Unexpected Modes of PDZ Domain Scaffolding Revealed by Structure of nNOS-Syntrophin Complex

See allHide authors and affiliations

Science  30 Apr 1999:
Vol. 284, Issue 5415, pp. 812-815
DOI: 10.1126/science.284.5415.812


The PDZ protein interaction domain of neuronal nitric oxide synthase (nNOS) can heterodimerize with the PDZ domains of postsynaptic density protein 95 and syntrophin through interactions that are not mediated by recognition of a typical carboxyl-terminal motif. The nNOS-syntrophin PDZ complex structure revealed that the domains interact in an unusual linear head-to-tail arrangement. The nNOS PDZ domain has two opposite interaction surfaces—one face has the canonical peptide binding groove, whereas the other has a β-hairpin “finger.” This nNOS β finger docks in the syntrophin peptide binding groove, mimicking a peptide ligand, except that a sharp β turn replaces the normally required carboxyl terminus. This structure explains how PDZ domains can participate in diverse interaction modes to assemble protein networks.

PDZ protein interaction domains (1) play a central role in organizing diverse signaling pathways (2–7). The scaffolding protein InaD (inactivation no after potential), which comprises five PDZ domains, tethers multiple proteins of theDrosophila phototransduction cascade (6). The postsynaptic density protein PSD-95, which has three PDZ domains, clusters and localizes N-methyl-d-aspartic acid (NMDA)–type glutamate receptors at synapses (2). By organizing such protein networks, PDZ scaffolding proteins increase the efficiency and specificity of signal transduction.

Most known PDZ-mediated interactions occur through recognition of short COOH-terminal peptide motifs (2). Peptide interaction screens indicate that the terminal carboxylate of these ligands is required for binding (8). This canonical mode of PDZ recognition is well characterized: structures of PDZ-peptide complexes reveal that recognition is achieved through an extended binding groove that terminates in a conserved carboxylate binding loop (9,10).

Nonetheless, an increasing number of PDZ-mediated interactions occur through modes that are not dependent on recognition of a COOH-terminal motif. A well-characterized example is the PDZ domain of neuronal nitric oxide synthase (nNOS), the neuron- and muscle-specific isoform of the enzyme that produces the second messenger nitric oxide (NO). The PDZ domain of nNOS specifically heterodimerizes with PDZ domains from PSD-95 and syntrophin in neurons and muscle cells, respectively (5). These interactions allow the integration of nNOS into specific signaling pathways. Association with PSD-95 in neurons is thought to couple NO production to NMDA receptor activation (11). Association with syntrophin in muscle cells localizes nNOS to the dystrophin complex (12), coupling NO production to muscle contraction. The resulting NO exerts a protective effect by increasing blood flow to match the heightened metabolic load of contracting muscle. Loss of this response in Duchenne muscular dystrophy may contribute to muscle degeneration (13).

These PDZ-PDZ interactions are distinct from canonical PDZ-peptide interactions in that they do not involve recognition of a COOH-terminal sequence on either partner protein. Moreover, the full tertiary structures of both PDZ partners are required, including 30 amino acids in nNOS that extend beyond the canonical PDZ region (5). These interactions are representative of an alternative class of PDZ domain interactions: the recognition of internal motifs. Such internal interactions have also been reported for PDZ domains from the scaffolding proteins GRIP and InaD (7,14).

We determined the crystal structures of the nNOS PDZ domain alone and in complex with the syntrophin PDZ domain to 1.25 and 1.9 Å resolution, respectively (15) (Table 1). These structures reveal the mechanism of PDZ heterodimerization and, more generally, of internal sequence recognition. The nNOS and syntrophin PDZ domains associate in a linear head-to-tail fashion that lacks the twofold rotational symmetry normally observed in dimer structures (Fig. 1). Linear association occurs because the nNOS PDZ domain has an unusual polarized structure with two distinct faces: a receptor face, comprising the canonical PDZ domain with its peptide binding groove (residues 1 to 100), and a ligand face, comprising a β-hairpin “finger” (residues 100 to 130). This β finger acts as a PDZ ligand, docking into the peptide binding groove of the syntrophin PDZ domain. The two-stranded nNOS β finger augments an antiparallel β sheet in syntrophin, forming a composite four-stranded sheet (16). In the complex, the peptide binding groove of the nNOS PDZ domain remains accessible to interact with additional proteins.

Figure 1

Linear head-to-tail heterodimer of nNOS-syntrophin PDZ domains. (A) The nNOS PDZ domain (orange) has a polarized structure with distinct receptor (peptide binding groove) and ligand (β-finger) faces. The nNOS ligand face docks against the syntrophin PDZ domain (purple) peptide binding groove. (B) Structure of the syntrophin PDZ domain (purple) in complex with a COOH-terminal peptide (orange) (10). The figure was generated with the program MOLSCRIPT (25).

Table 1

Summary of crystallographic analysis. Dashes indicate entries that are meaningless for native data sets alone.

View this table:

The first strand of the nNOS β finger (referred to as the pseudo-peptide motif) mimics a canonical COOH-terminal peptide ligand in its sequence-specific interactions with the syntrophin PDZ domain (Fig. 2) (17). However, the normally required terminal carboxylate of a peptide ligand is replaced by the sharp β turn at the tip of the nNOS β finger.

Figure 2

Presentation of the nNOS β finger as an internal sequence motif that mimics a COOH-terminal PDZ peptide ligand. (A) Sequence comparison of syntrophin PDZ domain preferred COOH-terminal peptide ligands (10) with the nNOS pseudo-peptide (18). Binding site nomenclature is from (9). (B) Stereo depiction of the nNOS β finger (orange) bound at the syntrophin (purple) binding groove. nNOS pseudo-peptide residues are indicated by the binding pocket number (in parentheses). All other numbers refer to the actual residue number. The GLGF/carboxylate binding loop (19) is highlighted. Hydrogen bonds are shown as dashed lines, and an ordered water molecule is shown as a red sphere. For clarity, nNOS strand H is shown as a Cα trace, and its backbone hydrogen bonds have been omitted. The image was generated with the program MOLSCRIPT (25). (C) Schematic depiction of heterodimer interface interactions. The nNOS pseudo-peptide segment is outlined.

The existence of this complex demonstrates that a terminal carboxylate group is not an absolute requirement for PDZ recognition. In canonical PDZ complexes, the ligand terminal carboxylate is cradled in the GLGF (18), or carboxylate binding, loop found at the end of the binding groove (19). Because the nNOS β finger lacks a carboxylate, yet still serves as a specific ligand, we propose that the GLGF loop serves another function aside from carboxylate recognition: acting as a steric block to prevent binding of peptides that extend beyond the recognition sequence (Fig. 3, A through C). The restriction imposed by such a block can be overcome only if the ligand chain either terminates, as do COOH-terminal peptide ligands, or folds back on itself, as does the nNOS β finger.

Figure 3

Recognition of internal motifs by PDZ domains. In (A) through (C), the GLGF loop acts as a steric block at the end of the binding groove, necessitating chain termination or a sharp β turn immediately after the recognition motif. (A) Interaction topology of a COOH-terminal peptide (orange) bound to PSD-95 PDZ3 (purple surface). (B) Interaction topology of the nNOS β finger (orange) with the syntrophin PDZ domain (purple surface). In (A) and (B), the hydrophobic ligand residue that packs at site 0 is shown in space-filling mode. Gray, carbon; red, oxygen. (C) Schematic of structural requirements for PDZ domain recognition of internal or COOH-terminal ligands. (D) Rigidly stabilized structure of nNOS β finger. Overlay of Cα traces of the uncomplexed (orange) and complexed (grey) nNOS PDZ domain structures, highlighting residues that stabilize the nNOS β-finger conformation. The main interaction is a salt bridge between Arg121 and Asp62, which is buried by the surrounding hydrophobic residues Ile16, Leu57, Pro100, Phe103, Thr105, Leu107, and Thr123. (E) Increased contact area in the PDZ heterodimer through tertiary interactions. Solvent excluded footprint of the nNOS PDZ domain (Cα trace shown in orange) bound to the syntrophin PDZ domain (purple surface, ∼800 Å2), compared to the footprint of a peptide ligand (pink surface, ∼400 Å2). Images were generated with the programs MOLSCRIPT (25) and WebLab Viewer Lite (26).

Thus PDZ domains can recognize internal motifs if these are presented within a secondary structure that is sterically compatible with the PDZ binding groove. The importance of a stable structural context is supported by the apparent conformational rigidity of the nNOS β finger. Despite extending far from the main body of the domain, the β finger adopts essentially the same structure in the crystals of the nNOS PDZ domain alone (Fig. 3D). This rigid conformation is stabilized by extensive interactions between the β finger and the main body of the nNOS PDZ domain. The core interaction, which pins the COOH-terminal base of the finger back against the main domain, is a buried salt bridge between Arg121 in the β finger (strand H) and Asp62 in the main PDZ domain (strand D).

Stabilization of the nNOS β-finger structure appears to be critical because the entire folded nNOS PDZ domain is required for interaction (5). Moreover, a cyclic peptide that bears an internal recognition motif can bind the syntrophin PDZ domain (20). The cyclic linkage may lock this peptide in a turn conformation similar to that of the nNOS β finger. Other internal PDZ ligands are also likely to present pseudo-peptide motifs in a constrained context (21).

The heterodimer structure also reveals the important role of tertiary interactions in PDZ recognition specificity. Many residues in nNOS and syntrophin that participate in hetereodimerization are distant in sequence from the primary pseudo-peptide contacts. As a result, heterodimerization buries twice the surface area on the syntrophin PDZ domain as compared to binding of a peptide ligand (Fig. 3E). Most of the additional tertiary contacts come from the most NH2-terminal strand of the nNOS PDZ domain (strand A), which in the complex packs against the syntrophin α helix. Residues that participate in this ensemble of primary and tertiary interactions (Ser95, Asn102, Ser109, His142, and Asp143) are uniquely conserved in the syntrophin and PSD-95 (second domain) PDZ domains, perhaps explaining why only these two domains heterodimerize with nNOS (22). Analogous tertiary interactions may also enhance the specificity of COOH-terminal peptide-mediated PDZ interactions (23).

The linear head-to-tail arrangement observed in the syntrophin-nNOS complex may allow PDZ proteins to achieve more than subcellular localization (Fig. 4A). Correct targeting of nNOS could be achieved by simply appending an appropriate COOH-terminal sequence. The bifunctional (that is, receptor and ligand) nNOS PDZ domain, however, in addition to localizing its cargo protein (the catalytic portion of NOS), adds on a new receptor site with altered specificity that can recruit additional proteins (24). Bifunctional PDZ domains could also oligomerize to yield heteropolymer filaments. Indeed, the polarized structure of the nNOS PDZ domain is typical of cytoskeletal filament monomers or viral capsid proteins, both of which form higher order repeating assemblies. Finally, these types of domains could mediate branching interactions between proteins with multiple PDZ domains. A search of known PDZ sequences revealed several that may have a nNOS type of β finger immediately COOH-terminal to the canonical PDZ domain (Fig. 4B).

Figure 4

Possible assembly of signaling networks through diverse PDZ interaction modes. (A) PDZ domains participate in at least three different classes of interactions: recognition of COOH-terminal motifs, recognition of internal motifs (stabilized by specific secondary and tertiary structure), and head-to-tail PDZ oligomerization. (B) Alignment of PDZ domains that have potential β-finger extensions similar to that of the nNOS domain (18). There are two classes of PDZ domain preferences: class I domains prefer -(S or T)-X-V- motifs and class II domains prefer -φ-X-φ- motifs (where φ is a hydrophobic residue). The aligned sequences contain either motif, spaced appropriately after the canonical PDZ fold, as well as a putative salt bridge pair corresponding to Arg121and Asp62 of nNOS.

These structures reveal that PDZ domains can interact in multiple ways—recognizing COOH-terminal or constrained internal motifs and forming head-to-tail oligomers. The ability of PDZ domains to participate in these diverse interactions and their potential to form higher order assemblies may explain why these modules so often mediate scaffolding of sophisticated signaling networks at synapses and other critical membrane sites.

  • * To whom correspondence should be addressed. E-mail: wlim{at}


Stay Connected to Science

Navigate This Article