Structural basis for conformational switching and GTP loading of the large G protein atlastin

Abstract

Atlastin, a member of the dynamin superfamily, is known to catalyse homotypic membrane fusion in the smooth endoplasmic reticulum (ER). Recent studies of atlastin have elucidated key features about its structure and function; however, several mechanistic details, including the catalytic mechanism and GTP hydrolysis-driven conformational changes, are yet to be determined. Here, we present the crystal structures of atlastin-1 bound to GDP·AlF(4)(-) and GppNHp, uncovering an intramolecular arginine finger that stimulates GTP hydrolysis when correctly oriented through rearrangements within the G domain. Utilizing Förster Resonance Energy Transfer, we describe nucleotide binding and hydrolysis-driven conformational changes in atlastin and their sequence. Furthermore, we discovered a nucleotide exchange mechanism that is intrinsic to atlastin's N-terminal domains. Our results indicate that the cytoplasmic domain of atlastin acts as a tether and homotypic interactions are timed by GTP binding and hydrolysis. Perturbation of these mechanisms may be implicated in a group of atlastin-associated hereditary neurodegenerative diseases.

Figure 1

Structures of atlastin-1. (A) Topology of atlastin-1. The globular G domain (orange) is connected to the middle domain (blue) by a short linker (green). This N-terminal, cytoplasmic unit is followed by two transmembrane α helices (black) and a C-terminal amphipathic helix (grey), which has been shown to interact with the lipid membrane. (B) Protomer structures of crystal forms 1 and 2. The GDP-bound structures of atlastin-1^1–446 were aligned relative to their G domains (Byrnes and Sondermann, 2011). Nucleotide and Mg²⁺ are shown as sticks and spheres, respectively. (C) Protomer structures of crystal form 3. The GDP·AlF₄⁻-bound form 3 structure of atlastin-1^1–446-N⁴⁴⁰T was superimposed on the GDP-bound form 1 structure with the G domain as the reference. The form 1 structure is shown in grey and the form 3 structure is coloured according to (A). Position of missense mutation in form 3 structure is indicated and its Cα is shown as a black sphere. An F_o−F_c omit map for the nucleotide (inset) is contoured at 4.0 sigma.

Figure 2

Crystallographic dimers of atlastin-1^1–446. (A) Atlastin-1 dimers. Crystallographic dimers observed in crystal form 2 (extended, P2₁2₁2₁ symmetry, left), form 1 (relaxed-parallel, P6₅22 symmetry, middle), and form 3 (tight-parallel, P2₁2₁2 symmetry, right) are shown. Bound nucleotides and metal ions are shown as sticks and spheres, respectively. (B) Crystal form 3 dimer interface. Interface residues are shown in colour on the surface of one half of the dimer.

Figure 3

Catalytic mechanism of atlastin-1. (A) Conformational changes upon GTP analogue binding. The conformational changes in the nucleotide binding site between GDP-bound form 1 (grey) versus GDP·AlF₄⁻-bound form 3 (coloured) are shown after superpositioning of the respective G domains. Nucleotide binding and switch motifs (motifs G1–G4) of the form 3 crystal structure are coloured in shades of blue. On the right, a sequence alignment of several dynamin superfamily members shows conserved regions involved in nucleotide binding and hydrolysis. Strictly conserved residues are highlighted in green. Residues shown in (B) are marked with an asterisk. (B) Residues involved in nucleotide hydrolysis. The nucleotide binding pocket of crystal form 3 (upper) is shown, with residues making direct or indirect contacts with the phosphate groups of the nucleotide analogue are shown as sticks. Mg²⁺ is shown as a green sphere. The attacking water molecule (labelled ‘aw’) and Mg²⁺-coordinating water molecules are shown as red spheres. Mutation of R⁷⁷ to alanine abolishes GTPase activity (lower panel). The GTPase activity was determined by measuring the production of inorganic phosphate over time at various protein concentrations. (C) Nucleotide-dependent oligomerization of atlastin-1^1–446-R⁷⁷A SEC-MALS data for wild-type (left panels) and mutant (R⁷⁷A, right panels) atlastin-1^1–446 is shown. The signal from the 90°-light scattering detector and refractive index detector is shown as coloured, solid lines (apo, red; GppNHp bound, green; GDP bound, purple; GDP·AlF_x bound, orange) and black, dashed lines, respectively (left Y axis). Average molecular weight calculations across the protein peak are shown as black circles (right Y axis). The theoretical molecular weight (based on primary sequence) for the monomer and dimer is shown as horizontal dashed lines. Proteins (30–40 μM) were incubated with nucleotides (2 mM) at least 30 min prior to SEC-MALS analysis.

Figure 4

Atlastin-1 middle domain FRET using C-terminal ECFP/EYFP fusions. (A) Experimental design of measuring middle domain dimerization. Cartoon depiction of fluorescent protein-labelled atlastin-1^1–446 in various crystallographic dimer conformations. Estimated length measurements are based on the crystal structures shown in Figure 2A. (B) Nucleotide-dependent dimerization of atlastin-1^1–446-ECFP/EYFP-fusion proteins. Molecular weight fractions for GppNHp- or GDP·AlF₄⁻-bound atlastin-1 were determined by SEC-MALS and fitted using the Multipeak Fitting Package in Igor Pro. Error bars correspond to calculated errors in the fitting function parameters. (C) GTPase activity atlastin-1^1–446-ECFP/EYFP-fusion proteins. GTPase activity was determined by measuring the production of inorganic phosphate over time upon GTP hydrolysis at various protein concentrations. (D) Emission spectra of atlastin-1^1–446-ECFP/atlastin-1^1–446-EYFP mixtures at equilibrium. Protein (20 μM total, 1:20 ratio of donor to acceptor) was mixed with either buffer or nucleotide for at least 20 min prior to measurement, and was excited at 445 nm (apo, red dashed line; GDP, purple; GppNHp, green). (E) FRET efficiencies. FRET efficiencies of wild-type (black) and R⁷⁷A mutant (white) atlastin-1^1–446-ECFP/EYFP in the presence of various nucleotides were calculated as stated in Materials and methods. (F) FRET efficiency versus time of wild type and mutant R⁷⁷A in the presence of GTP. Either wild-type (dark blue) or mutant R⁷⁷A (light blue) atlastin-1^1–446-ECFP/EYFP were mixed with an excess of GTP. Emission spectra were measured immediately after mixing, every 2 min up to 10 min, at 15 min, and at 30 min. The sample was then spiked with additional GTP, mixed, and measured again. Lamp shutters were closed in between measurements. A black dashed line represents the FRET efficiency of wild-type FRET pair in the presence of GppNHp at equilibrium.

Figure 5

Kinetics of atlastin-1 middle domain FRET. (A) Stopped-flow FRET measurements. A mixture of atlastin-1^1–446-ECFP/EYFP fusion proteins (1 μM each) was prepared in the absence of nucleotide, and mixed 1:1 with either buffer (apo) or nucleotide-containing buffer (GDP, GTP, GppNHp, or GDP·AlF_x; concentration: 2 mM) using a stopped-flow apparatus. A xenon arc lamp was set to 445 nm excitation, and two photomultiplier tubes with appropriate filters in place were used to measure emission output over time. (B) Simulations of FRET data. A three-state system was used for a hydrolysis competent reaction (left, green), whereas a two-state system (right, orange) was applied to model nucleotide-dependent processes in the absence of GTP hydrolysis. Both simulations start with a monomeric assembly, which is then allowed to interconvert to a tight-parallel (form 3-like) dimer. In the hydrolysis competent case, this dimer is then allowed to interconvert to a relaxed-parallel (form 1-like) dimer.

Figure 6

Atlastin-1 G domain FRET using Alexa 488/647 dye-labelled proteins. (A) Experimental design of measuring G domain dimerization. Cartoon depiction of fluorescently labelled atlastin-1^1–446 in various crystallographic dimer conformations. Estimated distances between Cβ atoms of K²⁹⁵, which is mutated to a cysteine for site-specific labelling, are based on the crystal structures shown in Figure 2A. (B) SDS–PAGE of dye-labelled atlastin-1^1–446. Atlastin-1 mutants were first labelled with Alexa 488-C₅ Maleimide as described in Materials and methods. Two gels run side-by-side were loaded with 2.5 μg of indicated atlastin-1 proteins. One gel was stained with Coomassie brilliant blue (protein dye). The second, unstained gel was imaged upon excitation of the fluorophore. (C) Nucleotide-dependent dimerization of dye-labelled atlastin-1^1–446 proteins. SEC-MALS experiments were carried out and analysed as in Figure 4B. (D) GTPase activity of atlastin-1^1–446 mutant proteins used for dye labelling. GTPase activity was determined as shown in Figure 4C. (E) Emission spectra of dye-labelled atlastin-1^1–446 mixtures at equilibrium. Alexa Protein (20 μM total, 1:20 ratio of donor to acceptor) was mixed with either buffer or nucleotide for at least 20 min prior to fluorescence measurement (apo, red dashed line; GDP, purple; GppNHp, green). (F) FRET efficiencies. FRET efficiencies of wild-type (black) and R⁷⁷A mutant (white) calculated for dye-labelled atlastin-1^1–446 donor/acceptor mixtures in the presence of various nucleotides were carried out as stated in Materials and methods. (G) G domain-mediated FRET efficiency versus time of wild type and mutant R⁷⁷A in the presence of GTP. Either wild-type (dark green) or mutant R⁷⁷A (light green), dye-labelled atlastin-1^1–446 was mixed with an excess of GTP essentially as in (E). Emission spectra were measured right after mixing, every 2 min up to 10 min, at 15 min, and at 30 min. The sample was then spiked with additional GTP, mixed, and measured immediately. Lamp shutters were closed in between measurements. A black dashed line represents the FRET efficiency of wild-type FRET pair in the presence of GppNHp at equilibrium.

Figure 7

Rapid-mixing, stopped-flow kinetics of G domain-mediated FRET. Alexa 488 and 647-labelled atlastin-1^1–446 mixtures were prepared in a 1:1 ratio of donor to acceptor in a nucleotide-free buffer. Samples were mixed 1:1 with either buffer (apo) or buffer containing the indicated nucleotides (2 mM) using a stopped-flow apparatus. A xenon arc lamp was set to 493 nm excitation, and two photomultiplier tubes with appropriate filters in place were used to measure emission output over time.

Figure 8

Middle domain-mediated control of atlastin’s GTPase activity. (A) Comparison of the G-middle domain interface in form 2 and form 3 crystal structures. The G domain of atlastin-1 is shown in orange except for switch regions (shades of blue) and residues 184–204 (dark red), which form a helix that changes its conformation between the form 2 and form 3 (or form 1) structures. In the GDP-bound form 2, the helix is bent to accommodate the first helix of the middle domain (blue with linker in green), while in the GDP·AlF₄⁻-bound form 3 (and GDP-bound form 1) structure, the helix straightens out, preventing this interaction. Nucleotide positions are indicated and shown as fuzzy spheres. (B) The G-middle domain interface in the form 2 crystal structure. The area of the structure boxed in red is shown, with the same colouring introduced in (A). Positions of point mutants used in (C) and (E) and the last residue of the truncated atlastin-1 construct (atlastin-1^1–366) are shown as sticks. (C) GTPase activity. The catalytic activity of atlastin-1 constructs 1–446, 1–339 (G), and 1–366 were measured. In addition, the effect of the isolated middle domain (M; residues 340–446) on the activity of the G domain was determined. Structure-guided mutants of the middle domain were included as well. (D) Effect of the middle domain on the nucleotide-dependent oligomerization of the G domain. SEC-MALS data for the isolated G domain (atlastin-1^1–339; left panel) alone or with addition of a 10 × molar excess of the isolated middle domain (atlastin-1^340–446; right panel) in the presence of GppNHp are shown. The signal from the 90°-light scattering detector and refractive index detector is shown as coloured, solid lines (G domain alone, green; G and M domains mixed, blue) and black, dashed lines, respectively (left Y axis). Average molecular weight calculations across the protein peak are shown as black circles (right Y axis). The theoretical molecular weight (based on primary sequence) for the monomer and dimer of the G domain, as well as the middle domain in the right panel, is shown as horizontal dashed lines. Proteins (30–40 μM) were incubated with nucleotides (2 mM) at least 30 min prior to SEC-MALS analysis. (E) Quantification of nucleotide-dependent dimerization of atlastin-1’s G domain. SEC-MALS data for the indicated atlastin-1 constructs in the absence or presence of isolated middle domain variants are shown. Data for samples incubated with GppNHp or GDP·AlF_x are shown. The experimental approach and presentation was as in Figure 4B.

Figure 9

Model for atlastin-mediated membrane fusion. Atlastin begins in a form 2-like, GTP-loading-competent state with the middle domain engaging the G domain. GTP binding and hydrolysis drive rapid disengagement of the middle domain from the G domain, immediately followed by G and middle domain dimerization. Once in this tethering complex, membrane curvature and stress caused by atlastin’s transmembrane domains and C-terminal amphipathic helix would allow fusion to occur spontaneously. Phosphate release follows, with relaxation and subsequent disassembly of the dimer. Other proteins that may interact with atlastin and contribute to membrane curvature, such as reticulons, are not shown.