MAGUKs end a tale of promiscuity.

Membrane-associated guanylate kinases (MAGUKs) constitute a family of scaffold molecules involved in diverse cellular processes, such as cell–cell communication, cell polarity, and signal transduction (1). MAGUKs are multidomain proteins with a core structure consisting of PSD-95/DLG/ ZO-1 (PDZ), Src homolgy 3 (SH3), and enzymatically inactive guanylate kinase (GK) domains. Although full-length MAGUKs generally display a high degree of binding specificity, binding studies with isolated PDZ domains often show the ability to interact with numerous partners (2). In PNAS, Li et al. (3) provide a conclusive structural answer to this discrepancy. The authors have solved the crystal structure of a PALS1/Crb complex that demonstrates why the PDZ– SH3–GK supramodule of PALS1, but not the isolated PDZ domain, binds with an extraordinarily high affinity to the C terminus of Crb (3). Domains in the supramodule are arranged in a way that a peptide comprising the last 17 residues of Crb binds simultaneously to the PDZ, the SH3 and, most surprisingly, to the GK domain using two distinct sites. The C-terminal fixation in the PDZ binding groove follows an expected main chain connectivity of PDZ/peptide complexes. Interestingly, residues of the SH3 domain collaborate in peptide binding by building a kind of a clasp. Using site-directed mutagenesis, Li et al. (3) prove that the clasp accounts for the high affinity and specificity of Crb binding to PALS1. This discovery has immediate consequences for other MAGUKs, for example at synapses in the brain, because the authors propose that sequence variations within the clasp-loop are responsible for defining their specificity. As proof-of-principle, Li et al. show that the C terminus of neurexin does not bind to the isolated PDZ but to the PDZ–SH3– GK supramodule of calcium/calmodulindependent serine protein kinase (CASK). Unexpectedly, the PALS1/Crb structure shows an additional participation of the GK domain in Crb binding (3). Crb binds in a manner comparable to the complex of a phosphorylated peptide p-LGN with the GK domain of DLG1/SAP97 (4). In SAP97/p-LGN, a phosphate group at Ser401 binds to conserved residues that coordinate guanosine monophosphate (GMP) in SAP95 (5). Mutations of Ser to Asp or to Glu are widely used as a tool to generate constitutively active kinases (6) because the carboxyl group of the acidic residues mimics a phosphate ion by negative charge and connectivity. This process is exactly what appears to happen naturally in the PALS1/Crb complex, in which Crb binds with Glu, instead of a phosphate, to the GMP binding pocket. As for Li et al.’s (3) example of synaptic MAGUKs, it is yet undetermined if binding of neurexin to CASK also involves the GK domain in addition to PDZ–SH3, but it has been shown that CASK phosphorylates the C terminus of neurexin (7). This finding implies that CASK may not only bind to the C terminus of neurexin via the PDZ–SH3 complex, but might strengthen its interaction by phosphorylation of a serine within the last 17 residues. The case of neurexin/CASK demonstrates that the structural data by Li et al. (3) not only provide insight into PALS1 interactions but have far-reaching implications on research into MAGUKs: additional clues for binding specificity can now be expected from the clasploop of particular SH3 domains and from presence of phosphorylatable Ser and Thr, or “constitutively active” Asp or Glu residues, upstream of the PDZ binding motif in Fig. 1. Crb-triggered oligomerization of PALS1. (A) Structural model of full-length PALS1 with the PDZ–SH3–GK supramodul in extended conformation (Left). On appearance of Crb C-terminal tail (black) with Glu at position −14 (red), PDZ and SH3 domains bind the tail with high affinity, and GK lifts off the SH3 domain to interact with Glu (arrow), accompanied by reorientation of the HOOK region (Right). (B) MAGUKs such as PALS1 can form oligomers, for example by using N-terminal L27 domains independent of a peptide binding to PDZ. (C ) Cross-linking of PDZ and GK as shown by Li et al. (3) results in higher-order complexes because of free SH3 and GK contact sites (arrow) for oligomerization. The PALS1/Crb has been modeled using the data from Li et al. (3) and published structures (PDB ID codes 3UIT, 3UAT, and 1JXM). Author contributions: C.R. and M.M. wrote the paper.