Crystal Structures of Nucleotide-Free and Glutathione-Bound Mitochondrial ABC Transporter Atm1

See allHide authors and affiliations

Science  07 Mar 2014:
Vol. 343, Issue 6175, pp. 1137-1140
DOI: 10.1126/science.1246729

Crossing the Membrane

Adenosine triphosphate (ATP)–binding cassette (ABC) transporters couple ATP hydrolysis to the translocation of a wide variety of substrates across cell membranes. Srinivasan et al. (p. 1137) describe the structure of a yeast mitochondrial transporter involved in Fe-S protein biogenesis. The structure reveals bound glutathione, which suggests that glutathione is part of the translocated substrate. J. Y. Lee et al. (p. 1133) describe the structure of a bacterial ABC transporter that confers protection against silver and mercury. This protein also binds glutathione derivatives. The structure provides insight into how ligand interactions are coupled to ATP hydrolysis.


The yeast mitochondrial ABC transporter Atm1, in concert with glutathione, functions in the export of a substrate required for cytosolic-nuclear iron-sulfur protein biogenesis and cellular iron regulation. Defects in the human ortholog ABCB7 cause the sideroblastic anemia XLSA/A. Here, we report the crystal structures of free and glutathione-bound Atm1 in inward-facing, open conformations at 3.06- and 3.38-angstrom resolution, respectively. The glutathione binding site includes a residue mutated in XLSA/A and is located close to the inner membrane surface in a large cavity. The two nucleotide-free adenosine 5′-triphosphate binding domains do not interact yet are kept in close vicinity through tight interaction of the two C-terminal α-helices of the Atm1 dimer. The resulting protein stabilization may be a common structural feature of all ABC exporters.

Mitochondria contain several adenosine 5′-triphosphate (ATP) binding cassette (ABC) transporters that are crucial for organellar communication with the cytosol and nucleus (14). These proteins are located in the mitochondrial inner membrane with the nucleotide binding domain (NBD) facing the matrix and hence are thought to function as exporters. Yeast Atm1, vertebrate ABCB7, and plant Atm3 are involved in the biogenesis of cytosolic-nuclear iron-sulfur (Fe/S) proteins, including DNA polymerases, primases, and helicases involved in genome integrity (514). The sulfur present in cytosolic-nuclear Fe/S proteins is generated from cysteine inside mitochondria by the cysteine desulfurase complex Nfs1-Isd11. This complex and other members of the mitochondrial iron-sulfur cluster (ISC) assembly machinery synthesize an unknown, sulfur-containing molecule that is exported by Atm1 and used by the cytosolic Fe/S protein assembly (CIA) system (15, 16). This molecule also signals the iron status to the nucleus (17). Upon ISC protein or Atm1 depletion, cellular iron uptake is increased, and mitochondria accumulate up to 30-fold higher iron concentrations (5, 9, 18, 19). Depletion of the sulfhydryl oxidase Erv1 or the redox molecule glutathione (GSH) shows a strikingly similar phenotype to Atm1 deficiency, suggesting that the three components cooperate in the export process (2023). In fact, Atm1-deficient mitochondria show an increased content of GSH (18). Moreover, the adenosine triphosphatase (ATPase) activity of Atm1 is stimulated by GSH, suggesting that this tripeptide may be directly involved in the export process (24).

Homologs of yeast Atm1 are found in virtually all eukaryotes and some bacteria and usually show ~50% sequence identity (fig. S1A). The function of the bacterial Atm1-like proteins is unknown, but they may also participate in the biogenesis of a metal-sulfur center (25). The human, mouse, and plant proteins can replace yeast Atm1, underlining the conserved function of this ABC transporter in eukaryotic Fe/S protein biogenesis and iron regulation. Mutations in human ABCB7 cause X-linked sideroblastic anemia and cerebellar ataxia (XLSA/A), an iron storage disease characterized by diminished cytosolic Fe/S protein function, iron-loaded mitochondria (sideroblasts), and defects in heme metabolism (2629). Recently, decreased expression of ABCB7 was claimed to play a role in refractory anemia with ring sideroblasts (RARS), a myeloid malignancy that can transform to acute leukemia (30, 31).

Structural information on Atm1 is essential for better understanding of its central function in cellular iron metabolism, identification of its substrate, and elucidation of the transport mechanism. To date, structural information on eukaryotic ABC transporters is available only for P-glycoprotein (P-gp), which is implicated in multidrug resistance (MDR), and the mitochondrial ABCB10, which is involved in heme metabolism in erythroid cells (3234); yet, these proteins differ from Atm1 in tissue specificity, intracellular localization, subunit composition, function, and substrate. We solved the three-dimensional structure of Saccharomyces cerevisiae Atm1 with x-ray crystallography using the single anomalous dispersion (SAD) method (Fig. 1 and fig. S1B) (35). The experimental phases were obtained from soaking the native crystals with tantalum bromide cluster (Ta6Br12) (fig. S2). Data collection and refinement statistics are summarized in table S1. A stereoview of the (2Fo-Fc) electron density map for the complete Atm1 dimer is shown in fig. S3. The final model encompasses the complete polypeptide sequence of Atm1 except for the first 98, nonessential residues, including the mitochondrial presequence (fig. S1A) (24). Two nucleotide-free structures were determined, one without and one with bound GSH to a resolution of 3.06 and 3.38 Å, respectively (Figs. 1 and 2A). Both structures were captured in an inward-facing, open conformation that conforms to the known eukaryotic ABC exporter fold and represents a pretranslocation state. The GSH-bound and -free structures were virtually identical, with an overall root-mean square deviation (RMSD) of 0.3 Å for the complete molecule (fig. S4).

Fig. 1 Crystal structure of the mitochondrial ABC transporter Atm1.

(A) Illustration of the nucleotide-free Atm1 homodimer (S. cerevisiae), with monomers colored in green and red and the NBDs facing the mitochondrial matrix. The approximate positioning of the lipid bilayer is indicated by dashed lines. (B) The Atm1 monomer depicted in rainbow colors shows the crossover of the TM4 and TM5 helices to the other monomer (not shown), resulting in the formation of a V-shaped molecule with a large cavity. The N and C termini, the NBDs, and the coupling helices ICL1 and ICL2 that form the transmission interface between the transmembrane helices and the NBDs are indicated. (C) Electrostatic surface potential representation of the Atm1 dimer (blue, positive; red, negative; and white, neutral). The substrate binding cavity located near the hydrophilic surface of the membrane is positively charged.

Fig. 2 Crystal structure of mitochondrial Atm1 in complex with glutathione.

(A) GSH represented by yellow spheres is well accommodated in the cavity formed by the Atm1 dimer. The residues corresponding to those mutated in XLSA/A disease are marked as blue spheres. (B) GSH (shown in stick representation) mainly interacts by forming hydrogen bonds with residues D398, R284, and R280, indicated by black dashed lines. The other residues that line the cavity are N390, N343, S394, and R397. (C) A section of the experimental (2Fo-Fc) electron density map contoured at 1σ (gray) shows Atm1-bound GSH (stick representation) in close vicinity to TM4, TM5, and TM6 (orange sticks). (Inset) A close-up view of the electron density for GSH from the same map.

Atm1 assembled as a dimer in the crystallographic lattice (fig. S5), and the molecules were packed in a “head to tail” arrangement. The ATPase activity of Atm1 was almost fully recovered after detergent solubilization of the crystals, suggesting that purification and crystallization did not affect function (fig. S6). The dimer consists of 12 transmembrane (TM) helices in two bundles of six helices per monomer. The six TM helices from each monomer do not form a single structural unit; rather, the long TM4 and TM5 helices reach out to the other monomer, thus forming an intertwined “domain-swapped” membrane-integrated arrangement (Fig. 1, A and B). This configuration creates a V-shaped molecule enclosing a 6900 Å3, positively charged cavity close to the hydrophilic phase of the matrix side of the inner membrane (Fig. 1C).

In several crystals, this cavity contained an electron density that fits best to GSH (Fig. 2). The ligand binding region corresponding to TM4, TM5, and TM6 was well resolved (fig. S7). Both purified Atm1 and the crystals contained GSH as verified by means of mass spectrometry, biochemical analysis, and equilibrium titration [apparent dissociation constant (Kd) = 23 μM] monitored with microscale thermophoresis (figs. S8 and S9A) (36). Verapamil, a molecule transported by the multidrug transporter P-gp, did not bind to Atm1 (fig. S9B). Because both purification and crystallization were performed without GSH, the compound must have associated with Atm1 after its synthesis in Escherichia coli, indicating stable complex formation during purification at low temperature. A bacterial homolog of Atm1 binds GSH in a similar location (37). The binding site of the Atm1-bound GSH ligand substantially differed from that of hydrophobic inhibitors of P-gp (32), despite the similar size and shape of the cavities of both transporters (fig. S10). These findings are consistent with the anticipated entry points of hydrophilic ligands such as GSH, as opposed to hydrophobic P-gp substrates that are thought to reach the cavity from the lipid phase (32). Last, GSH docking calculations with the Atm1 crystal structure coordinates supported a high-affinity binding site for GSH within the cavity, in close vicinity to the GSH binding site observed in the crystal structure (fig. S11).

The cavity in the Atm1 structure is lined with mostly charged and polar amino acid residues and likely serves as the substrate binding site. The two TM6 helices bend out of the membrane, forming a “kink” at residue S394 before connecting to the NBD through the linker region, offering flexibility to accommodate the GSH ligand (Fig. 2B). (Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.) GSH fits snugly to a positively charged cavity through interaction with residues R280 and R284 (TM4), N343 (TM5), N390, S394, R397, and D398 (TM6) (Fig. 2, B and C, and fig. S7). All GSH-coordinating residues are conserved in Atm1-like proteins, including human ABCB7 and bacterial homologs (fig. S1A). One of these residues (D398) corresponds to human E433, which is changed to lysine in patients with XLSA/A, clearly showing the physiological relevance of GSH ligand binding (Fig. 2A) (27). Both yeast Atm1 and human ABCB7 proteins carrying this mutation show diminished activity in the maturation of cytosolic Fe/S proteins. Likely, the site of GSH interaction is part of the substrate binding region of Atm1. GSH itself cannot be the physiological substrate because it is synthesized in the cytosol. What may be the identity of the Atm1 substrate? Cell biological data demonstrate that Atm1 exports a sulfur-containing molecule generated by the core mitochondrial ISC assembly machinery (15, 16). GSH may be part of the exported moiety because GSH, like Atm1, is essential for both cytosolic-nuclear Fe/S protein biogenesis and cellular iron regulation (2123). Therefore, in the simplest view GSH may be converted to glutathione persulfide (GSSH) and directly used as a sulfur carrier. Alternatively, a GSH-coordinated [2Fe-2S] cluster was reported to be stable in water (38). This or a similar compound may be synthesized by the mitochondrial ISC system and exported by Atm1. Atm1 structures determined from crystals soaked or cocrystalized with such compounds will help identifying its substrate.

Three of the XLSA/A patient mutations are located in the membrane domain either on the matrix [E208D in long TM2 helix; yeast E173 (29)] or intermembrane space sides (I400M between TM5 and TM6, yeast V365 (26); V411L in TM6, yeast V376 (28)] (Fig. 2A and fig. S1A). (In the mutants, other amino acids were substituted at certain locations; for example, E208D indicates that glutamic acid at position 208 was replaced by aspartic acid.) All of these amino acid exchanges are conservative (fig. S1A), indicating that even subtle changes in ABCB7 function may lead to the disease phenotype. Conspicuously, V376 in yeast Atm1 undergoes hydrophobic interactions with its counterpart residue in the other monomer, thus tightly closing the transmembrane channel toward the intermembrane space (Fig. 2A and fig. S12). Replacement of valine by the larger hydrophobic residue leucine is pathogenic (28). The mutation likely affects the tightness of the transport channel and/or the mobility of the transmembrane segments during the transport process.

One key feature of both Atm1 structures was the complete resolution of the C terminus that consists of an α-helix 24 amino acid residues long (Fig. 3A and fig. S1A). Such a structural feature has not been observed previously in any other ABC transporter crystal structure, although these proteins usually contain C-terminal segments that potentially could form α-helices. A representative electron density map (2Fo-Fc) contoured at 1σ for the C-terminal helix reveals an unambiguous density in this region (fig. S13). The helix is in close proximity to its counterpart helix from the second monomer, and both helices tightly interact in a crossover fashion, with distances ranging from 2.65 to 4.0 Å (Fig. 3B and table S2). The two helices form a connecting bridge between the two NBDs, thus locking the Atm1 dimer in the inward-facing, open conformation. Further, the α-helix from one monomer is long enough to make extensive contacts to residues G469 to S471 of the Walker A motif of the other monomer (Fig. 3B). It is evident that during the ABC transporter working cycle, the two α-helices either have to unlock and slide along each other or become more tilted to allow the tight interaction of the two NBDs after ATP binding. A superposition of the NBD structure of Atm1 with those of the known ABC exporters (mouse and Caenorhabditis elegans P-gp, bacterial multidrug-resistance protein Sav1866, and human mitochondrial ABCB10) clearly reveals the unprecedented character of the C-terminal α-helix of Atm1 (fig. S14). The short helix seen in C. elegans P-gp is present at the end of only one of the NBDs.

Fig. 3 The C-terminal α-helices stabilize the Atm1 dimer in the inward-facing, open conformation.

(A) The Atm1 NBDs are presented in gray, the Walker A loop in blue, and the two crossover C-terminal helices (table S2) in green and red as an illustration (side chains as sticks). The C-terminal helices include five residues of the strep-tag used for purification (side chains not shown for clarity). (B) The C-terminal residues E689 to L690 of one Atm1 monomer (green) almost touch the Walker A loop (G469 to S471, sticks) of the other monomer (blue). Residues of the strep tag were omitted here. (C) Wild-type (W303) or GalL-ATM1 yeast cells were transformed with plasmids encoding no protein (–), the model Fe/S protein Rli1-HA, and wild-type (WT) and mutated Atm1 (asterisk, truncated). Cells were grown for 64 hours in SD medium for depletion of Atm1, radiolabeled with 55Fe, and analyzed for 55Fe/S cluster insertion into Rli1-HA with immunoprecipitation and scintillation counting. (Bottom) Indicated proteins were visualized by the immunoblotting of cell extracts (pound sign, cross-reacting band).

Flexible interactions at the C termini might be a common feature of many if not all ABC exporters and may serve to stabilize the protein. To test this idea, we truncated the C-terminal part of Atm1 to remove half or all of the helix (mutants D676* and L666*, respectively). The mutant proteins were expressed from plasmids in regulatable GalL-ATM1 yeast cells in which the endogenous Atm1 can be depleted by growth in the presence of glucose (5). Both mutant proteins rescued the growth defect of Atm1-depleted cells at 30°C; yet at 37°C, a slightly slower growth was observed for Atm1-L666* (full deletion of α-helix) (fig. S15A). The growth defect was accompanied by substantially diminished protein levels of Atm1-L666* and a 50% reduced activity in the biogenesis of a representative cytosolic Fe/S protein Rli1 (Fig. 3C) (39). A similar diminution in the extent of Fe/S protein assembly can cause XLSA/A (27). Nevertheless, the increased iron levels of Atm1-depleted mitochondria could be normalized by both Atm1 mutant proteins (fig. S15B), suggesting that the truncation mainly affected protein stability rather than function.

On the basis of the two structures reported here, a model for substrate translocation is proposed. The Atm1 substrate enters the inward-facing, open structure from the matrix side to bind in the hydrophilic cavity (fig. S16). Because the GSH ligand was bound in the absence of ATP, substrate binding may occur before nucleotide binding. The scissor-like interaction of the two C-terminal α-helices of the Atm1 dimer may keep the two NBDs close together before their ATP-induced, tight association, stabilizing the protein. This effect may be relevant for other bacterial and eukaryotic ABC exporters, at least under adverse environmental conditions. The ATP-driven structural rearrangement leads to an outward-facing conformation that may be similar to a homology model calculated for Atm1 (fig. S16, right) by using a structure of the Sav1866 ABC transporter (40). As a result of the intimate NBD contact, the transmembrane helices rearrange and thereby create an exit pathway for substrate release. The pretranslocation state can be understood on the basis of the crystal structures, yet the complex ATP-driven alterations require further studies. Our structures will boost further mechanistic insights into Atm1-ABCB7 function, substrate interaction, and pathology.

Supplementary Materials

Materials and Methods

Figs. S1 to S16

Tables S1 and S2

References (4161).

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank U. Linne for mass spectrometry, S. Freibert for help with thermophoresis, M. Stümpfig and N. Herlerth for expert technical assistance, D. C. Rees for discussion, and the beamline staff for support at the X06SA beamline, Swiss Light Source, Paul Scherrer Institute, Villigen, Switzerland. R.L. acknowledges generous support from Deutsche Forschungsgemeinschaft (SFB 593, SFB 987, and GRK 1216), von Behring-Röntgen Stiftung, LOEWE program of state Hessen, Max-Planck Gesellschaft, and the Feldberg Foundation. Coordinates and structure factors have been deposited in the Protein Data Bank (PDB accession codes 4MYC and 4MYH).
View Abstract

Stay Connected to Science

Navigate This Article