Crystal Structures of the Scaffolding Protein LGN Reveal the General Mechanism by Which GoLoco Binding Motifs Inhibit the Release of GDP from Gαi *

Background: GoLoco (GL) motif binds to Gα and inhibits its guanine nucleotide dissociation. Results: Crystal structures of LGN-GL3(4)·Gαi1(3) complexes reveal a double Arg finger-mediated GDP recognition mechanism, which is distinct from that shown in the RGS14·Gαi1 complex. Conclusion: LGN-GL/Gαi interaction might represent a general binding mode between GoLoco motifs and Gαi. Significance: Our findings shed new light on the GoLoco motif-mediated G protein signaling regulation. GoLoco (GL) motif-containing proteins regulate G protein signaling by binding to Gα subunit and acting as guanine nucleotide dissociation inhibitors. GLs of LGN are also known to bind the GDP form of Gαi/o during asymmetric cell division. Here, we show that the C-terminal GL domain of LGN binds four molecules of Gαi·GDP. The crystal structures of Gαi·GDP in complex with LGN GL3 and GL4, respectively, reveal distinct GL/Gαi interaction features when compared with the only high resolution structure known with GL/Gαi interaction between RGS14 and Gαi1. Only a few residues C-terminal to the conserved GL sequence are required for LGN GLs to bind to Gαi·GDP. A highly conserved “double Arg finger” sequence (RΨ(D/E)(D/E)QR) is responsible for LGN GL to bind to GDP bound to Gαi. Together with the sequence alignment, we suggest that the LGN GL/Gαi interaction represents a general binding mode between GL motifs and Gαi. We also show that LGN GLs are potent guanine nucleotide dissociation inhibitors.

The ␣ subunit of the heterotrimeric G proteins (G␣) is a critical component of the G protein signaling pathway, in which G␣ cycles between the GDP-and GTP-bound states (1). In the canonical signaling model, ligand-mediated activation of G protein-coupled receptors (GPCRs) 4 catalyzes the exchange of GDP for GTP in binding to G␣ and subsequently results in the dissociation of G␣⅐GTP from G␤␥ heterodimer (2,3). The dissociated G␣⅐GTP binds to and activates downstream effectors, thus transducing signals from GPCR (4 -6). Because G␣ has intrinsic GTPase activity, the G␣ subunit subsequently returns to the G␣⅐GDP form, which marks the termination of the GPCR signaling. Many proteins have been discovered as regulators of the GTP-and GDP-bound forms of the G␣ reaction cycle. Among these, GoLoco motif proteins were discovered to bind specifically to GDP-loaded G␣ i or G␣ o and inhibit the spontaneous release of GDP from G␣. These GoLoco proteins are referred to as guanine nucleotide dissociation inhibitors (GDIs) (7)(8)(9)(10)(11).
The GoLoco motif (8,12,13) was first identified as a conserved sequence of 19 amino acids, occurring singly or as tandem repeats in a variety of signaling proteins across the animal kingdom (7). Our understanding of the molecular mechanism of the GDI function of GoLoco proteins is mainly based on the crystal structure of RGS14 GoLoco bound to G␣ i1 ⅐GDP (14), which shows that the conserved GoLoco motif and its variable C-terminal tail interact with the Ras-like and all-helical domains of G␣ i1 , respectively. A so-called "arginine finger" formed by the highly conserved (D/E)QR triad in the conserved GoLoco motif extends into the GDP-binding pocket and directly contacts the ␣and ␤-phosphates of GDP (14). This structure and the subsequent mutagenesis and structural studies (14 -18) suggested an appealing hypothesis: the highly variable C-terminal sequences following the conserved GoLoco motifs and the all-helical domain of G␣ subunits are likely the specificity determinants of interactions between GoLoco motifs and different G␣ subunits. However, because there no structures of GoLoco motifs in complex with G␣ other than the G␣ i1 ⅐RGS14 complex are available to date, the above hypothesis remains untested.
LGN is a multidomain scaffolding protein containing eight tetratricopeptide (TPR) repeats in its N-terminal region, a flexible linker sequence in the middle, and four GoLoco motifs in the C-terminal end (19,20).
LGN is an evolutionarily conserved protein (Pins in Drosophila, and GPR1/2 in Caenorhabditis elegans) that plays crucial roles in regulating spindle orientations during asymmetric cell division (19,21) and can be considered as an example member of the multiple GoLoco motif protein family. It forms a ternary protein complex with nuclear mitotic apparatus protein NuMA (Mud in Drosophila and Lin5 in C. elegans) and cortical membrane-bound G␣ i/o via its TPR repeats and GoLoco motifs, respectively (22)(23)(24)(25)(26)(27)(28). The central linker of LGN binds to the guanylate kinase domain of the DLG family scaffold protein in a phosphorylation-dependent manner (29 -31). In Drosophila neuroblast, loss of Pins or G␣ i affects cell polarity as well as mitotic spindle orientation (32). In mammals, overexpression or removal of LGN results in dramatic spindle rocking in metaphase and improper spindle pole organization (19,21,33). The binding of G␣ i through the GoLoco motifs was shown to regulate the cortical localization of LGN (33). Thus, the LGN GoLoco motifs can be viewed as scaffolding modules in tethering the TPR repeat partners (e.g. NuMA/Mud and mInsc/Insc) of LGN to the cell cortex via binding to membrane-attached G␣ i . Interestingly, the GoLoco motifs of LGN can directly bind to TPR repeats intramolecularly, thus keeping LGN in an autoinhibited conformation (22). G␣ i ⅐GDP binding to GoLoco motifs releases the autoinhibited conformation of LGN and renders LGN TPR repeats capable of binding to NuMA (22,34), although the mechanistic basis of the LGN autoinhibition is unknown.
In this study, we performed detailed biochemical and structural analyses of the interactions between LGN GoLoco motifs and G␣ i ⅐GDP. We demonstrate that in contrast to the RGS14/ G␣ i ⅐GDP interaction, only a few residues of the highly variable sequences C-terminal to the conserved GoLoco motifs of LGN are involved in binding to G␣ i ⅐GDP. The structures of two LGN GoLoco motifs in complex with G␣ i reveal a double Arg finger sequence (R⌿(D/E)(D/E)QR) within the GoLoco motif that is specifically involved in the GDP coordination. We further show that the LGN GoLoco⅐G␣ i ⅐GDP interaction observed in this study likely represents a general mode of GoLoco motif-mediated G␣ binding. We further demonstrate that the LGN GoLoco motifs are potent GDIs. Thus, the LGN GoLoco motifs can function as a G␣/LGN/NuMA/Insc scaffold as well as a regulator of G␣ signaling in asymmetric cell division.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The human G␣ i3 , G␣ i1 , mouse LGN GL fragments were individually cloned into a modified version of pET32a vector. All the mutations were created using the standard PCR-based method and confirmed by DNA sequencing. Recombinant proteins were expressed in Escherichia coli BL21 (DE3) host cells at 16 or 37°C and were purified by using a Ni 2ϩ -nitrilotriacetic acid-agarose affinity chromatography followed by size exclusion chromatography. For in vitro biochemical analysis, LGN GLs were expressed as the GST-fused proteins and purified by GSH-Sepharose affinity chromatography.
Isothermal Titration Calorimetry Measurements-ITC measurements were performed on an ITC200 Micro calorimeter (MicroCal) at 25°C. All protein samples were dissolved in the buffer containing 50 mM Tris, pH 8.0, 100 mM NaCl, and 1 mM EDTA. The titrations were carried out by injecting 40 l of G␣ i3 ⅐GDP aliquots (0.2 mM) into LGN GLs fragments fused to the C-terminal end of thioredoxin (0.02 mM) at time intervals of 2 min to ensure that the titration peak returned to the base line. The titration data were analyzed using the program Origin7.0 from MicroCal.
Fluorescence Polarization Assay-Fluorescence polarization assay were performed on a PerkinElmer LS-55 fluorimeter equipped with an automated polarizer at 25°C. Commercial synthesized peptides were labeled with fluorescein 5-isothiocyanate (Invitrogen) at the N termini. In a typical assay, the FITC-labeled peptide (ϳ1 M) was titrated with binding partners in a 50 mM Tris pH 8.0 buffer containing 100 mM NaCl, 1 mM DTT, and 1 mM EDTA. The K D values were obtained by fitting the titration curves with the classical one-site binding model, with or without invoking the Hill coefficient model.
GST Pulldown Assay-For GST pulldown assay, GST or GST-tagged proteins (60 l from 1 mg/ml stock solutions) were first loaded to 40 ml GSH-Sepharose 4B slurry beads in an assay buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 1 mM DTT, and 1 mM EDTA). The GST fusion protein-loaded beads were then mixed with potential binding partners, and the mixtures were incubated for 1 h at 4°C. After three times washing, proteins captured by affinity beads were eluted by boiling, resolved by 15% SDS-PAGE, and detected by Coomassie Blue staining.
Analytical Gel Filtration Chromatography-Analytical gel filtration studies were carried out on an AKTA FPLC system (GE Healthcare). Proteins at concentration of 10 -20 M in a volume of 100 l were loaded on a Superose 12 10/300 GL column 20 (GE Healthcare) equilibrated with the buffer containing 50 mM Tris, pH 8.0, 100 mM NaCl, 1 mM DTT, and 1 mM EDTA. Protein elution was detected by absorbance at 280 nm.
GDI Activity Assay-Measurements of AlF 4 Ϫ -induced increase of intrinsic tryptophan fluorescence were performed on the PerkinElmer LS-55 spectrometer with excitation at 292 nm and emission at 342 nm. Purified G␣ i3 protein was diluted in 2-ml cuvettes to 200 nM in a preactivation buffer (100 mM NaCl, 100 M EDTA, 2 mM MgCl 2 , 20 M GDP, 20 mM Tris-HCl, pH 8.0) and incubated at 30°C. At the time points 400 and 500 s after G␣ i3 dilution, 2 mM NaF and 30 M AlCl 3 (final concentrations), respectively, were added to the reaction mixture, and fluorescence intensity changes as a function of time were recorded. The GDI activities of GL peptides were assayed by repeating the above procedure except that the reaction mixtures contained defined concentrations of specific peptides.
The measurements of GTP␥S binding were also performed on PerkinElmer LS-55 spectrometer with excitation at 485 nm and emission at 530 nm (slit widths each at 2.5 nm). BODIPY FL-GTP␥S was diluted to 1 M in buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 10 mM MgCl 2 ) and equilibrated to 30°C in 2-ml cuvettes. Purified G␣ i3 was diluted to 100 nM in the buffer (100 mM NaCl, 100 M EDTA, 2 mM MgCl 2 , 20 M GDP, 20 mM Tris-HCl, pH 8.0) and preincubated with GL peptides (with different concentrations) at 30°C for 10 min before addition to the cuvette. Relative fluorescence levels were set to 0 at the average fluorescence reading over the first 70 s, and G␣ i3 /GL mixtures were added at the time point of 100 s.
Crystallography-Crystals of the G␣ i1(3) in complex with GL3/4 (diluted to 7.5 mg/ml in 50 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, 10 mM Mg 2ϩ , 20 M GDP buffer) were obtained by the hanging drop vapor diffusion method at 18°C. The crystals were grown in buffer containing 0.5 M ammonium sulfate, 1.0 M lithium sulfate monohydrate, 0.1 M sodium citrate tribasic dehydrate, pH 5.6. Crystals were soaked in crystallization solution containing a higher concentration (1.5 M) of lithium sulfate for cryo-protection. All the diffraction data were collected at Shanghai Synchrontron Radiation Facility BL17U at a wavelength of 0.9793 Å using a single crystal of each complex. The diffraction data were processed and scaled using HKL2000 (35). Molecular replacement was used to solve the structure of G␣ i1(3) ⅐GL4(3) with the program Molrep (36). The crystal structure of RGS14⅐G␣ i1 complex (Protein Data Bank code 1KJY) was used as a search model by removing the RGS14 peptide. The initial model was rebuilt manually and then refined using REFMAC (37) and PHENIX (38) against the whole data set. Further manual model building and adjustment were completed using COOT (39). The final refinement statistics are summarized in Table 1.

Mapping the Minimal G␣ i ⅐GDP Binding Sequences in LGN GoLoco
Motifs-The C-terminal region of LGN contains four GoLoco motifs, each of which consists of a conserved 19-residue fragment followed by a stretch of variable amino acid resi-dues with different lengths (Fig. 1A). We define the full-length GoLoco motif to be the conserved 19-residue fragment plus all of the following C-terminal sequence before the start of the next GoLoco motif core. With this definition, each GL1, 2, 3, and 4 motif of LGN consists of 54, 51, 34, and 51 residues, respectively ( Fig. 1A). Previous structural study of the G␣ i1 ⅐RGS14-GoLoco complex showed that the 16-residue sequence C-terminal to the GoLoco core motif make extensive contacts with G␣ i1 and thus are essential for the interaction between G␣ i1 and RGS14 (14). To understand the interaction between LGN and G␣ i , we set out to map the minimal G␣ i ⅐GDP binding sequence of each LGN GL. We first used GST-fused LGN GL with different lengths to pull down purified G␣ i3 ⅐GDP in our binding assay. This assay showed that each GL containing only the 19-residue core displayed only a background level of binding to G␣ i3 ⅐GDP (Fig. 1B). Obvious binding of G␣ i3 ⅐GDP to GL1 and GL4 was observed by extending the conserved 19-residue GL core by two residues (Fig. 1B). Any one of LGN GLs with length equal to or longer than 25 residues displayed comparable binding to their corresponding full-length motifs (Fig. 1B). We next measured the quantitative binding affinities of each of the four GLs to G␣ i3 ⅐GDP using isothermal titration calorimetry or fluorescence spectroscopy. Such quantitative binding assays revealed that the four full-length GLs share similar affinities (K D ϭ 54 -96 nM) in binding to G␣ i3 ⅐GDP (Fig.  1C). In agreement with the results derived from the pulldown binding assay, each LGN GL with a length of 25 residues has an essentially same binding affinity compared with the corresponding full-length motif (Fig. 1C), indicating that each of the 25-residue LGN GL contains the complete G␣ i ⅐GDP binding sequence. This finding is in sharp contrast to the interaction between G␣ i ⅐GDP and the RGS14 GoLoco motif, which requires a total length of 35 residues (14). Consistent with earlier studies (40), the LGN GLs bind to G␣ i3 ⅐GTP␥S with a ϳ100-fold weaker affinity than to G␣ i3 ⅐GDP (data not shown).
G␣ i ⅐GDP Can Simultaneously Bind to All Four LGN GLs-We next asked whether G␣ i ⅐GDP can simultaneously bind to the multiple GLs of LGN. We first tested the interaction between G␣ i ⅐GDP and the LGN GL34 tandem (aa 587-650), because the intervening sequence between the core sequences of GL3&4 is the shortest (15 residues to be exact; Fig. 1A). According to the structure of G␣ i1 ⅐RGS14 complex (14), two successive GL core sequences separated by a 15-residue linker cannot bind to two G␣ i because the bound G␣ i molecules would crash into each other. We examined the binding stoichiometry between LGN-GL34 and G␣ i3 ⅐GDP using analytical gel filtration chromatography. Upon addition of 2 or 3 molar ratios of G␣ i3 to GL34, a peak corresponding to a (G␣ i3 ⅐GDP) 2 ⅐GL34 complex was detected ( Fig. 2A), indicating that the two GLs in GL34 can simultaneously bind to G␣ i3 ⅐GDP. To further substantiate that the elution peak at ϳ11.60 ml in Fig. 2A represents the 2:1 stoichiometric complex formed between G␣ i3 ⅐GDP and GL34, we used two GL34 mutants (L594E and I628E), in which either the G␣ i3 ⅐GDP binding site on GL3 (the L594E mutant) or on GL4 (the I628E mutant) was disrupted. On the gel filtration column, the 1:2 mixtures of the two GL34 mutants with G␣ i3 ⅐GDP were eluted at a volume significantly larger than the wild type GL34, and a large portion of free G␣ i3 ⅐GDP was also detected (Fig. 2B); presumably the GL34 mutants only formed 1:1 stoichiometric complex with G␣ i3 ⅐GDP. This result also confirms that the wild type GL34 can form a 1:2 stoichiometric complex with G␣ i3 ⅐GDP. Further lengthening of the linker between GL3 and GL4 by inserting 10 flexible residues (five GS repeats, referred to as GL34Ins5GS) did not alter the elution profile of its complex with G␣ i3 ⅐GDP (data not shown), indicating that the 15-residue intervening sequence between GL3 and GL4 is sufficiently long for two molecules of G␣ i3 ⅐GDP to bind simultaneously to GL34. Similarly, two molecules of G␣ i3 ⅐GDP are capable of binding to LGN-GL12 (aa 483-586) or GL23 (aa 537-620). (Fig. 2, C and  D). Additionally, three molecules of G␣ i3 ⅐GDP were found to bind simultaneously to GL123 (aa 483-620) or GL234 (aa 537-650) of LGN (Fig. 2, E and F).
To characterize the binding stoichiometry more precisely, ITC analyses were performed. The titration profiles of G␣ i3 ⅐GDP to GL23 and GL34 can be well fitted with the model using one set of identical sites, yielding overall stoichiometries of 1.9:1 and 1.8:1, respectively (Fig. 3, A and B, and Table 2), consistent with the binding stoichiometry derived from the gel filtration analyses. The apparent binding affinity of GL23 was similar to those of the individual GoLocos, whereas GL34 had a weaker binding affinity than that of GL3 or GL4 (Table 2 and Fig. 1C). The titration profile of G␣ i3 ⅐GDP to the triple-GoLoco-containing protein GL234 was also fitted with the model with one set of binding sites, giving a weaker binding affinity of ϳ358 nM and a binding stoichiometry of 3.1:1 (Fig. 3C and Table 2). The purified GL234 protein underwent slight degradation, which might affect the accuracy of the binding affinity measurement. The titration profile of G␣ i3 ⅐GDP to GL12, however, was best fitted with the model that assumes two sets of binding sites (Fig. 3D), yielding one strong site (K D ϭ ϳ11 nM) and one weak site (K D ϭ ϳ188 nM) ( Table 2). The ITC titration profile of G␣ i3 ⅐GDP to GL123 was also fitted with the 'two sets of binding sites' model, giving rise to two strong sites (K D ϳ 4 nM) and one weak site (K D ϳ 186 nM) (Fig. 3E and Table 2). Similar atypical profiles of ITC titrations were also observed in the analyses of AGS3-GLs/G␣ i ⅐GDP interaction (41). It is worth  noting that these data analyses do not represent the complete description of the thermodynamics of the interactions between tandem LGN-GoLoco repeats and G␣ i3 ⅐GDP, in which intersite cooperativity likely exists. Because the full-length GoLoco region of LGN, i.e., GL1234, suffers from severe degradation, we did not analyze the binding property of GL1234 directly. However, the ITC titration data, consistent with the gel filtration analyses, strongly suggested that the full-length LGN binds G␣ i3 ⅐GDP with a stoichiometry of 1:4. The four GoLoco motifs of LGN have intrinsically similar binding affinities to G␣ i3 ⅐GDP. To explore the molecular details of the binding, we proceeded to determine the crystal structure of the G␣ i ⅐LGN-GoLoco complex.
Overall Crystal Structures of GL3 and GL4 in Complex with G␣ i ⅐GDP-Extensive efforts have been put to screen various constructs of the four LGN GLs in complex with GDP-loaded G␣ i3 or G␣ i1 , and we succeeded in obtaining well diffracting crystals for synthetic GL4 ( 621 DEDFFSLILRSQAKRMDEQRV-LLQRD 645 ) and GL3 ( 587 DEDFFDILVKCQGSRLDDQRCA-PPS 611 ) peptides in complex with G␣ i1/3 ⅐GDP. The G␣ i1 ⅐ GL4, G␣ i3 ⅐GL4, and G␣ i3 ⅐GL3 complexes diffracted to 2.9, 3.5, and 3.6 Å resolutions, respectively (Table 1). According to a previous structure-based protein design study, point mutations on G␣ i (E116L, Q147L, and E245L, respectively) can enhance its binding affinity to various GLs (15). We therefore constructed such three G␣ i3 mutants, hoping that the mutants might have higher affinities in binding to LGN GLs and thus yield better quality complex crystals. Opposite to our expectation, none of these mutants showed obviously enhanced binding to LGN GLs (data not shown). Nonetheless, the Q147L-G␣ i3 mutant⅐GL4 complex yielded better diffracting crystals (2.9 Å) than the wild type G␣ i3 ⅐GL4 complex.  Table 2.
The structures of G␣ i1(3) ⅐GL4 and G␣ i3 ⅐GL3 were solved by molecular replacement using the G␣ i1 ⅐RGS14 structure as the search model (Protein Data Bank code 1KJY) (14). The G␣ i ⅐GDP structure is well defined, and 21-22 amino acids of the GL3 or GL4 peptide are ordered in the structures of complexes (Fig. 4, A and B). The structures of G␣ i in the G␣ i1(3) ⅐GL4 and G␣ i3 ⅐GL3 complexes are highly similar to that in the G␣ i1 ⅐RGS14 complex (root mean square deviation of 0.67 Å), except for the Switch II region, which is shifted further away from the LGN-GL peptides because of the presence of two bulky hydrophobic residues in the GL peptides (Fig. 5, A and B). The GL peptides in the three complexes adopt highly similar structures (Fig. 4C). The N-terminal 10 residues of each LGN GL peptide (aa 623-632 of GL4 and aa 589 -598 of GL3), which corresponds to the first half of the conserved 19-residue GL core, forms an ␣-helix that occupies the cleft between Switch II and ␣3 of G␣ i (Fig. 4A). The following eight residues of the GL core (aa 633-640 of GL4 and aa 599 -606 of GL3) forms a "lid" in covering GDP. Only three or four residues C-terminal to the GL core (aa 641-643 of GL4 and aa 607-610 of GL3) were found to bind to the all-helical domain of G␣ i (Fig. 4). The structures of the LGN GL peptides in complex with G␣ i are entirely consistent with our biochemical data, showing that extending of the conserved GL core at the C-terminal end by three or four residues is necessary and sufficient for LGN GLs to bind to G␣ i (Figs. 1 and 2). The structures of the complexes also indicate that LGN GLs should function as GDIs by directly stabilizing the bound GDP as well as the interaction between the Ras-like domain and the all-helical domain of G␣ i (Fig. 4A).
A General Interaction Mode Revealed by the LGN GLs in Complex with G␣ i -Although the structures of G␣ i bound to the GLs of RGS14 and LGN are highly similar, the conformation of G␣ i -bound GLs of RGS14 and LGN are distinctly different (Fig. 5). First, a 16-residue fragment C-terminal to the conserved GL core of RGS14 is required for binding to G␣ i , and this 16-residue fragment forms ordered structure and has extensive interactions with the all-helical domain of G␣ i1 (14). In LGN-GL4/GL3, in contrast, only three or four residues C-terminal to the GL core are required for binding to G␣ i (Fig. 4A). Second, the orientation of the variable C-terminal tail of the RGS14 GL peptide is opposite to that of the LGN GL peptides (Fig. 5A). In the LGN GL4⅐G␣ i complex, the hydrophobic side chains of Val 641 , Leu 642 , and Leu 643 interact with Val 72 and Tyr 69 from the ␣A helix of the G␣ i all-helical domain; thus the C-terminal end of GL4 extends toward the N-terminal end of G␣ i ␣A (Fig.  5C). The residue corresponding to Val 641 in the RGS14 peptide is Gly 517 (Fig. 5D and Fig. 6A). The backbone carbonyl oxygen of Gly 517 forms two hydrogen bonds with side chains of Ser 75 and Gln 79 from G␣ i ␣A. The unique backbone dihedral angles ( ϭ 78°, ϭ Ϫ171°) of Gly 517 , which are not allowed by other amino acids, enable the C-terminal tail of the RGS14 GL peptide to take a sharp turn at this position and extend to the C-terminal end of G␣ i ␣A (Fig. 5, A and D). Sequence alignment of all known GLs from mammals reveals that only the GLs of RGS14 and RGS12 contain a Gly right after the conserved core motif, and the C-terminal residues of these two GLs share the identical sequence (Fig. 6A). The above structure-based amino acid sequence analysis suggests that the LGN GL/G␣ i interactions observed in this study represent the general mode of the interactions between GoLoco proteins and G␣ i . RGS14 and RGS12, instead, may represent a special subclass of GoLoco proteins in terms of G␣ i binding.
The Double Arg Finger-mediated GDP Binding of LGN GLs-The structure of the G␣ i1 ⅐RGS14 GL complex shows that a highly conserved (D/E)QR triad at the C-terminal end of the conserved GL core plays a critical role in binding to Mg 2ϩ -GDP (14). Similar to the G␣ i1 ⅐RGS14 GL interaction, the side chain of Arg 640 (Arg 606 ) of GL4 (GL3) in the (D/E)QR triad, which is equivalent to Arg 516 of RGS14, is inserted into the GDP-bind- LGN-GL4 in complex with G␣ i1 ⅐GDP. GDP is shown in the ball-and-stick model. All-helical domain and Ras-like domain of G␣ i1 is shown in wheat and light gray, respectively. The three switches are shown in violet, and the GL4 peptide is shown in cyan. B, the F o Ϫ F c density map of GL4 peptide is shown in green and contoured at 3.0 . C, comparison of the structures of the G␣ i1 ⅐GL4, G␣ i3 ⅐GL4, and G␣ i3 ⅐GL3 complexes by superimposing the backbone atoms in the three structures. G␣ i1 is shown the same as in A, whereas G␣ i3 in complex with GL3 (root mean square deviation of 0.76 Å) and GL4 (root mean square deviation of 0.51 Å) is not shown. GL3 peptide is shown in blue, and GL4 peptides bound to G␣ i1 and G␣ i3 are shown in red and green, respectively.

TABLE 2 Thermodynamic parameters of the bindings of LGN GoLoco motifs to G␣ i3 ⅐GDP determined by ITC titration
The titration data of GL12 and GL123 were fitted with the two sets of binding sites model, whereas the other data were fitted with the one set of binding sites model. N denotes the number of binding sites in each model. ing pocket and binds to ␣-phosphate of GDP (Fig. 6B). However, there is a distinct feature of GL4/GL3 in GDP binding with respect to RGS14 GL. Another highly conserved Arg five residues upstream of the Arg in the (D/E)QR triad in LGN GL peptides (Arg 635 in GL4 and Arg 601 in GL3) binds to the ␣ and ␤ phosphates of GDP (Fig. 6B). In RGS14 GL, the residue corresponding to this second Arg is a Gly, and a Mg 2ϩ ion was found to be necessary to stabilize the ␤ phosphates of GDP (14). Therefore, different from RGS14, LGN GLs use two Arg residues instead of one to bind to and stabilize GDP. The structures of the LGN GLs in complex with G␣ i further indicate that the LGN GLs can bind to GDP-bound G␣ i independent of the presence of Mg 2ϩ . This structure-based prediction is confirmed by direct binding experiment (data not shown). Sequence alignment analysis reveals that, except for RGS14 GL, the rest of GLs all contain a (R/K)X(D/E)(D/E)QR GDP-binding sequence (Fig.  6A), and we refer to this sequence as the double Arg finger. This sequence analysis further supports that the LGN GL/G␣ i interaction represents the general mode of GL-mediated binding to G␣.
The Double-arginine Fingers Are Critical to the GDI Activities of LGN-GLs-To confirm the functional importance of the two Arg in the double-arginine finger in LGN GLs, we performed point mutations of the two arginines and tested the G␣ i ⅐GDP binding affinities and GDI activities of these mutants. Single substitution mutations (R635G, R635A, and R640A) caused ϳ50-fold decrease in GL4 binding to G␣ i ⅐GDP, and the double mutation (R635A/R640A) led to ϳ500-fold G␣ i ⅐GDP binding affinity decrease (Fig. 6C). Similar results were also obtained from the other LGN GLs, indicating that the two conserved arginine fingers are critical for binding of G␣ i ⅐GDP to LGN-GLs. This finding is in contrast to the RGS14 GL, in which the substitution of the Arg in the finger with Ala or Leu did not decrease the binding affinity of RGS14 to G␣ i1 ⅐GDP (14). Careful examination of the crystal structures of G␣ i in complex with LGN GL peptides revealed that the side chains of the two Arg residues also form hydrogen bonds with Val 179 and Thr 181 from G␣ i (Fig. 6B). In contrast, the side chain of Arg 516 in RGS14 GL interacts exclusively with GDP (14).
The GDI activities of LGN GLs were evaluated by AlF 4 Ϫ -induced increase of intrinsic tryptophan fluorescence of G␣ i and by direct binding of BODIPY-GTP␥S to G␣ i . In agreement with the previous studies (40), the four GLs exhibited similar GDI activities (data not shown). Moreover, comparison of the GDI activities of GL peptides with different lengths showed that the 25-residue minimal G␣ i -binding GL fragments shown in Fig. 1 are also sufficient for their GDI activities (data not shown). Further quantification of the GDI activities using the association rate of BODIPY-GTP␥S binding revealed IC 50 values of a few M for LGN GLs, which is slightly weaker than that of showed a complete inhibition of GDP dissociation from G␣ i3 (Fig. 6D). The R635G-GL4 or the R635A-GL4 displayed obviously weakened GDI activities, whereas the R640A-GL4 and R635,640A-GL4 had essentially no detectable GDI activity (Fig.  6D). Substitution of the first Arg (Arg 601 ) in the double-arginine finger of GL3 with Ala or Gly also diminished its GDI activity (data not shown). Thus, we conclude that both arginines in the double-arginine finger of LGN GoLoco motifs are important for their GDI activity.

DISCUSSION
Both the binding to GDP-loaded G␣ subunits and the GDI activity of GL require residues beyond the 19-residue conserved core sequence (14,41). Because the C-terminal flanking sequences of GLs are highly diverse among GoLoco proteins (7), it has been hypothesized that the variable C-terminal tail sequences of GLs are the specificity determinants governing GL/G␣ interactions. In the present study, we demonstrate that only a few residues (3-4 aa) C-terminal to the conserved GL core are required for LGN GLs to bind to and to inhibit GDP dissociation of G␣ i ⅐GDP, a finding that is in sharp contrast to that of RGS14 GL. Sequence alignment analysis suggests that the conformation of the GL peptide in the G␣ i1 ⅐RGS14 structure is likely a unique example of GL/G␣ interaction. The LGN GL/G␣ i interaction described in the current study instead is likely a general binding mode between GLs and G␣. The structures of LGN GLs in complex with G␣ i ⅐GDP also suggest that FIGURE 6. The double arginine fingers of the LGN GLs play a crucial role in GDP coordination and GDI activity. A, sequence alignment of the GLs in mammalian GoLoco proteins. Absolutely and highly conserved residues are highlighted in red and yellow, respectively. The residue right behind (D/E)QR, which determines the C-terminal direction, is highlighted with a blue triangle. The residues involved in the interactions with G␣ i are labeled with red stars at the top. The di-arginine fingers are highlighted with black boxes. B, structural details of the GDP-binding pocket in the G␣ i1 ⅐GL4 complex and G␣ i1 ⅐RGS14 complex. Polar interactions are shown with dashed lines. Distances of polar interactions are shown with magenta numbers (Å). The color scheme is the same as in Fig.  4A. C, binding affinities of the GL4 mutants with single or double substitutions of its two arginine residues to G␣ i3 ⅐GDP derived from fluorescence-based assays. D, GDI activities of the wild type and mutant GL4 peptides measured with AlF 4 Ϫ -induced increase of intrinsic tryptophan fluorescence. E, structural model of the LGN⅐G␣ i ⅐GDP complex. The TPR domain, the TPR-binding NuMA peptide, and the GLs responsible for G␣ i ⅐GDP binding are shown in blue, red, and cyan, respectively.
the short variable C-terminal sequences of LGN GLs are unlikely to determine their binding specificity to G␣ subunits. Consistently, previous studies have shown that LGN GLs bind to all three forms of GDP-loaded G␣ i (i1, i2, and i3). As for the G␣ o ⅐GDP binding, discrepancies exist in the literature. An early study by McCudden et al. (40) reported that LGN GLs selectively bind to G␣ i ⅐GDP, but not to G␣ o ⅐GDP or G␣ s ⅐GDP. Recently, Kopein et al. (42) found that LGN, as well as its Drosophila homolog Pins, can bind robustly to both GDP-loaded G␣ o and G␣ i .
We have demonstrated in this study that every one of the four LGN GLs can bind to G␣ i ⅐GDP with high affinity. Additionally, although the LGN GL peptides are much shorter than their counterpart from RGS14, LGN GLs also act as potent GDIs. The structures of the LGN GL3 and GL4 in complex with G␣ i suggest that both the double-arginine finger and the short variable tail of the GL peptides are important for their GDI activities. The di-arginine finger makes extensive salt bridges with the phosphates of GDP, and the GDP in return makes contacts with both the Ras-like and all-helical domains of G␣ i . The variable C-terminal tail of the GL peptides further interacts with the all-helical domain of G␣ i . Thus, in addition to stabilizing GDP bound to G␣ i , the binding of GL peptide further promotes the closed conformation of G␣ i (i.e., by restricting the opening of the all-helical domain and subsequent dissociation of GDP from G␣ i ; Fig. 7).
The characteristic multiple GLs in LGN and its Drosophila homolog Pins have been implicated to play a role in regulating their intramolecular interactions between TPR repeats and GLs in response to the binding of G␣ i ⅐GDP and NuMA/Mud (34,43). In addition to this, the multiple GLs in LGN (Pins) also function as a scaffold in regulating the localization of related protein complexes and organizing signaling pathways mediating spindle orientations. The detailed characterizations of interactions between LGN-GLs and G␣ i ⅐GDP in this work demonstrate that in its open state the four LGN GLs have equal capacity to bind to G␣ i ⅐GDP (Fig. 6E). In another word, the stoichiometry of LGN/G␣ i ⅐GDP complex in vivo likely depends on the concentration of G␣ i ⅐GDP, which in turn regulates the cortical localization of LGN-bound proteins, such as NuMA.
Recently, it was found that the extrinsic GPCR Tre1 signaling determines the orientation of cortical polarity in the asymmetric cell division of Drosophila neuroblast (44). Tre1 was shown to activate G␣ o , and the GTP form G␣ o can specifically associate with the first GL of Pins (44). Thus, the presence of multiple GLs allows Pins to function as a scaffold to simultaneously engage G␣ o -and G␣ i -mediated signaling events during asymmetric cell division.