Mechanism of Assembly of a Substrate Transfer Complex during Tail-anchored Protein Targeting*

Background: Get4/5 is required for the efficient transfer of tail-anchored proteins to Get3. Results: The Get3·Get4/5 complex forms an intermediate mediated by electrostatic interactions. Conclusion: The rapid association of the Get3·Get4/5 intermediate complex is followed by a conformational change to the stable inhibited structure dominated by hydrophobic interactions. Significance: These results provide insight into the mechanism of tail-anchored protein targeting. Tail-anchored (TA) proteins, defined as having a single transmembrane helix at their C terminus, are post-translationally targeted to the endoplasmic reticulum membrane by the guided entry of TA proteins (GET) pathway. In yeast, the handover of TA substrates is mediated by the heterotetrameric Get4/Get5 complex (Get4/5), which tethers the co-chaperone Sgt2 to the targeting factor, the Get3 ATPase. Binding of Get4/5 to Get3 is critical for efficient TA targeting; however, questions remain about the formation of the Get3·Get4/5 complex. Here we report crystal structures of a Get3·Get4/5 complex from Saccharomyces cerevisiae at 2.8 and 6.0 Å that reveal a novel interface between Get3 and Get4 dominated by electrostatic interactions. Kinetic and mutational analyses strongly suggest that these structures represent an on-pathway intermediate that rapidly assembles and then rearranges to the final Get3·Get4/5 complex. Furthermore, we provide evidence that the Get3·Get4/5 complex is dominated by a single Get4/5 heterotetramer bound to one monomer of a Get3 dimer, uncovering an intriguing asymmetry in the Get4/5 heterotetramer upon Get3 binding. Ultrafast diffusion-limited electrostatically driven Get3·Get4/5 association enables Get4/5 to rapidly sample and capture Get3 at different stages of the GET pathway.

The targeting of membrane proteins to their correct location in the cell is a highly regulated process (1). The majority of membrane proteins are targeted via the signal recognition particle, which typically recognizes the initial hydrophobic transmembrane domain as it emerges from the ribosome (2). However, the ubiquitous TA 4 proteins, defined topologically by a single transmembrane domain near the C terminus, are unable to access the signal recognition particle pathway as their targeting signal, the single transmembrane domain, emerges from the ribosome only after protein synthesis is complete and must be targeted to the endoplasmic reticulum post-translationally (3,4). In eukaryotes, TA proteins account for 1-2% of the proteome and are involved in many essential cellular processes such as apoptosis, vesicle fusion, and protein trafficking (5,6).
A series of genetic and biochemical experiments in yeast and mammalian cells identified members of a dedicated pathway for delivery of TA proteins to the endoplasmic reticulum. In yeast, these factors comprise the guided entry of TA proteins (GET) pathway that consists of six proteins, Get1-5 and Sgt2, all with homologs in higher eukaryotes (7,8). The first committed step in TA targeting is the formation of an Sgt2⅐TA complex (9), which may be assisted by chaperones. TA substrate is transferred to Get3 in a Get4/5-dependent manner (9). Extensive structural characterization of Get3, the central TA targeting factor, demonstrated that it undergoes ATP-dependent conformational changes from an open to closed form required for capturing the TA substrate (10 -15). Once formed, this Get3⅐TA complex is then localized to the endoplasmic reticulum by the membrane proteins Get1 and Get2, which stimulate release of the TA protein and subsequent insertion into the endoplasmic reticulum membrane (16 -19).
Efficient delivery of a TA substrate to Get3 requires the heterotetrameric Get4/5 that provides the link between Sgt2 and Get3 (9,20,21). Get4/5 stabilizes the ATP-bound state of Get3, delaying ATP release and inhibiting ATP hydrolysis, thereby locking Get3 in a conformation competent for TA substrate binding (22). Recently, the first crystal structure of a Get3⅐ Get4/5 complex provided insight into the role of nucleotide in complex formation where Get4 binds to two functionally distinct binding interfaces that govern anchoring and ATPase regulation on a closed Get3 (23). Importantly, mutations introduced at these interfaces demonstrated that Get4-mediated regulation of ATP hydrolysis by Get3 were critical for efficient TA targeting. A recent study has now shown that Get4/5 binding to Get3 occurs on rapid diffusion-limited time scales, suggesting an electrostatic interaction (24). However, the nature of this interaction cannot be explained by the aforementioned Get3⅐Get4/5 structure.
This report describes the crystal structure of a Get3⅐Get4/5 intermediate complex from Saccharomyces cerevisiae in two crystal forms, at 2.8 and 6.0 Å, respectively. The structure represents an initial binding interaction mediated by electrostatics that facilitates the rate of subsequent complex formation. This is supported by kinetic analysis of Get3⅐Get4/5 complex formation that provides direct evidence for the two-step complex formation mechanism. Finally, mass spectrometry and multiangle light scattering were used to demonstrate that, under physiological steady-state conditions, a single Get4/5 heterotetramer is bound to one monomer of a Get3 dimer. This work allowed us to generate a refined model for Get3⅐Get4/5 complex formation.

Experimental Procedures
Protein Cloning, Expression, and Purification-The sequences of Get4 and Get5 were cloned as described previously (25). To generate the Get4/5N used in this study, this construct was further modified by truncating the C terminus of Get4 (residues 291-312) and by the addition of a stop codon after residue 54 within Get5 and verified by DNA sequencing. All Get4/5 proteins were overexpressed in BL21-Gold(DE3) (Novagen) grown in 2ϫ YT (16 g/liter tryptone, 10 g/liter yeast extract, and 5 g/liter NaCl) medium at 37°C and induced for 3 h by the addition of 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside. Cells were lysed using a Microfluidizer (Microfluidics) and purified as a complex by nickel affinity chromatography (Qiagen). The affinity tag was removed by an overnight tobacco etch virus protease digest at room temperature while dialyzing against 20 mM Tris, pH 7.5, 30 mM NaCl, and 5 mM ␤-mercaptoethanol (BME). A second nickel-nitrilotriacetic acid column was used to remove any remaining His-tagged protein, and then the sample was loaded onto a 6-ml Resource Q anion exchange column (GE Healthcare). The peak containing the Get4/5N complex was collected and concentrated to 15-20 mg/ml. Initial purifications of the Get4/5N complex were verified to be a single monodispersed species by size exclusion chromatography (SEC) using a Superdex 200 16/60 column (GE Healthcare) equilibrated with 20 mM Tris, pH 7.5, 100 mM NaCl, and 5 mM BME. Full-length Get4/5 used in biochemical studies was further purified using a Superdex 200 16/60 column (GE Healthcare) equilibrated with GET buffer (50 mM K-HEPES, pH 7.5, 150 mM KOAc, 10 mM Mg(OAc) 2 , 10% (v/v) glycerol, and 5 mM BME). Fractions containing Get4/5 were pooled and concentrated to ϳ5 mg/ml. The S. cerevisiae Get3 coding region was cloned as described previously (10). A His 6 tag followed by a tobacco etch virus protease site was fused to the N terminus, and a stop codon was placed in front of the C-terminal His 6 tag. All S. cerevisiae Get3 mutants were generated using the QuikChange method. All Get3 proteins were expressed in BL21-Gold(DE3), grown in 2ϫ YT (16 g/liter tryptone, 10 g/liter yeast extract, 5 g/liter NaCl) medium, and induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 16 h at 22°C. Cells were lysed using a Microfluidizer and purified by nickel affinity chromatography. The affinity tag was removed by an overnight tobacco etch virus protease digest at room temperature while dialyzing against 20 mM Tris, pH 7.5, 100 mM NaCl, and 5 mM BME. A second nickel-nitrilotriacetic acid column was used to remove any remaining His-tagged protein, and then the sample was run on a Superdex 200 16/60 column (GE Healthcare) equilibrated with dialysis buffer (crystallization trials) or GET buffer (biochemical studies). Fractions corresponding to a dimer of Get3 were pooled and concentrated to 15-20 mg/ml. Get3⅐Get4/5N complex was formed by equilibrating 105 mol of Get4/5N with 100 mol of Get3 at room temperature in 500 l of 20 mM Tris, pH 7.5, 10 mM NaCl, 5 mM BME, 1 mM MgCl 2 , and either 1 mM ADP, AMP-PNP, or ATP. Prior to complex formation, Get3 had been pre-equilibrated with 1 mM MgCl 2 and either 1 mM ADP or 1 mM AMP-PNP for 5 min at room temperature. Get3⅐Get4/5N complex was further separated from free Get4/5N using Superdex 200 10/300 (GE Healthcare) equilibrated with 20 mM Tris, pH 7.5, 10 mM NaCl, and 5 mM BME. Get3⅐Get4/5N complex using full-length Get4 was formed as above but in the absence of nucleotide and MgCl 2 . This complex was further separated from free Get4/5N using Superdex 200 16/60 (GE Healthcare) equilibrated with 20 mM Tris, pH 7.5, 10 mM NaCl, and 5 mM BME. All complexes were concentrated to 10 -12 mg/ml before use in crystallization experiments.
Crystallization-Get3⅐Get4/5N crystal trials were carried out using the sitting drop vapor diffusion method at room temperature by equilibrating equal volumes of the protein complex solution and reservoir solution using a TTP LabTech Mosquito robot and commercially purchased kits (Hampton Research, Qiagen, and Molecular Dimensions Ltd.). The 2.8-Å Get3⅐ Get4/5N intermediate complex crystals grew in the presence of 17% PEG 3350, 0.24 M sodium citrate, and 30 mM tris(2-carboxyethyl)phosphine. Crystals were cryoprotected by transferring directly to 10 l of a reservoir solution supplemented with 20% glycerol, 1 mM ADP or AMP-PNP, and 1 mM MgCl 2 and incubated for Ն10 min before being flash frozen in liquid nitrogen. The 6.0-Å Get3⅐Get4/5N intermediate complex crystals grew in the presence of 12% PEG 3350 and 0.1 M sodium malonate, pH 5.0, and were cryoprotected by transferring directly to 10 l of a reservoir solution supplemented with 20% glycerol before being flash frozen in liquid nitrogen.
Data Collection, Structure Solution, and Refinement-All structures were solved using data sets collected on Beamline 12-2 at the Stanford Synchrotron Radiation Lightsource. Each structure was solved from a single data set that was integrated using MOSFLM (26) or XDS (27) and scaled and merged using SCALA (28,29). The crystal of Get3⅐Get4/5N in the intermedi-ate conformation diffracted to 2.8 Å, and phases were obtained by molecular replacement with Phaser (30) as implemented in PHENIX (31) using a monomer of the open (apo) form of Get3 (Protein Data Bank code 3A37) and one Get4/5N heterodimer (Protein Data Bank code 3LKU) as search models. Coordinates were refined using REFMAC v6.3 (28,29) with NCS and B-factor restraints. Manual rebuilding was performed using Coot (32). Translation/libration/screw groups were included in the later refinement stages and were determined using the TLSMD web server (33). These associated regions within the structure were allowed a certain degree of variation based on predicted motions they might have during data collection. The final model refined to an R-factor of 22.4% (R free ϭ 26.1%). The second structure in the intermediate conformation was solved to 6.0 Å using methods similar to those described above. This structure lacked the Get4 C-terminal truncation engineered for the other structures and refined to an R-factor of 27.4% (R free ϭ 30.3%). Full crystallographic statistics are in Table 1.
Structure Analysis and Figures-Schematic representations of protein structures were prepared using PyMOL (Schrödinger, LLC), whereas surface representations were prepared using UCSF Chimera (34). Surface figures were made in Chimera. Electrostatic surface potentials were calculated using Adaptive Poisson-Boltzmann Solver with default values as implemented in the PDB2PQR web server (35,36).
Isothermal Titration Calorimetry-Get3⅐Get4/5N binding experiments were carried out using the MicroCal iTC200 system (GE Healthcare). Binding affinities were measured by titrating 350 M Get4/5N into a sample cell containing 50 M Get3 D57V. All samples were diluted in GET buffer containing 1 mM ATP. For the first titration, 0.4 l of Get4/5N was injected. For all following titrations, 2 l of Get4/5N was injected for 20 intervals spaced 120 s apart at 25°C. The stirring speed and reference power were 1000 rpm and 5 cal/s. Affinity constants were calculated from the raw data using Origin v7.4 software (MicroCal).
Multiangle Light Scattering-100 l of 35 M Get3 and 100 l of 35 M Get4/5 were prepared in GET buffer containing 1 mM ATP and either loaded separately or together onto a Superdex 200 10/300 (GE Healthcare) column equilibrated in GET buffer. Samples were analyzed using a Dawn HELEOS II multiangle light scattering unit (Wyatt Technology).
Alkylation and Mass Spectrometry-Get4/5 S48C/C177T was reduced with 2.5 mM tris(2-carboxyethyl)phosphine in GET buffer (without reducing agent) at room temperature for 2 h. For each alkylation reaction, 1 M Get4/5 alone or in the presence of 2 M Get3 was preincubated in 2 mM ATP for 10 min to allow for complex formation followed by the addition of 30 M N-ethylmaleimide. The reaction was quenched with 9.0 mM DTT at the indicated time points, concentrated under vacuum, and redissolved in 0.2% formic acid, and 50 pmol of protein was analyzed on an LC-MSD SL 1100 series (Agilent) mass spectrometer. The samples were chromatographed on a 2.1 ϫ 150-mm Zorbax 300SB-C3 column (Agilent) using a gradient consisting of 0.2% formic acid and 0.2% formic acid in 89.8% acetonitrile and 10% methanol. The m/z values of the intact proteins were measured in the single quadrupole, deconvoluted, and quantified using the ChemStation software (Agilent). Control experiments in which different ratios of unalkylated and alkylated proteins were mixed and subjected to MS analysis showed the quantification of the ratio of alkylated species to be reliable (37).
Get4/5 PEGylation Assay-Get4/5 Q34C/C177T containing a His 6 tag was reduced with 2.5 mM tris(2-carboxyethyl)phosphine in GET buffer (without reducing agent) at room temperature for 2 h. For each PEGylation reaction, 0.45 M Get4/5 alone or in the presence of 2 M Get3 was preincubated in 2 mM ATP for 10 min to allow complex formation followed by the addition of 60 M PEG maleimide (10-kDa conjugates; Sigma). The reaction was quenched with 9.0 mM DTT at the indicated time points, and the extent of Get4/5 PEGylation was followed by Western blotting with an anti-His antibody (Qiagen). The ratio of unmodified Get4/5 to PEGylated Get4/5 was determined using ImageQuant software (GE Healthcare).
The reaction was incubated overnight at 4°C and stopped by adding 2 mM DTT. Excess dye was removed by gel filtration using Sephadex G-25 (Sigma) equilibrated with GET buffer.
Association and Dissociation Rate Measurements-All rate measurements were performed on a KinTek Corp. stoppedflow apparatus using sample prepared in GET buffer. For association rate measurements, acrylodan-labeled Get4/5 was held constant at 0.2 M, Get3 concentration was varied as indicated, and ATP was present at 2 mM. Observed rate constants (k obsd ) were plotted as a function of Get3 concentration and fit to the following equation (Equation 1).
in which k on is the association rate constant and k off is the dissociation rate constant. For dissociation rate measurements, a pulse-chase experiment was used. A complex between acrylodan-labeled Get4/5 (at 0.15 M) and Get3 (at 0.3 M) was preformed by incubation in 2 mM ATP for 10 min followed by addition of unlabeled Get4/5 at 6 M as the chase to initiate Get3⅐Get4/5 dissociation. The time course for change in fluorescence (F obsd ) was fit to the following double exponential function (Equation 2).
in which F e is the fluorescence when the reaction reaches equilibrium, ⌬F 1 and k fast are the magnitude and rate constant of the fluorescence change in the fast phase, and ⌬F 2 and k slow are the magnitude and rate constant of the fluorescence change in the slow phase.
Kinetic Modeling-The kinetic model was derived using Kin-Tek Explorer Professional Software (KinTek Corp.) (38,39). Kinetic and equilibrium simulations were carried out using the rate constants listed in Fig. 1H, and values for the experimental data were taken from Rome et al. (24).
Kinetic Simulations-Simulation of Get3⅐Get4/5 association was carried out using the KinTek Explorer association function. Equations used to model the associations are listed in Table 2 where G3 denotes Get3, G45 denotes fluorescently labeled Get4/5, Get345 denotes the intermediate Get3⅐Get4/5 complex, G345* denotes the inhibited Get3⅐Get4/5 complex after rearrangement, and G345*G45 denotes the saturated complex with two Get4/5 bound to the Get3 dimer (refer to Fig. 1H). Modeling used the following initial conditions: [G45] 0 ϭ 0.2 M, [G3] 0 ϭ 0.1, 0.18, and 0.26 M. The following fluorescence values (F in arbitrary units) were assigned to each species: F ϭ 5 for G45, F ϭ 10 for G345 and G345*, F ϭ 20 for G345*45. For the equation G345 N G345*, the forward and reverse rate constants were varied until the simulation results matched experimental data.
Simulation of the dissociation kinetics of the Get3⅐Get4/5 complexes was carried out using the KinTek Explorer two-step mixing function, which accurately reflects the experimental pulse-chase setup. In the first time period (T 1 ), 0.3 M G3 and 0.15 M G45 were mixed for 3 s allowing for equilibrium to be reached. During the second time period (T 2 ), 6 M unlabeled Get4/5 (chase; designated C) was added to this mixture. Dissociation of the Get3⅐Get4/5 complexes during T 2 involves a collection of exchange reactions between G45 and the chase (unlabeled Get4/5) represented by equations in Table 3.
To simulate the equilibrium of complex formation between Get3 and Get4/5, the KinTek Explorer equilibrium titration function was used. For these simulations, 0.05 M G45 was held constant, and the fluorescence at equilibrium was simulated with increments of 0.005 M G3 until a final concentration of 0.5 M G3 was reached.

Results and Discussion
Formation of the Get3⅐Get4/5 Complex Follows an Electrostatically Driven Intermediate-To understand the mechanism of assembly for the Get3⅐Get4/5 complex, we characterized the assembly pathway using pre-steady-state kinetics. Complex formation was monitored using acrylodan-labeled Get4/5 whose fluorescence is enhanced upon binding Get3 (24). The Get3⅐Get4/5 complex has been demonstrated to be highly stable but dynamic, and its formation follows rapid biphasic association at near diffusion-limited time scales (24). The observed association rate constants in both kinetic phases are linearly dependent on Get3 concentration (24), indicating that both phases represent bimolecular association of Get3 with Get4/5. Although the initial report used full-length Get4/5, the same biphasic association was observed with a truncated Get4/5N that lacks the homodimerization domain of Get5 and hence forms a heterodimer (Fig. 1, A and B) (23,25). This rules out the possibility that the second kinetic phase arises from binding of Get3 to the second Get4 molecule in a Get4/5 heterotetramer.
Additional characterizations strongly suggest the presence of an additional rearrangement step in the assembly pathway. First, the rate of the initial "burst" phase is strongly dependent on ionic strength, consistent with an electrostatically driven association (Fig. 1C) (24), whereas the second association phase is independent of ionic strength (Fig. 1D) (24). This suggests a structural rearrangement to adopt a more stable interface prior to the second association phase. Second, despite the large changes in the relative association rates of the salt-sensitive and -insensitive phases, the amplitudes of these two phases are unaltered by changes in ionic strength (Fig. 1E). This argues against the possibility that the biphasic association arises from parallel pathways, i.e. that the second association phase arises from a subpopulation of Get3 (or Get4/5) that binds the interaction partner with slower kinetics. Third, the dissociation rate constant extrapolated from the fast association phase (Fig. 1B) is 8 -10 s Ϫ1 (k obsd ϭ k on [Get3] ϩ k off ); this k off value is 10 -100fold faster than those of the final Get3⅐Get4/5 complex determined directly in pulse-chase experiments (Fig. 1F) (24), suggesting that the initial association gives rise to an unstable The Get3⅐Get4/Get5 Intermediate Complex DECEMBER 11, 2015 • VOLUME 290 • NUMBER 50

The Get3⅐Get4/Get5 Intermediate Complex
intermediate that must subsequently rearrange to adopt more stable interfaces. Further in support of this model are results from the dissociation rate measurements. Although dissociation of the Get3⅐Get4/5 complex also exhibits biphasic kinetics, neither the rates nor the magnitude of the phases is dependent on ionic strength (Fig. 1F). Finally, the magnitude of the slow dissociation phase increases successively when apo-Get3 is compared with ADP-and ATP-bound Get3 (Fig. 1G), consistent with previous data showing that ATP strengthens binding between Get3 and Get4/5 by inducing Get3 into a more closed conformation (22)(23)(24)(25)31). These new data, coupled with kinetic simulation, support a multistep mechanism for the Get3⅐Get4/5 interaction as the simplest model to explain all the available observations ( Fig.  1H) (24). In this model, association of Get3 with Get4/5 involves the following steps. (i) Resting Get3 bound to ATP is biased toward a "closed" conformation. (ii) Initial binding of Get4/5 to Get3 is rapid and generates an intermediate dominated by electrostatic interactions. (iii) This is followed by a conformational change to the final stable structure dominated by hydrophobic interactions. (iv) Under the experimental conditions, the second subunit in the Get3 dimer can bind another Get4/5 to give rise to the second phase in association kinetics. Using the experimentally observed k on and k off values from both the salt-sensitive and salt-insensitive phases, the equilibrium binding data, and kinetic modeling, we were able to completely assign the rate constant for individual microscopic steps in this model. Analytical simulations based on this model reproduced the experimentally observed association/dissociation kinetics and equilibrium titrations ( Fig. 2A). Furthermore, the equilibrium titration data with full-length Get4/5 (24) could only be fit to a 1:2 stoichiometry in which a Get3 dimer is bound to two different Get4/5 heterotetramers (Fig. 2B), consistent with the biphasic nature of complex formation.
Although binding of a second Get4/5 to Get3 was observed (Fig. 1H), multiple observations suggest that the stable inhibited Get3⅐Get4/5 complex (species iii) dominates under most conditions, and accumulation of the saturated Get3⅐Get4/5 complex (species iv) is modest. First, previous assays of Get3 activity show that binding of one full-length Get4/5 is sufficient to inhibit the ATPase activity of Get3 (22). Second, kinetic simulations show that during complex assembly with relatively stoichiometric amounts of proteins the saturated Get3⅐Get4/5 complex accumulates to Ͻ20% of all the complexes formed (Fig. 2C). Coupled with the different Get3⅐Get4/5N complexes in various crystal structures (23), it appears that binding of Get4/5 to Get3 has a preferred but not obligatory stoichiometry.
The Structure of a Get3⅐Get4/5 Intermediate Complex-The recent crystal structure of an ATP-bound Get3⅐Get4/5N complex demonstrated a stable interface that included conserved hydrophobic interactions (23). The structure allowed rationalization of how Get4 stabilized an ATP-bound Get3; however, it was unclear how the structure could account for the multistep Get3⅐Get4/5 assembly seen experimentally. To address this question, we set out to obtain a structure of Get4/5N bound to a Get3 dimer under conditions that would stabilize an initial intermediate complex. As before (23), Get3 and Get4/5 were expressed and purified separately, then combined, and purified as a 1:1 complex in low ionic strength conditions by SEC (23). Initial crystals were obtained using a Get4/5N construct similar to that used for solving the structure of the Get4/5N heterodimer (25). This resulted in a crystal that diffracted to 6.0 Å in space group P2 1 . Phases for this crystal form were obtained  (Table 1). To improve the diffraction limit, purification and crystallization trials were performed in the presence of either ADP or the non-hydrolyzable ATP analogue AMP-PNP. In addition, 22 residues from the C terminus of Get4 were truncated as they are disordered in previous structures (12,25). This resulted in a new crystal form that diffracted to 2.8 Å in space group P2 1 2 1 2. The final model was refined to an R/R free of 22.4/26.1% (Table  1). Although ADP and AMP-PNP produced identical crystal forms, the highest resolution data set reported here was collected from a crystal that grew in the presence of AMP-PNP, although the nucleotide was not visible in the density. The Get3⅐Get4/5 Intermediate Complex Is Dominated by an Electrostatic Interface-In both crystal forms, Get4 binds in a similar orientation on Get3 (Fig. 3, A and B). The 2.8-Å structure contains a 2:1 complex of a Get3 dimer bound to a single Get4/5N heterodimer (Fig. 3A). This crystal form is related to another Get3 crystal form (14) in which limited space in the crystal lattice allows for only one Get4/5N molecule despite the 1:1 stoichiometry of the complex put into crystallization trays (as measured by SEC). Although nucleotide was present throughout purification and crystallization, the nucleotidebinding pocket appears empty. For the 6.0-Å structure (Fig. 3B), the asymmetric unit contained a 1:1 complex with eight copies of Get3 bound to eight Get4/5N heterodimers (Fig. 3B, top  right).
In both cases, Get3 is in an open conformation, and Get4/5N binds in the same interface and orientation, defined here as the intermediate complex. Despite the differences in crystallization and space group, there is little difference in the interface between the two structures, only a slight 10°rotation at the furthest point (Fig. 3C). The higher resolution structure will be used for reference for the rest of the text. In this intermediate complex, Get4 interacts with a single monomer of Get3 in an orientation compatible with both the open and closed structures, burying ϳ970 Å 2 of surface area (Figs. 3D and 4A). The interactions in this interface are electrostatic and involve the positively charged surface of Get4 binding to the negatively charged surface of Get3 (Fig. 4, A and B).
Get5 Get4 Get3 Get3   (wheat and magenta). B, left, one subunit from the 6.0-Å structure containing two Get4/5N molecules (Get4, salmon; Get5N, gray) bound to a Get3 dimer (wheat and magenta). Top right, the asymmetric unit of the 6.0-Å structure of Get4/5N (Get4, red; Get5N, gray) bound to Get3 (wheat and magenta). The black box outlines one Get3 dimer (wheat and magenta) bound to two Get4/5N heterodimers (Get4, salmon; Get5N, gray). The rest of the asymmetric unit is colored as follows: Get3, light gray; Get4, dark gray; Get5N, black. C, two views showing the relative orientation of Get4 monomers from the 2.8-(green) and 6.0-Å (red) crystal forms. Alignment was based on helices ␣10 and ␣11 from the contacting Get3 monomer. D, view of the interface from the 2.8-Å structure showing interactions between Get4 (green) and Get3 (wheat and purple). Residues Asp 263 (red) and Glu 307 (blue) were tested for interaction and colored based on phenotype. These residues are unique to this interface and are not found in the ATP-bound structure (23). Dashed lines represent potential interactions and are labeled with their atomic distances.
In this orientation, Get4 ␣1 sits on top of a groove formed by the loop before Get3 ␣1 and the loop following Get3 ␣11 such that the N terminus of Get4 ␣1 packs against the loops following Get3 ␣9 and ␣10, resulting in an interaction between Lys 23 on Get4 ␣1 and Glu 307 on Get3 (Fig. 3D). In addition, Get4 ␣2 packs roughly perpendicularly to Get3 ␣10 and ␣11 such that the N terminus contacts Get3 ␣11 and the C terminus contacts Get3 ␣10. This results in interactions of the invariant residues Arg 42 on Get4 ␣2 with Asp 308 on Get3 and Arg 45 on Get4 ␣2 with both Glu 253 and Gln 257 on Get3 ␣10 (Fig. 3D). The N terminus of Get4 ␣3, which is oriented between ␣1 and ␣2, contacts the top of Get3 ␣10, allowing His 51 on Get4 ␣3 to interact with Asp 263 on Get3 ␣10 (Fig. 3D).
Comparison of the Two Distinct Get3⅐Get4/5 Complexes-In the previously reported ATP-bound Get3⅐Get4/5N complex, Get4/5N is bound across the dimer interface of a closed Get3, forming interactions with both monomers, which we define as the inhibited complex (Fig. 4C). This is in contrast to the current structure, which is bound to an open Get3 and interacts with only one monomer (Fig. 4C). Because the Get4-binding surfaces do not undergo a significant conformational change between the open and closed Get3 structures, we aligned the helices (␣10 and ␣11) involved in Get4 binding from our intermediate complex onto the inhibited complex. Both structures have overlapping binding sites and predominantly utilize the N terminus of Get4 for the interaction (Fig. 4B). The difference in orientation involves a rotation centered on ␣2 of Get4. This rotation allows Get4 ␣2 to form the majority of the interactions in the inhibited complex, whereas Get4 ␣1 is responsible for many interactions in the intermediate complex. Based on the multistep assembly observed from kinetic analysis, we believe that the new structure dominated by electrostatics represents the intermediate Get3⅐Get4/5 complex (Fig. 1H, species ii) prior to rearrangement to the final structure stabilized by hydrophobic interactions.
Mutational Analysis of the Intermediate Complex-The intermediate complex interface involves numerous electrostatic residues on both Get3 and Get4. Of these interactions, only Get3 Asp 263 -Get4 His 51 and Get3 Glu 307 -Get4 Lys 23 are unique to the intermediate interface and are not found in the inhibited complex (Fig. 5A). The first interaction (Asp 263 -His 51 ) is not conserved; however, in higher eukaryotes, there is conservation of a likely salt bridge (Fig. 5A). The conserved Get3 Glu 307 also interacts with the cytoplasmic domain of Get2 (16,17), whereas Get4 Lys 23 is not conserved (Fig. 5A). Substitution of Get3 Asp 263 leads to a loss of observable binding to Get4/5N in isothermal titration calorimetry measurements (Fig. 5, B and C), whereas substitution of Get3 Glu 307 had a more modest loss in affinity (Fig. 5, B and C).
If the intermediate complex is on-pathway to the final Get3⅐Get4/5 complex, one would predict that disrupting the interactions unique to this intermediate would slow down Get3⅐Get4/5 association. Disrupting the putative Get3 Glu 307 -Get4 Lys 23 interaction (E307A) resulted in similar association rates compared with wild type Get3 and, together with the isothermal titration calorimetry results, suggests that the interaction is not essential for complex formation (Fig. 5, B-D). In contrast, significantly reduced rates of Get3⅐Get4/5 association were observed with Get3 D263A (Fig. 5, B-D). Although the association rates are slightly higher than previously reported, the differences are small (ϳ2-fold) and do not change the interpretation. Both mutants had virtually identical dissociation rates compared with wild type Get3; therefore, the difference in affinity for D263A is specific for complex formation kinetics (Fig. 5, B-D).
If Get3 D263A follows the same assembly pathway as wild type Get3, one would expect that once the Get3 D263A⅐Get4/5 complex is formed it has the same equilibrium stability as the wild type complex because Asp 263 does not participate in interface interactions in the final complex (Fig. 5E). However, the weakened equilibrium stability of the Get3 D263A⅐Get4/5 complex indicates that the changes induced by D263A are more complex, and this mutant likely follows a different assembly pathway leading to a non-native complex. Because a reaction follows the fastest pathway, this observation in turn suggests that the D263A mutation slows the native assembly pathway more extensively than was experimentally measured. Collectively, mutational analysis of the intermediate complex pro- The Get3⅐Get4/Get5 Intermediate Complex DECEMBER 11, 2015 • VOLUME 290 • NUMBER 50 vides independent evidence that the structure observed here represents an on-pathway intermediate during Get3⅐Get4/5 assembly. Stoichiometry between Get3 and Get4/5-The stoichiometry of the Get3⅐Get4/5 interaction is not fully determined. A recent SEC-multiangle light scattering analysis using an ATPase-deficient Get3 (D57N) suggested that the complex contains one Get4/5 heterotetramer bound to one Get3 dimer (40). Using wild type Get3, SEC-multiangle light scattering analysis with equimolar concentrations of Get3 and Get4/5 confirmed that the size of the complex is consistent with a single Get4/5 heterotetramer bound to a single Get3 dimer (Fig. 6A) (23).
One puzzling observation arises from these data: despite the presence of two Get4 molecules in full-length Get4/5, there has been no evidence for binding of Get3 to the second Get4 in the Get4/5 heterotetramer. To directly test this asymmetry of Get4/5 during complex formation, we developed an alkylation protection assay. An engineered cysteine at the Get3⅐Get4/5 interface (Get4 S48C (24)) is allowed to react with N-ethylmaleimide in the absence and presence of various factors (Fig. 6B). In the free Get4/5, Get4 Cys 48 is solvent-exposed (25) and rapidly alkylates to completion (Fig. 6, B and C). Consistent with binding at a single Get4 interface, only 50% of Get4 Cys 48 was protected from alkylation by Get3 (Fig. 6C). This protection pattern was observed at Get3 concentrations nearly 1000-fold above the dissociation constant for the Get3⅐Get4/5 complex, indicating that the 50% protection did not arise from incomplete Get3⅐Get4/5 binding. These results strongly suggest that only one site of a Get4/5 heterotetramer is able to bind Get3.
To further support these results, accessibility of Get4/5 was probed using a 10-kDa PEG maleimide. A second site (Get4 Q34C) that is also occluded in the Get3⅐Get4/5N structure (Fig.  6D) was chosen to further validate the interface. Importantly, substitution of Gln 34 to alanine does not impair binding to Get3 (23). In this reaction, Get4 Q34C with PEG maleimide forms a covalent adduct that gives a 10-kDa increase in mass that can be detected by SDS-PAGE and Coomassie staining. In the absence of Get3, this reaction resulted in 100% PEGylation of Get4/5 over a 2-min time course, whereas addition of saturating Get3 resulted in 50% PEGylation (Fig. 6E). This provides corroborating evidence that only one Get4 molecule in a heterotetramer binds Get3. Together, these data indicate that once Get3 binds to Get4/5 the other Get4 molecule in the Get4/5 heterotetramer is inhibited from further interaction with another Get3 dimer. A . .  structural rearrangement to a more stable complex (24) (Fig.  2A). This conformational change is consistent with previous results showing that Get4/5 binding induces Get3 into an "occluded" state, leading to Get3 ATPase inhibition and delayed ATP dissociation kinetics (22). Based on this model, once Get3 has transitioned to the occluded conformation, complex dissociation would be insensitive to buffer ionic strength ( Fig. 1F) but dependent on nucleotide (Fig. 1G). The initial Get3⅐Get4/5 complex presented here (Fig. 3, A  and B) reveals a novel binding interface composed of electrostatic interactions (Fig. 3D). This structure provides the molecular basis for the initial salt-dependent association seen in our kinetic description of the Get3⅐Get4/5 interaction (Fig. 1). Although Get3 is in an open conformation, these structures likely represent the true on-pathway intermediate because the Get3-Get4 interface does not change between the open and closed forms. This is corroborated by the D263A mutation introduced within this interface that drastically slows down complex formation (Fig. 5). These data coupled with the kinetic experiments argue that these structures represent the initial intermediate complex during Get3 and Get4/5 binding.
As we reported previously, Get4/5 is able to precisely discriminate between nucleotide states of Get3, enabling Get4/5 to regulate Get3 activity and prime it for efficient capture of the TA substrate (22)(23)(24). The structure described here provides evidence for an additional interaction dominated by electrostatic interactions and characterized by fast association rates. This ultrafast diffusion-limited association of Get4/5 to Get3 (24) may function as an additional mechanism for Get4/5 to select for the appropriate conformational state of Get3 or inversely enables Get3 to constantly sample numerous Get4/5 until transfer of TA substrate occurs. Interestingly, this may