Inscuteable and NuMA proteins bind competitively to Leu-Gly-Asn repeat-enriched protein (LGN) during asymmetric cell divisions

Abstract

Coupling of spindle orientation to cellular polarity is a prerequisite for epithelial asymmetric cell divisions. The current view posits that the adaptor Inscuteable (Insc) bridges between Par3 and the spindle tethering machinery assembled on NuMA∶LGN∶Gαi^GDP, thus triggering apico-basal spindle orientation. The crystal structure of the Drosophila ortholog of LGN (known as Pins) in complex with Insc reveals a modular interface contributed by evolutionary conserved residues. The structure also identifies a positively charged patch of LGN binding to an invariant EPE-motif present on both Insc and NuMA. In vitro competition assays indicate that Insc competes with NuMA for LGN binding, displaying a higher affinity, and that it is capable of opening the LGN conformational switch. The finding that Insc and NuMA are mutually exclusive interactors of LGN challenges the established model of force generators assembly, which we revise on the basis of the newly discovered biochemical properties of the intervening components.

Asymmetric cell divisions regulate the position and the fate choice of daughter cells, with impact on numerous phenotypes of multicellular organisms. During development, asymmetric divisions coordinate cell growth with cell specification to determine tissue morphogenesis, while in adult life they sustain tissue homeostasis and regeneration (1). In asymmetric divisions specific cortical landmarks instruct the orientation of the mitotic spindle to promote unequal partitioning of fate determinants in cellular systems as diverse as Caenorhabditis elegans zygotes, Drosophila neuroblasts, as well as vertebrate skin and neural progenitors (2). Spindle coupling to polarity cues involves the recruitment at cortical sites of molecular devices, known as force generators, whose main task is to capture astral microtubules emanating from the spindle poles and to establish pulling forces. Core components of force generators are the evolutionary conserved NuMA∶LGN∶Gαi complexes, termed Mud∶Pins∶Gαi in flies. Topologically, tetratricopeptide repeats (TPR) present in the N-terminal portion of LGN mediate the interactions with NuMA, while GoLoco motifs at the C terminus serve as a docking platform for four Gαi^GDP subunits anchored at the plasma membrane via a myristoyl group (3). LGN exclusively binds to GDP-loaded Gαi (4). The association of cortical NuMA with the microtubule motor Dynein/Dynactin (5) provides a sliding anchorage for depolymerizing microtubules, whose shrinkage pulls towards the cortex the connected spindle pole. FRET studies revealed that LGN behaves as a conformational switch held in a closed form in interphase by head-to-tail interactions (3).

An issue intimately related to how force generators are assembled is how they are recruited at sites of polarization. In polarized asymmetric divisions, the apico-basal polarity axis is established by the asymmetrical distribution of Par3∶Par6∶aPKC at the apical cortex, which are able to recruit NuMA∶LGN∶Gαi^GDP via an adaptor named Inscuteable (Insc) (6). Insc was first identified in larval Drosophila neuroblasts as a partner of Par3 (Bazooka in flies) colocalizing with polarity proteins right after delamination (7–9), and later shown to bind Pins (10). Mammalian Insc homologues endowed with similar properties have been discovered in rat retina (11) and in mouse developing skin (12). Insc overexpression is sufficient to induce apico-basal divisions in cells that normally divide in a planar fashion such as Drosophila embryonic ephitelial cells (7), vertebrate neuroepithelial progenitors (13), and mouse skin progenitors (6). The portion of fly Insc encompassing residues 252–615 recapitulates Insc functions in neuroblasts (hence termed “asymmetric domain”), and directly interacts with Pins (9). Based on its interaction with both Par3 and LGN, Insc has been considered the molecular link between cortical Par proteins and NuMA∶LGN∶Gαi^GDP complexes. This notion is mainly substantiated by imaging analyses conducted in fly neuroblasts and mouse skin progenitors showing that during asymmetric metaphases Par3, Insc, LGN, and NuMA colocalyze in a cortical region underlying one of the spindle poles (1, 6, 12). However, no proof has been provided for the simultaneous association of LGN with Insc and NuMA, which remains a key question to elucidate how force generators work. We thus set out to study how LGN interacts with Insc and NuMA.

Results

Determination of the Crystal Structure of Pins∶Insc.

Human Insc∶LGN∶Gαi^GDP was refractory to crystallization. We thus took advantage of the functional information available on fruitfly proteins (Fig. 1A), and generated a dicistronic vector for coexpression in bacteria in which we cloned the TPR repeats and part of the linker region of Pins (Pins^TPR-LINKER) fused with a GST moiety in the first cassette, and a fragment corresponding to the asymmetric domain of Insc (Insc^ASYM) in the second cassette. This strategy yielded a soluble Pins^TPR-LINKER∶Insc^ASYM complex eluting from a size exclusion chromatography (SEC) as a monodisperse sample (Fig. 1B). In order to obtain diffraction quality crystals, we had to further shorten these Pins and Insc constructs by limited proteolysis. When subjected to trypsinization, both subunits of Pins^TPR-LINKER∶Insc^ASYM were trimmed to smaller fragments (Fig. S1), which were assigned by mass spectrometry to Pins^25–406 and Insc^303–340 respectively. SEC analysis confirmed that Pins^25–406 stably associates with Insc^303–340 with a 1∶1 stoichiometry (Fig. 1B). To gain an understanding of the organizational principles of the interaction, we determined the structure of the proteolitically defined Pins^25–406∶Insc^303–340 complex, referred to as Pins^TPR∶dInsc^PEPT (where “d” stands for Drosophila).

Fig. 1.

Generation of the Pins:Insc complex. (A) Scheme of LGN and Insc. (B) SEC elution profiles of Pins^TPR-LINKER∶Insc^ASYM and Pins^TPR∶dInsc^PEPT.

To determine the crystallographic structure, we used SAD phases from selenomethionine-substituted protein crystals. The structure has been refined to 2.1 Å resolution with an R_free of 25.6% (Table S1). The final model includes residues 307 to 335 of dInsc peptide, and residues 39–386 of Pins.

General Architecture of the Pins∶Insc Assembly.

The Pins^TPR domain folds into a cradle-shaped arch harboring the extended dInsc^PEPT into its inner concave surface (Fig. 2 A–C). In such arrangement, the Pins and dInsc polypeptide chains run with opposite direction, with the helical N terminus of dInsc^PEPT contacting the C-terminal portion of Pins^TPR. Spiraling of the Pins^TPR domain around dInsc^PEPT generates an extensive interaction surface of about 1,900 Å², corresponding to approximately 50% of the dInsc^PEPT surface area.

Fig. 2.

Structure of the Pins^TPR∶dInsc^PEPT dimer. (A–C) Ribbon models of the Pins^TPR∶dInsc^PEPT complex viewed at the indicated orientations. (D) Surface representation of Pins^TPR with the dInsc^PEPT lining along the conserved asparagines (purple). (E) Sequence and secondary structure of Pins^TPR and dInsc^PEPT colored according to the conservation.

The Pins^TPR domain consists of 17 antiparallel helices forming 8.5 TPR repeats (Fig. S2), with a pattern of hydrophobic amino acids consistent with the canonical TPR consensus (14). The sequence 39–76 adopts a TPR-like arrangement despite deviating from the consensus. In addition, the fourth TPR repeat presents an insertion between the first and the second helix (Fig. S2) resulting in longer α4A and α4B helices joined by a 8-residue loop. Consecutive TPR units stack in much the same way as reported for the TPR-containing domain of O-linked N-actylglucosamine transferase (15), giving rise to a right-handed superhelical twist of the molecule. The unusual sequence composition of the TPR4 repeat distorts the superhelix by an outward displacement of the subsequent TPR units with respect to the helical axis (see Fig. S3).

A peculiar aspect of the Pins^TPR structure is the presence within the groove of a set of asparagines positioned on the third and fourth turn of the αA helix of each TPR repeat, which generate a ridge extending throughout the concave surface of the molecule (Fig. 2D). These asparagine residues occupy the positions 6 and 9 of the TPR consensus, within an invariant NLGN motif found in most of the Pins TPR repeats. Not surprisingly, in the binary complex dInsc^PEPT lines up with this asparagine ladder. This binding mode is reminiscent of the binding of similarly extended ligands to multiple ARM-repeat proteins such as the NLS peptides bound to karyopherin α (16), and the E-cadherin peptide bound to β-catenin (17), thus suggesting a general recognition mechanism common to all these topologically diverse helical scaffolds.

Determinants of Interactions Between Pins and Insc.

We next addressed the structural determinants of the Pins^TPR∶dInsc^PEPT assembly. As expected for TPR repeat-containing proteins, the hetero-dimer interface is contributed by the amino acids of the αA helices, facing the inner side of the domain. To facilitate the description, we will divide the interactions into three major modules roughly matching the dInsc^PEPT N-terminal α-helix, the central E322-P323-E324^Insc motif, and the elongated C-terminal stretch. The first module involves the helix αA of dInsc^PEPT that packs against the TPR repeats 6 and 7 of Pins^TPR with an orientation almost perpendicular to the repeats’ axis (Fig. 3A). This helix contains the invariant Trp313^Insc that is hydrogen-bonded to Ser303^Pins and further stabilized by stacking interactions with Asn306^Pins. The correct orientation of the αA helix is ensured on one end by a hydrogen bond between Glu308^Insc and Asn346^Pins, and on the other end by the hydrogen bond between the Trp319^Insc and the carbonyl group of Lys176^Pins located in the TPR4 insertion. At Trp319^Insc the dInsc^PEPT chain turns sharply to progress parallel to the αA5 helix of Pins^TPR until Met326^Insc, crossing the whole surface of the TPR domain (Fig. 3B). Hydrogen bonds from the conserved asparagines Asn223^Pins, Asn226^Pins, and Asn263^Pins anchor the dInsc^PEPT main chain to the TPR scaffold, and a number of salt bridges between Glu322^Insc and Glu324^Insc and Arg244^Pins, Arg258^Pins and Arg259^Pins greatly contribute to strengthen the interaction. The electrostatic potential of the Pins^TPR surface in this area confirms that the negatively charged EPE^Insc triplet is accommodated into a positively charged patch of the central portion of the Pins^TPR domain (Fig. S4), thus suggesting that charge complementarity might constitute the major factor dictating the ligand specificity here. The third area of interactions involves the C-terminal end of dInsc^PEPT and the first three TPR units of Pins^TPR. This portion of dInsc^PEPT inserts the side chains of Lys332^Insc and Ile334^Insc into two adjacent pockets at opposite sites of the helix (Fig. 3C). Lys332^Insc snugly fits into the negatively charged cavity formed by Ser81^Pins, Ser123^Pins, and Asp107^Pins, whereas Ile334^Insc sits into a predominantly hydrophobic environment contributed by Leu44^Pins, Phe66^Pins, and Ile83^Pins.

Fig. 3.

Analysis and conservation of the Pins^TPR∶dInsc^PEPT dimer interface. (A–C) Enlarged views of the dimer interface. (D) GST-dInsc^PEPT, mutated as indicated above, on GSH beads was incubated with Pins^TPR. After washing species on beads were analyzed by SDS-PAGE. (E) Binding of Pins^TPR mutants to wt-dInsc^PEPT. (F and G) Same assays as in boxes D and E performed with human LGN^TPR and hInsc^PEPT.

To address the relevance of each interaction module for the dimer assembly, we substituted key residues identified by our structural analysis. The binding ability of the mutated proteins was tested in a pull-down assay performed with GST-dInsc^PEPT immobilized onto glutathione sepharose (GSH) beads and purified Pins^TPR constructs in solution. We first asked if dInsc^PEPT mutants were able to bind wild-type Pins^TPR (Fig. 3D). As expected, double replacement of E322-E324^Insc with alanine or arginine fully abrogates the binding. The same effect is caused by substitution of the Trp313^Insc with alanine or glutamic acid, underscoring a fundamental role of this triptophane in docking the dInsc^PEPT helical fragment onto Pins^TPR. Mutations of Trp319^InscAla and Met326^InscAla, as well as substitution of the invariant residues Leu329^Insc, Lys332^Insc, and Ile334^Insc on the dInsc^PEPT C terminus do not severely affect binding. We next tested the structural determinants of the binding on the Pins^TPR side, focusing our mutational analysis on the central EPE^Insc module. In keeping with the role predicted for the invariant asparagines of Pins^TPR in anchoring the dInsc^PEPT backbone, removal of the Asn226^Pins side chain is sufficient to abolish binding (Fig. 3E). Single mutations of Arg244^Pins and Arg259^Pins to alanine or glutamic acid strongly reduce binding, while double substitution of Arg258^Pins and Arg259^Pins with alanine results in a Pins^TPR mutant unable to associate with dInsc^PEPT. In summary, our results indicate that; (i) all three dInsc^PEPT portions including the N-terminal α-helix, the central EPE^Insc motif, and the extended C-terminal tail synergize in binding Pins^TPR; (ii) the polar interactions between the negative charges of the EPE^Insc motif and the positively charged inner surface of Pins^TPR is a hallmark of dInsc^PEPT recognition; (iii) besides Glu322^Insc and Glu324^Insc, a key determinant of the dimeric assembly is the invariant Trp313^Insc.

The Pins∶Insc Interface Is Evolutionary Conserved.

One of the open questions in the stem cell field is to which extent the molecular mechanisms underlying Drosophila neuroblast divisions are conserved in vertebrates. To gain insight into this issue, we generated constructs of human LGN and Insc corresponding to the fruitfly domains we had used for the structure determination, and repeated the GST pull-down assays with an analogous battery of mutants. Sequence analysis revealed that the counterparts of the dInsc^PEPT spans residues 23 to 58 of the human Insc, to which we refer as hInsc^PEPT, where “h” stands for human. As predicted by the strong sequence conservation of the ortholog proteins in these regions (Fig. 2E, Fig. S5), equivalent mutations on hInsc^PEPT and dInsc^PEPT impair the association with the cognate LGN^TPR and Pins^TPR domains (Fig. 3 F and G). In particular, replacement of Trp31^hInsc or of the glutamic acid couple E40-E42^hInsc fully abrogates the interaction with LGN^TPR. We next assessed the contribution of the conserved Asn203^LGN, Arg221^LGN, Arg235^LGN, and Arg236^LGN toward hInsc binding. Simultaneous substitution of Arg235^LGN and Arg236^LGN with alanine, as well as charge reversal of Arg236^LGN to Glu severely perturbed the interaction, mirroring the effect observed for the corresponding Pins^TPR mutations (Fig. 3G). Only upon replacement of Asn203^LGN with a bulky phenylalanine we could score a binding reduction recapitulating the impairment observed with Asn226^Pins. Finally, to evaluate whether equivalent chemical interactions sum up to similar overall binding affinities, we measured the strength of the binary interaction in both species by Isothermal Titration Calorimetry. The dissociation constants were 5 nM for the Pins^TPR∶dInsc^PEPT interaction and 13 nM for the LGN^TPR∶hInsc^PEPT one (Fig. S6). Collectively these results indicate that the LGN∶Insc interface is evolutionary conserved.

NuMA Binds to LGN with an EPE-Motif.

In mitosis the association of LGN with NuMA is fundamental to sustain the spindle orientation process. Previous studies have mapped the LGN-binding site of NuMA within a C-terminal portion of about 20 KDa encompassing residues 1,878–1,910 (18). This fragment of NuMA has been shown to enter a complex with LGN^TPR, raising the question as to whether it is compatible with the concomitant presence of Insc on the same domain. Primary sequence analysis of NuMA revealed the existence of a conserved EPE motif at position 1,896–1,898, strongly suggesting that NuMA could interact with LGN with the same binding mode displayed by Insc (Fig. 4A). To test this hypothesis, we expressed NuMA^{1,886–1,914} (referred to as NuMA^PEPT hereon) fused to a GST moiety, and checked its ability to interact with LGN^TPR by GST pull-down assay. Indeed, NuMA^PEPT binds LGN^TPR in an EPE-dependent manner (Fig. 4B). Furthermore, all asparagine and arginine residues of LGN^TPR found to be in contact with EPE motif of hInsc^PEPT are also important for NuMA^PEPT recognition. Single substitutions on LGN^TPR abrogate binding to NuMA^PEPT due to the lower binding affinity displayed by NuMA^PEPT for LGN^TPR with respect to the one between hInsc^PEPT and LGN^TPR, with dissociation constants of about of 50 nM and 13 nM respectively (Fig. S6). We conclude that NuMA^PEPT and hInsc^PEPT associate with LGN^TPR via a conserved EPE motif.

Fig. 4.

NuMA competes with Insc for the binding to LGN via the invariant EPE motif. (A) Alignment of hInsc^PEPT and NuMA^PEPT around the conserved ECE/EPE motif. (B) Validation of the NuMA^PEPT:LGN^TPR interface. (C) Fluorescence polarization-based competition assays showing the decrease in polarization signal upon titration of unlabeled hInsc^PEPT into a mixture of LGN^TPR and fluorescein-NuMA^PEPT, or titration of unlabeled NuMA^PEPT into a mixture of LGN^TPR and fluorescein-hInsc^PEPT. (D) SEC analysis of the competitive interactions of Insc and NuMA^{1,807–1,987} with LGN∶Gαi^GDP.

NuMA and Insc Are Competitive Interactors of LGN.

To ascertain whether hInsc^PEPT and NuMA^PEPT are mutually exclusive interactors of LGN^TPR, we developed a fluorescence polarization-based competition assay. Fluorescein-labeled NuMA^PEPT at a concentration of 15 nM was mixed with 250 nM of LGN^TPR. Based on their reciprocal affinity, under these conditions about 85% of the fluorescein-NuMA^PEPT are in complex with LGN^TPR (Fig. S7). We then titrated into the reaction increasing amounts of unlabeled hInsc^PEPT, and measured the residual fluorescein-NuMA^PEPT polarization at steady-state. The monotonic decrease of the polarization signal upon hInsc^PEPT addition indicates that hInsc^PEPT effectively competes with NuMA^PEPT, and implies that the simultaneous binding of the two peptides on LGN^TPR is precluded (Fig. 4C). We confirmed this result by performing an analogous competition experiment in which we titrated unlabeled NuMA^PEPT to compete the binding of fluorescein-labeled hInsc^PEPT to LGN^TPR. The fourfold lower affinity displayed by NuMA^PEPT towards LGN^TPR with respect to the one of hInsc^PEPT predicts that higher doses of NuMA^PEPT are required to fully displace hInsc^PEPT from LGN^TPR. The shift observed in the NuMA-inhibited curve as compared to the Insc-inhibited one fully satisfies this prediction.

We next asked if the competition observed for the short NuMA^PEPT and hInsc^PEPT peptides reflects a genuine inability of full-length Inscutable and NuMA to concomitantly enter a complex with LGN. We purified a homogeneous form of human LGN∶Gαi^GDP from insect cells following a previously described colysis method (4). When injected on a SEC column, the LGN∶Gαi^GDP complex elutes at an apparent molecular weight compatible with the expected 1∶4 stoichiometry (Fig. 4D). We also adapted the purification protocol to generate an Insc∶LGN∶Gαi^GDP complex by coexpression of Insc and LGN in insect cells followed by colysis with Gαi. We then mixed LGN∶Gαi^GDP and Insc∶LGN∶Gαi^GDP in roughly equimolar amounts and analyzed the sample by SEC. Despite the significant difference in the theoretical molecular weights caused by the presence of Insc, the two complexes coelutes in about the same fractions. To test the effect of NuMA, we additioned the LGN∶Gαi^GDP plus Insc∶LGN∶Gαi^GDP mixture with a 20-fold molar excess of NuMA^{1,807–1,987}, and separated the resulting species by SEC. NuMA^{1,807–1,987} readily bound to free LGN∶Gαi^GDP and eluted in a well separated peak distinct from the Insc∶LGN∶Gαi^GDP one. In agreement with the competition model, the NuMA^{1,807–1,987} excess was unable to dissociate Insc from LGN∶Gαi^GDP due to its lower binding affinity. Of note, NuMA∶LGN∶Gαi^GDP elutes earlier than the Insc containing complex in spite of its smaller theoretical mass. We suspect that such behavior might be ascribed to specific hydrodynamical properties of the NuMA assembly, although we cannot rule out a dimerization of the NuMA-containing hetero-trimer. Collectively these findings demonstrate that Insc and NuMA utilize the ECE/EPE recognition motif to competitively associate with LGN.

Both NuMA and Insc Open the LGN Conformational Switch.

One of the key events in the cortical force generators assembly is the release of the head-to-tail interactions holding LGN in an inhibited state. Previous studies have shown that the association of NuMA to LGN^TPR is capable of opening the conformational switch (3). Given the similarity in the binding mode of NuMA and Insc toward LGN^TPR, we enquired if Insc would also induce a similar conformational transition. To this end, we first reproduced the displacement the C-terminal LGN^GoLoco region from LGN^TPR observed upon NuMA addition (Fig. S8). We next took advantage of the stable LGN^TPR∶LGN^GoLoco assembly mimicking the closed conformation of LGN to investigate the effect of Insc binding on the LGN conformational state. Our previous results proved that hInsc^PEPT forms a tight complex with LGN^TPR. When we repeated a similar GST pull-down experiment using LGN^TPR∶LGN^GoLoco in solution, only the LGN^TPR half of the sandwich remained associated with the beads (Fig. 5A), providing compelling evidence that Insc opens the LGN switch. An analogous behavior is observed upon incubation of LGN^TPR∶LGN^GoLoco with NuMA^{1,807–1,987}, that we included as a positive control.

Fig. 5.

Both NuMA and Insc open the LGN conformational switch. (A) Upon incubation of 2 μM GST-hInsc^PEPT with 4 μM of preformed LGN^TPR∶LGN^GoLoco, only LGN^TPR is retained on beads, demonstrating that the simultaneous presence of LGN^GoLoco and Insc on the TPR domain is excluded. (B) Revised model of force generators functioning during asymmetric divisions.

Discussion

In this study we report the characterization of the Pins^TPR∶dInsc^PEPT complex, and provide a molecular explanation for the mutual exclusive interaction of Insc and NuMA to LGN. While this manuscript was in preparation, Zhu and coworkers arrived to similar conclusions analyzing the structure of the LGN∶NuMA complex (19).

A 38-residue fragment of Drosophila Insc encompasses the Pins^TPR binding region. This fragment of Insc shares a high sequence similarity to functional homologues recently discovered in mammals, fully supporting the notion that the basic mechanism responsible for the recruitment of force generators at polarity sites is evolutionary conserved. With the exception of a short N-terminal α-helix, the Insc^PEPT is intrinsically unstructured, and lines on the scaffold provided by the superhelical TPR arrangement of Pins with an extended conformation. The interaction surface is organized around a core module involving the EPE motif of Insc^PEPT and the central TPR5-6 of Pins, whose specificity is primarily dictated by charge complementarity. The binding is further stabilized by polar and hydrophobic interactions contributed by the αA helix of Insc^PEPT. Not surprisingly, the large interaction surface characterizing the topology of the Pins^TPR∶Insc^PEPT heterotypic dimer accounts for an elevated binding affinity (of about 5 nM for the fly proteins and 13 nM for the human counterparts). The structure of mouse LGN^191–350, corresponding to what we name TPR5-8, with Insc^19–40 suggests that vertebrate proteins assemble with organizational principles similar to the fly ones (19). However, the short mouse constructs only depict the interaction of LGN^TPR with the αA helix of Insc^PEPT, up to the first Glu of the EPE motif. Intriguingly, the mouse LGN∶Insc interaction seems to be characterized by lower affinity compared to human and fly ones (with K_D of 63 nM for LGN^TPR5–8∶Insc^19–40, and of 47 nM for LGN^TPR1–8∶Insc^20–57).

The evidence that NuMA forms a complex with the same LGN^TPR domain associating to Insc raised the question of whether it binds in a similar manner. Indeed, comparison of the primary sequence of Insc^PEPT with the known LGN-binding portion of NuMA revealed the presence of an EPE triplet that turned out to be essential for the LGN recognition, with a similar molecular signature of the EPE motif of the Insc^PEPT. Notably, the NuMA ortholog in fly (Mud) codes for two consecutive EDE-EGE motifs in the Pins-binding region, whose interplay remains to be clarified (Fig. S9). The structure of LGN in complex with NuMA^PEPT fully supports the notion that the EPE-interaction module represents a common region required for docking unstructured ligands on the LGN^TPR scaffold (19). In the case of NuMA, the interface is further contributed by a helical fragment forming a bundle with helices αA2 and αA3 of LGN^TPR. The consequence of the partial overlap between the Insc and NuMA binding sites is that their concomitant loading on LGN is excluded.

A key step during the assembly of the force generators is the opening of the LGN conformational switch that keeps the molecule in an inactive state (3). Binding of NuMA to LGN^TPR induces the release of the intramolecular interactions holding the molecule in a closed form. In agreement with the similarity in the binding modes, we demonstrated that also Insc disengages the LGN GoLoco motifs from the TPR domain. Together these findings imply that the GoLoco motifs contact the TPR repeats in the same region occupied by Insc and NuMA. Primary sequence inspection revealed that the GoLocos of both Pins and LGN do not contain EPE triplets, suggesting that either the head-to-tail interaction involves alternative TPR patches sterically occluded by the presence of Insc and NuMA, or that less conserved negatively charged triplets are accommodated on the same TPR5-6 of LGN.

The well established model for force generators recruitment at polarity sites rests on the assumption that Insc and NuMA can be part of the same apically localized multisubunit complexes containing Par proteins. This model stems from colocalization experiments showing that in asymmetric mitoses Par3, Insc, LGN, and NuMA cluster together in apical crescents (6), complemented by coimmunoprecipitation assays in which LGN∶Gαi were found in association with Par3∶Insc (12) and NuMA (18). The finding that Insc and NuMA are mutually exclusive partners of LGN is both unexpected and puzzling. In particular, the higher affinity characterizing the Insc binding to LGN shifts the balance of the unmodified proteins towards the Insc∶LGN complex formation, which is instrumental in recruiting LGN with Par proteins at the onset of mitosis but cannot account for microtubule-pulling forces. What is the possible mechanism for transferring LGN from Insc to NuMA? The architecture of the Insc^PEPT∶Pins^TPR structure whereby an extended ligand is accommodated on a large domain allows a high degree of regulation of the interaction strength. Posttranslational modifications on either side of the dimer might locally alter the contacts without affecting the rest of the interface, as it has been demonstrated for the similarly organized complex between the cytoplasmic domain of E-cadherin and β-catenin (17). Such modulating modifications can in principle occur on Insc, NuMA, or on LGN. To date, no experimental information is available regarding putative Insc or NuMA modifications. More controversial is the literature relative to LGN phosphorylations. In mitotic Drosophila neuroblasts, Pins has been found phosphorylated by Aurora-A on Ser436 at about half of the linker connecting the TPR domain with the GoLoco motifs (20). Using an “induced polarity” assay in S2 cells, phospho-Ser436^Pins was shown to trigger a redundant NuMA-independent spindle orientation pathway engaging the membrane associated Dlg protein. It is to date unclear if such pathway is conserved in vertebrates. Notably, during oriented symmetric cell divisions of MDCK cells, phosphorylation on a similarly positioned Ser401 of LGN functions in excluding force generators from the apical cortex in order to prevent apico-basal spindle orientation (21). In this context, phospho-Ser401^LGN would selectively prevent binding of LGN to apical Gαi. Based on our structural and biochemical results, it is difficult to provide a molecular explanation as to whether these LGN phosphorylations could also impact on the Insc and NuMA binding. Recent observations support the notion that the pool of NuMA∶LGN∶Gαi colocalizing with Par3∶Insc in embryonic mouse skin progenitors is tightly regulated to set the balance between symmetric and asymmetric divisions, though no mechanism for this has been put forward (22). In summary, we reckon more has to be learnt to understand what brings LGN from Insc to NuMA.

An additional question relates to the mechanism maintaining effective NuMA∶LGN∶Gαi^GDP species at the correct cortical sites in the absence of Insc. Based on the knowledge we acquired, we propose a step-wise model that can be schematized as follows (Fig. 5B): (i) in the early phases of mitosis LGN is brought to the apical membrane in conjunction with Par proteins by the high-affinity interaction with the preformed Par3∶Insc complex. Binding of LGN to Insc triggers the conformational switch transition enabling the relocation of Gαi^GDP moieties previously distributed all around the plasma-membrane with Gβγ; (ii) upon mitotic progression, when LGN is already at the correct sites, a yet unidentified molecular event alters the relative affinities of Insc and NuMA for LGN to shift the balance between the Insc-bound and the NuMA-bound LGN pools. We hypothesize that the four Gαi subunits present on LGN at this stage are sufficient to transiently hold cortical NuMA∶LGN∶Gαi^GDP in proximity of Par proteins to allow directional microtubule pulling. We speculate that NuMA∶LGN∶Gαi^GDP is a short-lived complex and disassemble, possibly assisted by a specialized GEF for Gαi such as as Ric-8A (23), releasing apo-LGN in the cytoplasm to start a new cycle. Such a dynamical interaction network would allow for a continuous regulation of the force exerted on astral microtubules throughout mitosis. Future attempts to validate the model in vivo will greatly benefit from the biochemical tools presented in this study.

Experimental Procedures

Detailed descriptions of crystallization, structure determination, and other experimental procedures are available in the SI Materials and Methods. Atomic coordinates and structure factors for the Pins∶Insc structure have been deposited to the Protein Data Bank under accession code 4A1S.