|Home | About | Journals | Submit | Contact Us | Français|
In the cytoplasm, the correct delivery of membrane proteins is an essential and highly regulated process. The post-translational targeting of the important tail-anchor membrane (TA) proteins has recently been under intense investigation. A specialized pathway, called the GET pathway in yeast and the TRC pathway in vertebrates, recognizes ER targeted TA proteins and delivers them through a complex series of handoffs. An early step is the formation of a complex between Sgt2/SGTA, a co-chaperone with a presumed ubiquitin-like-binding domain (UBD), and Get5/UBL4A, a ubiquitin-like domain (UBL) containing protein. We structurally characterize this novel UBD/UBL interaction for both the yeast and human proteins. This is supported by biophysical studies that demonstrate that complex formation is mediated by electrostatics generating an interface that has high-affinity with rapid kinetics. In total, this work provides a refined model of the interplay of Sgt2 homologs in TA targeting.
Two homologous pathways have been elucidated for the targeting of TA proteins to the ER (recently reviewed in (Chartron et al., 2012a; Denic, 2012; Hegde and Keenan, 2011)). The best characterized is the fungal GET pathway (Guided Entry of TA proteins). In its simplest form, a TA sorting complex comprising Sgt2, Get4, and Get5 (alternatively name Mdy2) facilitates the loading of a TA substrate onto the Get3 ATPase. The Get3/TA complex is targeted to the ER where the TA is released for insertion by the membrane proteins Get1 and Get2. Vertebrates have a related system, referred to as the TRC pathway (Transmembrane domain Recognition Complex). In this case, the sorting complex similarly contains the proteins TRC35 and UBL4A (alternatively named GDX), homologs of Get4 and Get5 respectively; however, they form a three-component complex with the protein BAG6 (alternatively named BAT3 or Scythe), referred to as the BAG6 complex. From here, the TA is handed to a Get3 homolog, TRC40, which delivers the protein to WRB, a homolog of Get1.
Sgt2, a heat-shock protein (HSP) co-chaperone, facilitates the first committed step in TA protein targeting. It has been linked genetically and physically to the GET pathway in multiple studies (Battle et al., 2010, Constanzo et al., 2010; Liou et al., 2007; Wang et al., 2010). It recruits a variety of HSP families via an internal tetratricopeptide repeat (TPR) domain (Chartron et al., 2011; Wang et al., 2010). The Sgt2 C-terminal domain can bind to sequences of six or more hydrophobic residues (Liou and Wang, 2005) including ER-destined TA proteins, which are then handed to Get3 dependent on the Get4/Get5 complex (Wang et al., 2010). Mitochondrial TA proteins, which can also co-purify with the Sgt2/Get4/Get5 complex, are associated with bound HSPs and are not transferred to Get3 (Wang et al., 2010). The role of Sgt2 in sorting between target organelles is supported by the mislocalization of ER resident TA proteins to the mitochondria in Δsgt2 cells (Costanzo et al., 2010). The C-terminal domain, rich in asparagine, glutamine and methionine, contains only a short conserved sequence and is weakly predicted as helical. A small N-terminal homodimerization domain (Sgt2-N) mediates the association with a single copy of Get5, providing the link to the rest of the GET pathway (Kohl et al., 2011; Chang et al., 2010; Chartron et al., 2011; Liou et al., 2007). By immunoprecipitation from yeast lysate, the majority of Get5 is associated with Sgt2 (Wang et al., 2010).
Get4 and Get5 form the adaptor required for the transfer of TA proteins from Sgt2 to Get3 (Wang et al., 2010). Get4 forms a complex with the N-terminal domain of Get5 and sequesters a nucleotide-bound Get3 (Bozkurt et al., 2010; Chang et al., 2010; Chartron et al., 2010). Get5 has a central ubiquitin-like domain (Get5-UBL) that binds Sgt2-N, and a C-terminal homodimerization domain, resulting in an extended Get4/Get5 heterotetrameric complex (Chartron et al., 2010; Chartron et al., 2012b). The complex between Get5 and Sgt2 can be disrupted in vitro by a pair of mutations to the UBL domain that also lead to incomplete rescue of Δget5 growth defects under stress conditions (Chartron et al., 2010). Moreover, the TA protein transfer reaction is competed by excess Get5, which cannot alone form a productive complex with Get3 (Wang et al., 2010). These results underscore the importance of the physical interaction between Sgt2 and the Get4/Get5 complex. The details of the transfer of TA proteins, including energetic requirements and the in vivo stoichiometry of Sgt2, Get4/Get5 and Get3 over the course of the hand off, remain to be established.
SGTA, the human homolog of Sgt2, associates with the BAG6 complex through its N-terminus (Winnefeld et al., 2006) and UBL4A is postulated to bridge this interaction (Chartron et al., 2012a; Hegde and Keenan, 2011). The BAG6 complex is required for loading TA proteins onto TRC40 (Leznicki et al., 2010; Mariappan et al., 2010) along with being involved in the degradation of membrane proteins in the cytoplasm (Hessa et al., 2011; Minami et al., 2010; Wang et al., 2011).
In this report, we use a combination of structural biology and biochemistry to define the interaction between Sgt2 and Get5, conserved from yeast to humans. We present the first structures of the N-terminal homodimerization domains from two homologs, yeast Sgt2 and human SGTA, and characterize their fold as a new class of UBD. Further, we solve the structure of the central domain of Get5, demonstrating it as a novel UBL. Finally, we determined the structure of the complex between an Sgt2-N homodimer and the Get5-UBL, revealing an interaction strongly influenced by electrostatics. We use a variety of methods to demonstrate that this interaction has high-affinity with rapid binding kinetics. This provides critical context for understanding the Sgt2/Get4/Get5 complex. Similarly, this work provides a mechanism for membrane protein entry in the seemingly more complicated mammalian system.
The Sgt2-N domain is both a homodimerization domain and a binding platform for the Get5-UBL domain. This provides the physical link between the Sgt2 chaperone complex and the GET pathway. Sgt2-N does not have sequence homology to known structures or other characterized UBDs. To understand how these dual functions are accomplished, we first determined the structure of the S. cerevisiae Sgt2-N homodimer using solution NMR. Statistics for NMR structure calculations are in Table S1. A monomer of Sgt2-N consists of three helices (Figure 1A–C). The first two helices are of similar length and mediate homodimerization, forming a four-helix bundle with two-fold symmetry, consistent with the postulated coiled-coil (Tobaben et al., 2003). A third, shorter helix packs against either side of the bundle away from the dimer interface. The residues C-terminal to this helix, which form the linker to the TPR domain, give weaker signals than the helical region in all NMR experiments and only partial chemical shift assignments were possible. This is due to different rates of motion relative to the rest of the protein rather than proteolysis, as signals for the terminal residues are observed. The few NOE derived contacts observed in this sequence restrain it in partially folded conformations (Figure 1C).
The symmetry axis places the equivalent helices from each subunit head-to-tail resulting in two unique surfaces. We designate the surface composed of the α2 helices the “Get5 binding surface”, and the opposite composed of the α1 helices the “α1 surface”, (Figure 1B & C, S1A & B). The dimer contacts between α2 helices are made by conserved small residues (Ser32, Ala36, Cys39, and Ala43), resulting in close packing between the main chains of the two subunits (Figure 1B). Cys39 and Val35 make a small hydrophobic patch at the center of the Get5 binding surface (Figure 1A, D). The partially exposed Cys39 is strictly conserved across eukaryotes and Val35 is fully solvent exposed, conserved as either valine or isoleucine (Figure 1G). Despite the close proximity of the Cys39 sulfhydryl groups (5.3 ± 0.4 Å sulfur-sulfur distance), the Cβ shifts of 27.819 ppm argue against disulfide bond formation, even after incubation for several months (Sharma and Rajarathnam, 2000). The conserved acidic residues Asp28 and Glu31 and the conserved positioning of Glu42, Glu47 and Glu49 result in a negatively charged ring surrounding the hydrophobic patch (Figure 1A & E).
The dimer contacts between α1 helices are held further apart than the α2 helices by the interlocking of large hydrophobic side chains (Figure 1B & S1A). This results in an exposed hydrophobic face at the α1 surface that is protected, to some extent, from solvent by the partially folded carboxyl terminal linker (Figure 1C & S1B). When an Sgt2-N variant with this linker deleted is purified, it forms a higher order oligomer, likely due to aggregation at this face (data not shown).
We also investigated the N-terminal domain of human SGTA (SGTA-N). The domain was co-expressed and purified with an affinity-tagged UBL domain of UBL4A (UBL4A-UBL), demonstrating a complex that could be separated from excess UBL4A (Figure S1C). We attempted to crystallize this complex; however, in the three resulting distinct crystal structures, only density for SGTA-N was observed. Structures were determined by molecular replacement using Sgt2-N at 1.35-1.45 Å resolutions. The structures differ predominantly by a post-purification oxidation of the conserved cysteine with buffer components (Figure S1E & F). Crystallographic data collection and refinement statistics are provided in Table S2. SGTA-N shares the four-helix bundle topology with Sgt2-N (Figure 1F). As expected from the Sgt2-N solution data, the cysteine sulfhydryl groups do not form a disulfide bond between subunits.
As was the case for structures of the Sgt2/SGTA-TPR domains (Chartron et al., 2011; Dutta and Tan, 2008), Sgt2-N and SGTA-N have very similar architecture with an RMSD of 1.24 ± 0.07 Å over equivalent Cα atoms (Figure 1F). The sequences have high homology (Figure 1G) and all of the general features are conserved including the hydrophobic patch surrounded by charge at the binding face (Figure 1F & S1D). One difference is that SGTA-N does not have a third α-helix. The residues that would correspond to α3 are disordered or involved in non-physiological crystallographic contacts.
The solution NMR structure of the UBL4A-UBL domain showed that it has the expected ubiquitin fold (Figure 2A) (PDBID:2DZI, RIKEN Structural Genomics Initiative). Get5-UBL has several small sequence insertions suggesting some structural differences (Figure 2F); therefore, to fully characterize the yeast system, we determined the structure of the Get5-UBL domain. Initially, we solved a structure by solution NMR (Figure S2A). Simultaneously, we obtained crystals of Get5-UBL that diffracted to 2.4 Å resolution. We were able to obtain phases by molecular replacement using the Get5-UBL solution structure. Three copies of Get5-UBL were present in the asymmetric unit (Figure 2B & S2B). They have an average RMSD of 0.75 Å over main chain atoms. Statistics for the solution structure calculations and crystallographic data collection and refinement are presented in Tables S1 and S2, respectively. The solution and crystal structures are very similar with an average main chain RMSD of 1.23 Å (Figure S2A), with most variation in Loops 1 and 5.
While overall the structures from Get5 and UBL4A are similar, there are a few differences due to insertions (Figure S2C). Based on the structures and sequences of other animal homologs, Get5-UBL has a two-residue insertion around Pro84 in Loop 1. Pro84 is cis allowing for a tight turn, and there is no detectable cis to trans isomerization by NMR. There are two additional insertions in Get5-UBL, His113 in Loop 3, and Ala141 that causes a short coil-like turn to extend at the end of Loop 6. Despite these differences, the surface elements on the face of the β-sheet are conserved between Get5 and UBL4A (Figure 2A & B).
The UBL of Get5 and UBL4A have features that distinguish them from other UBLs with high homology to ubiquitin. UBLs occur either as independent units known as Type I UBLs that can be conjugated onto other proteins or as Type II UBLs that are domains in larger proteins that frequently mediate binding to the proteasome (Jentsch and Pyrowolakis, 2000). Get5 falls into the latter class but does not associate with the 20S proteasome or poly-ubiquitin chains (Hu et al., 2006; Saeki et al., 2002). Therefore, one would expect that there are features that distinguish Get5 homologs from other UBLs. Comparing Get5-UBL to ubiquitin, the two most significant structural differences are the conformations of Loop 1 and Loop 6 (Figure 2C). Get5 binds Sgt2 via its UBL through a hydrophobic patch formed by the conserved Leu120 (Chang et al., 2010; Chartron et al., 2011), which is equivalent to the conserved Ile44 of ubiquitin that forms the “I44 patch” (Hicke et al, 2005). In Get5, the conformations of Loop 1 and Loop 6 significantly reduce the size of the patch compared to ubiquitin (Figure 2D). Moreover, the surface around the I44 patch has a positive charge for both proteins; however, it is significantly more pronounced in Get5 (Figure 2E).
We previously demonstrated that only a single copy of Get5-UBL binds to dimeric Sgt2-N and the double mutation L120A/K122A prevents complex formation (Chartron et al., 2011). We decided to probe the interaction further using isothermal titration calorimetry (ITC) of wild type and mutant proteins (Figure 3 & S3). All variants behaved similarly to wild type during purification (data not shown). Sgt2-N and Get5-UBL interact with 10−8 M affinity independent of which protein is used as a titrant. The Get5-UBL L120A mutation has a thousand-fold lower binding affinity consistent with the significance of this position in UBLs. The Get5-UBL L120I mutant bound Sgt2-N with similar affinity to wild type. This is surprising considering that a leucine at this position is completely conserved, in contrast to the isoleucine in most UBLs. Mutations of three other nearby hydrophobic residues that compose part of the hydrophobic patch, G123Y, V125A, and M147A also significantly lowered the binding affinity.
The complementary surface charge of the two proteins suggests that a significant component of the interaction involves electrostatics. As expected for this type of binding, affinity drastically decreases as salt concentration increases. Moreover, mutations of any of the lysines on this face of Get5-UBL to alanine (79, 118, 122, and 124) reduce binding affinity 5–10-fold. Interestingly, ubiquitin, which contains most of the residues tested, binds Sgt2-N with negligible affinity (data not shown).
We made reciprocal mutations to the conserved face of Sgt2-N and tested their binding to Get5-UBL (Figure 3). Similar to Get5, mutation of the two exposed hydrophobic residues to alanine had the strongest effect with over 100-fold lower affinity (V35A and C39A). The presence of the completely conserved cysteine was curious. We decided to do a series of typically minimal changes at this position to test for effect on complex formation. A slightly bulkier hydrophobic side chain, C39V, resulted in a 100-fold lower affinity while removing the sulfhydryl to a smaller amino acid, C39A, resulted in a nearly 300-fold lower affinity. The strongest effect was conversion of the sulfhydryl to the more polar hydroxyl, C39S, resulting in a ~700-fold lower affinity. Alanine mutations of the acidic residues that comprise the charged face have a similar effect on affinity as mutations of the basic residues of Get5-UBL. One exception is that mutation of the peripheral Glu47, located at the beginning of α3, had no effect on affinity.
Multiple lines of evidence demonstrate complex formation between Get5 and Sgt2 (Chang et al., 2010; Chartron et al., 2011; Liou et al., 2007; Wang et al., 2010) and the complex formed by the minimal Get5-UBL and Sgt2-N domains appears stable by size exclusion chromatography. Based on ITC, the complex has high-affinity; however, investigations using co-immunoprecipitation find variable amounts of Sgt2 or SGTA associated with the Get5/UBL4A partners. Although there are many possible reasons why the in vivo stoichiometry is variable, we hypothesized that fast binding kinetics could explain how co-precipitation might be dependent on experimental conditions. To measure the association and disassociation rate constants we turned to surface plasmon resonance (SPR). In this experiment, polyhistidine-tagged Get5-UBL or Get5-UBL-C was immobilized and Sgt2-N was used as the analyte (Figure 4A). Get5-UBL-C includes the C-terminal dimerization domain of Get5 that is separated from the UBL domain by a flexible linker (Chartron et al., 2010). We previously demonstrated that the less restrictive Get5-UBL-C dimer can bind two Sgt2-N domains (Chartron et al., 2011); therefore, we expect that each subunit will act independently. The proteins were well behaved on the chip giving stable concentration dependent saturation (Figure 4B). The rapid saturation of response after analyte injection (<1 s) and then rapid reduction after the injection is stopped are indicative of fast on and off-rates, consistent with our hypothesis.
The equilibrium dissociation constant can be calculated by plotting response units as a function of Sgt2-N concentration after response units reach equilibrium (Figure 4B-D). For Get5-UBL-C, the plot fit to a Kd of 7.49 × 10−7 M. Compared to ITC, this is nearly two orders of magnitude lower affinity. Although the different techniques are not expected to give identical results due to experimental conditions, the relative results from mutants are consistent. We suspect that steric constraints based on interactions with the antibody affect the measured rates. This is seen when we used immobilized Get5-UBL, which is expected to bring the Sgt2 binding face into closer proximity to the immobilizing antibody. This set-up had consistently lower binding affinities, for example wild type binding is reduced a further five-fold to a Kd of 3.78 × 10−6 M. Mutants in Sgt2-N or Get5-UBL show similar changes in binding affinity compared to ITC. Charged mutants Get5-UBL K124A, and Sgt2-N D28A and D31A all had similar affinities on the order of 3–4-fold weaker than the wild type protein. Also, residues in the hydrophobic interface had the strongest effect, reducing the affinity of Get5-UBL L120A or Sgt2-N C39A to below what could be accurately determined in this experimental set-up.
The importance of charge complementarity is consistent with two preformed interfaces that are electrostatically steered toward complex formation. These types of protein interactions are known to have fast binding kinetics (Sheinerman et al., 2000). Here, the association phase of the SPR data were used to determine a k of 1.06 × 107 M−1 s−1 on and a dissociation rate constant of ~4 s−1 (Figure 4E). The dissociation rate constant could be independently estimated from the equilibrium Kd and kon values, which gives an approximate koff of 7.9 s−1. Consistent with both the electrostatic mechanism and the ITC data, increasing salt concentration from 100 to 300 mM salt resulted in nearly an order of magnitude change in Kd (Figure 4D). All of this data points to a highly specific interface that has rapid on- and off-rates.
For independent verification of the kinetics, we measured exchange rates by NMR using EXSY spectra (Perrin and Dwyer, 1990). A 2:1 ratio of 15N-labeled Get5-UBL to unlabeled Sgt2-N homodimer was prepared, allowing detection of approximately equimolar free and complex forms of Get5-UBL. The complex dissociates and reforms within the time scale of the experiment (mixing times, τm, of 60 or 120 ms). As a result, exchange cross peaks are observed for every proton that has a change in chemical shift between the two states of Get5-UBL (Figure 4F & S4A). The side chain amide protons of Gln82 and the main chain amide protons of Val125, Asn129, Met147, Ile148 and Lys149 produce exchange peaks with enough resolution for volume integration. The magnitudes of these peaks were used to calculate exchange rates (Figure S4B). Overall exchange rates of 7.5 ± 2.5 s−1 (τm = 60 ms) or 5.7 ± 2.4 s−1 (τm = 120 ms) were obtained and represent koff,,NMR. Using the ITC-derived Kd of 34 nM, kon,NMR is approximately 2.0 × 108 M−1 s−1, 20-fold faster than measured by SPR. Because the solution and immobilized Get5 off-rates are similar, the lower binding affinity seen by Get5 immobilization in the SPR experiment is from a slower kon. Finally, a high-affinity complex with fast kinetics is consistent the single peak seen by size exclusion chromatography (Stevens, 1989).
Compared to the proteins alone, the NMR spectra of the Sgt2-N/Get5-UBL complex have a dramatic reduction in resolution. This is a result of peak broadening, due to both slower tumbling and the fast kinetics of complex formation and dissociation. Therefore, rather than generate a uniformly 13C/15N-labeled sample of 235 residues, we opted to investigate two asymmetrically labeled complexes to reduce the amount of chemical shift overlap. Chemical shift perturbations (CSP) on the 2D 1H-15N-HSQC spectra were examined for both proteins (Figure 5A). On Sgt2, the most drastic CSPs occur at the Get5 binding surface, with very little change occurring in the rest of the protein (Figure 5B). Binding of a single Get5-UBL is anticipated to break the symmetry of Sgt2-N, and indeed several residues on the Get5 binding surface split into two cross peaks with reduced peak height (Figure 5A & S5A). Residues without perturbation maintain a single cross peak, indicating that symmetry remains away from the binding site. The Get5-UBL domain is more broadly affected by binding to Sgt2-N, but the most intense CSPs occur at Ile81 and the loop consisting of residues 123–129, including Gly123 and Val125.
Inspection of NOESY spectra failed to conclusively identify enough new cross peaks resulting from intermolecular contacts to determine the structure by NOE distance restraints alone. This is not surprising, based on the presumed interface. Electrostatic interactions between a glutamate or aspartate and a lysine yield weak proton NOE cross peaks. Additionally, the expected hydrophobic interactions occur in crowded regions of the spectra. We proceeded to determine the individual solution structures of Sgt2-N and Get5-UBL as they are in complex. Data collection for the full assignments of the proteins in complex was performed at 37°C, which reduced the severity of line broadening effects. This had the additional effect of averaging the two states of Sgt2-N; we therefore treated Sgt2-N as symmetric in these calculations.
Overall, the structures of Sgt2-N and Get5-UBL while in complex are not significantly different from the free solution structures (Figure S5B & C). A few intermolecular NOEs could be identified and were used in the calculation of the complex structure (Figure 5D). In the absence of substantial numbers of NOE-derived distance restraints, the structures of complexes can be determined by molecular docking driven by other experimental data introduced as ambiguous interaction restraints (AIRs) (Dominguez et al., 2003). We defined 14 AIRs using the CSP and mutagenesis data, the details of which are provided in the Extended Experimental Procedures. Moreover, we collected residual dipolar couplings that restrict rotational freedom between the models during docking. We performed docking between the two separate complex structures to generate a full model of Sgt2-N and Get5-UBL (Figure 5E). Statistics for the structure calculation of each component, as well as the complex, are provided in Tables S3 and S4.
The experimentally restrained docking returns a well-converged structure where the predicted binding faces interact (Figure 5E). All of the residues that were identified to be involved experimentally are found at the interface (Figure 5F). The interface is comprised of the hydrophobic patch on Get5-UBL (Ile81, Leu120, Val125, Gly143 and Met147) that docks against the reciprocal patch that contains two each of Cys39 and Val35 from the Sgt2-N dimer (Figure 5G). Additionally, Thr145 packs against one copy of Val35. The conserved lysines 79, 85, 118, 122 and 149 all make electrostatic contacts to the charged face of Sgt2-N (two each of aspartates 28, 31, 38 and Glu42).
Ubiquitin and UBL domains are abundant in cells and all share common features (Winget and Mayor, 2010). Perhaps more abundant are the different UBD motifs found in a wide variety of frameworks (Hicke et al., 2005; Husnjak and Dikic, 2012). The combinatorial use of UBLs and UBDs results in a wide diversity of interactions that are utilized in many different contexts. The complex between Sgt2 and Get5 introduces a novel interaction.
As is typical for the most commonly observed UBD interactions, Sgt2-N binds at the face that contains the I44 patch (Figure 6A & B). Different from other UBDs, Sgt2 uses a symmetrical dimer interface to interact with its UBL. The only other example of a UBD dimer is the swapped CUE motif found in Vps9 that binds similar to other CUE domains (Figure 6B, PDBID:2P3Q) (Prag et al., 2003). Most other of the characterized UBDs bind primarily with a single polypeptide. Comparing UBD binding interfaces, Sgt2 buries a relatively large surface area (682.8Å2) (Figure 6A & B). The next closest interface for a yeast UBD is that of Ufd2 bound to the UBL of Rad23, which has an interface of 614.3Å2 (Hänzelmann et al., 2010).
Most of the I44 patch-binding UBDs use α-helical motifs, as does Sgt2. These proteins interact with similar groups of residues on their respective UBLs (Figure 6C and D). UBLs have a number of conserved residues around the I44 patch. For ubiquitin, these include: Leu8, Arg42, Ile44, Gly47, His68 and Val70. These interactions are conserved on Get5 in the interface with Sgt2 (Ile81, Lys118, Leu120, Gly123, Thr145 and Met147). For these residues, most are similar in nature to the canonical residues except for Thr145, which is a histidine in most UBLs.
Of the Sgt2/Get5 interactions, the most interesting are the residues that are unique in the interface relative to all other UBLs (Figure 6C & D). Four residues fit this description. The most provocative, the highly conserved Ile44 of ubiquitin, is a leucine in all Get5 homologs. Surprisingly, mutating this to isoleucine had no affect on binding affinity (Figure 3). The second is Gln82 in Get5, which forms a conserved H-bond network with negative charges on Sgt2. In ubiquitin, this residue is a threonine that likely cannot contribute to a similar network and, in fact, is pointed away from the I44 patch. Next is Lys122, a positive charge that is conserved in Get5 homologs adjacent to the I44 patch yet is missing in other UBLs. This lysine forms salt bridges with the conserved Asp28 and Asp31 on Sgt2. In ubiquitin, this position is an alanine that points its side chain away from the interface. The final residue is Thr145 whose equivalent in ubiquitin is a histidine (His68). In UBL4A, the position is an asparagine, a more polar residue. For ubiquitin, His68 is typically described as a component of the hydrophobic pocket lining the edge of the I44 patch. For the Get5/Sgt2 complex, the smaller threonine is likely required to accommodate the tight interface.
The biogenesis of TA membrane proteins requires sorting in the cytoplasm followed by targeting to the membrane then subsequent insertion into the bilayer. For TAs destined for the ER, the conserved GET pathway governs this process. Sgt2 and Get5 are members of the so-called yeast TRC that includes Get4 and HSPs. This complex is responsible for binding and then sorting of ER-destined TAs to the targeting factor Get3 (Wang et al., 2010). Sgt2 contains three domains (Figure 7A). The C-terminal domain binds hydrophobic peptides with varying affinities. Mitochondrial TA proteins are typically less polar compared to ER-destined substrates and it is thought that this feature allows Sgt2 to selectively bind the latter with higher affinity (Wang et al., 2010). The central domain contains three TPR repeats that bind multiple classes of HSP proteins whose structure was recently solved (Chartron et al., 2011). At the N-terminus is a homodimerization domain that we report the first structure of here. In solution, Sgt2 forms an extended dimeric complex with the C-terminal domain moving freely at the end. Get5 forms an obligate heterotetramer with Get4 (Figure 7A). Get5 contains three domains: an N-terminal domain that wraps around Get4 forms the heterodimer interface (Chang et al., 2010; Chartron et al., 2010), a C-terminal domain forming a small, stable dimerization motif whose structure was also recently solved (Chartron et al., 2012b) and the central domain is a novel UBL domain whose structure we report for the first time here. The Get4/Get5 complex is also extended in solution (Chartron et al., 2011; Chartron et al., 2010).
The initial identification of Get4 and Get5 as bona fide members of the GET pathway did not reveal the connection to Sgt2 (Jonikas et al., 2009). In hindsight, this is surprising as previous biochemical and genetic links had been reported (Liou et al., 2007). Subsequent studies clearly linked the N-terminal dimerization domain of Sgt2 to the UBL domain of Get5, solidifying the role of Sgt2 in TA targeting (Battle et al., 2010; Chang et al., 2010; Chartron et al., 2011; Wang et al., 2010). We previously demonstrated that the complex formed was stable enough for co-purification, yet the interface was sensitive to mutation (Chartron et al., 2011; Chartron et al., 2012b). The structures reveal a strong electrostatic component to the interface between Get5 and Sgt2, which results in a complex with fast on- and off-rates.
The affinity of Sgt2 to Get5 is remarkably high (Figure 3). In the cell, every Get4/Get5 heterotetramer will, on average, be bound by an Sgt2, as one falls off another quickly replaces it, consistent with experimental results (Wang et al., 2010). Our previous work demonstrated that only a single Sgt2 dimer could bind to the Get4/Get5 heterotetramer at one time (Chartron et al., 2012b); therefore, Get5 has two potential binding sites for Sgt2 that are rapidly sampling. This makes sense in a model for TA targeting (Figure 7B), as a Get3/Get4/Get5 complex would be stable in the cytoplasm. This complex would screen, over multiple rounds, Sgt2 proteins to find one that stably bound an ER-destined TA protein. This interaction would lead to a hand off of the TA to Get3, which would then be released from Get4 to find its ER receptors. How Sgt2 finds TA proteins is a matter of conjecture. The simplest model is that it captures free TAs in the cytoplasm; however, it remains seductive to imagine that the highly abundant HSPs provide at least one route into the pathway. In that context, Sgt2 would bind HSPs transiently allowing for multiple rounds of binding to find appropriate substrates. Once a TA protein was bound it would be a stable complex that could be found by Get4/Get5.
In metazoans the picture becomes more complicated (Figure 7C). In addition to homologs for the yeast GET proteins, TA targeting includes a large multidomain protein called BAG6. BAG6 contains an N-terminal UBL domain that has features characteristic of typical UBLs (Figure 6D) and a C-terminal BAG domain. It is linked to TA targeting by forming a stable complex with TRC35 and UBL4A (Mariappan et al., 2010). TRC35 and UBL4A lack the features necessary for direct complex formation in yeast (Chartron et al., 2012a; Chartron et al., 2010) and it is likely that they both bind BAG6 directly. The first structure solved from this complex is the UBL domain from UBL4A, which has not been described in the literature (PDBID:2DZI). Although it has yet to be experimentally demonstrated, it seems very likely that SGTA performs a similar role in TA targeting to its yeast counterpart (Figure 7C). The structure of the dimerization motif of SGTA supports this, as it is a highly conserved domain, like UBL4A-UBL, with all of the features that are important for heterodimer formation. BAG6 is demonstrated to be a dimer and forms a complex with SGTA via the UBL domain of UBL4A (Yihong Ye, personal communication). This then would mirror all of the components of the yeast system including two each of SGTA, UBL4A and TRC35 (Figure 7C).
In mammalian cells, the BAG6 complex is linked to the degradation of mislocalized membrane proteins and dislocated ER products (Hessa et al., 2011; Wang et al., 2011). TA proteins can be redirected to the degradation pathway suggesting that the two pathways intersect (Hessa et al., 2011). SGTA has now been shown to be an important component for targeted degradation of ERAD substrates transiently binding hydrophobic membrane proteins prior to hand-off to BAG6 (Figure 7C) (Yihong Ye, personal communication). This additional role would also benefit from rapid sampling of SGTA to the BAG6 complex, raising the possibility that, in yeast, Sgt2 could have multiple roles as well.
Where the proteins are linked to degradation pathways, it seems plausible that the additional role for UBLs may have co-evolved. This is clear for the BAG6-UBL, which has all of the features of UBLs involved in degradation and is critical for targeting to the proteasome (Hessa et al., 2011; Wang et al., 2011). The Get5/UBL4A-UBL represents a unique class with features that clearly distinguish them from other UBLs (Figure 6). In fact, when the conserved histidine of the BAG6-UBL is converted to an asparagine, it completely loses its ability to bind to standard UBDs (Yihong Ye, personal communication). A similar effect is not seen when the conserved Leu120 is replaced by an isoleucine, although this could be a change to prevent other UBDs from binding Get5.
Sgt2 and SGTA are novel UBDs that have very specific binding partners. An interface dominated by electrostatics is unique amongst UBL/UBD complexes. This presumably allows high-affinity, rapid binding while strongly rejecting unfavorable interactions with other UBLs and UBDs. One interesting side note is that in mammals, both UBL4A and SGTA have tissue specific isoforms, Ubl4B and SGTB (Tobaben et al., 2003; Yang et al., 2007). While SGTA and SGTB have similar conserved sequence elements, UBL4B is missing a number of the residues that likely form the SGTA/UBL4A interface (Figure 6D). This suggests it may have lower affinity or perhaps an unknown UBD.
In this report we demonstrate a novel, conserved UBD/UBL interaction critical for TA targeting. The complex is in a dynamic equilibrium that allows for rapid sampling of the various components. This attribute is likely essential for the various roles that Sgt2 homologs must play. This work opens the door to understanding the steps of target selection and discrimination that are required for a regulated process. The finer details of the process of TA targeting continue to be resolved at a rapid pace; however, each new insight leads to unexpected elaborations. With the recent link to regulated proteolysis, TA targeting is becoming part of the greater picture of homeostasis in the cell.
Sgt2-N and Get5-UBL were overexpressed and purified from E. coli. Chemical shift assignments were made using uniformly 13C/15N-labeled samples. Data were collected on a Varian INOVA 600 MHz spectrometer with a triple resonance probe or a Bruker Avance 800 MHz with a TCI cryoprobe. Structures were calculated using NOE-derived distance restraints.
Double-labeled proteins were mixed with excesses of natural abundance protein. The asymmetrically labeled complexes were then purified by size exclusion chromatography. Chemical shifts were assigned and structures were calculated separately for each protein in the complex. Chemical shift perturbation and mutagenesis data were used to define AIRs that were used along with 14 intermolecular NOE distance restraints and RDCs to calculate the complete structure of the complex (Dominguez et al., 2003). Details of solution NMR experiments are provided in Extended Experimental Procedures.
Human SGTA-N and UBL4A-UBL were co-expressed in E. coli, purified by affinity and size exclusion chromatography. All crystallizations were performed by the vapor diffusion technique. Diffraction data were collected at beamline 12-2 at the Stanford Synchrotron Radiation Lightsource. Phases were recovered by molecular replacement using the solution NMR structures presented here. Additional details of crystal growth, data collection and model refinement are provided in Extended Experimental Procedures.
Data were collected using a MicroCal iTC-200 calorimeter. Details of sample preparation are provided in Extended Experimental Procedures. Proteins in the sample cell were at 30–100 μM and proteins in the injection syringe were concentrated to 800– 1000 μM. For each experiment, an initial injection of 0.4 μl was followed by 19 injections of 2 μl each. The cell was allowed to equilibrate for 120 seconds in between titrations. Data were processed using Origin v7.0 (OriginLab) software using a single-site model.
Sample and running conditions are described in Extended Experimental Procedures. Data were collected using a Biacore T-100 system upgraded to T-200 sensitivity (GE healthcare). Mouse anti-pentahistidine antibody (Qiagen) was covalently linked to a CM5 dextran chip using standard amide coupling chemistry. Hexahistidine tagged Get5-UBL (wild type or mutants) or Get5-UBL-C were then immobilized. Equilibrium binding analysis was performed using BIAevaluation software (GE healthcare). Kinetic analysis between wild type Sgt2-N and Get5-UBL was performed using Kaleidagraph (Synergy Software). The slope values ks at each concentration were determined by linear regression fitting of the association phase to the integrated 1st order rate equation. Values of ks were plotted against concentration and the association rate kon was determined from the linear fit to equation (1).
This fit results in a koff of 3.9 s−1, but since this may not accurately be determined by this method (Karlsson et al., 1991), we additionally estimate koff using the equilibrium dissociation constant as the ratio of koff to kon.
We thank Yihong Ye, NIH, for providing unpublished data and discussion. We thank Shu-Ou Shan, Douglas Rees, Axel Müller, Meera Rao, Christian Suloway and Michael Rome for critical reading of the manuscript. We thank members of the laboratory for support and useful discussions. We thank Graeme Card, Ana Gonzalez and Michael Soltice for help with data collection at SSRL BL12-2. We thank Robert Peterson and Juli Feigon for use of the 800 MHz NMR spectrometer at UCLA. We thank Jost Vielmetter, Harry Gristick, Claude Rogers and Abigail Pulsipher for assistance with SPR and ITC experiments. We are grateful to Gordon and Betty Moore for support of the Molecular Observatory at Caltech. Operations at SSRL are supported by the US DOE and NIH. W.M.C. is supported by NIH grant R01GM097572. The atomic coordinates, and structure factors or NMR restraints, have been deposited in the RCSB Protein Data Bank, www.pdb.org, with PDB ID codes 2LXB, 4GOD, 4GOE, 4GOF, 2LXA, 4GOC & 2LXC for the Sgt2-N (NMR), three SGTA-N (X-ray), Get5-UBL (NMR), Get5-UBL (X-ray) and Get5-UBL/Sgt2-N structures, respectively). Chemical shifts have been deposited in the Biological Magnetic Resonance Bank, http://www.bmrb.wisc.edu/, with accession numbers 18669, 18670, and 18671 for Get5-UBL, Sgt2-N, and Get5-UBL/Sgt2-N, respectively.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.