Structural Basis for the Phosphorylation-regulated Interaction between the Cytoplasmic Tail of Cell Polarity Protein Crumbs and the Actin-binding Protein Moesin*

Background: Crumbs functions in cell growth control via its short cytoplasmic tail. Results: The structure of the crumbs cytoplasmic tail in complex with moesin protein 4.1/ezrin/radixin/moesin (FERM) domain is determined. Conclusion: Phosphorylation of crumbs cytoplasmic tail disrupts its binding to moesin but not to protein associated with Lin7–1 (PALS1). Significance: Our study suggests a model for the role of crumbs in sensing cell growth and apical-basal polarity establishment. The type I transmembrane protein crumbs (Crb) plays critical roles in the establishment and maintenance of cell polarities in diverse tissues. As such, mutations of Crb can cause different forms of cancers. The cell intrinsic role of Crb in cell polarity is governed by its conserved, 37-residue cytoplasmic tail (Crb-CT) via binding to moesin and protein associated with Lin7–1 (PALS1). However, the detailed mechanism governing the Crb·moesin interaction and the balance of Crb in binding to moesin and PALS1 are not well understood. Here we report the 1.5 Å resolution crystal structure of the moesin protein 4.1/ezrin/radixin/moesin (FERM)·Crb-CT complex, revealing that both the canonical FERM binding motif and the postsynaptic density protein-95/Disc large-1/Zonula occludens-1 (PDZ) binding motif of Crb contribute to the Crb·moesin interaction. We further demonstrate that phosphorylation of Crb-CT by atypical protein kinase C (aPKC) disrupts the Crb·moesin association but has no impact on the Crb·PALS1 interaction. The above results indicate that, upon the establishment of the apical-basal polarity in epithelia, apical-localized aPKC can actively prevent the Crb·moesin complex formation and thereby shift Crb to form complex with PALS1 at apical junctions. Therefore, Crb may serve as an aPKC-mediated sensor in coordinating contact-dependent cell growth inhibition in epithelial tissues.

The type I transmembrane protein crumbs (Crb) plays critical roles in the establishment and maintenance of cell polarities in diverse tissues. As such, mutations of Crb can cause different forms of cancers. The cell intrinsic role of Crb in cell polarity is governed by its conserved, 37-residue cytoplasmic tail (Crb-CT) via binding to moesin and protein associated with Lin7-1 (PALS1). However, the detailed mechanism governing the Crb⅐moesin interaction and the balance of Crb in binding to moesin and PALS1 are not well understood. Here we report the 1.5 Å resolution crystal structure of the moesin protein 4.1/ezrin/radixin/moesin (FERM)⅐Crb-CT complex, revealing that both the canonical FERM binding motif and the postsynaptic density protein-95/Disc large-1/Zonula occludens-1 (PDZ) binding motif of Crb contribute to the Crb⅐moesin interaction. We further demonstrate that phosphorylation of Crb-CT by atypical protein kinase C (aPKC) disrupts the Crb⅐moesin association but has no impact on the Crb⅐PALS1 interaction. The above results indicate that, upon the establishment of the apicalbasal polarity in epithelia, apical-localized aPKC can actively prevent the Crb⅐moesin complex formation and thereby shift Crb to form complex with PALS1 at apical junctions. Therefore, Crb may serve as an aPKC-mediated sensor in coordinating contact-dependent cell growth inhibition in epithelial tissues.
The establishment of epithelial cell polarity is a complicated and dynamic process that involves cell-cell adhesions, assembly of junction complexes, reorganization of cytoskeleton, directional transportation of vesicles, and specific localization of proteins and lipids (1)(2)(3)(4). Over the last two decades, genetic and cell biology studies have identified three tripartite protein complexes, namely the crumbs (Crb) 3 complex composed of Crb⅐PALS1⅐PALS1-associated tight junction (PATJ), the partition-defective (PAR) complex composed of PAR3/PAR6/ aPKC, and the Disc large (DLG) complex composed of DLG/ lethal giant larvae (LGL)/Scribble, as the principle cell polarity regulators (5)(6)(7)(8)(9)(10). Among these regulators, only Crb is a transmembrane protein and necessary for both the apical-basal cell polarity and the assembly of the zonula adherens in Drosophila epithelia (11)(12)(13). In early Drosophila embryo development, dysfunction mutations of Crb lead to the loss of the apicalmembrane identity, and overexpression of Crb causes expansion of the apical-membrane size at the expense of the basolateral membranes (14,15).
As a type I transmembrane protein, Drosophila Crb is composed of an extracellular region, a transmembrane domain, and a 37-residue cytoplasmic tail (Crb-CT) that contains a FERM binding motif (FBM), a PDZ binding motif (PBM), and several potential aPKC phosphorylation sites (16 -19) (Fig. 1A). Expression of transmembrane domain-tethered Crb-CT alone can rescue most of the embryonic polarity defects in crb mutant (15,16), indicating that the cytoplasmic tail plays a crucial role for the functions of Crb. Mechanistically, the PBM of Crb-CT is known to specifically bind to the PDZ-Src homology 3 (SH3)guanylate kinase supramodule of PALS1 and stabilizes the apical Crb complex (20), whereas the FBM is considered not to be directly engaged in the polarity complex formation but instead to be involved in the Hippo signaling pathway and the apical membrane-cytoskeleton regulations (21)(22)(23)(24)(25). Coincidently, the aPKC phosphorylation sites in Crb-CT are located near or within the Crb-FBM (Fig. 1A), implying that the interaction(s) between Crb-FBM and its target(s) may be regulated by aPKCmediated phosphorylation of Crb-CT. The ezrin-radixin-moesin (ERM) proteins are a family of widely distributed membrane-associated proteins that provide a structural linkage between plasma membranes and cortical cytoskeletons (26 -30). These three proteins share similar domain organizations and high sequence identities, all having an N-terminal FERM domain and a C-terminal domain that can fold back to bind to the FERM domain forming an autoinhibited conformation (Fig. 1A) (31)(32)(33). The autoinhibited ERMs can be activated by phosphorylation on a conserved Thr (Thr-567 in ezrin, Thr-564 in radixin, Thr-558 in moesin) at their respective inhibitory C-terminal domain and/or via binding to phosphatidylinositol 4,5-bisphosphate (PIP 2 ) (34 -37). The activated ERMs then bind to membranes via their FERM domains (directly to membrane lipids and/or transmembrane proteins) and to actin filaments via their C-terminal actin binding motifs. In Drosophila epithelia, Crb exhibits close co-localization with moesin and stabilizes the apical membrane-cytoskeleton, leading to reinforcement of the zonula adherens and effective coupling between epithelial morphogenesis and cell polarity (25). However, the detailed mechanism governing the Crb⅐moesin interaction and Crb-mediated regulation between apical-basal cell polarity and apical cytoskeleton reorganization are not well understood.
Here we biochemically and structurally characterized the interaction between Crb-CT and moesin FERM domain. The 1.5 Å resolution crystal structure of the moesin-FERM⅐Crb-CT complex reveals a typical FERM/FBM binding mode in which the FBM of Crb-CT forms a short ␤-strand and fits into the canonical F3 lobe target binding site. To our surprise, the PBM of Crb-CT also contributes to the binding to moesin-FERM by occupying the PIP 2 binding site at the F1/F3 cleft, implying that Crb-CT may mimic the role of PIP 2 in the activation of the ERM family proteins. We further show by NMR spectroscopy that phosphorylation of Crb-CT by aPKC disrupts the Crb⅐moesin association, likely by preventing the formation of the ␤-strand conformation of the Crb-FBM. The biological implications of these findings are discussed.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The mouse moesin FERM domain (residues 1-297) and fly Crb-CT (residues 2110 -2146) were amplified by PCR using their respective fulllength cDNA as the templates. Each PCR product was individually cloned into a modified pET-32 M vector. Various mutants were created using standard two-step PCR-based methods and confirmed by DNA sequencing. Recombinant proteins each with an N-terminal Trx-His 6 tag were expressed in Escherichia coli BL21(DE3) cells at 16°C. The expressed proteins were purified by a nickel-nitrilotriacetic acid-agarose affinity chromatography followed by a size-exclusion chromatography.
Isothermal Titration Calorimetry (ITC) Assay-ITC was carried out on a MicroCal VP-ITC at 25°C. All proteins were dis-solved in a buffer containing 50 mM Tris, pH 7.5, 400 mM NaCl, 1 mM EDTA, and 1 mM DTT. The titration processes were performed by injecting 5-10-l aliquots of protein samples in a syringe (concentration of 200 M) into protein samples in cell (concentration of 20 M) at time intervals of 120 s to ensure that the titration peak returned to the baseline. The relatively high salt concentration is used to ensure that the FERM domains used in the binding studies behave well in solution. The data were analyzed using the Origin 7.0 program and fitted by the one-site binding model.
Crystallization-For crystallization, the tagged proteins were treated with a small amount of human rhinovirus 3C protease at 4°C overnight to cleave the fusion tags and further purified by a step of size-exclusion chromatography. Crystals of the Moesin-FERM⅐Crb-CT complex were obtained by the hanging drop vapor diffusion method at 16°C within 2 days. To set up a hanging drop, 1 l of concentrated protein mixture at a 1:1 stoichiometric ratio was mixed with 1 l of crystallization solution with 20% PEG3350 and 0.2 M ammonium iodine. Before diffraction experiments, crystals were soaked in the crystallization solution containing an additional 30% glycerol for cryoprotection. The diffraction data were collected at the Shanghai Synchrotron Radiation Facility and were processed and scaled using HKL2000 (39).
Structure Determination-The initial phase was determined by molecular replacement using the apo form of moesin-FERM (PDB code 1EF1) as the searching model. The model was refined in Phenix (40) against the 1.5 Å dataset. The Crb-CT peptide was built subsequently in COOT (41). In the final stage, an additional TLS refinement was performed in Phenix. The final model was further validated by using MolProbity (42). The refinement statistics are listed in Table 1. All structure figures were prepared using PyMOL. The sequence alignments were prepared and presented using ClustalW (43).

Crb-CT Specifically Binds to Moesin FERM Domain-Unlike
its extracellular region, the Crb-CT, especially the PBM, FBM, and aPKC phosphorylation sites, is highly conserved across different species and isoforms (Fig. 1A). Specifically, the fly Crb-CT is essentially the same as mammalian Crb2-CT, and therefore, we continued to use the fly Crb-CT after our recent study of the Crb-CT/PALS1 interaction (20). To investigate the Crb⅐moesin interaction, we used highly purified recombinant Crb-CT and moesin-FERM proteins. Quantitative binding assays showed that Crb-CT directly binds to moesin-FERM with a disassociation constant (K d ) of ϳ5 M in the presence of 400 mM NaCl in the assayed buffer (Fig. 1B). Interestingly, although sharing a high amino acid sequence identity with moesin-FERM, the FERM domain of merlin had no detectable binding to Crb-CT (Fig. 1C), indicating that moesin-FERM encodes its intrinsic target binding specificity. A moesin-FERM chimera (termed as moesin-FERM F3/Merlin ), in which the F3 lobe was replaced by the corresponding F3 lobe of merlin, failed to bind to Crb-CT (Fig. 1D), indicating that the F3 lobe is chiefly responsible for the binding of moesin-FERM to Crb-CT.

Overall Structure of the Moesin-FERM⅐Crb-CT Complex-
To understand the molecular basis governing the moesin⅐Crb interaction, we determined the structure of moesin-FERM in complex with Crb-CT by x-ray crystallography ( Table 1). The moesin-FERM⅐Crb-CT complex crystals were diffracted to up to 1.5 Å resolution. In the crystals the moesin-FERM⅐Crb-CT complex forms a heterodimer with one complex per asymmetric unit. The final structural model contains most of residues from the complex, except for a flexible loop (residues 2127-2135) of Crb-CT, which connects its FBM and PBM (Fig. 1A and Fig. 2). Moesin-FERM adopts a typical cloverleaf architecture composed of three lobes (F1, F2, and F3). The overall fold of the FERM domain in the moesin⅐Crb complex structure is essentially identical to the apo moesin-FERM structure (the overall RMSD of 1.6 Å with the 285 aligned residues). The two highly conserved motifs in Crb-CT, FBM and PBM, bind to moesin-FERM at the F3 lobe and a cleft between the F1 and F3 lobes, respectively (Fig. 2).
The Crb-FBM Binding Site in the F3 Lobe of Moesin-FERM-In FERM-containing proteins, F3 lobes act as the major target binding site. A groove (known as the ␣␤-groove) mainly formed by ␤5 F3 and ␣1 F3 is a well characterized target binding region in FERM domains including those of moesin (44), radixin (45)(46)(47)(48), talin (49,50), Merlin (51), and myosin-X (52,53). Most of the previously characterized targets, especially those interacting with ERM proteins, bind to the ␣␤-groove in a ␤-strand or ␤-strand-like structures (Fig. 3B). In the moesin-FERM⅐ Crb-CT complex, the FBM of Crb also adopts a ␤-strand structure and binds to the ␣␤-groove in the F3 lobe of moesin-FERM (Fig. 3A), extending the anti-parallel ␤-sheet formed by ␤5 F3 , ␤6 F3 , and ␤7 F3 . The FBM/F3 interaction is mainly mediated by hydrogen bonds. Some hydrophobic interactions (e.g. Pro-2121 Crb inserts its aliphatic side chain into a hydrophobic cleft formed by Ile-245 F3 and Ile-248 F3 ) also contribute to the FBM/F3 interaction. The three consecutive and strictly conserved residues in the FBM, Gly-2117, Thr-2118, and Tyr-2119 (termed as the GTY motif and known as the iconic sequence of FBMs), are intimately involved in the FBM/F3 interaction (Fig.  3A). The very tight packing between Gly-2117 and the ring of Phe-250 F3 is afforded by the lack of side chain of Gly-2117. The side chain of Thr-2118 forms a pair of hydrogen bonds with Asp-247 F3 . Tyr-2119 forms a strong hydrogen bond with His-288 F3 (bond length of 2.7 Å) in addition to interacting with Met-285 F3 through their hydrophobic side chains. Consistent with our structural analysis, the substitution of Tyr-2119 with Ala abolished Crb-CT binding to moesin-FERM (Fig. 3C).
The PBM of Crb Binds to the F1/F3 Cleft of Moesin-FERM-An unexpected finding in the moesin⅐Crb complex structure is that the Crb-PBM directly contacts the moesin FERM domain  ( Fig. 2A), a mode that has not been observed in any other FERM domains characterized biochemically and structurally. The last seven residues of Crb-CT, which contains its PBM, fit snuggly into the cleft formed by highly conserved residues from the F1 and F3 lobes (Fig. 4A). The C-terminal tail carboxyl group of Crb-PBM forms a salt bridge with Lys-278 F3 . Leu-2145 Crb at the Ϫ1 position of PBM inserts into a hydrophobic pocket formed by Leu-281 F3 , Phe-250 F3 , and the aliphatic part of Lys-278 F3 . Glu-2143 Crb at the Ϫ3 position of PBM forms two salt bridges with Lys-60 F1 and Lys-83 F1 . The residues immediately N-terminal to Crb-PBM also play a role in binding to the F1/F3 cleft (Fig. 4A). Glu-2142 Crb makes two hydrogen bonds with the main chains of Leu-61 and Asn-62 in the ␤4/␤5 loop of the F1 lobe. The two proline residues, Pro-2140 Crb and Pro-2141 Crb , are involved in hydrophobic interactions with residues from the F1 and F3 lobes. Consistent with the above structural analysis, removal of the last four residues from Crb-CT weakened its binding to moesin-FERM by ϳ5-fold assayed in high salt buffers (Fig. 4B). We expect that the contribution of the PBM to the Crb-CT/moesin-FERM is larger at physiological salt concentrations, given the heavy involvement of the charged residues in the interaction. It is important to note that the residues involved in the moesin-FERM⅐Crb-CT interaction are highly conserved in ERM proteins as well as Crb homologs across spe- cies (Fig. 1A), suggesting that this interaction is conserved throughout metazoan.
ERM proteins are believed to associate with membranes either by binding to transmembrane proteins or directly to phospholipids. The association of ERMs with PIP 2 potentiates the activation of ERM proteins (34,37). Interestingly, the Crb-PBM binding site in F1/F3 cleft overlaps with the PIP 2 head group binding site on moesin-FERM (54) and the Crb-PBM mimics PIP 2 binding to moesin by engaging positively charged residues in the F1/F3 cleft (Fig. 4, C and D), indicating that the bindings of Crb and PIP 2 to moesin-FERM are mutually exclusive. Therefore, it is hypothesized that the association of Crb with ERMs may potentiate the activation as well as membrane localization of ERMs better than phospholipids alone.
aPKC Phosphorylates Crb and in Turn Abolishes the Moesin⅐CrbInteraction-AlthoughaPKC-dependentphosphorylation of Crb is essential for epithelial apical-basal cell polarity in Drosophila (19), the molecular basis underlying this phosphorylation-dependent cell polarity establishment and maintenance is not clear. The aPKC phosphorylation sites (Thr-2115 and Thr-2118) are located near/in Crb-FBM (Fig. 3A). We demonstrated earlier on that Crb-FBM is not involved in the binding of Crb-CT to PALS1 (20). Therefore, we focused our investigation on the potential phosphorylation-regulated interaction between Crb-CT and moesin-FERM. Perhaps it is not too surprising that a phosphorylation-mimic mutation of Crb-CT (Crb-CT_T2118E) only moderately weakened its binding to moesin-FERM (Fig. 5A), as the side chain of Thr-2118 is solvent-exposed and makes relatively minor contribution to the binding by forming hydrogen bond with Asp-247 F3 . Surprisingly, a synthetic phosphor-peptide of Crb-CT, in which Thr-2118 of the FBM (i.e. GT2118Y) was specifically phosphorylated, displayed no detectable binding to moesin-FERM (Fig. 5B), indicating that aPKC-mediated phosphorylation can completely disrupt the Crb⅐moesin interaction. Based on the structure shown in Fig. 3A, the large difference between the phosphor-mimetic mutation of Crb-CT (Crb-CT_T2118E) and the corresponding phosphor-Crb-CT (Crb-CT_Thr(P)-2118) in their bindings to moesin-FERM cannot be explained solely by their charge differences. Instead, the phosphorylation of Thr-2118 might alter the conformation of Crb-CT, and such phosphorylation-induced conformational alterations cannot be fully mimicked by the T2118E mutation. To test this hypothesis, we compared the structural changes of Crb-CT induced by Thr-2118 phosphorylation by NMR spectroscopy. The NOE pattern of WT Crb-FBM peptide derived from the 1 H homonuclear NOESY spectrum indicates that the peptide adopts a largely extended structure (Fig. 5D) and thus can easily form the observed ␤-strand structure upon forming complex with moesin-FERM. In contrast, a stretch of (i, iϩ2) NOEs surrounding Thr(P)-2118 were detected for the Crb-FBM_Thr(P)-2118 peptide (Fig. 5, E and F), indicating that phosphorylation of Thr-2118 induces the formation of a turn-like structure between Arg-2116 and Ser-2120, and this turn-like structure is likely stabilized by the interaction of the side chains between Arg-2116 and Thr(P)-2118 (Fig. 5, E and F). As a consequence, the formation of the turn-like structure of the phosphor-Crb-FBM likely prevents Crb-CT from binding to moesin-FERM.

DISCUSSION
The interaction between Crb-CT and moesin characterized here is ϳ50-fold weaker than the Crb-CT/PALS1 interaction that we demonstrated earlier on (20). Would the crumbs⅐ moesin interaction even occur if PALS1 is present? We believe that both interactions can exist in cells, but the two interactions likely occur in different regions/time points/growth conditions in living cells. There are several possible scenarios that can occur in cells. Although the binding between crumbs and moesin is not very strong, the enrichment of moesin by actin filaments (via the C-terminal actin binding domain of moesin) can increase the binding avidity between crumbs and moesin. In polarized epithelial cells, PALS1 is normally enriched in the apical cell cortex together with aPKC; the PALS1/crumbs interaction would dominate under such condition. On the other hand, the moesin/crumbs complex may have the advantage of forming in regions with enriched moesin/actin filaments (e.g. at leading edges of non-polarized migrating cells).
Actin cytoskeleton reorganization is a major step during the establishment of apical-basal cell polarity. A series of recent studies has shown that the actin cytoskeleton organization is closely linked to the Hippo signaling pathway, as numerous Hippo-mediated cellular processes (e.g. cell growth and differentiation, cell-cell, and cell-matrix contact-induced tissue morphogenesis/homeostasis, cell migrations, etc.) involve changes in the F-actin structures (55)(56)(57)(58)(59)(60). We demonstrate in this study that aPKC-mediated phosphorylation of Crb-CT disrupts its binding to moesin and thus can result in weakening of the moesin-tethered coupling of Crb-harboring plasma membranes to cortical actin cytoskeleton and consequent reduction of cortical tensions in polarized epithelia. This cortical tension change may then trigger the activation of the Hippo signaling pathway and prevent cells from further growth. Because aPKC-mediated phosphorylation of Crb-CT does not affect the interaction between Crb-CT and PALS1 (Fig. 5C), aPKC phosphorylationmediated release of Crb from moesin is likely to be accompanied by increased formation of the Crb⅐PALS1 complex at the tight junctions, a process favorable for cell polarity establishment and stabilization. As such, a plausible picture of the role of Crb in connecting apical-basal cell polarity establishment and contact-induced cell growth inhibition emerges (Fig. 6). In this picture, formation of the apical-basal polarity leads to apical enrichment and activation of aPKC and subsequent phosphorylation of Crb-CT. The phosphorylated Crb dissociates from moesin and in turn promotes Crb to form complex with PALS1, thereby stabilizing apical-basal polarity with concomitant cell growth inhibition possibly via activation of the Hippo signaling pathway. In addition to its role in the regulation of apical-basal cell polarity and apical membrane/cytoskeleton interactions, Crb-CT has also been reported to directly function in the Hippo signaling pathway by binding to the FERM domain-containing protein Expanded (22,23). Both Crb/Expanded and Crb⅐ moesin interactions involve the conserved FBM with aPKC phosphorylation sites. Understanding the relations of Crb-CT to its diverse targets may provide further insights for the regulatory roles of Crb in distinct cellular processes. Our study here indicates that aPKC plays a vital regulatory role in determining the functions of Crb in cell growth or cell polarity establishment.