Structural insights into the aPKC regulatory switch mechanism of the human cell polarity protein lethal giant larvae 2

Significance

Epithelial cells have a spatially polarized organization. For example, one surface of an intestinal epithelial cell, called the apical side, faces the lumen of the gut and has a membrane composition distinct from those of the basolateral sides. Several proteins that control the development and maintenance of apical-basolateral polarity have been identified, but their molecular mechanisms are poorly understood. Lethal giant larvae (Lgl) is a basolateral polarity protein that is lost selectively from the apical membrane during development, due to its phosphorylation by atypical protein kinase C. Here, we describe the 3D structure of Lgl in both its unmodified and phosphorylated states, and show that phosphorylation of Lgl mediates a structural switch that controls its association with the plasma membrane.

Abstract

Metazoan cell polarity is controlled by a set of highly conserved proteins. Lethal giant larvae (Lgl) functions in apical-basal polarity through phosphorylation-dependent interactions with several other proteins as well as the plasma membrane. Phosphorylation of Lgl by atypical protein kinase C (aPKC), a component of the partitioning-defective (Par) complex in epithelial cells, excludes Lgl from the apical membrane, a crucial step in the establishment of epithelial cell polarity. We present the crystal structures of human Lgl2 in both its unphosphorylated and aPKC-phosphorylated states. Lgl2 adopts a double β-propeller structure that is unchanged by aPKC phosphorylation of an unstructured loop in its second β-propeller, ruling out models of phosphorylation-dependent conformational change. We demonstrate that phosphorylation controls the direct binding of purified Lgl2 to negative phospholipids in vitro. We also show that a coil–helix transition of this region that is promoted by phosphatidylinositol 4,5-bisphosphate (PIP₂) is also phosphorylation-dependent, implying a highly effective phosphorylative switch for membrane association.

The development of structurally and functionally distinct cell surfaces is essential for the proper function of most animal cells. The establishment and maintenance of such cell polarity require a set of so-called polarity proteins, whose core components are conserved throughout metazoa. In epithelial cells, apicobasal polarity is controlled by the spatial and temporal cross-talk between polarity complexes located on the apical or basolateral membranes, as well as the cell–cell junctions that delineate these membrane domains (1, 2). A central modulator of this process is atypical protein kinase C (aPKC) (3). aPKC robustly interacts with partitioning-defective 6 (Par-6), and they are found together in a subapical epithelial region in complex with the aPKC substrate Par-3 (bazooka in Drosophila) (4, 5). The aPKC/Par-6/Par-3 complex is important for the establishment of apical-basal polarity, as well as for the maturation of epithelial junctions in Drosophila and mammals. It also has roles in asymmetric cell division (6–8). In Caenorhabditis elegans zygotes, a homologous PKC3/Par-6/Par-3 complex, which localizes to the anterior half following the entry of the sperm cell, is essential for anterior/posterior polarity (6). The aPKC/Par-6 complex also colocalizes with the crumbs/PALS1/PATJ (crumbs/stardust/discs-lost in Drosophila) polarity complex located at the apical domain. This complex is also crucial for the establishment of epithelial apicobasal polarity, and it is likewise regulated by aPKC phosphorylation (9).

In addition to its roles in the apical polarity complexes, aPKC is responsible for the phosphorylation of other polarity proteins, including lethal giant larvae (Lgl) (3). First identified in Drosophila genetic screens, lgl mutant flies develop cancer-like defects, leading to uncontrolled growth of larval brain neuroblasts and imaginal discs (10–13). Lgl is conserved in eukaryotes and can be found in two isoforms in mammals (14). Its mutations result in polarity defects in mice and other animals (15–17), and down-regulation of Lgl occurs in various human cancers (18). Lgl has important roles in all aspects of cell polarity, including in the development of epithelial apical-basal polarity, asymmetrical cell division, and cell migration.

The molecular basis of Lgl function in cell polarity is poorly understood, but it is clear that it depends upon aPKC phosphorylation-dependent cellular localization. In early stages of Drosophila epithelial development, when polarity is being established, Lgl is both cytoplasmic and uniformly localized at the cell cortex. At later stages, aPKC phosphorylation excludes Lgl from the apical domain, where Par-6 and aPKC concentrate. This apical exclusion does not occur in a nonphosphorylatable mutant Lgl, which results in aberrant polarity (19). In fully polarized cells, Lgl is located mostly at the lateral membrane, where it actively excludes Par-6 from the cell cortex (19). A similar spatial and temporal localization of Lgl in relation to the aPKC/Par-6 complex is also observed in mammalian epithelial cultures (20).

In a polarized epithelial cell, Lgl colocalizes with the polarity proteins scribble (Scrib) and discs large (Dlg) at the apical margin of the lateral membrane. Experiments in Drosophila have demonstrated a strong genetic interaction among these three genes, indicating that they act together in a common pathway in the regulation of cell polarity (21). Lgl and Dlg mutations have been shown to produce similar effects on fly development (21), and a direct low-affinity interaction between the guanylate kinase domain of human Dlg4 and a Lgl2 peptide containing phosphorylated aPKC target sites has been characterized biochemically and structurally (22). Interactions between Scrib and Lgl have also been demonstrated (23), but strong biochemical evidence for the existence of a ternary Lgl/Dlg/Scrib complex is lacking. Interaction of Lgl, controlled by aPKC phosphorylation, has been reported with additional targets, including nonmuscle myosin II (NMII) (15, 24, 25), syntaxin4 (26, 27), and others (28–32). In addition to aPKC, cytoplasmic Lgl is phosphorylated by the aurora A and B kinases at mitosis in epithelial cells, which promotes its mitotic relocalization (33–35).

Budding yeast express the Lgl homolog Sro7, a 1,033-residue protein that is essential for polarized exocytosis in bud growth (36). The Sro7 structure (37) comprises 14 WD40 repeats arranged in two seven bladed β-propeller barrels. The distant Lgl/Sro7 sequence homology (∼10% identity) suggests them to be structurally and functionally related. Mouse Lgl1 was shown to partially rescue low salt tolerance and temperature sensitivity associated with the loss of Sro7 and its homolog, Sro77 (38, 39). The role of Lgl in exocytosis in vertebrates has yet to be solidly established, however.

The target serine-rich sequence for aPKC phosphorylation resides in a peptide region predicted to connect two adjacent WD40 repeats in Lgl (Fig. 1A). The equivalent sequence is missing from Sro7, consistent with the absence of its regulation by aPKC. Phosphorylation of Lgl by aPKC changes its cellular localization and its protein–protein interaction preferences. A major structural rearrangement promoted by phosphorylation has been proposed as a mechanistic explanation for this switch (40). On the other hand, the electrostatic alteration of this region by multiple serine phosphorylations in the aPKC target site could be the major determinant of membrane targeting and/or protein–protein interaction.

Fig. 1.

Human Lgl2 protein. (A) Cartoon representation of the human Lgl2 primary structure. N and C β-propeller blades, numbered 1–14, appear in blue and yellow, respectively. The region containing the aPKC-targeted serines (residues 641–680) is marked in black, with the seven serine residues detected by mass spectrometry to be phosphorylated by aPKCι shown in red. Each β-propeller blade is counted from the N terminus of strand A to the C terminus of strand D (including internal loops). The C propeller loop regions discussed in the main text are shown in parentheses. Regions excluded from the crystallized Lgl2(13–978) construct are shown in gray. The dashed arrow represents the spatial proximity of β14C and β14D, which are both part of blade 14. The drawing is scaled for the relative sequence lengths of each structural domain. (B) Cartoon representation of the Lgl2 structure (top view of the pLgl2 crystal form 2 structure is presented). The structure is rainbow-colored to better illustrate the peptide chain continuity (N and C termini are blue and red, respectively). The β-propeller blades are numbered in black, and the individual A–D strands are noted in blade 1. The 10D and 11A ends of the unstructured 10–11 loop at the top side of Lgl2 are indicated.

Here, we report crystal structures of human Lgl2 in both its phosphorylated and unphosphorylated forms. Our structural results confirm the double β-propeller core structure for Lgl2. There are no apparent structural variations between the unphosphorylated and phosphorylated states of the protein. We demonstrate the preferential interaction of Lgl2 protein with phosphatidylinositol 4,5-bisphosphate (PIP₂)-containing membranes that is abolished by aPKC phosphorylation, as previously demonstrated with a polybasic peptide fragment of this loop (41, 42). We also show that a coil–helix transformation of this region, promoted by PIP₂, is phosphorylation-dependent, which implies an effective membrane switch mechanism. These results set a firm basis for understanding the mechanistic roles of Lgl in cell polarity.

Results

Structure of Lgl2.

To explore the structure of Lgl in its different functional forms, we crystallized human Lgl2 both in its unphosphorylated state and in its in vitro aPKCι-phosphorylated state. The human Lgl2 gene encodes a 1,020-amino acid protein (Fig. 1A). Based on a multiple sequence alignment (MSA) of Lgl homologs with the structured protein regions of Sro7 (37), the Lgl2 construct used in this study was slightly trimmed at its N and C termini; the deleted regions were also predicted to be unstructured by Foldindex (43). Crystals of the unphosphorylated Lgl2(13–978) appeared in two forms that yielded useful diffraction data. Crystal form 1 (44) diffracted to 3.2 Å, whereas form 2 crystals (45) diffracted anisotropically with a maximum limit of 2.2 Å. Phosphorylated Lgl2 (pLgl2) also yielded crystals, similar to the unphosphorylated Lgl2 form 2 crystals (46), and diffracted to 1.9 Å. Another crystal form (form 3) of pLgl2 (47) was also obtained, and diffracted to 3.9 Å.

Phase determination proved exceptionally challenging for the Lgl2 crystals. Traditional molecular replacement methods using Sro7 as a search model consistently failed, which was not surprising, given the 10% sequence identity between these proteins. Experimental phasing was also problematic in all of the Lgl2 crystal forms. However, we were able to determine the structure of crystal form 1 using a computational method that is described in an accompanying paper (48).

The refined form 1 Lgl2 structure was used as a search model to solve the higher resolution form 2 crystal structures of both unphosphorylated Lgl2 and pLgl2. The electron density maps from these crystals were used to build models that were very similar to the search model, but the higher resolution of crystal form 2 revealed additional details. These included water molecules and other small molecules present in the crystallization buffer, as well as an α-helix at the Lgl2 N terminus (residues 14–20) that is absent in the form 1 model (48). This α-helix contributes to a crystal packing interface in both of the crystal forms 2 and 3 that is mediated mostly by hydrogen bonds. The crystal form 1 Lgl2 structure includes 827 of 979 amino acids of the Lgl2 protein construct; there was no electron density for residues 13–17, 471–484, 635–707, and 853–858. Both the unphosphorylated and phosphorylated form 2 structures are missing residues 261–263, 472–485, and 631–708, and the loop residues 554–555 are missing in the unphosphorylated form 2 structure. The C-terminal sequence beyond S937 is also missing from all of the structures. Other than these slight variations in the unstructured regions, the structures are highly similar (Fig. 2 A and B).

Fig. 2.

Lgl2 structures. Alignment between the unphosphorylated Lgl2 crystal form 2 structure (A and B; magenta), unphosphorylated Lgl2 crystal form 1 (rmsd = 0.72 Å, 97.94% structure overlap) (A, cyan), and pLgl2 crystal form 2 structure (rmsd = 0.30 Å, 99.88% structure overlap) (B, blue). (C) Structural alignment between the unphosphorylated Lgl2 crystal form 1 (cyan) and Sro7 (orange, Protein Data Bank ID code 2OAJ) (rmsd = 2.3 Å, 57.99% structure overlap). Top views are shown on the left side, and side views are shown on the right side.

The Lgl2 structure comprises 14 antiparallel β-sheets, which fold into a couple of seven-bladed β-propeller structures that, together, resemble an open clamshell (Figs. 1B and 2). The first β-propeller of Lgl2 comprises residues 33–379, and residues 380–923 form the second β-propeller. The last β-strand of the second propeller (14D) is formed by residues 24–29 near the N terminus of the protein, thereby connecting the first and last strands of the structure with a short peptide link (Fig. 1B). This topology is also observed in other double-barrel β-propeller proteins, including Sro7 (Fig. 2C), and is assumed to contribute to the structural stability of the protein fold and the tight interaction between the two propellers (37, 49, 50). The seven antiparallel β-sheets that form the blades of each propeller have a simple connectivity with neighboring strands contiguous in sequence, with the first strand (A) closest to the barrel axis and the last at the periphery. This topology orients the AB and CD loops to the “bottom” of the clamshell, whereas the BC and loops connecting strand D to strand A of the next blade are oriented to the “top” (Fig. 1B). Whereas the structural framework, including the relative orientation between the β-propellers, is relatively conserved between Lgl2 and Sro7, the loops at the top and bottom surfaces of the β-propellers, important for specific interactions, are unique to each protein (Fig. 2C).

The second β-propeller contains significantly more of the Lgl2 protein sequence than does the first propeller, and presents some irregular features relative to the first propeller. In particular, blades 8–11 contain several long insertions between and within the blades (Figs. 1 and 3). The largest insertion lies between blades 10 and 11 (residues 627–713). This peptide region, which includes stretches of highly conserved residues, including the aPKC-targeted serines, is mostly missing from the structure (residues 631–708 in crystal form 2 and residues 635–707 in crystal form 1). Both ends of this region face the top side of the protein, where the ordered structure of the loop ends 6 amino acids (or 10 amino acids for the crystal form 2) N-terminal to the first aPKC phosphorylation site at S641 discussed below (Fig. 3C). A second region deviating from the canonical β-propeller fold is the 9CD loop that includes residues 534–576 (Figs. 1A and 4). This elongated loop folds back along the lateral side of the second β-propeller toward the top and includes a β-strand (residues 543–547) that pairs in a parallel orientation with strand D of the blade. Following this strand, a sequence-conserved part of this loop (residues 548–564, partly missing in the form 2 unphosphorylated structure) sits above the top surface level of the second β-propeller (Fig. 3). This forms a noticeable protrusion at the top surface. Interestingly, this part of the 9CD loop is in close proximity to the observed termini of the regulatory 10–11 loop (Fig. 3 C and D).

Fig. 3.

Large insertion elements in the Lgl2 C β-propeller. Views from the top (A), bottom (B), and sides (C and D) are shown. The different insertion elements are colored as follows: 8AB, yellow; 8–9, cyan; 9CD, blue; 10–11, green. The C-terminal region is colored orange.

Fig. 4.

Sequence conservation in Lgl2. (A) Side view. (B) Surface representation. The top (Upper) and bottom (Lower) surfaces are shown. Sequence conservation was determined by ConSurf based on MSA of 150 Lgl orthologs.

In the three different crystal forms, Lgl2 maintains the same relative angle between the two β-propellers (Fig. 2 and SI Appendix, Fig. S1), which indicates that the interface between them is quite rigid. The interface is large, 2,422 Ų, and features extensive van der Waals and hydrogen bond interactions that are mediated principally by peptide connections that meander between the domains, such as the connection between blades 14 and 1 near the N terminus. The elongated connection between strands 8A and 8B (Fig. 1A) makes a major contribution to this interface. Just after emerging from strand 8A, this segment folds into a long α-helix (α5) that docks between blades 8 and 9 and the bottom side of the 9CD loop (Fig. 3 and SI Appendix, Fig. S2). Following this helix, a long loop stretches over the bottom side of the first propeller, where it contacts residues from blade 7, blade 1, and the C-terminal tail before connecting back to strand 8B (Fig. 3 and SI Appendix, Fig. S2). This ring-shaped structure thus appears to contribute to the structural rigidity of the interpropeller interface. In addition, an exceptionally long loop between blades 8 and 9 (Fig. 1A, partly disordered in the structures) contacts blades 6 and 7 in the first β-propeller. Finally, the C-terminal tail of Lgl2, after exiting β-strand 14C, goes back to the first propeller and interacts with the bottom side of blades 1 and 2 (Fig. 3 and SI Appendix, Fig. S2). In accord with their major role in the interdomain interaction, these three peptide regions contribute 74% of the buried interface between domains; without them, the interface would only be 627 Ų.

In their roles as protein–protein interaction hubs, the key interactions of the WD40 family reside at their protein surface (51, 52). Alignment of metazoan Lgl sequences reveals a striking disparity in the level of conservation of residues located close to the top side versus those located at the bottom of the two β-propellers (Fig. 4). The bottom surface is poorly conserved, whereas moving toward the top, the conservation becomes stronger. Such a pattern is found in other WD40 β-propeller proteins and originates, in part, from the signature WD40 sequence located at the C-terminal part of the C and D strands (52). In addition, most of the characterized WD40 β-propeller interactions involve residues at the top surfaces (52). Indeed, the loops on the top surface of Lgl2 show the highest conservation scores, with the most conserved segments residing on the loops of the second β-propeller, including the protrusion formed by the unique 9CD loop and the highly conserved 8–9 connection. In addition, loops from blades 5–7 in the first propeller contribute to the highly conserved surface. The unstructured 10–11 loop that contains the conserved phosphoserine region, as well as other highly conserved positions, sits at one end of this conserved surface (Fig. 4B). We speculate that this conserved surface is important for interactions with binding partners that are regulated by phosphorylation on the nearby phosphoserine loop.

The Effect of aPKC Phosphorylation on Lgl2.

aPKC targets a region on the 10–11 loop that includes several highly conserved serine residues (SI Appendix, Fig. S3). Specific aPKC phosphorylation has been reported for the different orthologs of Lgl at three to five of these serine residues (20, 27, 53, 54). We examined the phosphorylation of purified Lgl2 treated with purified aPKCι by mass spectrometry (the same sample that was used for crystallization), and found seven phosphorylated serine residues on the 10–11 loop (S641, S645, S649, S653, S660, S663, and S680), which include the previously reported sites (Fig. 1 and SI Appendix, Figs. S3 and S4). The unphosphorylated Lgl2 control sample showed no phosphorylation in the same analysis. Previous work has suggested that a large structural rearrangement of Lgl occurs upon aPKC phosphorylation (40). However, the unphosphorylated Lgl2 and pLgl2 structures in crystal form 2 are virtually identical (Fig. 2B). Moreover, the structure of phosphorylated Lgl2 solved in an additional crystal form (form 3) is similar to the others, indicating that there is no influence of crystallization conditions or crystal packing on these results (SI Appendix, Fig. S1). Importantly, the loop between blades 10 and 11, containing the target serine phosphorylation sites, is missing in both the pLgl2 and unphosphorylated Lgl2 structures. Correspondingly, circular dichroism (CD) spectroscopy analysis of Lgl2(626–680), a protein fragment covering most of the sequence of this loop, indicates that this region is a random coil in solution for both its aPKC-phosphorylated and unphosphorylated forms (Fig. 5 A and B).

Fig. 5.

Coil–helix transition of the Lgl2 10–11 loop. (A) SDS/PAGE of the purified control and aPKC phosphorylated Lgl2(626–680). Due to its highly basic nature, Lgl(626–680) migrates abnormally slowly in SDS/PAGE. This is moderated by aPKC phosphorylation. (B) The 10–11 loop is unstructured in solution regardless of its phosphorylation state. CD spectra of Lgl2(626–680) vs. pLgl2(626–680). deg, degrees. (C) PIP₂ binding promotes an α-helical conformation of the 10–11 loop. CD spectroscopy of Lgl2(626–680) with increasing 10% PIP2/90% PC membrane vesicles. (D) aPKC phosphorylation significantly reduces α-helix propensity of the 10–11 loop. The α-helix percentage values were calculated from the mean residue molar ellipticity values of CD signals at 222 nm of Lgl2(626–680) and pLgl2(626–680) at increasing trifluoroethanol (TFE) contents.

Previous work has demonstrated the importance of a polybasic region in the 10–11 loop in Lgl for membrane targeting (41, 42). This region contains multiple highly conserved, positively charged residues located within the aPKC phosphorylation target region of Lgl (SI Appendix, Fig. S3). The direct interaction of Lgl with the plasma membrane (PM) was shown to be independent of an intact cortical actin network but, instead, depends on the presence of lipids with negatively charged head groups (41). A preference was demonstrated for PIP₂ and, to a lesser extent, phosphatidylinositol 4-phosphate (PIP₄), both of which are enriched in the inner leaflet of the PM (41, 42). The in vivo PM localization of Lgl was shown to decrease, or completely vanish, when multiple lysines and arginines of this region were mutated to alanine, or when the inner PM PIP₂ and PIP₄ were electrostatically blocked or specifically depleted. This was also shown for isolated polybasic regions of Lgl, which displayed reduced binding to negatively charged membrane vesicles both when their positively charged residues were mutated and when these peptides were phosphorylated by aPKC (41, 42). To test whether these properties are also found in the intact protein, we tested purified Lgl2(13–978) for its in vitro binding of membrane vesicles. Unphosphorylated Lgl2 showed direct binding to membranes, with preference given to negatively charged PIP₂-containing vesicles (Fig. 6 A and B). In contrast, pLgl2 bound PIP₂-containing vesicles weakly and did not bind 1-oleoyl-2-palmitoyl-sn-glycero-3-phosphocholine (PC) vesicles at all (Fig. 6B). In addition, phosphorylation by aPKC releases the already bound Lgl2 from PIP₂/PC membrane vesicles (Fig. 6C). These results support a mechanism in which the phosphorylation-dependent dissociation of Lgl from the apical membrane is due to neutralization of the positively charged residues of this loop by neighboring phosphoserines, thereby ablating the electrostatic association with the membrane. Thus, it appears that local chemical modification of the 10–11 loop by aPKC, rather than extensive structural changes in Lgl, drives its membrane association.

Fig. 6.

Lgl2 binds negatively charged membranes and is released by aPKC phosphorylation. Pelleting of Lgl2 with vesicles containing PC and the indicated fraction of negatively charged PIP₂ or PS is shown. Membrane-bound protein is detected in the pellet (P), while the unbound protein is in the supernatant (S). (A) Unphosphorylated Lgl2 binds preferentially to negatively charged vesicles, with a higher preference for PIP₂-containing vesicles compared with PS-containing vesicles with similar charge density (compare lanes 2 and 3 and lanes 4 and 5). (B) Binding of unphosphorylated or aPKC pre-pLgl2(13–978) to PC or 90:10 (PC/PIP₂) vesicles. (C) Vesicle-bound Lgl2 is released from the pellet by aPKC phosphorylation. Lgl2-bound vesicle mixtures were incubated with assay buffer with or without ATP (1 mM final concentration) and aPKCι kinase domain (1:100 kinase/Lgl2 ratio) before membrane pelleting.

To test the structural consequences of membrane interactions on the 10–11 loop, we measured the effect of adding PIP₂-containing vesicles (10% PIP₂/90% PC) to Lgl2(626–680) using CD spectroscopy. A short peptide region harboring the C. elegans Lgl phosphorylation sites (residues 656–681, corresponding to human Lgl2 640–665) was previously shown to fold into an α-helix in the presence of phosphatidylserine (PS)-containing negatively charged vesicles (42). Our results similarly show a clear transition from random coil to α-helix with increasing amounts of added PIP₂ vesicles (Fig. 5C). A helical wheel projection of the residues around the polybasic region of Lgl2 reveals that most of the positively charged amino acids lie on one side of the α-helix (SI Appendix, Fig. S5), which would form an electrostatically favorable site for binding the negatively charged membrane. We next compared the coil–helix transitions of the phosphorylated and unphosphorylated Lgl2(626–680) fragments by addition of trifluoroethanol (Fig. 5D). Our results show that aPKC phosphorylation significantly reduces the α-helix propensity of this protein region. Thus, phosphorylation both neutralizes the positive charges and disrupts formation of a favorable spatial distribution of positively charged residues needed for the interaction with negatively charged membranes.

Discussion

Although its important role in cell polarity is well established, the mechanism of Lgl in this process is poorly understood. Without any evidence for enzymatic activity, Lgl presumably functions as a dynamic scaffold protein that participates in various protein–protein interactions. These interactions depend on its phosphorylation state and its cellular localization. Our data present the high-resolution structure of a metazoan Lgl. The structure is characterized by a stable core with a highly conserved surface that is a likely site for multiple protein–protein interactions regulated by phosphorylation.

The peptide that connects Lgl blades 10 and 11 contains conserved serine phosphorylation sites and positively charged amino acids that, upon mutation, disrupt Lgl cellular localization and lead to loss of polarity. Previous work has suggested that this region could act as a flexible hinge between the N- and C-terminal regions of Lgl, which, upon phosphorylation, would be made rigid by the interaction of the negative phosphate groups with positively charged amino acid side chains. This molecular rearrangement in Lgl was proposed to promote dissociation of Lgl from the actin cytoskeleton (40). The structures of Lgl2 in both its unphosphorylated and aPKC-phosphorylated states rule out major structural changes produced by aPKC phosphorylation. Rather, it appears that association with the membrane is mediated by the α-helical polybasic region of the 10–11 loop sequence, which is flexibly connected to the double-barrel core of Lgl. Moreover, modeling suggests that interaction of the helix with the membrane would place the core of the protein away from the membrane (SI Appendix, Fig. S6), making it unlikely that there would be an influence of the membrane itself on the rest of the structure.

Although a phosphorylation-induced structural switch is not supported by our results, the arrangement of multiple serine phosphorylation sites flanked by multiple positively charged residues likely produces a major electrostatic on/off switch. Induced by the charge neutralization effect of phosphorylation, this switch controls the interaction and dissociation of Lgl from negatively charged lipid head groups, such as PIP₂, that are enriched at the intracellular surface of the PM. Consistent with this model, deletion of this charged region of the loop results in loss of membrane localization, and multiple serine-to-alanine mutations in the 10–11 loop break the switch and the aPKC-dependent exclusion of Lgl from the apical domain (41). Our results with in vitro purified Lgl2 protein support this previously described model for the PM localization of Lgl by direct binding to the polar membrane. Similar to Lgl, mira, and numb, other aPKC substrate proteins that are excluded from the apical cortex by the Par complex also contain positively charged motifs that mediate their cortical localization. As in Lgl, these motifs preferentially bind to negatively charged phospholipids and have proximal aPKC phosphorylation sites (55). This suggests a common scheme in the mechanism of apical cortical exclusion by aPKC.

At early stages of development, Lgl is uniformly located at the cell cortex. At a certain time point, aPKC phosphorylation is triggered, which results in turning off binding of Lgl to the apical membrane. Enrichment of PIP₂ phospholipids at the apical membrane is induced by the PIP₃ phosphatase PTEN during epithelial morphogenesis (56). Thus, the laterally localized Lgl observed at later stages of development might be due to the binding to downstream membrane-localized targets, such as Dlg or Scrib, rather than the PM. Moreover, pLgl peptides have been shown to bind to Dlg4. Further work is required to verify the phosphorylation state of Lgl at the lateral membrane and its functional consequences.

Methods

DNA Constructs.

The DNA encoding the human Lgl2 residues 13–978, followed by a tobacco etch virus (TEV) protease site and a protein-A tag, was cloned into the pVL-1393 insect cell expression vector (57). The DNA of human aPKCι kinase domain (248–596) was cloned into the pAcHLT-A (BD Biosciences) baculovirus expression vector, with an N-terminal 6xHis tag and a TEV site preceding the kinase sequence. The DNA-coding Lgl2(626–680) was cloned into the MCS1 of the pCDF-Duet (Novagen) Escherichia coli expression vector, with an N-terminal a 6xHis tag, a maltose-binding protein (MBP), and a TEV site sequence preceding the Lgl2 sequence.

Protein Expression and Purification.

For protein expression in Sf9 cells [Lgl2(13–978), aPKCι], recombinant baculovirus was obtained using BestBac 2.0 Linearized Baculovirus DNA (Expression Systems), following standard procedures. Cells were grown in SF9 media infected by the recombinant baculovirus and kept shaking at 27 °C for 48 h before being harvested by centrifugation. Cell pellets were suspended in their appropriate buffers (discussed below) and stored at −80 °C.

For expression in E. coli [Lgl2(626–680)], BL21(DE3) cells transformed with the expression vector were grown in LB media at 37 °C until reaching an OD₆₀₀ of 0.5. Isopropyl β-d-1-thiogalactopyranoside (0.5 mM) was then added, and cells were grown for additional 12–16 h at 16 °C before being harvested by centrifugation.

For purification of unphosphorylated Lgl2, thawed cells were suspended in Lgl2 buffer 1 [20 mM Tris (pH 8.5), 300 mM NaCl, 10 mM MgCl₂, 10% glycerol] containing 1 mM DTT, 1 mM ATP, 1 mM PMSF, 150 nM aprotinin, 1 μM leupeptin, 1 μM E-64 protease inhibitor, and 5 units of DNase, and lysed using an Emulsiflex homogenizer. After a 90-min centrifugation at 185,000 × g, clarified lysate was incubated for 2–4 h with 10 mL of IgG-Sepharose resin, followed by a wash with 300 mL of Lgl2 buffer 1 with 1 mM DTT, 1 mM ATP, and protease inhibitors. The resin was then incubated at 30 °C in 50 mL of the same buffer for 30 min to maximize the dissociation of endogenous Sf9 aPKC from Lgl2. After overnight incubation with Lgl2 buffer 1 with TEV protease, cleaved Lgl2 protein was collected from the column flow-through. The NaCl concentration was then diluted threefold, and the protein was loaded onto a Mono-Q column with 20 mM Tris (pH 8.5), 100 mM NaCl, 10 mM MgCl₂, 10% glycerol, and 1 mM DTT. After a wash with the same buffer, the protein was eluted by an increasing NaCl gradient, where Lgl2 peaked at around 188.7 mM NaCl (13.22 mS/cm). Next, the protein was dephosphorylated by overnight incubation with λ-phosphatase (New England Biolabs) in the manufacturer’s supplied buffer with 1 mM MnCl₂ at room temperature. The dephosphorylated protein was purified using a Superdex 200 size exclusion chromatography (SEC) column in 20 mM Tris (pH 8.0) and 200 mM NaCl (Lgl2 SEC buffer).

For pLgl2 preparation, the purified Lgl2 protein at 2.5 μM was incubated with aPKCι kinase domain (1:10 kinase/Lgl2) in Lgl2 SEC buffer containing 1 mM ATP and 10 mM MgCl₂ for 8 h at room temperature. The protein was then diluted twofold with 20 mM Tris (pH 8.0) and purified on a Mono-Q column preequilibrated with 20 mM Tris (pH 8.0) and 100 mM NaCl. The pLgl2 was eluted by an increasing NaCl gradient peaking at around 213.5 mM NaCl (19.37 mS/cm). Selected peak fractions were then further purified by Superdex 200 SEC in Lgl2 SEC buffer. Phosphorylation of Lgl2 was assayed by Pro-Q Diamond phosphoprotein gel stain (Thermo Fisher Scientific) (SI Appendix, Fig. S4) and mass spectrometry (discussed below).

The aPKCι kinase domain was purified as described (58), with the following modifications. TALON resin was used instead of HiTrap HP. Protein eluted from the TALON resin was dialyzed against 20 mM Tris (pH 8.0), 100 mM NaCl, 10 mM MgCl₂, 10% glycerol, and 1 mM DTT at 4 °C overnight. Protein was then directly loaded onto Mono-Q column preequilibrated with the same buffer with 0.5 mM EDTA and eluted with an NaCl gradient. Protein at 2 mg/mL was then incubated overnight with 15 mM MgCl₂, 1 mM ATP, and PDK1 at a 1:50 (PDK1/aPKC) mass ratio. PDK1 phosphorylated protein was then purified by a second Mono-Q column with 20 mM Tris (pH 8.0), 100 mM NaCl, 10% glycerol, and 1 mM DTT. Protein was eluted by an NaCl gradient, peaking at around 300 mM NaCl. Eluted protein was then loaded onto a Superdex 200 SEC column with 20 mM Tris (pH 8.0) and 200 mM NaCl.

For Lgl2(626–680) purification, cells expressing a 6xHis–MBP fusion of Lgl2(626–680) were suspended in Lgl2 buffer2 [20 mM Tris (pH 8.0), 200 mM NaCl, 5% glycerol, 5 mM 2-mercaptoethanol (BME)], with 1 mM EDTA, protease inhibitors, and DNase as described above. Cells were lysed as described above. The lysate was incubated for 30 min with 10 mL of amylose resin and eluted in Lgl2 buffer 2 with 10 mM maltose. The eluted protein was purified by preparative Superdex 75 SEC in 20 mM Hepes (pH 8.0), 200 mM NaCl, 5% glycerol, and 5 mM BME, and incubated with TEV protease and λ-phosphatase overnight, as described above. Protein was then loaded onto a Mono S cation exchange column and eluted by an increasing NaCl gradient. Selected fractions containing the free Lgl2(626–680) were then further purified by Superdex peptide SEC in Lgl2 SEC buffer. Since Lgl2(626–680) does not have tryptophan residues in its sequence, concentration of the purified protein was determined by the BCA assay (Pierce).

For comparison of the unphosphorylated Lgl2(626–680) vs. pLgl2(626–680) CD spectroscopy signal, an identical amount of the protein at 170 μM was incubated with aPKCι kinase domain (1:100 kinase/Lgl2) for 3–4 h at room temperature in Lgl2 SEC buffer containing 10 mM MgCl₂ with or without 1 mM ATP. Complete phosphorylation of the protein fragment was detectable by a band migration shift in SDS/PAGE (Fig. 5A). The proteins were then purified from the kinase using Superdex peptide SEC in 20 mM Tris (pH 8.0) and 150 mM NaCl.

Protein Crystallization and Data Collection.

Lgl2 crystals were grown by vapor diffusion at 22 °C. The unphosphorylated Lgl2(13–978) protein was crystallized in two conditions, using protein concentrated to 3–5 mg/mL: Condition A [19.8% PEG 3350, 0.29 M Na₂SO₄, 0.1 M Bis-Tris propane (pH 7.5), 3% methanol] yielded mostly plates with a P422 lattice that were not suitable for data collection and, occasionally, square pyramid-shaped crystals (crystal form 1) in space group P2₁2₁2₁, and condition B [18–21% PEG 2000 MME (methyl ether 2000), 80–100 mM SPG (2:7:7 succinic acid/sodium dihydrogen phosphate/glycine at pH 6.0)] yielded crystals (crystal form 2) in space group C2.

The pLgl2(13–978) was crystalized in two conditions, using protein concentrated to 6–12 mg/mL: condition C [25% PEG 1500, 100 mM SPG (pH 8.0) with crystals similar to the form 2 dephosphorylated Lgl2(13–978) crystals] and condition D [1.4 M ammonium sulfate, 100 mM Hepes (pH 7.5), 150 mM NaCl, 1.07% 1,6-hexanediol (crystal form 3, space group P4₁2₁2)]. Harvested crystals were soaked in cryoprotectant solutions (25% ethylene glycol, 10% glycerol, and 15% glycerol in crystallization solutions for crystal forms 1, 2, and 3, respectively) before being frozen in liquid nitrogen for data collection. Diffraction data were collected under cryogenic conditions at the Stanford Synchrotron Radiation Laboratory. Unit cell parameters, data collection, and refinement statistics are shown in SI Appendix, Table S1.

Structure Solution and Refinement.

Diffraction data were integrated by XDS (59) and scaled by Aimless (60). Due to the anisotropic diffraction, the unphosphorylated Lgl2 and pLgl2 crystal form 2 data were subjected to the STARANISO Server (Global Phasing Limited) (staraniso.globalphasing.org/cgi-bin/staraniso.cgi) to perform an anisotropic cutoff and to apply an anisotropic correction to the data.

Phases for the crystal form 1 data were obtained using the Sro7 structure (Protein Data Bank ID code 2OAJ) as described in an accompanying paper (48). The resulting model was further refined using the Phenix-Refine program (61). For the solution of the other Lgl2 and pLgl2 data, the refined form 1 Lgl2 model was used as a search model for molecular replacement in Phenix-Phaser (61). Structure refinement was done using Phenix-Refine and Buster (62). Refinement statistics are provided in SI Appendix, Table S1.

Structure Analysis.

Lgl2 interface size and content were analyzed by jcPISA (63). For doing this, the peptide chains between L379/A340 and H22/P33 were disconnected, saving a model having each β-propeller as a different chain.

Structure 3D alignments were performed using the CLICK server for topology-independent comparison of bimolecular 3D structures (cospi.iiserpune.ac.in/click).

MSA for homologs of the human Lgl2 gene was done by ConSurf in a sequence-only mode (64) with default parameters [UniRef90, MAFFT (multiple alignment using fast Fourier transform); E-value = 0.0001, iterations = 1, maximal %ID between sequences = 95, minimal %ID for homologs = 35, 150 sequences]. ConSurf in full mode (65–67) was then run using the MSA files from the previous run.

Structure Modeling.

The 10–11 loop missing in the crystal structure was modeled by the ab initio protocol implemented in Rosetta (68) using pairwise contact restraints predicted by our in-house–developed code for amino acid contact prediction. The contact map is predicted by a deep convolutional neural network that has a similar architecture and performance as published models (69).

Mass Spectrometry Analysis.

Samples were processed and analyzed on an Orbitrap Elite Hybrid Ion Trap-Orbitrap Mass Spectrometer. The Top3 method that cycled between CID (collision induced dissociation), ETD (electron transfer dissociation) and HCD (high-energy collisional dissociation) fragmentation modes was used. The data were searched against a Spodoptera frugiperda database, along with the human Lgl2 sequence and common contaminants from the CRAPome (contaminant repository for affinity purification) using Proteome Discoverer 2.2 with Sequest HT, Percolator, and ptmRS.

Membrane Pelleting Assay.

For vesicle preparation, 18:1–16:0 PC, porcine brain l-α-PIP₂, and porcine brain l-α-PS (all from Avanti Polar Lipids, Inc.) were mixed at the desired molar ratio in a glass tube and then kept under argon flow to evaporate the solvent chloroform or chloroform/methanol. The evaporated lipid mixtures were stored overnight in a desiccator under vacuum and then resuspended in 20 mM Tris (pH 8.0), 1 mM DTT, and 200 mM NaCl to reach a total lipid concentration of 5 μM. The solutions were then subjected to 10 cycles of freezing in liquid nitrogen + 2 min of thawing in a water bath. The vesicles were then passed 10 times through an extruder with a 100-nm pore size. Vesicle solutions were kept for few days at 4 °C.

For pelleting assays, purified phosphorylated or unphosphorylated Lgl2(13–978) was diluted in assay buffer [20 mM Tris (pH 8.0), 200 mM NaCl, and 1 mM DTT to 6.3 μM (Fig. 6A); 20 mM Tris (pH 8.0) and 300 mM NaCl to 5.4 μM (Fig. 6B); and 20 mM Tris (pH 8.0), 300 mM NaCl, and 1 mM MgCl₂ to 5.4 μM (Fig. 6C)]. Diluted Lgl2 was mixed with the vesicle preparations at a 4:1 (protein/vesicles) volume ratio. After 30 min of incubation at room temperature, the mixture was centrifuged at 4 °C for 40 min at 200,000 × g. For the experiment in Fig. 6C, the Lgl2/vesicles mixture was supplemented with assay buffer (1:5 buffer toLgl2/vesicles mixture), with or without 6 mM ATP and 270 nM aPKCι kinase domain, and incubated at room temperature for an additional hour before centrifugation. The supernatant of each sample was then removed, and the pellet was resuspended in the same volume of assay buffer + 1% Triton X-100. Samples of the separated phases were analyzed by SDS/PAGE.

CD Spectroscopy.

Lgl2(626–680) or pLgl2(626–680) samples were added with increasing amounts of trifluoroethanol or 10% PIP₂ (90% PC) vesicles, mixed well, and incubated for >2 min before each reading. To minimize the solvent scattering, buffer conditions were adjusted to 10 mM Tris (pH 8.0) and 75 mM NaCl. The relative concentrations of Lgl2(626–680) and pLgl2(626–680) between the CD experiments were determined using the Bio-Rad Protein Assay. To avoid the effect of the altered charge due to phosphorylation on the assay, this was done for a sample of each treated with λ-phosphatase (New England Biolabs).

Data were collected using a Jasco-j815 CD spectrometer. CD spectra at 195–260 nM were averaged from three to six sequential readings of each sample. Following the appropriate buffer spectra subtraction, values were converted to mean residue molar ellipticity as described (70). Helix propensity (%helix) was determined by the spectra readings at 222 nM as described (71).