Get3 is a holdase chaperone and moves to deposition sites for aggregated proteins when membrane targeting is blocked

The cytosol of cells is a crowded aqueous solution enclosed by membranes. In this environment, the exposure of strongly hydrophobic molecules is dangerous because it can precipitate the aggregation of proteins that have not yet achieved, or transiently lost, their native state or are terminally misfolded. Different molecular chaperones assist with the shielding, refolding or targeting to degradation of proteins that transiently or inappropriately expose regions of hydrophobicity (Tyedmers et al., 2010; Hartl et al., 2011). In contrast, membrane proteins permanently and appropriately exhibit strongly hydrophobic segments that are eventually embedded in a cellular membrane. Co-translational membrane targeting minimizes the risk for membrane protein aggregation by coupling translation directly to membrane targeting (Shao and Hegde, 2011); however, there are many membrane proteins that are subject to post-translational targeting, e.g. mitochondrial membrane proteins, substrates of the post-translational Sec translocon or tail-anchored membrane proteins destined for the secretory pathway. For these proteins, translation is fully completed before they are targeted to the appropriate membrane and their biogenesis involves the transient exposure of nascent transmembrane segments in the cytosol.

Tail-anchored membrane proteins of the secretory pathway use the recently characterized GET machinery (Hegde and Keenan, 2011; Chartron et al., 2012) comprising several cytosolic factors, most prominently the Get3/TRC40 ATPase, and a receptor at the membrane of the endoplasmic reticulum (ER). Extensive biochemical and structural analysis delineated the following steps during tail-anchored protein biogenesis: engaging the tail-anchored protein when it emerges from the ribosome (Get4, Get5, Sgt2), shuttling the tail-anchored protein through the cytosol (Get3) and mediating its release from Get3 into the ER membrane (Get1, Get2). The nucleotide state of Get3 controls its conformation and hence its ability to interact with tail-anchored proteins. Repeated cycles of tail-anchored protein insertion via the GET pathway are coupled to the ATP hydrolysis cycle of the Get3 protein. Studies focusing on the localization of Get3 have visualized Get3-GFP as being localized to the cytosol as well as the ER during mid-logarithmic growth, consistent with its known role in tail-anchored protein insertion. However, in the absence of the ER membrane receptor comprised of Get1 and Get2, the GFP fusion protein accumulates in distinct foci that also contain mislocalized tail-anchored protein substrates (Schuldiner et al., 2008). Despite the aggregation potential of such precursors that fail to become correctly membrane integrated, the cellular pathways that modulate any resulting toxic species are poorly understood. In order to address the physiological relevance of Get3-positive foci as designated deposition sites for accumulating tail-anchored proteins, we have characterized the localization of Get3-GFP under various forms of cellular stress with the potential to perturb tail-anchored protein membrane integration.

Wild-type cells expressing Get3-GFP from its endogenous promoter were exposed to a variety of stressful conditions including oxidative and reductive challenge or treatment with CaCl₂. Amongst these Get3-GFP was only observed in foci under glucose starvation, i.e. after incubation in synthetic complete (SC) medium lacking a carbon source (−D; Fig. 1A). Get3-GFP localization to foci occurred within 60 minutes of glucose withdrawal (Fig. 1B) and reversed shortly after glucose re-addition (Fig. 1C), which led to the re-localization of Get3 into cytosol and perinuclear ER (Fig. 1D). Get3 re-localization into the cytosol also occurred when the synthesis of new proteins was blocked after glucose starvation and before glucose re-addition (Fig. 1E). Steady-state protein levels of Get3-GFP (Fig. 1F) or endogenous Get3 (Fig. 1G) were unaffected by glucose withdrawal or re-addition even in the absence of new protein synthesis, suggesting that Get3 shuttles in and out of the foci as cellular metabolism changes (Wilson et al., 1996; Ashe et al., 2000). Get3-HA behaved indistinguishably from the GFP fusion protein as determined by indirect immunofluorescence (Fig. 1H) excluding GFP-based artifacts. GFP fusions of the GET receptor proteins Get1 and Get2 did not change localization and remained stable under glucose starvation (Fig. 1I,J). Taken together this data suggests a specific response of Get3 to glucose availability.

Next, we tested whether Get3 entering foci depended on other GET pathway proteins (Fig. 2A,B) such as the two ER membrane receptor proteins Get1 and Get2 (Schuldiner et al., 2008), or the cytosolic GET proteins Get4, Get5 (Jonikas et al., 2009) and Sgt2 (Battle et al., 2010; Chang et al., 2010; Wang et al., 2010; Chartron et al., 2011; Kohl et al., 2011), which are implicated in loading Get3 with tail-anchored protein clients. We found that both, Get4 and Get5, were required for Get3-GFP to enter foci upon glucose withdrawal and that the upstream GET pathway components, GFP-Get5, which was necessary for Get3′s redirection to foci under glucose withdrawal (Fig. 2A,B), and Sgt2-GFP, which was not necessary (Fig. 2A,B) but can interact with Get5 (Chang et al., 2010; Wang et al., 2010; Chartron et al., 2011), also accumulated in Get3-positive foci (Fig. 2C). In contrast, absence of the ER membrane receptor proteins Get1/Get2 substantially increased the number of cells with Get3-GFP-positive foci (Fig. 2A,B), even in the presence of glucose. Since Get4/5 are implicated in substrate loading and Get1/2 are required for substrate release it seems plausible that Get3 is mostly loaded with clients in the foci. Consistent with this notion the number of Get3-GFP-positive foci was reduced to 15% of the number observed under glucose starvation alone when translation was inhibited at the onset of glucose starvation in wild-type cells (Fig. 2D).

We thus asked whether tail-anchored protein substrates that cannot be inserted colocalize with Get3 in the foci. To impair Get3-dependent membrane targeting and hence create a large population of non-inserted clients we employed a Δget1get2 deletion strain (Fig. 2E). Indeed, when we quantified the colocalization of the tail-anchored protein Cherry-Sed5, a client of the GET pathway, with Get3-GFP in foci we observed colocalization in medium containing glucose (11% of all foci) that became extensive upon glucose withdrawal (87% of foci; Fig. 2F). In conclusion, Get3 and tail-anchored proteins are found in cytoplasmic foci when membrane integration of the clients is not possible and glucose depletion significantly increases the colocalization.

Glucose starvation leads to a transient drastic reduction of cellular ATP levels that recover as the cell adapts to the lack of this carbon source (Wilson et al., 1996; Ashe et al., 2000). AMPK kinase is at the center of a major signal transduction pathway that mediates this adaptation. Thus we asked whether deletion of AMPK kinase subunit-encoding genes SNF1 or SNF4 affected Get3-GFP localization to foci upon glucose starvation (Fig. 3A). However, Get3-GFP shuttled into and out of the foci in a Δsnf1 or Δsnf4 deletion mutant when the cells were glucose-starved and re-exposed to glucose. This indicates that the major pathway of glucose sensing involving Snf1/Snf4 kinase does not regulate Get3-GFP localization to foci.

ATP hydrolysis is required for repeated cycles of tail-anchored protein insertion (Stefanovic and Hegde, 2007; Favaloro et al., 2008). To investigate whether ATP hydrolysis affected localization of Get3 to foci, we employed a Get3_D57E-GFP mutant variant, which was modeled on a prokaryotic Get3 homologue ArsA with a fivefold lower affinity for Mg²⁺ and a drastically reduced ATPase activity (Zhou and Rosen, 1999). Interestingly, the Get3_D57E variant, which has 8.5-fold lower ATPase activity as compared to wild-type Get3 (Fig. 3B) but is unaffected in its ability to bind tail-anchored protein clients (Fig. 3C), localized to the foci even in the presence of glucose (Fig. 3D). Similarly, and in contrast to wild-type Get3-GFP (Fig. 2E,F), colocalization of Get3_D57E-GFP and the tail-anchored protein Cherry-Sed5 in a Δget1get2get3 deletion strain was hardly sensitive to glucose availability (Fig. 3E,F). Hence the reduced ability to hydrolyze ATP may direct Get3 to foci or trap it there whereas the interaction with the ER membrane receptor was shown to be unaffected when D57 was mutated (Mariappan et al., 2011). As for wild-type, Get3 localization to foci was dependent on Get5 (c.f. Fig. 3D; Fig. 2A,B) implying that the mutant Get3_D57E enters the same foci as wild-type Get3 under glucose starvation. These results strongly suggest that the ability to hydrolyze ATP influences the cellular localization of Get3.

To understand the nature of the foci we performed a thorough colocalization study with markers for a range of cellular structures that are known to be punctate (Fig. 4). Analysis of Get3-GFP-positive foci compared to proteins that label these compartments excluded the following localizations for Get3-GFP under glucose starvation (Fig. 4A): Golgi (Anp1), endosomes (Chc1), prevacuole (Snf7), peroxisomes (Pex3), ER-mitochondrial exchange sites (Mdm34), nucleus-vacuole junction (Nyv1), lipid droplets (Erg6), P bodies (Dcp2), actin patches (Sac6) and autophagosomes (Ape1).

In contrast, when we tested strains expressing GFP fusions of chaperones Hsp42, Hsp104, Sis1, or the Hsp70 homologue Ssa2 we observed clear colocalization with the foci (Fig. 4B). Hsp42 and Hsp104 have been characterized as markers for deposition sites of aggregated proteins (Kaganovich et al., 2008; Liu et al., 2010), in particular under heat stress conditions (Specht et al., 2011). The scaffolding protein Sgt2 (Chang et al., 2010; Wang et al., 2010; Chartron et al., 2011; Kohl et al., 2011) that was present in the foci (Fig. 2C) can interact with both, Hsp70 chaperones and Get3 (via Get5). Hence the cytosolic GET complex may be one component of a larger chaperoning platform sequestering and possibly triaging proteins that expose hydrophobic surfaces.

To obtain better insight into the morphology of the deposition sites, we employed two complementary electron microscopy techniques to visualize Get3-positive foci. First, we prepared cryo-sections of glucose-starved and chemically fixed Δget1get2 cells expressing Get3-GFP (Fig. 5A–C) a strain likely to yield sections with abundant foci (cf. Fig. 2B). We then used anti-GFP immuno-labeling to mark the foci. This analysis revealed electron-dense and hence protein-rich structures devoid of membranes. The foci were sometimes associated with the vacuole (Fig. 5C,E), a characteristic of insoluble protein deposits (IPODs; Kaganovich et al., 2008). Second, we employed high-pressure freezing and Lowicryl-embedding (Fig. 5D–F), which reduced the efficiency of immuno-labeling but revealed the fibrous nature of the foci and demonstrated the exclusion of ribosomes from the deposition sites. Morphologically similar structures were observed in wild-type cells under glucose-starvation (Fig. 5D,E) and in Δget2 cells in the presence of glucose (Fig. 5F). As in Δget1 and Δget1get2 strains, tail-anchored proteins cannot be delivered to the ER membrane in this background resulting in the cytosolic accumulation and aggregation of tail-anchored protein clients (Schuldiner et al., 2008). In conclusion, Get3, when unable to deliver tail-anchored protein clients to the ER membrane, is consistently recruited to chaperone-rich foci, which are free of membranes or ribosomes and most probably represent deposition sites for aggregated hydrophobic proteins.

We found that Hsp104-GFP also moved to foci in a Δget3 strain in the presence of glucose (Fig. 6A). This implies that the chaperone reacts to the presence of aggregation-prone proteins supporting our finding that the observed foci are deposition sites for protein aggregates. Interestingly, the colocalization of Hsp104-GFP with a tail-anchored protein client, Cherry-Sed5, in the foci formed in either Δget3 (Fig. 6B) or Δget1get2 (Fig. 6C) strains increased substantially upon glucose withdrawal. This indicates tuning of the cellular response to different threats to proteostasis.

Under heat shock, Hsp104 is well established as a factor that associates with protein aggregates at deposition sites (Kaganovich et al., 2008; Liu et al., 2010; Specht et al., 2011). To address the interplay between Hsp104 and Get3 during different forms of proteotoxic stress, we asked whether Get3-GFP was recruited to foci under heat shock (Fig. 6D) and found that, in contrast to Hsp104-GFP, Get3-GFP remained soluble. This implies that Get3 is not involved in sequestering heat-denatured clients to deposition sites. On the other hand Hsp104-GFP was acutely sensitive to the absence of Get3 (Fig. 6A). Furthermore, Hsp104-GFP colocalized with Get3-RFP upon glucose starvation but moved to distinct Get3-negative foci upon additional heat shock (Fig. 6E). These data suggest that Get3 activity modulates both, the demand for and the capacity of the cell's cytosolic chaperoning activity, when aggregation-prone proteins challenge the biogenesis or function of cytosolic proteins.

Based on the presence of several chaperones in the Get3-positive foci we investigated whether Get3 displays any chaperoning activity itself. We thus purified Get3 and tested its influence on the aggregation of chemically denatured citrate synthase (CS; Fig. 7A,B) or thermally denatured luciferase (Fig. 7C). Get3 was highly efficient in preventing the aggregation of both model clients exhibiting a maximum inhibition of aggregation when present at a fourfold molar excess of Get3-dimers or a twofold molar excess of Get3-tetramers (Suloway et al., 2011) to CS. The chaperone activity of Get3 was robust when compared to the prokaryotic chaperone Hsp33, which is highly efficient when oxidized. It is of note that addition of either 2 mM ATP or ADP reproducibly reduced the holdase activity of Get3, implying that Get3 acts most efficiently on these clients in the absence of nucleotides (Fig. 7D). This result is fully consistent with our observations involving mutant Get3_D57E (Fig. 3) and the hypothesis that the chaperone holdase activity of Get3 becomes relevant under conditions of energy depletion when Get3 is unable to deliver its tail-anchored protein clients to the ER membrane and when the capacity for protein degradation may also be reduced. Intriguingly, Get3 and its mammalian homologue TRC40 have previously been shown to bind tail-anchored protein clients in the absence of nucleotides (Favaloro et al., 2008; Stefer et al., 2011) and this was true for ATP-hydrolysis deficient Get3_D57E (Fig. 3B,C). Most holdase chaperones protect cells from the toxic potential of hydrophobic proteins by preventing their aggregation without the requirement of ATP (Eyles and Gierasch, 2010).

Auld et al. (Auld et al., 2006) found the GET3 gene to be co-regulated with the proteasome consistent with a functional relationship to ubiquitin-dependent protein degradation. We addressed the fate of the tail-anchored protein client after glucose re-addition by monitoring the steady-state levels of Cherry-Sed5 and endogenous Sed5 in cycloheximide chase experiments. During glucose starvation Sed5 was stable (Fig. 8A,B). In the absence of tail-anchored protein insertion (Δget1get2 and Δget1get2get3 strains; Fig. 8A,B) Sed5 was less stable in the presence of Get3 (Δget1get2 strain) upon glucose re-addition. This raises the possibility that chaperoning by Get3 is a prerequisite for efficient degradation when glucose becomes available again. This scenario is further supported by our observation that in the absence of Get3, Get4 or Get5 (Fig. 8C,D), Sed5 remained stable after glucose re-addition consistent with the notion that Get3 keeps the tail-anchored protein client soluble for other factors to act on it.

Integrating the results presented here, we propose that the Get3 ATPase has a second cellular function in addition to its well-characterized role as a factor that targets newly synthesized tail-anchored proteins to insertion sites in the ER membrane (Fig. 8E). This second function as a holdase exploits the chaperone capacity of Get3 to bind hydrophobic regions of proteins and becomes relevant under conditions that interfere with the tail-anchored protein insertion cycle and the delivery of short secretory proteins to the translocon (Johnson et al., 2012), either because the ER membrane receptor for Get3 is absent or because cellular energy levels are significantly depleted.

Our characterization of Get3 localization under glucose starvation substantially extends the previous observation that Get3 and tail-anchored protein substrates are found in aggregates in cells lacking the GET receptor (Schuldiner et al., 2008). First, the kinetics and reversibility of the process (Fig. 1) and its dependence on other cytosolic GET components (Fig. 2) argue against an uncontrolled aggregation process. Second, even in the absence of the GET receptor, Get3 remains sensitive to glucose starvation, which maximizes its colocalization with a model tail-anchored client (Fig. 2). This glucose sensitivity in subcellular localization is intimately connected to Get3’s ability to hydrolyze ATP (Fig. 3) suggesting that Get3 may directly respond to cellular energy status. In support of this hypothesis, we have excluded a major glucose-sensing signal transduction cascade that involves the yeast AMP-activated kinase Snf1/4 as a putative pathway responsible for the glucose-dependent relocalization of Get3 (Fig. 3).

Cellular quality control is aimed at preventing protein aggregation. However, it becomes increasingly clear that cells possess a second line of defense once the factors involved in counteracting aggregation are overwhelmed (Tyedmers et al., 2010). Live-cell imaging has been key to obtaining insight into the cellular strategies by which aggregating proteins are directed to specific deposition sites and hence spatially segregated from processes that they might interfere with (Kaganovich et al., 2008). Subsequent work has shown that different molecular chaperones are targeted to different types of deposition sites in a complex and dynamic pattern (Specht et al., 2011). Two complementary methods of immuno-electron microscopy unequivocally demonstrate the presence of Get3-GFP at protein-dense foci with fibrillar appearance (Fig. 5). The cytosolic GET complex comprising Sgt2, Get4 and Get5, is implicated in transferring nascent tail-anchored protein precursors from the ribosome to Get3. Because the details of this transfer and loading process are poorly understood it was critical to determine whether ribosomes are present in the Get3-GFP-positive foci observed under glucose starvation. High-pressure freezing allows the detection of ribosomes by electron microscopy and in combination with immunolabeling Get3-GFP, we were able to reveal that ribosomes are absent from Get3-GFP-positive sites (Fig. 5E,F). This result suggests that the cytosolic GET complex is more than a ribosome-associated transfer complex and exerts control over Get3-client complexes even after loading.

At the deposition sites, Get3 strikingly colocalizes with molecular chaperones that can prevent the aggregation of proteins exposing hydrophobic regions and/or reactivate aggregated proteins (Fig. 4B). The dual function of Get3 as a holdase chaperone and a targeting factor for post-translational membrane integration may enable fast adaptation of the cell once cellular energy status has improved: Precursor proteins can remain associated with Get3 unless they require the activity of another factor present at the deposition site. The organization of various types of chaperones into larger assemblies controlling aggregation-prone proteins implies that clients can be transferred between effector proteins that decide and execute their fate. The problem of sorting hydrophobic clients between different chaperones that recognize precisely this quality – albeit with very different outcomes for the client protein – has previously been recognized and addressed by elegant in vitro assays (Wang et al., 2010; Hegde and Keenan, 2011). However, our study is the first to define the physiological conditions and cellular structures that integrate the capacity of the GET pathway for hydrophobic protein sorting into global cellular proteostasis. Furthermore, our finding that Get3 acts as a holdase chaperone for clients other than tail-anchored proteins (Fig. 7) raises the possibility that this activity can protect other classes of aggregation-prone proteins. Our findings also question the notion that membrane integration is the only fate available to a client protein once it has been sorted into a complex with Get3 (Fig. 8). A broader contribution of Get3 to cellular quality control is strongly supported by the intricate functional relationship between Get3 and the ubiquitin-proteasome system (Auld et al., 2006) and the wider interaction networks formed by GET associated factors such as Sgt2 in yeast and SGTA and Bag6 in mammalian cells (Chang et al., 2010; Leznicki et al., 2010; Mariappan et al., 2010; Wang et al., 2010; Chartron et al., 2011; Hegde and Keenan, 2011; Hessa et al., 2011; Kohl et al., 2011; Chartron et al., 2012; Leznicki and High, 2012).

We have used two model clients, citrate synthase and luciferase, to monitor Get3’s ability to prevent protein aggregation upon chemical or thermal denaturation (Fig. 7). Both proteins contain stretches of ca. 20 amino acids that are rich in hydrophobic residues (http://dgpred.cbr.su.se/; Hessa et al., 2007). Whilst these hydrophobic regions will normally be buried in the hydrophobic core of the enzymes and interruptions by polar residues clearly distinguish them from transmembrane segments they will be exposed upon denaturation and might be recognized by Get3. Interestingly, maximal protection of either client from aggregation required at least a fourfold molar excess of Get3 consistent with the idea that higher-order complexes of Get3 are involved in the holdase function. The recent demonstration that a tetramer of an archeal Get3 homologue forms a hydrophobic chamber capable of accommodating a transmembrane segment (Suloway et al., 2011) raises the possibility that the observed holdase activity involves the occlusion of aggregation-prone hydrophobic surfaces in a cage structure formed by a Get3 tetramer. The observed inhibitory effect of adenine nucleotides on the holdase activity (Fig. 7D) is compatible with the intricate structural interplay of the nucleotide-binding domain and the helices involved in forming the hydrophobic binding site of Get3 (Hegde and Keenan, 2011; Chartron et al., 2012). Furthermore, the negative impact of adenine nucleotides on Get3 holdase activity provides an important link to our observations regarding the localization of Get3 in vivo: Get3 changes its localization when cells are glucose-starved or when alteration of a critical residue by sited-directed mutagenesis interferes with its ATPase activity (Fig. 3). Whilst the tail-anchored protein insertion cycle requires ATP hydrolysis, Get3 will be in a position to exert its holdase function under conditions of energy depletion when hydrophobic proteins accumulate.

It has not been rigorously determined how the threat to cellular proteostasis from different classes of aggregation-prone proteins adds up. The heat sensitivity of the get mutants (Shen et al., 2003; Metz et al., 2006; Schuldiner et al., 2008) may be explained by the accumulation of membrane protein precursors that jeopardize the capacity of the cell to withstand the load of misfolded proteins caused by heat shock. Our results clarify this point by clearly demonstrating that the chaperone Hsp104-GFP senses the presence of cytosolically accumulating tail-anchored protein precursors (Fig. 6) and that its localization with respect to clients and Get3 depends on the precise nature of the proteotoxic stress. In conclusion we show that the capacity of Get3 to shield hydrophobic proteins is integrated into a spatially controlled and intricately regulated network of activities that probe, retain, reactivate or degrade hydrophobic clients. The possibilities for sorting clients between these individual activities are only beginning to emerge.

Supplementary material Table S1 lists all the strains, supplementary material Table S2 all plasmids, and supplementary material Table S3 all oligonucleotides used in this study. General molecular biology and basic yeast methodology followed protocols provided (Ausubel et al., 1997). p415MET25 mCherry-Sed5 (pAA1307) was constructed by fusing the mCherry ORF to the 5′ end of the SED5 ORF via an engineered NotI site coding for three alanines, which was cloned into p415MET25.

Cells were diluted to an OD600 of 0.2 from an overnight pre-culture and grown to mid-log phase (OD600 of 0.6) in synthetic complete medium (SC). Cells were then harvested by low speed centrifugation, washed twice, and resuspended in SC or medium lacking glucose (SC-D).

Images of live cells were acquired at room temperature on a Delta Vision RT (Applied Precision) restoration microscope using a 100× 0.35–1.5 Uplan Apo objective and specific band pass filter sets for FITC or RD-TR-PE. The images were collected using a Coolsnap HQ (Photometrics) camera and representative images from several independent experiments are shown as the results.

Whole cell lysates were prepared by harvesting 2 ml of culture, treated as indicated, by low speed centrifugation. The pellet was resuspended in 0.1 M NaOH and incubated at room temperature for 10 minutes (Yaffe and Schatz, 1984). Samples were then harvested and resuspended in non-reducing SDS-PAGE loading buffer. 10 µl of each sample (equivalent of 0.01 OD₆₀₀ unit per lane) was resolved by SDS-PAGE and transferred to nitrocellulose membrane. For protein detection the following antibodies were used: anti-Get3 guinea pig antiserum (Metz et al., 2006), anti-GFP mouse monoclonal B-2 (Santa Cruz Biotechnology), anti-Sed5 antiserum (Schuldiner et al., 2008) and anti-Pgk1 mouse monoclonal 22C5 (Molecular Probes, Invitrogen) followed by either anti-guinea pig-HRP or anti-mouse-HRP conjugated secondary antibodies (Jackson ImmunoResearch). Membranes were washed between antibody incubations in blocking solution (1×TBS, 5% milk powder, 0.02% IGEPAL CA-630) and visualized using a 1∶1 mixture of enhanced chemiluminescence (ECL) western blotting reagents (Thermo Scientific). Alternatively, IRDye-conjugated secondary anti-guinea pig, anti-rabbit or anti-mouse antibodies (Li-Cor) were used and visualized employing an Odyssey SA (Fig. 8).

Where indicated, translation was inhibited by the addition of cycloheximide to mid-log cells incubated in SC or SC-D for 1 hour at a final concentration of 0.5 mM from a stock solution of 50 mM cycloheximide in DMSO. Samples were taken at the indicated time points before collection of whole cell lysate and analysis by western blotting as described above.

Yeast cells were grown to mid log phase in SC medium at 30°C as described above and glucose-starved for 1 hour in SC-D if indicated. For classical immuno-EM, cells were concentrated by filtration and then chemically fixed and processed for cryo-ultrasectioning according to the Tokuyasu method (Tokuyasu, 1973). In detail, concentrated cells in medium were mixed 1∶1 with 4% paraformaldehyde/PBS and incubated for 30 minutes at room temperature. The fixative was replaced by 2% paraformaldehyde/PBS and cells incubated overnight at 4°C. Cells were postfixed for 2 hours in 4% paraformaldehyde/0.2% glutaraldehyde/PBS on ice and washed three times with 0.02 M glycine/PBS prior to embedding in 10% gelatine. Small blocks were infused with 2.3 M sucrose/PBS at 4°C overnight, mounted on metal pins and frozen in liquid nitrogen. Ultrathin cryo-sections (75 nm) were cut using a cryo-ultramicrotome (Ultracut UCT with EM FCS, Leica) and a diamond knife (Diatome) at −110°C and placed on formvar-coated nickel grids.

For immunolabeling, sections were incubated with polyclonal rabbit anti-GFP antibodies (Abcam) for 20 minutes, followed by incubation with protein A-10 nm gold (CMC, Utrecht) for 20 minutes. Sections were contrasted with 0.4% (w/v) uranyl acetate in 2 M methyl-cellulose for 10 minutes on ice and embedded in the same solution.

For Lowicryl embedding, cells were grown as described above, rapidly filtered and vitrified using an EM HPM100 (Leica) high-pressure freezer. Fixed cells were further processed by freeze substitution in a Leica EM AFS and embedded in HM20 (Polysciences) according to Kukulski et al. (Kukulski et al., 2011). 75 nm ultrathin sections were cut with an ultramicrotome (Ultracut UCT, Leica), used for immuno-gold labeling employing antibodies as described above and contrasted with 1% (w/v) uranyl acetate/H₂O for 20 minutes and lead citrate for 1 minute.

All sections were examined with a Philips CM120 transmission electron microscope and micrographs were acquired with a CCD camera (Megaview III, Olympus Soft Imaging Systems). Image processing was performed using iTEM software (Olympus Soft Imaging Systems).

Get3 was expressed in BL21 (DE3) from a pQE80 derivative as a fusion of two Z domains (IgG binding domain of proteinA) to Get3 (Metz et al., 2006). Briefly, cultures were grown in autoinduction medium (50 mM NH₄Cl, 25 mM Na₂HPO₄, 25 mM KH₂PO₄, 5 mM Na₂SO₄, 2 mM MgSO₄, 0.5% tryptone, 0.25% yeast extract, 0.05% glucose, 0.2% lactose, and 0.5% glycerol supplemented with 200 µg/ml ampicillin) for 22 hours. Pelleted cells were lysed by sonification in extraction buffer (50 mM Tris-HCl pH 7.5, 50 mM NaCl, 2 mM Mg-acetate and 1 mM imidazole). Cleared lysates were subjected to ultracentrifugation. Get3 was immobilized on a Ni-NTA column and washed with extraction buffer, eluted with 500 mM imidazole, and dialyzed against TEV cleavage buffer (50 mM Tris-HCl pH 8, 1 mM DTT). After cleavage with TEV protease the Z domain tag and TEV protease were removed by passage over a Ni-NTA column and the Get3-containing flow-through was dialyzed against extraction buffer. For ATPase assays immobilized Get3 ZZ fusion protein was sequentially washed in extraction buffer containing 1% Triton-X100, 1 M NaCl, 5 mM Mg-ATP and 0.5 M Tris-HCl pH 7.5, respectively, in order to remove contaminating ATPases such as prokaryotic chaperones. The protein was eluted with extraction buffer containing 500 mM imidazole and buffer-exchanged on a PD-10 column into the same buffer without imidazole. Sec61β was purified as described in Leznicki et al. (Leznicki et al., 2010).

The ATP-hydrolysis rates of Get3 and Get3 D57E (uncleaved fusions to two Z domains) were measured with an NADH-coupled ATP-regenerating system at 25°C. The reaction medium consisted of 50 mM Tris-HCl pH 8.0, 100 mM NaCl, 10 mM MgCl₂, 2 mM phosphoenolpyruvate, 25 units/ml of pyruvate kinase and lactate dehydrogenase, 2 mM ATP, 200 µM NADH and 120 µg/ml Get3.

Binding of recombinant Get3 protein (uncleaved fusions to two Z domains) to immobilized Sec61β was tested as described in Leznicki et al. (Leznicki et al., 2010): Sec61β with and without the tail-anchor segment was immobilised on Ultra Link Biosupport beads according to manufactures instructions. 1 µl of 0.5 mg/ml Get3, wild-type or D57E, was added to 50 µl of beads resuspended in 50 mM Tris-HCl, pH 7.5. Binding reactions were allowed to proceed at 4°C for 1 hour with constant agitation. Beads were washed 3 times with 50 mM Tris-HCl, pH 7.5, 50 mM NaCl and the bound fractions eluted during 30 minutes of incubation at 37°C in SDS sample buffer. Unbound and bound fractions were analysed by non-reducing SDS-PAGE and western-blotting.

Chaperone assays were performed as previously described (Buchner et al., 1998). Briefly, 12 µM citrate synthase (Roche) was denatured overnight at room temperature in 4.5 M Guanidinium-HCl, 40 mM HEPES pH 7.5. Aggregation was initiated by addition of 75 nM citrate synthase to 40 mM HEPES pH 7.5 at 30°C with or without purified Get3 at the indicated concentrations relative to citrate synthase. Light scattering was monitored at 360 nm in a Hitachi F4500 fluorescence spectrophotometer. Where indicated 2 mM ATP or ADP was added to Get3 and incubated for 1 minute prior to addition of citrate synthase.

Aggregation of luciferase was initiated by the addition of 120 nM luciferase to 40 mM Mops pH 7.5, 50 mM KCl at 43°C, with or without the indicated concentrations of Get3. Aggregation was monitored for 900 seconds as described in the text.

We thank P. March (FLS Bioimaging Facility) for support with live cell imaging and D. Riedel, D. Wenzel and G. Heim for technical advice regarding electron microscopy; A. Clancy for expert help with microscopy; P. Leznicki for generously providing purified Sec61β; W. Voth for a control experiment using the chaperone assay; Y. Elbaz, J. Metz and V. Schmidt for the construction of yeast strains; M. Ashe and J. Weissman for generously sharing yeast strains and Schwappach lab members for critical reading of the manuscript.