates based on carboxyl-terminal epitopes, although it remains to be established whether this binding was driven by the bacterial PDZ-like domain. This is apparently the case, because the PDZ-like domain of a related protease (Tsp protease) retains the ability to recognize C-terminal sequences (25). Thus, the distinct topology of this class of PDZ-like domains does not preclude binding in a fashion similar to that of metazoan PDZ domains.

Some interactions require multiple PDZ domains. In the case of syntenin, two adjacent PDZ domains are required for optimal interaction with the ligand. Mutational analysis showed that the PDZ2 domain of syntenin has higher affinity for the syndecan FYA C-terminal sequence than does the PDZ1 domain and that both PDZ domains are required for high-affinity binding (cooperative binding) (26). Finally, in some cases, internal motifs in proteins (that is, peptide sequences without a free carboxyl-terminus) form binding sites for PDZ domains. This is the case for an S or TXV motif buried in the sequence of the transient receptor potential (TRP) Ca2+ channel, which is able to interact with one of the PDZ domains of InaD (27). Phage display screening of a peptide library with the PDZ domain of syntrophin (28) identified cyclized peptides that bound with high affinity to this domain. Peptides show the predicted consensus sequence of syntrophin PDZ domain ligands, but only bind when their carboxyl-terminal sequence is constrained by cyclization.

Although most PDZ domains display a single binding specificity, a subset of PDZ domains has dual binding specificities. PICK1 has a single PDZ domain that interacts with a large panel of peptides, including protein kinase C (class I QSAV sequence) and ErbB2 receptor (class II DVPV sequence) (29, 30). Erbin has a classical class I PDZ domain that shows a high-affinity interaction with the class I p0071/δ–catenin peptide sequence (DSWV) and weaker-affinity binding (10 times less) to the class II ErbB2 peptide (DVPV) (31, 32). Recently determined structures of both complexes highlight the plasticity of this PDZ domain that enables it to recognize two peptide sequences (33, 34). Indeed, the isopropyl group of valine (position -2) of the ErbB2 peptide displaces the peptide backbone away from the α-helix of the PDZ domain, which allows recognition of a class II ligand.

With an understanding of these modalities of PDZ interaction in hand, it is now possible to generate PDZ domains with novel binding specificities by changing some residues within the domain (35-37). A computer-based approach was recently undertaken to enhance the diversity of peptide recognition, starting with the PSD-95 PDZ3 domain as a backbone and designing mutations that forced the domain toward novel binding specificities. A panel of PDZ domains that recognized diverse carboxyl-terminal sequences was successfully produced and was assayed in biochemical experiments. Because variations in the last four amino acids of proteins are sufficient to determine specific binding (38), engineered PDZ domains could theoretically substitute for antibodies directed against the carboxyl-termini of these proteins (37). It is only necessary that the targeted peptides terminate with hydrophobic residues that point toward the hydrophobic pocket of the PDZ domain.

Dimerization of PDZ Domains

The best-described examples of PDZ-PDZ dimerization involve PSD-95 (PDZ2) and syntrophin PDZ domains, which both bind to the nNOS PDZ domain in a "head-to-tail" fashion (39) (Fig. 3). The nNOS PDZ domain has a 30-residue carboxyl-terminal extension that folds in a β-hairpin structure (β finger) and docks in the peptide-binding sequence of the syntrophin or PSD-95 PDZ domain, mimicking a free carboxyl-terminal (40, 41). Accordingly, mutations within the nNOS β finger abrogate interactions with PDZ domains (41, 42). Such dimerization of PDZ domains precludes interaction with peptide ligands. Alternatively, a PDZ domain can directly form a complex with itself through homodimerization. For example, NHERF has two PDZ domains that homodimerize, which allows this short protein to create scaffolds for receptors and cytoskeleton-associated proteins (43). The PDZ6 domain of GRIP also engages in homodimerization. The dimeric interface between the two PDZ domains involves a βA strand and an αA-βD loop from each domain, which leaves the peptide-binding pocket oriented in the opposite direction and free for binding (20). Therefore, this dual binding increases the likelihood that the PDZ domains will simultaneously interact with several ligands to form a network.

Fig. 3.

Possible PDZ interaction modes. PDZ domains participate in at least four different classes of interaction: recognition of carboxyl-terminal motifs in peptides, recognition of internal motifs in peptides, PDZ-PDZ dimerization, and recognition of lipids.

Binding to Lipids

PDZ domain proteins are frequently associated with the plasma membrane, a compartment where high concentrations of phosphatidylinositol 4,5-bisphosphate (PIP2), a ligand for PH, C2, and ENTH protein interaction domains, are found. Using gel-filtration assays and surface plasmon resonance, Zimmermann and colleagues recently demonstrated direct interaction between PIP2 and a subset of class II PDZ domains (syntenin, CASK, Tiam-1) with an affinity comparable to the PH-PLCδ domain interaction with PIP2 (10 to 50 μM) (Fig. 3) (44). This subset of PDZ domains showed weaker-affinity binding to PIP3 (phosphatidylinositol 3,4,5-bisphosphate). Hydrolysis of PIP2 at the plasma membrane disrupts the membrane localization of syntenin, which suggests that this interaction is required for plasma membrane targeting. As with peptide binding, cooperative binding to PIP2 took place when the two PDZ domains of syntenin were incubated with PIP2, which suggests similar modalities of binding for peptides and lipids. It is interesting that excess PIP2 interfered with the interaction between the syntenin PDZ domain and the FYA syndecan peptide (and vice versa), which reinforces the idea that binding sites for both ligands overlap within the domain. Further questions arise from this study: Is PIP2 binding restricted to class II syntenin, CASK, and Tiam-1 PDZ domains or can it be mediated through a larger set of PDZ domains? Are the key residues within the PDZ domain required for binding to peptides identical to the ones required for PIP2 binding? A more systematic analysis of the ability of PDZ domains to bind to diverse lipids (PIP2, PIP3, etc.), as well as resolution of the structure of the complex between PIP2 and PDZ domains, will obviously be critical to understanding these novel interactions. Although there is no doubt that PDZ domains show more versatility in their interactions than first anticipated, the great majority of interactions mediated by PDZ domains known to date lie in the classical scheme of a PDZ domain bound to a carboxyl-terminal free peptide.

An interesting similarity exists among the PDZ, PTB, and PH domains in the context of structural and binding specificity. First, all three domains fold in a very similar fashion, despite the absence of sequence homology and the lack of any common peptide binding specificity (45). Second, all three domains have the ability to interact with peptides and lipids, albeit with different preferences (46, 47). It is currently difficult to determine which of these domains evolved first, although PTB domains, which are not found in bacterial genomes, are probably not the most likely candidates. A seductive idea is that these domains arose from a common ancestral domain (a "professional binder") and acquired increased binding specificities toward peptides and lipids during evolution through a diversification of the primary peptide sequence that conserved the optimal three-dimensional structure of their ancestor.

Functions of PDZ Proteins

PDZ domain proteins were initially thought to be mainly involved in the scaffolding of proteins into functional units (sometimes called transducisomes) and the aggregation of receptors at the plasma membrane. However, PDZ proteins clearly do more than simply plugging different proteins together. They direct supramolecular complexes to specific subcellular compartments, thereby contributing to the signaling specificity of many receptors, including receptor tyrosine kinases and glutamate receptors. Given the tremendous number of PDZ domain interactions described in the literature, it is beyond the scope of this review to exhaustively list all biological processes that involve PDZ proteins. We will focus here on some evolutionarily conserved PDZ protein complexes described recently and on some available mammalian genetic models that shed light on the role of PDZ domains in human diseases.

Adaptors for Tyrosine Kinase Receptors

In addition to cell adhesion molecules and GPCR (G protein-coupled receptors), receptor tyrosine kinases are frequently associated with PDZ proteins. Pioneering work by Stuart Kim's laboratory has suggested ways to think about the regulation of these receptors. Let-23, the C. elegans epidermal growth factor (EGF)-receptor (EGFR), is targeted (or retained) basolaterally through an interaction with the Lin-7-Lin-2-Lin-10 PDZ protein complex (35). Loss of function of any of these three proteins leads to Let-23 mislocalization, rendering it inaccessible to its ligand Lin-3, which is secreted by the apical gonad. In these mutants, Let-23 is as inactive as when its signaling machinery (Let-60, the homolog of Ras, and Sem-5, the homolog of Grb2) is ineffective. Evolution has faithfully retained the Lin-7-Lin-2-Lin-10 (also named Veli-Cask-Mint) protein complex in mammals, using it to target glutamate--instead of EGF--receptors (48, 49). Nevertheless, direct interaction between EGFR and PDZ molecules is also observed in vertebrates. ErbB2 and ErbB4, two human EGFR relatives, bind to Erbin and PSD-95, respectively, through a direct PDZ domain interaction (50-52).

Another example involves the clustering and autophosphorylation of the platelet-derived growth factor (PDGF)-receptor, which is more efficient when it interacts with NHERF, and that leads to more potent MAPK activation (43). The PDZ protein Dishevelled, an important component of the Wnt-β-catenin signaling pathway, also participates in signaling from the MUSK tyrosine kinase receptor and may therefore play a role in the synaptogenesis of the neuromuscular junction (53). The IGF-1 receptor and c-Kit were recently proposed as targets for other classes of PDZ proteins (54, 55). Most RTKs (if not all) use similar transduction machinery (for instance, Grb2, Shc, and so on) through their binding sites for SH2 and PTB domain proteins. This "core signaling machinery" cannot, however, explain how RTKs achieve their exquisite functional specificity. One route to this specificity probably lies in the recruitment of adaptors such as PDZ proteins that (i) have great molecular, and therefore functional, diversity; (ii) bind to a restricted group of receptors; (iii) are involved in scaffolding various (positive and negative) regulators of receptor activity; and (iv) are usually localized to very specific subcellular compartments, where they nucleate receptors and their modulators. For example, at the basolateral membrane of epithelial cells, ErbB2 is the only EGFR able to bind to the PDZ protein Erbin through a PDZ domain interaction (50). In turn, Erbin interacts with β-catenin and β4-integrin, two substrates of ErbB2 that amplify tyrosine kinase-mediated cell transformation (32, 56). Thus, through a unique four-amino acid carboxyl-terminal sequence, a very specific complex is brought to ErbB2 in a specific subcellular compartment.

The Let-23 story in C. elegans tells us how precise routing is crucial for the activity of some RTKs in polarized cells (35). Similarly, some ligands of RTKs engage in PDZ interactions that regulate their activity. Transforming growth factor (TGF)α, a ligand for EGFR, is produced as a precursor (pro-TGFα) that is tethered at the plasma membrane and then shed into the extracellular medium to act as a diffusible mitogenic factor. Syntenin is required for the proper trafficking of pro-TGFα along the early secretory pathway, through interactions with the C-terminal ETVV sequence of the ligand (57). EphB tyrosine kinase receptors are activated by glycosylphosphatidylinositol anchored (EphrinA) or transmembrane (EphrinB) ligands. The latter have an 80- to 90-amino acid cytoplasmic domain involved in "reverse" signaling events (for instance, EphrinB is tyrosine phosphorylated by intracellular tyrosine kinases and subsequently triggers signaling events in the cell that produced it). Appropriate positioning and clustering of EphrinB and their respective EphB receptors must be achieved to allow receptor activation in adjacent cells. PDZ proteins probably play a fundamental role in these processes. Indeed, the Eph-related receptor EphB3 is connected to AF-6 through a PDZ domain interaction, which promotes clustering of the receptor (13, 58). On the side of the cell expressing EphrinB, presence of a PDZ domain-binding site within the cytoplasmic regions enables interaction with syntenin, GRIP1 or RGS-PDZ3 PDZ (59, 60). EphrinB-EphB2 receptor interaction promotes the formation of lipid raft patches containing GRIP1 and associated signaling molecules in EphrinB-expressing cells (61). In conclusion, PDZ interactions present a versatile option for specifying the function and finely tuning the activities of tyrosine kinase receptors and their ligands.

Epithelial Polarity

As mentioned above, proteins that bear PDZ domains are often localized at specific subcellular sites near the plasma membrane of polarized cells, such as epithelial and endothelial cells, and neurons. Epithelial cell membranes are organized into two different regions with distinct biochemical and functional properties: an apical membrane that is in contact with fluids and a basolateral membrane that is in contact with the extracellular matrix (on the basal side) and with adjacent cells (on the lateral sides). Adherens junctions (AJ) of the lateral membranes regulate cell-cell adhesion, whereas tight junctions (TJ) preclude mixing of components of the apical and lateral compartments (Fig. 4). Epithelial polarity is highly conserved in Metazoans. It is required for building most tissues and uses a group of proteins structurally and functionally conserved from worms and flies to humans (62). During recent years, several laboratories have deciphered the role of PDZ proteins in establishing and maintaining epithelial polarity. PDZ proteins play a central role in each step of the process that conveys signals from cell surface molecules to the interior, where they participate in signaling cascades and construction of the cytoarchitecture.

Fig. 4.

Schematic diagram of a polarized epithelial cell. Diverse PDZ protein complexes targeted to the apical and subapical, basolateral, and junctional domains are represented.

Two evolutionarily conserved protein complexes that contain PDZ proteins, Crumbs-PALS1 (stardust)-PATJ (discs lost) and Cdc42-Par6-Par3 (bazooka)-atypical protein kinase C (aPKC), were implicated in the assembly of tight junctions and the development of epithelial polarity in Drosophila, C. elegans, and mammals (63-66). Located in the more basal regions of polarized epithelial cells (associated with septate junctions and basolateral membranes) lies the Dlg (discs large)-Lgl (lethal giant larvae)-Scribble protein complex, which is also required for the development of epithelial polarity (67). Genetic approaches have demonstrated that the PDZ proteins PALS1, PATJ, Par6, Par3, Dlg, and Scribble participate in signaling pathways necessary for epithelial apico-basal polarity, in collaboration with proteins having different functional activities, including a myosin II-binding protein (Lgl), a small GTPase (Cdc42), and a serine-threonine kinase (aPKC). Recent reports have further dissected the fine-tuning and dynamics of these functions and the functional interactions between these protein complexes during fly tissue morphogenesis. These studies led to the conclusion that these proteins act in a single pathway devoted to apical protein localization (68, 69). It is interesting that in mammals, Hurd and colleagues demonstrated a direct protein interaction between Par6 and PALS1 within the tight junction (70) that unified two of these protein complexes into a single complex. Other cell surface molecules of the tight junctions, including Claudins, Occludins, JAM (junctional adhesion molecules), and Nectins cell adhesion molecules, are bound to AF-6, MUPP1, CASK, or ZO-1 proteins that link them to the cytoskeleton or to signaling complexes. In adherens junctions, the E-cadherin-β-catenin complex and Nectins interact directly or indirectly with a subset of PDZ proteins including Veli (Lin-7) and AF-6 (71, 72). More functional data are now needed to determine the respective role of these PDZ proteins in the establishment and maintenance of cell polarity in mammals.

The Synapse

The synaptic junction between nerve cells serves as another model for understanding the biology of PDZ proteins. The synapse is an exquisite example of a cell junction between polarized subcellular compartments--with its presynaptic terminal, containing the neurotransmitter vesicles, connected to the postsynaptic region, where neurotransmitter receptors activate numerous signaling pathways. Dense collections of proteins that were observed in early electron microscopic studies of synapses are now known to contain an impressive array of PDZ proteins, a subset of which are shown in Table 2. The prototypical PDZ protein PSD-95 was isolated from the postsynaptic density, which is observed immediately beneath the postsynaptic membrane (1, 73, 74). On the presynaptic side, the presynaptic web (75) is a set of proteins, including PDZ proteins, that surround the active zone where neurotransmitter vesicles are released. It is beyond the scope of this review to describe all synaptic PDZ proteins, and here we have chosen to focus primarily on signaling complexes found on the postsynaptic side of the synapse.

Fig. 5.

Assembly of synaptic PDZ proteins. A schematic representation of the presynaptic and postsynaptic terminal illustrating PDZ domains (red) in synaptic proteins and their interaction partners.

Table 1.

Classification of PDZ domains according to their specificity for C-terminal peptides. A list of PDZ domain interactions. More interactions are available on the World Wide Web. The C-terminal sequence is the last four amino acids of the ligand proteins. ψ, hydrophobic amino acid; X, unspecified amino acid. NMDAR2A, NMDA receptor 2 subunit; ARCVF, a member of the p120(ctn) subfamily of armadillo-like proteins.

Table 2.

A subset of PDZ domains and their binding partners at mammalian synapses. This table shows a representative subset of synaptic proteins containing PDZ domains. The synonyms for each protein are listed, and the asterisk (*) indicates that the protein is found at presynaptic terminals, whereas all others are postsynaptic. To illustrate the potential for PDZ proteins to contribute toward the assembly of large networks of proteins, the number of published primary (direct) protein-binding partners is shown with the names of those proteins. The number of secondary proteins (proteins that bind to primary binding partners) is also shown. A given PDZ protein is capable of interacting with more than 100 proteins within two protein-protein interactions.

Neurotransmitter receptors are targeted to both pre- and postsynaptic sites, where they mediate signaling by changing transmembrane ion flow or by activating second-messenger pathways. In the mammalian central nervous system, where glutamate is the major excitatory neurotransmitter, there are three main subclasses of glutamatergic receptors, each of which interacts with cytoplasmic PDZ proteins. The transmission of excitatory postsynaptic potentials is primarily mediated by AMPA receptors, whereas signaling cascades are initiated by NMDA receptors and metabotropic (mGlu) receptors. The first evidence that PDZ domain proteins bound glutamate receptors arose from yeast two-hybrid studies using the cytoplasmic carboxyl-terminal domains of NMDA receptors (8), which bound PSD-95 and other MAGUK proteins. It was later observed that AMPA receptors also interact with PDZ proteins (GRIP, ABP, PICK1, SAP97) (76-79), as do mGlu receptors (CIPP1, PICK1) (80-82).

Biochemical studies of glutamate receptor complexes revealed that NMDA receptors and mGluR were components of ~2000-kD complexes with their cognate PDZ proteins and other proteins, whereas AMPA receptors were found in physically separate complexes (83, 84). The role of PDZ domain proteins in synapse signal transduction has been studied by genetic manipulation in the mouse. The NMDA receptor, which allows Ca2+ influx, binds PSD-95 through two of its PDZ domains, and a third PDZ domains binds a GTPase called SynGAP (83, 85, 86). Thus, with reference to the NMDA receptor, PSD-95 can be considered a primary interactor and SynGAP a secondary interactor in steps into or stages of an intracellular network. Mouse mutations of PSD-95 (87) and SynGAP (88) show that neither of these proteins are required for the synaptic localization of the NMDA receptor, but that they are important for synaptic plasticity. Moreover, studies of double-mutants reveal that PSD-95 couples two pathways to the NMDA receptor, one that enhances synaptic strength and another that depresses it. The pathway that enhances synaptic strength is coupled through SynGAP, which in turn regulates a Ras-ERK pathway. A detailed comparison of the effects on synaptic function of different patterns of stimulation (short bursts of action potentials of varying frequencies) shows that these two pathways are recruited independently by distinct frequencies of neuronal firing. This has general importance because it indicates that the network of PDZ interactions is involved in discriminating the different levels of receptor activation and effectively triaging these inputs and directing the distinct downstream effector pathways. At the level of the whole animal, mutations in PSD-95 and SynGAP result in specific impairments to spatial learning, implying that integrity of the synapse PDZ network is essential for cognitive function.

Protein Networks

The study of intracellular networks comprised of hundreds of proteins is a new area of research both for metabolic networks and protein interactions. A cell-wide study of yeast interactions found that a single large network of 2358 interactions between 1538 proteins was made, and the next largest network contained only 19 proteins (89, 90). Using an assignment of function, based on the cellular role the protein was engaged in and not its precise biochemical activity, a regional architecture within this large network emerged. Proteins of similar function were close to each other and organized into clusters separated by no more than two other proteins. Moreover, the daunting complexity of these interaction networks has been simplified by the illustration that the networks follow a simple mathematical description, known as scale-free topology. This topology, which describes a network where a small number of hub proteins with large numbers of connections are among a large number of proteins with very few connections, has been shown to predict some biological processes (91). In an important way, these networks have features that produce robustness and tolerance of error--a property of many biological systems.

The PDZ proteins at the synapse provide an illustration of the potential for a large network of proteins. Using a database of mammalian binary proteins interactions (92), we have listed the number of direct (primary interaction) binding partners for some synaptic PDZ proteins. In addition, the number of secondary binding proteins (in other words, those that interact with the primary binding partners) are also shown. These numbers indicate that a potential network of hundreds of proteins lies within two protein-protein interaction steps of these PDZ proteins. This example of receptor signaling into a more extensive network of PDZ proteins requires modeling of protein networks. The magnitude of this problem is illustrated by summing the reported binary interactions from the NMDA receptor (which shows 17 primary interactions leading to 385 secondary proteins) or other receptors, such as CD44 (10 primary and 125 secondary interactions), or integrin-α1 (5 primary and 137 secondary interactions), or the EphB3 receptor (8 primary and 74 secondary interactions). The average number of primary interaction proteins for synaptic PDZ proteins was 10 and for secondary proteins 108. The understanding of the importance of this complexity will be enhanced by the addition of other data, such as functional information (for instance, enzyme function and protein domain architecture).

The model that large networks of proteins exist within a few interactions of any protein presents new challenges to biology and bioinformatics. In recent years, the accumulation of literature on receptor-mediated signal transduction has led toward a view of multiple linear pathways interconnected or "cross-talking" through common interactions at different levels. At the same time, it is now recognized that multiprotein nanomachines consisting of more than 100 proteins are commonplace [for instance, the splicesome (93)]. As schematically illustrated in Fig. 5, the postsynaptic NMDA receptor can be used as a model for receptor-mediated signaling in a discrete cellular compartment, which allows us to examine some general principles of PDZ domain functions. One of the future challenges will be to build models of multiprotein networks and explain the complexity of multiprotein machines.

Genetic Models and Diseases Involving PDZ Proteins

Although redundancy is a common theme in the world of PDZ proteins (one PDZ domain usually has a spectrum of partners, each of which interacts with various PDZ proteins), genetic studies in invertebrates have emphasized the central role of PDZ proteins during development and in adult life. Many biological functions are more complicated in mammals because of the existence of multigenic families (for example, MAGUKs and LAPs), but, nevertheless, interesting knockout mice and human diseases have shed light on the importance of PDZ proteins. For example, loss of Shroom, a PDZ protein bound to actin and localized in the adherens junctions, leads to failure of the neural tube to close, causing exencephaly, acrania, cleft palate, and spina bifida in mice (94). Analysis of af-6 knockout mice confirms the critical role of AF-6 for regulation of cell-cell junctions. Loss of neuroepithelial polarity is evident in mutant embryos that die by 10 days post coitum (95). GRIP1 has seven PDZ domains that interact with synaptic glutamate receptors, but is obviously crucial for more than neuronal functions: embryonic lethality occurs in grip-/- mice because of defects in the dermo-epidermal junction that mimick epidermolysis bullosa (96). Interesting genetic models totally confirmed the biological relevance of some PDZ protein complexes. Indeed, disruption of cask or dlg, two genes encoding MAGUK proteins, results in craniofacial dysmorphogenesis with cleft palate in mice (97, 98). Consistent with these effects, these proteins form a tight protein complex at the basolateral side of epithelial cells (99). Also impressive is the case of Harmonin. This PDZ protein is expressed in the inner ear sensory hair cells, where it interacts with specific classes of myosins and cadherins. Defects of harmonin or of its partners cause sensorineural deafness and blindness (100).

The central role played by PDZ proteins in maintaining tissue integrity suggests that their disruption would have a profound effect on signaling pathways or on the cytoarchitecture, as found in cancers. Loss of polarity is apparent during epithelial cell transformation after dysregulation of oncogenes or tumor suppressors, but also when some key components necessary for epithelial integrity are missing (101). Consistent with this, ablation of function of the PDZ protein Scribble in Drosophila not only affects cell polarity, but also provokes cell proliferation and tumorigenesis in imaginal discs (67). Scribble mutant mice develop a severe neural tube defect (102), which makes it difficult to comment on the activity of this LAP protein in adults. The high conservation of Scribble from flies to humans nevertheless suggests a probable implication of this protein in tumorigenesis (103). It is interesting that decreased levels of Scribble are apparent in cells infected by human papillomavirus type 16 (HPV-16), a process thought to promote higher tumorigenic activity of the virus (104). The E6 oncoprotein produced by the virus binds with high affinity to PDZ proteins that are important for cell-cell junctions, including Scribble, and also MUPP-1 and discs large, and promotes their polyubiquitinylation (an E3-ubiquitin ligase is associated with E6) and subsequent proteasome-mediated degradation (104-106). In addition to perturbations of cell cycle perpetrated by E6, loss of cell-cell adhesion may improve the ability of infected cells to escape from the epithelial monolayer and invade adjacent tissues.


With the list of PDZ proteins from several genomes in hand, a major goal will certainly be to unravel the identity of their ligands through in silico and experimental approaches and to decipher the function of these interactions in vivo. In addition to understanding the binary interactions, it will also be necessary to develop new models and tools to understand the central role of PDZ domain proteins in assembling and regulating protein networks. In parallel, the design of PDZ domain inhibitors could produce novel pharmacological tools of therapeutic interest (32, 107).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
View Abstract

Navigate This Article