Structural insight into the membrane insertion of tail-anchored proteins by Get3


  • Atsushi Yamagata,

    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    2. Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8501, Japan
    Search for more papers by this author
  • Hisatoshi Mimura,

    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    Search for more papers by this author
  • Yusuke Sato,

    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    2. Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8501, Japan
    Search for more papers by this author
  • Masami Yamashita,

    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    2. Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8501, Japan
    Search for more papers by this author
  • Azusa Yoshikawa,

    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    2. Department of Biological Information, Graduate School of Bioscience and Technology, Tokyo Institute of Technology, Yokohama 226-8501, Japan
    Search for more papers by this author
  • Shuya Fukai

    Corresponding author
    1. Structural Biology Laboratory, Life Science Division, Synchrotron Radiation Research Organization and Institute of Molecular and Cellular Biosciences, The University of Tokyo, Tokyo 113-0032, Japan
    2. Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8501, Japan
    Search for more papers by this author

  • Communicated by: Toshio Hakoshima



Tail anchored (TA) proteins, which are important for numerous cellular processes, are defined by a single transmembrane domain (TMD) near the C-terminus. The membrane insertion of TA proteins is mediated by the highly conserved ATPase Get3. Here we report the crystal structures of Get3 in ADP-bound and nucleotide-free forms at 3.0 Å and 2.8 Å resolutions, respectively. Get3 consists of a nucleotide binding domain and a helical domain. Both structures exhibit a Zn2+-mediated homodimer in a head-to-head orientation, representing an open dimer conformation. Our cross-link experiments indicated the closed dimer-stimulating ATP hydrolysis, which might be coupled with TA-protein release. Further, our coexpression-based binding assays using a model TA protein Sec22p revealed the direct interaction between the helical domain of Get3 and the Sec22p TMD. This interaction is independent of ATP and dimer formation. Finally, we propose a structural mechanism that links ATP hydrolysis with the TA-protein insertion mediated by the conserved DTAPTGH motif.


Targeting of membrane proteins to their destinations is not only an important but also a challenging task for cells. Membrane insertion of the majority of membrane proteins in eukaryotic cells occurs through a co-translational pathway assisted by the signal recognition particle (SRP) and the Sec61 translocon in endoplasmic reticulum (ER) membrane. The N-terminal transmembrane domain (TMD), which emerges from the ribosome in the initial stage of polypeptide synthesis, is recognized by the SRP, and served as a targeting signal for the membrane insertion by the Sec61 translocon (Shan & Walter 2005; Rapoport 2007). The subsequent membrane insertion is coupled with elongation of the polypeptide chain by the ribosome.

Different from the canonical co-translational membrane insertion pathway, the unique post-translational insertion has been proposed for a class of membrane proteins termed ‘tail anchored (TA) protein’, which is anchored to the membrane by a single TMD near the C-terminus with the N-terminal domain remaining in the cytoplasm. TA proteins localize to intracellular and cellular membranes and play critical roles in numerous cellular processes (Wattenberg & Lithgow 2001; Borgese et al. 2003). Examples of TA proteins include the Sec61β/γ subunit in Sec61 translocon, the various SNAREs for membrane trafficking and the apoptosis-related Bcl2 proteins. Although the localization of TA proteins is regulated in the membrane insertion step, its molecular mechanism has been poorly understood. The SRP/Sec61-dependent co-translational mechanism appears highly unlikely for the insertion of TA proteins, as the TMD close to the C-terminus does not emerge from the ribosomal tunnel until the translation is terminated. Instead, the membrane insertion of TA proteins has been known to occur post-translationally with ATP hydrolysis (Kutay et al. 1995; Steel et al. 2002).

The highly conserved ATPase Get3 was recently identified as a key factor for the post-translational TA protein insertion pathway (Stefanovic & Hegde 2007). Human Get3 (also known as TRC40 and Asna-1) specifically binds to the TMD of the post-translationally synthesized model TA protein Sec61β in the cytosol. This cytosolic complex is then delivered to the ER membrane, and the subsequent membrane insertion occurs dependent on ATP hydrolysis by Get3. A similar result was reported independently using another TA protein, RAMP4 (Favaloro et al. 2008). In contrast, the molecular chaperone pair Hsc70 and Hsp40 has been also identified as ATP-dependent factors for the post-translational insertion pathway distinct from the Get3-dependent pathway (Rabu et al. 2008). These two distinct insertion pathways might function in a complementary fashion with each other.

Yeast Get3 (also known as Arr4p) binds to the TMD of TA proteins and is targeted to the ER membrane via the interaction with the membrane-bound receptor complex Get1/2 to form the ternary complex named Get complex (Schuldiner et al. 2005, 2008; Auld et al. 2006). Loss of the Get complex results in not only a drastic reduction in the TA protein insertion but also the mislocalization of TA proteins (Schuldiner et al. 2008). The failure of the proper TA protein insertion could explain the metal- and heat-sensitive phenotypes of the Get3 deletion strain (Shen et al. 2003; Metz et al. 2006; Schuldiner et al. 2008).

Here we report the crystal structure of yeast Get3 in ADP-bound and nucleotide-free forms at 3.0 Å and 2.8 Å resolutions, respectively. Both structures exhibit a homodimer that is bridged by a Zn2+, which is coordinated by the conserved CXXC motif. We show the conformational transition from the open dimer state observed in our structure to the closed dimer state detected by our cross-link experiments. The closed dimer-stimulating ATP hydrolysis might be coupled with TA-protein release. Further, we introduced a new binding assay based on the coexpression of Get3 with the model TA protein Sec22p to identify the TA-protein binding site. Together with the structural and functional analyses, we proposed the mechanism that links ATP hydrolysis and TA protein insertion.


Overall structure

Yeast full-length Get3 produced in Escherichia coli was purified and crystallized as described in Experimental procedures. The crystals are in the space group of P21212 containing two Get3 molecules in the asymmetric unit. The structure was determined by the single wavelength anomalous dispersion (SAD) method using selenomethione (Se-Met) substituted Get3. A strong residual electron density between two Get3 subunits was observed in the anomalous difference Fourier map for Se-Met Get3, suggesting a bound metal ion. This was identified as a Zn2+ from the anomalous difference Fourier maps at two wavelengths flanking an absorption edge for Zn2+ (Fig. 1B). Despite the presence of high concentration of AMP-PNP (∼5 mm) in crystallization solution, no electron density was observed consistent with the bound nucleotide. The final model of the nucleotide-free Get3 was refined to R/Rfree factors of 0.236/0.281 at 2.8 Å resolution (Table 1). By contrast, in the Get3 structure co-crystallized with ADP, an ADP molecule in each subunit was clearly evident in the electron density map (Fig. S1 in Supporting Information). The ADP-bound Get3 structure was refined to R/Rfree factors of 0.252/0.282 at 3.0 Å resolution. The nucleotide-free and ADP-bound Get3 structures are essentially the same with each other (an rmsd value of 0.4 Å for Cα atoms). Therefore, we will hereafter describe the ADP-bound Get3 structure.

Figure 1.

 Overall structure of the Get3 dimer. (A) The ribbon diagram of the Get3 dimer structure with the schematic diagram of the Get3 sequence motifs. Each subunit consists of the NBD (green in subunit A and orange in subunit B) and the helical domain (cyan in subunit A and beige in subunit B). ADP (shown as red spheres) is bound in each NBD. A Zn2+ (shown as a magenta ball) bridges two subunits. The P-loop and DTAPTGH motif are colored in yellow and pink, respectively. (B) Enlarged view of the upper interface (top) and the lower interface (bottom). The anomalous difference Fourier map (contoured at 5.0 σ) at the wavelength of 1.2500 Å shows a bound Zn2+. Residues involved in the dimer interaction and Zn2 are shown as sticks and a magenta ball, respectively. The subunits A, B and P-loop are colored in green, orange and yellow, respectively.

Table 1.   Data collection and refinement statistics
Data collectionUnliganded formADP
  1. †Overall/last shell.

  2. Rsym is the unweighted R value on I between symmetry mates.

  3. §= ∑hkl||Fobs(hkl)| − |Fcalc(hkl)||/∑hkl|Fobs(hkl)|.

  4. ¶Subunit A/Subunit B.

Wavelength (Å)1.000000.979201.00000
Total reflections217 034235 025128 972
Unique reflections31 95524 69625 728
Resolution (Å)50.0–2.8050.0–3.0550.0–3.00
Completeness (%)†99.1 (92.6)99.3 (97.5)97.8 (94.6)
I/σ†38.2 (3.6)15.7 (2.3)23.2 (2.5)
Rsym†‡0.061 (0.263)0.126 (0.443)0.061 (0.342)
Refinement statistics
Reflections (work/free)27863/3117 24371/1233
No. of atoms4899 4927
 Protein4898 4872
 ADP0 54
 Zn2+1 1
Average B-factor
 Protein¶86.3/71.4 96.5/89.9
 ADP¶  103.6/99.6
 Zn2+80.63 95.7
 Bond length (Å)0.0089 0.0087
 Bond angle (°)1.307 1.372
R/Rfree§0.237/0.281 0.252/0.282
Ramachandran plot
 Most favored (%)86.2 86.3
 Disallowed (%)0 0

The two Get3 subunits in the asymmetric unit form a dimer in a head-to-head orientation (Fig. 1A). Each Get3 monomer structure consists of the core nucleotide binding domain (NBD, green and orange in Fig. 1A) and the protruding helical domain (cyan and beige). The NBD comprises one short anti-parallel and seven parallel strands flanked by a total of seven helices on both sides to form the core α/β structure (Fig. S2). The helical domain (residues 170–233) is composed of two long helices with a short helix between them. Almost half of the helical domain in subunit B (residues 190–209) was completely disordered (Fig. 1A), while the main chain could be traced in the equivalent region of subunit A, suggesting the intrinsic flexibility of the helical domain. Another flexible region in Get3 structure is that between β4 and α5 (Fig. S2). In subunit B, the single helix (α4) in this region was clearly visible, which lies in the similar direction as the helical domain. However, the equivalent region was disordered in subunit A. This helix, together with the helical domain, constitutes the lower half of the dimer, which are apparently more flexible than the upper half of the dimer composed of NBDs. These flexible regions are less conserved in amino acid sequences, compared with the highly conserved NBD cores (Fig. S2).

ADP occupies the nucleotide binding site (NBS) in each subunit. A Zn2+ is bound between the two subunits near the top of the NBD. The overall structure of the Get3 dimer closes in the Zn2+-binding site and opens toward the helical domains to form an open dimer, as discussed below.

Dimer interface and Zn2+ binding site

Get3 dimer interface buries a total 1380 Å2 of the surface area. Only between the NBDs, two dimer interfaces are apparently formed (Fig. 1B): the upper Zn2+-mediated interface and the lower protein-mediated interface. In the upper interface, a Zn2+ is tetrahedrally coordinated by the Sγ atoms of Cys285 and Cys288 from each subunit (Fig. 1B). The Cys285-X-X-Cys288 motif of Get3 is completely conserved among all organisms (Fig. S2). To confirm a stoichiometry between Get3 and the bound Zn2+, we performed inductively coupled plasma mass spectrometry (ICP-MS), which can quantify a metal ion with a very high sensitivity (ppb–ppt level) by combining the ionization in plasma and the mass spectrometry. ICP-MS measurement revealed nearly 50% zinc occupancy in the purified Get3, indicating single zinc per Get3 dimer, in perfect agreement with our crystal structure. Get3 exists as a dimer in solution analyzed by size exclusion chromatography (Fig. S3) (Shen et al. 2003). A single substitution of Cys285 to serine residue (C285S) gave an elution profile different from that of the wild type protein, suggesting that it was no longer able to form a stable dimer (Fig. S3). ICP-MS measurement confirmed no zinc bound in the C285S mutant. Although there is another highly conserved cysteine motif Cys240-X-Cys242, it is located in the β-strand and not involved in either metal ion binding or dimer interaction (Fig. S2). This supports a previous biochemical study indicating that Cys285 and Cys288 are involved in dimer formation, but either Cys240 or Cys242 is not (Metz et al. 2006). In addition to the Zn2+ bridges, the protruding Arg287 side chain contributes to the dimer interaction by forming hydrogen bonds with the main chain oxygen of Leu316.

The lower interface between α9 and P-loop backbones (yellow in Fig. 1B) is stabilized largely by hydrophobic interactions. Phe246 forms a hydrophobic interaction with the backbone of Gly27 and Gly28, which are the N-terminal residues in P-loop of the adjacent subunit. The side chains of Leu247 from each subunit interlock with each other by hydrophobic interactions. In addition to these hydrophobic interactions, there is a hydrogen bond between the Oη atom of Tyr250 in subunit A and the Nζ atom of Lys26 in subunit B, though the corresponding residues on the opposite subunit pair do not interact. The residues described here are highly conserved among all organisms (Fig. S2). We investigated the size exclusion chromatography profile of the mutant proteins for these residues (F246A, L247E, L247R, F246A/L247E, F246A/L247R, Y250A, Y250K, F246G/L247G/Y250A as well as R287A), but none of them showed any significant difference from that of the wild-type protein (data not shown). These indicate that the Zn-thiolate interaction, which is much more robust than a hydrogen bond or a hydrophobic interaction, can still retain to bridge two subunits even if the lower protein-mediated interface is disrupted.

Nucleotide binding site

ADP occupies the NBS of each subunit without any contact from the adjacent subunit. The structure of the ADP binding site in subunit A is shown in Fig. 2A. The main chain nitrogens from Val29 to Thr32 in the P-loop form hydrogen bonds with the β-phosphate group of ADP. The side chain of Lys31 forms an electrostatic interaction with the β-phosphate group. Further, the side chains of Thr32 and Thr33 form hydrogen bonds with the β- and α-phosphate groups, respectively. Three hydrophobic residues, Leu316, Ile321, and Phe330, surround the adenine moiety of ADP. The N6 and N7 atoms of the adenine ring form hydrogen bonds with the Oδ and Nδ atoms of Asn272, respectively. The N6 atom also forms a hydrogen bond with the main chain oxygen of Pro315. Further, the N1 atom forms a hydrogen bond with the main chain nitrogen of Cys317. In addition to these hydrophilic interactions of the NBD with the phosphate and adenine moieties of ADP, the 2′-OH group of the ribose moiety forms a hydrogen bond with the main chain nitrogen of Ile321. The residues interacting with ADP via their side chains are highly conserved among all organisms (Fig. S2).

Figure 2.

 Nucleotide binding site. (A) The ADP binding site of the Get3 subunit A. ADP-interacting residues are labeled. Hydrogen bonds and electrostatic interactions are shown as dotted lines. The P-loop and DTAPTGH signature motif are highlighted in yellow and pink, respectively. (B) The ADP binding site of the first ArsA NBS. A Mg2+ is shown as a green ball. The water molecule, by which Asp142 can coordinate Mg2+, is shown as a red ball. Coloring schemes of the P-loop and DTAPTGH signature motif are the same as (A).

TA protein binding site

Binding of Get3 to the substrate TA proteins has been detected by yeast two-hybrid assay (Schuldiner et al. 2008) or chemical cross-link using in vitro translated TA proteins (Stefanovic & Hegde 2007; Favaloro et al. 2008; Schuldiner et al. 2008). In this study, we further introduced another simple approach to detect the binding on the basis of the coexpression of glutathione-S-transferase (GST)-tagged Get3 with the substrate TA proteins. We here selected the full-length Sec22p as a model TA protein. Sec22p is an R-SNARE protein that functions in membrane trafficking between the ER and Golgi (Cao & Barlowe 2000; Liu et al. 2004). Specific recognition of Sec22p by Get3 was shown previously using yeast two-hybrid assay, and the ATP-dependent membrane insertion was demonstrated by an in vitro reconstitution experiment (Schuldiner et al. 2008). Sec22p consists of three domains: the N-terminal longin domain (residues 6–117), the SNARE motif (132–192) and the C-terminal TMD (193–213). The crystal structure of the TMD-truncated human Sec22 in complex with Sec23/24 revealed that the globular longin domain is folded together with a part of the SNARE motif (Gonzalez et al. 2001; Mancias & Goldberg 2007). However, the linker connecting both domains is poorly ordered.

As shown in Fig. 3A, when GST-Get3 and the full-length Sec22p were coexpressed in E. coli, both proteins were present in soluble fraction and copurified by glutathione affinity chromatography. In contrast, only small amounts of the TMD-truncated mutant of Sec22p (residues 1–193; Sec22pΔTMD) were copurified with Get3, compared with the full-length Sec22p (Fig. 3A, see also Fig. 5B). These results showed that our coexpression-based binding assay successfully detected a specific interaction between Get3 and the Sec22p TMD. A residual binding ability of Sec22pΔTMD to Get3 is likely owing to an interaction between Get3 and the cytoplasmic region of Sec22p in agreement with a reported cross link between Get3 and the cytoplasmic region of Sec61β (Favaloro et al. 2008).

Figure 3.

 The helical domain of Get3 binds to the TMD of Sec22p. (A) The direct interaction between Get3 and Sec22p. The coexpressed GST-Get3 and Sec22p were present in soluble fraction (lane 3), and the complex was coeluted from glutathione sepharose column (lane4). Sec22p cannot interact with GST alone (lanes 1 and 2). Truncation of the TMD drastically reduced the copurified protein (lane 6). Partial truncation of the helical domain (Get3Δ190–209) also reduced the copurified protein (lane 8). (B) The DTAPTGH signature motif physically links the ATPase catalytic site and the TA protein-binding site. The DTAPTGH motif of Get3 (pink sticks) starts at Asp166 and continues through the helical domain (cyan). Residues 190–209 (truncated in the Get3Δ190–209 mutant protein) are highlighted in yellow. The ArsA signature sequence involves Asp142 and His 148, which bind to a Mg2+ ion (shown as a green ball) and one of three Sb3+ ions (shown as a magenta ball), respectively.

Figure 5.

 TA protein binding by mutant Get3. (A) Binding of the Get3 mutants to Sec22p. The Sec22p copurified with the K31A, D166N and C285S mutants of Get3 was analyzed by SDS-PAGE. (B) Quantification of the interaction between Get3 and Sec22p. Each densitometry of the Sec22p or Sec22pΔTMD bands was compared after the normalization. Results from at least two independent experiments are shown.

The best candidate for the TA protein-binding region is the protruding helical domain that is unique to Get3. From the structural viewpoint, the 35 Å-long helical domain of Get3 might be able to cover the Sec22p TMD, which comprises 21 amino acid residues and is predicted to fold into a 30 Å-long helix. This idea was further supported by the cross-link experiments between human Get3 and the TA proteins Sec61β and RAMP4 using the Cys-specific cross-linker BMH (Stefanovic & Hegde 2007; Favaloro et al. 2008). Among eight cysteine residues in human Get3, five are conserved in yeast Get3 and located inside of the NBD core (Fig. S2). When we mapped the remaining human Get3 cysteine residues on the yeast Get3 structure, only two residues located outside were favorable for binding to TA proteins. These residues (i.e. Cys205 and Cys268 in human Get3; Ser203 and Tyr262 in yeast Get3) are predicted to locate within or near the helical domain, respectively (Figs S2,4).

Replacement of the flexible region of the helical domain by a Gly-Ala-Ala-Gly linker in Get3 (Get3Δ190–209) drastically reduced the amount of the copurified Sec22p (Fig. 3A,B, see also Fig. 5B). This result shows that the helical domain of Get3 constitutes the primary binding site for TA proteins. The amino acid sequence of the helical domain is amphipathic and less conserved than the NBD of Get3 (Fig. S2). This low sequence conservation may correspond to the broad specificity to a wide variety of TMD sequences. Finally, we propose a plausible docking model between Get3 and a 20-residue TMD helix on the basis of the aforementioned Cys-specific cross-link experiment (Favaloro et al. 2008), where the TMD was placed outside the helical domain so that Cys205 and Cys268 of human Get3 (corresponding to Ser203 and Tyr262 in yeast Get3, respectively) might be cross-linked with a cysteine residue in the TMD and that in the cytoplasmic domain, respectively (Fig. S4).

Dimer conformation-dependent ATP hydrolysis

In many Walker type ATPases, two acidic residues in the vicinity of the NBS play a critical role to coordinate Mg2+ for ATP hydrolysis. In our Get3 structure, no electron density corresponding to Mg2+ was observed despite the presence of Mg2+ in the crystallization solution. The NBD of Get3 shows sequence homology with the equivalent domain of the arsenic transport ATPase, ArsA from E. coli (PDB code 1F48) (Zhou et al. 2000; Bhattacharjee et al. 2001; Shen et al. 2003). ArsA has duplicated NBDs and metal-binding domains in a single polypeptide chain. Each of the ArsA NBDs is structurally similar to the Get3 NBD (Figs S2,5). In the first NBD of E. coli ArsA, Mg2+ is coordinated by Asp45 and a water molecule hydrogen bonding with Asp142 (Fig. 2B). These two aspartate residues in E. coli ArsA correspond to Asp57 and Asp166 in yeast Get3 (Figs 2A,S2). D57N and D166N mutants of Get3 completely lost their ATPase activities (Table S1), suggesting that Asp57 and Asp166 also play critical roles in Mg2+ coordination in Get3. However, in our Get3 structure, the side chain of Asp57 flip away from the putative Mg2+ coordination site, while Asp166 forms a hydrogen bond with Lys31 in Walker A type motif required for ATP hydrolysis. Concomitantly, compared to E. coli ArsA, the N-terminal part (i.e. residues 27 and 28) of the Get3 P-loop is apart from the phosphate moiety of the bound ADP. As a result, the Get3 NBS is loosely packed and appears unfavorable for ATP hydrolysis (Fig. 2A,B).

The structurally homologous N- and C-terminal halves of E. coli ArsA form a pseudo dimer, the lower half of which is tightly closed by the interaction with the substrate metalloid (Fig. S5), in striking contrast to the open dimer of Get3 (Fig. 4A). Previous biochemical and structural studies of ArsA have suggested that the substrate metalloid binding brings together the N- and C-terminal halves to form the closed dimer enhancing the ATPase activity of ArsA (Rosen et al. 1999; Zhou et al. 2000). This raises the possibility that Get3 might alternate its dimer conformation from the open form to the closed form to stimulate ATPase activity. Indeed, we could build a reasonable closed-dimer model of Get3, guided from the pseudo closed dimer structure of ArsA (Fig. 4B).

Figure 4.

 Chemical cross-link of mutant Get3. (A) Get3 dimer model from the crystal structure. As part of the helical domain in subunit B is disordered in our structure, we superposed subunit A to B in this model. (B) The closed dimer model of Get3. The model is based on the superposition to ArsA pseudo dimer. (C) Cross-linked dimer band of the purified S132C and M200C mutant Get3. The upper band corresponding to the cross-linked dimer by the engineered cysteine residue pair has disappeared or is decreased in intensity in the presence of DTT. Western blotting analysis using the anti-polyhistidine antibody verified that these bands are Get3. (D) Chemical cross-link between the cysteine residues was stimulated by the oxidant CuCl2. Get3 incubated in the presence or absence of either ATP, AMP-PNP or ADP was allowed to react with CuCl2. The production of the cross-linked Get3 dimer was most enhanced in the presence of ATP.

To test if this is the case, chemical cross-linking experiments were performed with the engineered cysteine residue pairs in the presence of ATP, AMP-PNP or ADP or in the absence of any nucleotides (Wada et al. 2009). We mutated Ser132 (in the NBD) and Met200 (in the helical domain), which protrude from the inside wall of the dimer (Fig. 4A). The inter-subunit distances calculated from the positions of Ser132 and Met200 are 28 and 32 Å in the open dimer conformation (Fig. 4A) and 6.0 and 6.2 Å in the predicted closed dimer conformation, respectively (Fig. 4B). When analyzed by nonreducing SDS-PAGE, the S132C and M200C mutants of Get3 purified without any reducing reagents showed bands corresponding to the cross-linked dimer in addition to those corresponding to the monomer (Fig. 4C,D). These additional bands corresponding to the cross-linked dimer disappeared or decreased in intensity in the presence of the reducing reagent, dithiothreitol (Fig. 4C). In contrast, addition of the oxidant CuCl2 increased the intensity of the bands corresponding to the cross-linked dimer (Fig. 4D). In both mutants, particularly in the S132C mutant, production of the cross-linked dimer was most enhanced in the presence of hydrolyzable ATP. This result suggests that the closed dimer is stabilized in the transition state during ATP hydrolysis but not in the initial ATP-binding or posthydrolysis state (equivalent to the nonhydrolyzable AMP-PNP-bound or ADP-bound state, respectively). Similar transition state-specific stabilization of the protein–protein interaction has been observed in the Ras–RasGAP complex, as described below (Scheffzek et al. 1997). Further, the cross-link of the M200C mutant indicates that the helical domains from each subunit associate to stabilize the closed dimer conformation, as shown in Fig. 4B. ATPase activity was drastically reduced in the Get3Δ190–209 mutant that lacks the closed dimer interface in the helical domain (Table S1). This suggests that the stable closed dimer conformation is necessary for effective ATPase activity. Thus, the helical domain not only binds to the substrate TA proteins but also likely stabilizes the closed dimer conformation for ATP hydrolysis.

ATP and dimer-independent TA protein binding

In order to test if the binding to TA proteins is affected by the dimer formation or conformation, we performed the aforementioned TA protein-binding assays using K31A, D166N and C285S mutant Get3 proteins. Lys31 is the conserved lysine residue in the Walker A box, whose mutation has been known to abolish the ATP binding in many Walker-type ATPases (Saraste et al. 1990; Yamagata & Tainer 2007). Asp166 is a putative Mg2+ coordinating residue and D166N mutant of Get3 cannot hydrolyze ATP (Table S1). Therefore, both mutant proteins are likely unable to form a stable closed dimer. C285S mutant is deficient in the Zn2+ coordination and could not form a dimer (Fig. S3). As shown in Fig. 5A,B the amount of the full-length Sec22p copurified with either K31A, D166N or C285S mutant was comparable to that with wild type Get3. This result clearly shows that TA protein binding is totally independent of both dimer formation and ATP-induced conformational change. In addition, the membrane recruitment of Get3 by the interaction with the receptors, Get1 and Get2, are also independent of ATP (Schuldiner et al. 2008). Therefore the dimer conformation-dependent ATP hydrolysis might be coupled with TA protein release.


Two groups have recently reported the structures of yeast Get3 (Mateja et al. 2009; Suloway et al. 2009). Especially, Mateja et al. (2009) showed the tightly closed dimer form of S. cerevisiae Get3 which is trapped by ADP-AlF4. The closed dimer structure of Get3 represents apparently more favorable structure for ATP hydrolysis than our open dimer structure. At the NBS in the closed conformation, the bound ADP forms additional interactions with the adjacent subunit. In addition, the bound AlF4, which mimics a γ-phosphate moiety in the transition state during ATP hydrolysis, interacts with the lysine residue from the adjacent subunit, similar to the interaction between RasGAP and AlF3 bound to Ras in the transition state complex of Ras and RasGAP (Scheffzek et al. 1997). These indicate that the closed dimer conformation is essential for ATP hydrolysis and is most stabilized in the hydrolysis transition state, consistent with our idea based on the cross-link experiments. The closed dimer-dependent ATP hydrolysis also provides an important implication for the conserved DTAPTGH motif (Fig. S2). As shown in Fig. 3B, the Get3 D166TAPTGH172 motif starts at Asp166, which is essential for ATPase activity, and continues through the helical domain. Thus, the interaction between helical domains in the closed dimer will affect the NBS to stimulate ATPase activity via the DTAPTGH motif. A similar concept has been proposed for the ArsA structure (Zhou et al. 2000). The ArsA ATPase activity is enhanced by substrate metalloid binding (Rosen et al. 1999). The ATPase catalytic (Mg2+ binding) site and the substrate (Sb3+) binding site of ArsA are connected by the D142/447TAPTGH148/453 motif, where Asp142/447 coordinates the catalytic Mg2+ via a water molecule and His148/453 is responsible for the substrate Sb3+ binding (Zhou et al. 2000) (Fig. 3B). Thus, the signal by the substrate binding is transmitted to the ATPase catalytic site by the DTAPTGH motif. In addition, as the ATPase activity of Get3 has been considered to be coupled with TA protein release (Stefanovic & Hegde 2007), the DTAPTGH motif might also modulate ATP hydrolysis and TA protein release.

Mateja et al. (2009) proposed that the TA protein is bound in the conserved hydrophobic cluster near the root of the helical domain and α5 in our structure. In their model, TA protein recognition is dependent on the tight dimer formation induced by ATP binding. Currently we would not like to conclude which site is the correct TA protein binding site, as no structural data about the complex between Get3 and TA protein are available. However, it should be noted that the initial TA protein recognition is independent of ATP from our coexpression experiments and the biochemical studies by other groups (Stefanovic & Hegde 2007). Thus, we propose the working model of the membrane insertion mechanism by Get3 as follows: Get3 alone is in the dynamic equilibrium between the open dimer conformation shown in our crystal structure and the closed dimer conformation detected by our cross-link experiment (Fig. 6A). After Get3 specifically binds to TA proteins (Fig. 6B), the complex will be recruited to the membrane via an interaction with membrane-bound receptors, which are identified as Get1/2 in S. cerevisiae (Schuldiner et al. 2005, 2008; Auld et al. 2006) (Fig. 6C). In contrast to the first two steps, the last insertion step requires ATP hydrolysis (Stefanovic & Hegde 2007; Schuldiner et al. 2008). We propose that the interaction with the receptor might stabilize the closed dimer conformation of Get3. This proposed mechanism satisfies the following three requirements: (i) Get3 forms a closed dimer during ATP hydrolysis (shown in our cross-link experiments); (ii) Get3 interacts with the receptor in order to insert TA proteins to the membrane (Schuldiner et al. 2008); (iii) TA protein release is dependent on ATP hydrolysis (deduced from our study). Given the stabilization of the closed-dimer conformation by binding to the receptor, the interaction between helical domains would stimulate ATP hydrolysis via the DTAPTGH motif (Fig. 6D). Then, ATP hydrolysis might induce the local conformational change in the NBS, which should be transmitted back to the helical domain to release the substrate TA proteins via the DTAPTGH motif (Fig. 6E).

Figure 6.

 Model for the membrane insertion mechanism by Get3. (A) Get3 alone is in both open and closed conformations. (B) Get3 specifically binds to TMD of TA protein via the helical domain. One Get3 dimer binding to one TA protein molecule is simply depicted here for clarity, though no experimental data for a stoichiometry between Get3 and TA protein are available. (C) The complex is recruited to the membrane by the Get1/2 receptors. (D) The closed dimer conformation stimulates ATPase activity. (E) ATP hydrolysis induces TA protein release and subsequent membrane insertion.

Experimental procedures

Cloning, expression and purification of yeast Get3

Get3 gene was amplified from S. cerevisiae genomic DNA (strain NRRL Y-53, ATCC) using standard PCR method, and cloned into pET-21a vector. The amino sequence translated from the cloned DNA sequence has a single mutation against that of strain S266C, of which the genome was sequenced. The mutated site G155D is a nonconserved residue among all organisms, suggesting that this mutation is due to the difference in yeast strain. Initially, we overexpressed and purified the full-length Se-Met Get3 with the C-terminal hexahistidine tag under the nonreducing condition. The Se-Met-labeled Get3 was expressed in E. coli Rosetta-gami 2 (DE3) cells (Stratagene, Santa Clara, CA, USA). Cells were grown in CoreTM medium (Wako, Osaka, Japan) without methionine to an OD600∼0.6, then Se-Met (final concentration of 0.5 g/L) was added and protein expression was induced with 0.4 mm isopropyl β-d-thiogalactopyranoside. Cells were thawed in 20 mm Tris-HCl (pH 8.0) buffer containing 500 mm NaCl and lyzed by sonication. After centrifugation, the supernatant was applied onto Ni2+-chelating affinity column, and the eluted protein was further purified by anion exchange and size exclusion chromatographies without any reducing reagents. After the structure was determined, we found that a Zn2+ is coordinated by cysteine residues. Thus, the proteins used later were purified under the reducing condition. Under the reducing condition, Get3 was overexpressed in E. coli Rosetta (DE3) cells and purified with 2 mm DTT. The Get3 mutants D57N, D166N and, C285S were generated using Quickchange site-directed mutagenesis kit (Stratagene), and proteins were purified in the same manner as the wild type Get3 under the reducing condition.

Crystallization of Get3

All crystals were obtained by sitting drop vapor diffusion method at 20 °C. The crystals of the nucleotide-free Get3 were obtained by mixing equal volumes of Get3 (400 μm of Get3 with 250 μm ZnSO4, 5 mm AMPPNP, 10 mm MgCl2 and 2 mm DTT) with the reservoir solution [13.5% PEG 3350, 0.18 m tri-sodium citrate, 9% MPD and 90 mm MES (pH 6.5)]. Crystals were directly flash cooled at 100 K. The diffraction data set was collected at BL17A in PF (Tsukuba, Japan). Crystals belong to the space group P21212 with cell dimensions of = 114.9 Å, = 221.6 Å, = 49.3 Å. The Se-Met-labeled Get3 with 1 mm AMP-PNP, 10 mm MgCl2 and 1 mm ZnCl2 was crystallized with the reservoir solution (15% polyethylene glycol 3350, 0.2 m tri-sodium citrate, 10% MPD). The SAD data set was collected at BL41XU in SPring-8 (Hyogo, Japan). For the identification of Zn2+, the diffraction data sets for the native Get3 at two wavelengths flanking the Zn2+ absorption edge (i.e. 1.25000 Å for high energy remote and 1.29000 Å for low energy remote) were collected up to 3.35 and 3.50 Å resolutions, respectively.

The ADP-bound Get3 (400 μm Get3 with 2.5 mm ADP, 5 mm MgCl2 and 100 μm ZnSO4) was crystallized in the same condition as the nucleotide-free Get3, except that HEPES (pH 7.5) was used instead of MES. The diffraction data set was collected at BL41XU in SPring-8 (Hyogo, Japan). The crystal was in the same space group with the similar unit cell dimensions of a = 115.3 Å, = 222.6 Å, = 49.3 Å, respectively. All diffraction data sets were processed with HKL2000 program suite (Otwinowski 1993). The data collection statistics are shown in Table 1.

Structure determination of Get3

The nucleotide-free Get3 structure was determined by the SAD method. The identification of Se sites and the subsequent phase calculation were performed with the programs shelx97 (Sheldrick 2008) and sharp (de la Fortelle & Bricogne 1997). The phases were improved by solvent flipping with the program solomon (Abrahams & Leslie 1996). We manually built the initial model by the program coot (Emsley & Cowtan 2004). The phases were further improved by noncrystallographic symmetry averaging with DM using a mask produced from ∼80% of the atomic model of Get3 (Cowtan & Main 1996) A residual high peak in the anomalous difference Fourier map calculated from the Se-Met-labeled Get3 data set was interpreted as a Zn2+. The anomalous signal at the site of Zn2+ was clearly evident in the anomalous difference Fourier map at a high-energy remote wavelength (1.25000 Å) for Zn2+, whereas no signal was observed in that at a low-energy remote wavelength (1.29000 Å). The atomic model was refined to the final R/Rfree values of 0.237/0.281 with the program cns (Brunger et al. 1998).

For the ADP-bound Get3 structure, the initial model was first obtained by rigid body refinement. The electron density of the bound ADP was readily visible in both 2Fo-Fc and Fo-Fc electron density maps (Fig. S1). The model was refined with the program cns (Brunger et al. 1998) to the final R/Rfree values of 0.252/0.283. Stereochemistry was assessed by the program procheck (Laskowski et al. 1993). Refinement statistics are shown in Table 1. All molecular graphics were prepared with PyMOL (DeLano Scientific;

ICP-MS measurement

The zinc occupancy was quantified by ICP-MS. 58 μm of Get3 purified under the reducing condition was diluted 10-fold by MilliQ (Millipore, Billerica, MA, USA) in order to be measured by HP-4500 ICP-MS (Agilent Technologies, Santa Clara, CA, USA). The zinc occupancy in wild type Get3 was estimated to be 46.6 ± 2.1%, whereas that in C285S mutant Get3 was not detectable.

Coexpression and copurification of the complex between Get3 and Sec22p

The GST, GST-Get3/Get3Δ190-209 and SEC22/SEC22ΔTMD genes were cloned into the pET-Duet1 vector. The expression of the protein was performed in the same manner as the C-terminal histidine-tagged Get3. The cells were suspended in 1× PBS buffer and disrupted by sonication. After centrifugation, 40 mL of the supernatant (8.5 mg/mL of total proteins) was applied to 5 mL of glutathione sepharose FF column (GE Healthcare, Buckinghamshire, UK). After twice washing by 25 mL of 1× PBS buffer, the protein was eluted in 20 mL of 20 mm Tris-HCl (pH 7.7) buffer containing 150 mm NaCl and 15 mm reduced glutathione. The protein complex was analyzed by SDS-PAGE and stained with Coomassie Brilliant Blue. The densitometry of the protein band was performed with the software Imagej ( Although the hexahistidine tag was fused to the C-terminus of Sec22p for detection by Western blotting, the full-length Sec22p-(His)6 represented a weak signal slightly above the nonspecific signals. Thus, we confirmed Sec22p by the N-terminal amino acid sequencing with the perfect match of the N-terminal 10 residues. By contrast, Sec22pΔTMD-(His)6 was clearly demonstrated by Western blotting with the monoclonal anti-polyhistidine antibody (Sigma, St Louis, MO, USA).

Chemical cross-linking of S132C and M200C mutant Get3

S132C and M200C mutant Get3 and wild type Get3 were expressed in Rosetta (DE3) E. coli cells and purified without the reducing reagents. A 10 μm protein in 20 mm Tris-HCl (pH 7.5), 100 mm NaCl, was incubated with or without 1 mm nucleotides and 10 mm MgCl2 at 30 °C. Disulfide bond formation was stimulated by adding CuCl2 at the final concentration of 10 μm. After 20 min incubation at 25 °C, samples were analyzed by SDS-PAGE.


We thank C. Toyoshima for the support of this research and critical advice. We thank L. Craig for critical reading of the manuscript. We are grateful to Y. Setsuda and R. Matsumoto for the ICP-MS measurement. We are also grateful to S. Tomioka for the N-terminal amino acid sequencing. We thank the beam-line staffs at BL-5A, BL17A of Photon Factory (Tsukuba, Japan) and BL41XU of SPring8 (Hyogo, Japan) for technical help during data collection. This work was supported by grants from MEXT to A.Y., S.F. and H.M. Y.S. and M.Y. are supported by JSPS research fellowships for young scientists.

Accession numbers

The coordinates and structure factors have been deposited in the Protein Data Bank with the accession codes 3A36 (nucleotide-free Get3) and 3A37 (ADP-bound Get3).