Improved crystallization of Escherichia coli ATP synthase catalytic complex (F₁) by introducing a phosphomimetic mutation in subunit ∊

A phosphomimetic mutation in subunit ∊ dramatically increases reproducibility for crystallization of Escherichia coli ATP synthase catalytic complex (F₁) (subunit composition α₃β₃γ∊). Diffraction data were collected to ∼3.15 Å resolution using synchrotron radiation.

Abstract

The bacterial ATP synthase (F_OF₁) of Escherichia coli has been the prominent model system for genetics, biochemical and more recently single-molecule studies on F-type ATP synthases. With 22 total polypeptide chains (total mass of ∼529 kDa), E. coli F_OF₁ represents nature’s smallest rotary motor, composed of a membrane-embedded proton transporter (F_O) and a peripheral catalytic complex (F₁). The ATPase activity of isolated F₁ is fully expressed by the α₃β₃γ ‘core’, whereas single δ and ∊ subunits are required for structural and functional coupling of E. coli F₁ to F_O. In contrast to mitochondrial F₁-ATPases that have been determined to atomic resolution, the bacterial homologues have proven very difficult to crystallize. In this paper, we describe a biochemical strategy that led us to improve the crystallogenesis of the E. coli F₁-ATPase catalytic core. Destabilizing the compact conformation of ∊’s C-terminal domain with a phosphomimetic mutation (∊S65D) dramatically increased crystallization success and reproducibility, yielding crystals of E. coli F₁ that diffract to ∼3.15 Å resolution.

1. Introduction  

F-type ATP synthases compose an evolutionarily related family of energy-coupling, ion-transporting enzymes which is responsible for the synthesis of most cellular ATP in plants, animals and many bacteria. The ATP synthase functions as a dual-engine rotary motor (Duncan, 2004 ▶). A membrane-embedded complex (F_O, composition a ₁ b ₂ c ₁₀) acts as a turbine to transport protons (H⁺; Na⁺ in some bacteria). A peripheral stator stalk and a central rotor stalk connect F_O to an extrinsic catalytic complex (F₁, composition α₃β₃γ₁δ₁∊₁) in which rotation of the asymmetric central stalk coordinates the conformational changes of three alternating catalytic nucleotide sites during net ATP synthesis or hydrolysis. High-resolution structures of F₁-ATPases have been solved over the past two decades, almost exclusively by crystallographic studies of mitochondrial F₁ (MF₁) isolated from bovine heart (Abrahams et al., 1994 ▶; Bowler et al., 2007 ▶) or yeast (Kabaleeswaran et al., 2006 ▶). Structures of a detergent-solubilized MF₁/c-ring complex from yeast (Stock et al., 1999 ▶) and bovine heart (Watt et al., 2010 ▶) were also determined at medium resolution. On the other hand, the bacterial ATP synthase of Escherichia coli (EF_OF₁) has provided the predominant system for genetic and biochemical studies of the functional mechanism of ATP synthases for over 30 years (Gibson, 2000 ▶). Likewise, single-molecule studies developed to characterize the rotary mechanics of the enzyme have relied almost exclusively on bacterial enzymes (Noji et al., 2011 ▶; Börsch, 2011 ▶).

Compared to mitochondrial homologues, bacterial F₁-ATPases have proven very difficult to crystallize and their crystals usually diffract X-rays weakly, to resolutions often not sufficient to build a complete atomic model. For instance, a 4.4 Å map of EF₁ (Hausrath et al., 1999 ▶, 2001 ▶) computed using 64% complete crystallographic data was reported in 1999. More recently, the structure of Caldalkali­bacillus thermarum F₁ with all empty nucleotide-binding pockets was reported to 3.3 Å resolution (Stocker et al., 2007 ▶). A possible reason for the poorer propensity of bacterial F₁ to crystallize as compared with MF₁ lies in the ∊ subunit’s C-terminal domain (CTD), which, in F₁ of bacteria and chloroplasts, is capable of dynamic conformational changes. Increasing evidence suggests this ‘structural plasticity’ of ∊’s CTD serves a regulatory role(s), being modulated by signals from the catalytic sites on F₁ (different nucleotides) and from F_O (proton-motive force, or inhibitor binding to F_O) (Duncan, 2004 ▶; Feniouk et al., 2006 ▶). Last year, we reported a complete atomic model of EF₁ lacking the δ subunit (EF₁-δ) refined to an R _work/R _free ≃ 24.3/26.4%, at 3.26 Å resolution (Cingolani & Duncan, 2011 ▶). In this structure, the ∊-CTD adopts a highly extended conformation which inserts deeply into the central cavity of the enzyme and engages both rotor and stator subunits in extensive contacts that are incompatible with functional rotation. In this paper, we present a strategy to improve the crystallization of EF₁ based on introducing a phosphomimetic mutation in subunit ∊.

2. Material and methods  

2.1. Cloning, expression and purification of Escherichia coli ATP synthase catalytic complex  

The genes encoding wild-type EF_OF₁ subunit were cloned in a pUC-based vector pJW1 (Wise, 1990 ▶). The ∊S65D mutation was created by site-directed mutagenesis using the plasmid p3DC (Duncan, Zhou et al., 1995 ▶) as a template; DNA sequencing confirmed that ∊S65D was the only mutation introduced within the atpDC genes. Then p3DC+∊S65D was cleaved with HindIII and ligated with a 5.5 kB HindIII fragment from wild-type p3U (Duncan, Zhou et al., 1995 ▶). This p3U+∊S65D was used to express EF_OF₁ including the ∊S65D subunit in E. coli strain LE392ΔatpI-C (Schaefer et al., 1989 ▶). For a typical EF₁ preparation, 10 l of E. coli were grown aerobically at 310 K in a Bioflo-2000 fermentor (New Brunswick Scientific) using 10 mM glucose and 1%(v/v) glycerol as carbon sources. Cells were harvested 6–8 h after inoculation and E. coli membranes were prepared as described (Duncan, Bulygin et al., 1995 ▶). Soluble EF₁ and EF₁(∊S65D) were dissociated from membranes and purified at 277 K, as previously described (Senior et al., 1979 ▶; Cingolani & Duncan, 2011 ▶). Typical yields of purified EF₁ are 50–100 mg per 50 g of wet cells, and specific ATPase activities of 40–45 units mg⁻¹ at 303 K under ‘routine’ conditions (pH 8, 2 mM ATP, 1 mM Mg acetate), for both wild-type and EF₁(∊S65D). The δ subunit was completely removed from EF₁ by gel filtration in the presence of detergent LDAO (lauryldimethylamine oxide) (Hausrath et al., 1999 ▶) at room temperature (∼295 K). This procedure provided >80% yield of EF₁-δ. Final samples of purified EF₁-δ and EF₁-δ(∊S65D) (≥10 mg ml⁻¹) were quick-frozen in liquid N₂ and stored at 193 K. Protein concentrations were determined by a modified Lowry assay (Peterson, 1977 ▶). ATPase activity was assayed at 303 K by a coupled-enzyme assay of ADP produced (Pullman et al., 1960 ▶) with typical conditions: 20 mM MOPS–Tris [3-(N-morpholino)propanesulfonic acid–Tris], pH 8.0, 50 mM KCl, 1 mM phosphoenol pyruvate, 0.3 mM NADH, 0.2 mg pyruvate kinase per ml, 0.1 mg lactate dehydrogenase per ml, 2 mM ATP, 1 mM magnesium acetate.

2.2. Detection of phosphorylation in subunit ∊  

Samples of purified EF₁-δ were analysed by SDS–PAGE (Duncan, Bulygin et al., 1995 ▶) and stained with SYPRO Orange (Life Technologies); gels were scanned with a Typhoon-9410 imager [GE Healthcare Life Sciences; 488 nm laser, 526 nm short-pass (SP) emission filter] and sample purity was analysed with ImageQuant TL software. For detection of putative subunit phosphorylation, SDS–PAGE gels were first stained with Pro-Q Diamond (Life Technologies) and scanned [Typhoon-9410, 532 nm laser, 580 band-pass (BP) emission filter], then stained with SYPRO Orange and scanned for total protein bands as above.

2.3. Crystallization of the ATP synthase catalytic complex (F₁) bearing a phosphomimetic mutation  

Prior to crystallization, frozen samples of EF₁-δ(∊S65D) at >10 mg ml⁻¹ were thawed quickly and extensively dialysed at room temperature against dialysis buffer (50 mM Tris–HCl, 0.1 mM Na–EDTA, pH 7.5, ±5 mM β-mercaptoethanol in a 10 kD cut-off Pierce Slidalyzer (3 ml size). Dialysed EF₁ samples were concentrated to ∼30 mg ml⁻¹ by ultrafiltration (Vivaspin, 10 kD cut-off). Crystallization trials were carried out using the hanging-drop vapour diffusion method in Linbro 24-well plates (Hampton Research) in a range of concentrations between 10 and 30 mg ml⁻¹, although the most reproducible and well diffracting crystals were obtained using EF₁-δ at 20 mg ml⁻¹. The most successful crystallization buffer was 0.1 MOPS–NaOH, pH 7.0, MgSO₄ 75 to 150 mM and PEG 8K, 6–10%(w/v). Droplets set up by mixing 4 µl of protein with an equal volume of reservoir solution and equilibrated against 600 µl of reservoir solution at 293 K gave the largest EF₁-δ(∊S65D) crystals, usually 2–5 d after setting up crystallization. Additives were screened with EF1-δ(∊S65D) at a protein concentration of 15 mg ml⁻¹ in crystallization buffer; crystallization was carried out by the hanging-drop method in Crystal Quick 96-well plates (Hampton Research) using a HYDRA II crystallization robot, at the Kimmel Cancer Center X-ray Crystallography and Molecular Characterization Shared Resource Facility (Thomas Jefferson University). Concentrated additive (7 µl, typically 1 M) was mixed with 53 µl of crystallization buffer (+5 mM βME) and 0.4 µl of this mixture was added to 0.4 µl of protein sample.

2.4. X-ray data collection and analysis  

Single crystals of EF₁-δ(∊S65D) were cryocooled after slow addition of cryoprotectant (glycerol) to 25%(v/v). Several data sets were collected at National Synchrotron Light Source (NSLS, Upton, NY) beamlines X29 and X6A. Diffraction data were reduced to intensities using the programs DENZO and SCALEPACK (Otwinowski & Minor, 1997 ▶) of the HKL-2000 package. A complete summary of diffraction statistics is presented in Table 1 ▶.

Table 1. Summary of diffraction data statistics for EF₁-δ(∊S65D).

The numbers in parentheses refer to the statistics for the outer-resolution shell (3.25–3.15 Å).

Crystallization condition	9%(w/v) PEG 8000, 0.1 M MOPS pH 7.0, 100 mM MgSO4, 4.0%(v/v) 1-propanol
Beamline	NSLS X29
Wavelength (Å)	1.075
Space group	C2
Reflections (total/unique)	4371954/262052
Unit cell	a = 433.3, b = 181.5, c = 224.2 Å, α = 90.0, β = 108.4, γ = 90.0°
Resolution (Å)	60–3.15
Completeness (%)	91.4 (64.3)
Redundancy	2.6 (2.1)
R sym † (%)	13.9 (76.5)
〈I〉/〈σ(I)〉	12.4 (2.3)

^† R _sym = where I_i(hkl) is the ith intensity measurement of reflection hkl, including symmetry-related reflections, and 〈I(hkl)〉 is its average.

3. Results and discussion  

3.1. The troublesome crystallogenesis of EF₁-δ  

Crystallization of wild-type EF₁-δ was achieved over eight years by a ‘brute force’ effort that included biochemical procedures and post-crystallization treatments (Cingolani & Duncan, 2011 ▶). Several factors were empirically found to promote crystallization. Removing the δ subunit was essential to reduce sample heterogeneity and obtain ordered crystals that diffracted to ∼5–7 Å resolution. Dehydration of EF₁-δ crystals in the presence of glycerol improved diffraction to ∼4.3 Å, greatly reducing the fall-off in diffraction intensity at higher resolution. Finally, addition of ∼1 mM AMPPNP (adenyl-imido­diphosphate) during dehydration dramatically improved the diffraction quality and resolution of EF₁-δ crystals. In October 2008, a single, extensively dehydrated crystal was used to collect a complete data set to ∼3.3 Å resolution, at NSLS beamline X25. This crystal belongs to space group C2 with unit-cell parameters a = 435.9, b = 183.1, c = 225.4 Å and β = 108.9°, and contains four EF₁-δ complexes in the asymmetric unit. These data were used to determine the structure of the auto-inhibited EF₁-δ, which was refined to an R _work/R _free ≃ 24.3/26.4%, at 3.26 Å resolution (Cingolani & Duncan, 2011 ▶). Despite three years of effort after obtaining this first data set, all attempts to reproduce EF₁-δ crystals diffracting past 3.5 Å resolution were unsuccessful. EF₁-δ crystallization was extremely variable. On average, one dehydrated crystal for every 30–40 tested diffracted past 4 Å resolution, making it difficult to perform co-crystallization studies with other co-factors. Crystallization was also greatly preparation dependent, and even the same EF₁-δ sample that had crystallized in one droplet failed to yield crystals in another drop set up under identical conditions. The success of crystallization could be improved by microseeding of older crystals in solution of freshly purified EF₁-δ, although this still relied on obtaining initial crystals of the enzyme.

3.2. Engineering ∊ subunit by a phosphomimetic mutation  

After determining the structure of EF₁-δ (Cingolani & Duncan, 2011 ▶), it became evident that the conformation adopted by ∊-CTD was likely the limiting factor in growing reproducibly well diffracting crystals of EF₁-δ. In the structure of EF₁-δ, the ∊-CTD adopts a highly extended conformation (denoted as ∊_X in Fig. 1 ▶ a), which inserts deeply inside the catalytic F₁ core. This conformation of ∊-CTD is drastically different from that seen in the only other bacterial F₁ structure determined so far (Stocker et al., 2007 ▶) and in all structures of isolated bacterial ∊ (Wilkens & Capaldi, 1998 ▶; Uhlin et al., 1997 ▶; Yagi et al., 2007 ▶), in which ∊-CTD folds as a compact helical hairpin (denoted as ∊_C in Fig. 1 ▶ b). Likewise, ∊-CTD is also compact in ∊ subunits of all mitochondrial homologues, due to a mitochondria-specific subunit that stabilizes the ∊_C state (Gibbons et al., 2000 ▶; Kabaleeswaran et al., 2006 ▶). Thus, we hypothesized that the co­existence of inhibitory (∊_X) and noninhibitory (∊_C) conformations of ∊-CTD in EF₁-δ samples could be the intrinsic source of heterogeneity hampering reproducible crystallization. In an attempt to optimize both the success rate and reproducibility of EF₁-δ crystallization, we focused on an unexpected feature revealed by the EF₁-δ structure. In the crystal structure, a 4–8σ peak of positive density was seen coordinating the hydroxyl group of Ser65 in ∊ (∊S65); this site was originally modelled as a sulfate ion (Fig. 1 ▶ c) (Cingolani & Duncan, 2011 ▶). This putative ion is located at hydrogen-bonding distance from the ∊S65 hydroxyl group and would sterically hinder conversion to the ∊_C conformation: superimposing the ∊_X and ∊_C conformations using the ∊-NTD (N-terminal domain), it was found that this density would clash with atoms of Thr82, Ala83 and Ile84, which move to become part of the ∊-NTD β-barrel in the ∊_C state (Fig. 1 ▶ b, yellow strand). We hypothesized that this ion or possible phosphorylation of ∊S65 would destabilize the ∊_C state, thus favouring the inhibitory ∊_X state. In support of this hypothesis, selective staining of purified EF₁-δ with Pro-Q Diamond Phosphoprotein Gel Stain suggests that a fraction of ∊ subunit is phosphorylated (Fig. 1 ▶ d), although amounts recovered from bands were insufficient to confirm this by mass spectrometry. Therefore, to mimic the ion or possible phosphorylation at ∊S65 and increase occupancy of the ∊_X state in EF₁-δ samples, we replaced ∊S65 with an aspartic acid, whose carboxyl group serves as a phosphomimetic.

Figure 1.

Stabilization of ∊-subunit CTD with a phosphomimetic mutation. Ribbon diagrams of ∊ subunit with the CTD in an extended (pdb 3oaa) (a) or compact conformation (b) (pdb 1aqt). In both panels, the ∊-subunit N-terminal β-barrel (∊1–80) is coloured in grey, while α-helices, β-strands and random coiled regions in ∊-CTD are in red, yellow and green, respectively. (c) A σ_A-weighted F _o − F _c electron-density map (blue mesh) computed at 3.26 Å resolution is overlaid on the putative ion interpreted as sulfate (SO₄). The density was computed after omitting the ion from the final refined model and is displayed at 5σ above background. The side chain of ∊S65 coordinating the putative sulfate ion is also shown. The electron-density figure was generated using PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.4, Schrödinger, LLC). (d) Phosphoprotein staining suggests that a fraction of ∊ subunit in EF₁ may be phosphorylated. A 20 µg sample of wild-type EF₁-δ was separated on 12.5% SDS–PAGE and subunit bands quantified after first staining with Pro-Q Diamond (phosphoprotein-selective, lane Q), and then after staining total protein with SYPRO Orange (lane S) (see §2). Q/S ratios represent the relative subunit staining intensities for the two dyes. Control samples with phosphorylated and nonphosphorylated proteins (PeppermintStick standards, Life Technologies) were run on the same gel; for a dilution (∼0.07 µg each band) that showed phospho-staining intensities similar to those of the lower ∊ band of EF₁-δ, Q/S ratios were 2.0 and 5.9 for phosphoprotein standards (ovalbumin and β-casein, respectively), and 0.14 for a nonphosphorylated protein (bovine serum albumin).

3.3. Crystallization of EF₁-δ(∊S65D)  

EF₁-δ(∊S65D) was expressed and purified as for the wild-type enzyme. In crystallization trials, EF₁-δ(∊S65D) had dramatically greater success of crystallization. Large plate-like crystals could be obtained reproducibly within 36–48 h in the presence of 9–11%(w/v) PEG 8K, 150 mM MgSO₄, 0.1 M MOPS–NaOH, pH 7.0. In contrast to EF₁-δ, microseeding did not significantly improve crystallogenesis, and crystals obtained without microseeding (Figs. 2 ▶ a and 2 ▶ b) were larger than those previously obtained with wild-type EF₁-δ. The success of EF₁-δ(∊S65D) crystallization was close to 90%, with nearly every drop showing large rod-like crystals. The ability to reliably obtain crystals allowed us to screen crystallization additives. Several compounds were found to increase the size of EF₁-δ(∊S65D) crystals and reduce clustering when used at ≥4%(v/v) final concentration. The most significant and reproducible were 1-propanol (Figs. 2 ▶ c and 2 ▶ d), acetone (Figs. 2 ▶ e and 2 ▶ f) and 1-butanol (Figs. 2 ▶ g and 2 ▶ h). In diffraction trials, most EF₁-δ(∊S65D) diffracted X-rays to ∼5 Å resolution even without dehydration. Dehydration in the presence of nucleotide dramatically improved diffraction quality and reduced radiation-induced diffraction decay. In a screen of ∼20 dehydrated crystals at NSLS beamline X29, a few EF₁-δ(∊S65D) surpassed 3.5 Å resolution. A complete data set to 3.15 Å resolution was measured at beamline X29 (Table 1 ▶), and subsequent diffraction analysis revealed this crystal form belongs to space group C2 with four EF₁-δ(∊S65D) in the crystallographic asymmetric unit, corresponding to 32 polypeptide chains and approximately 13 250 residues. Although the best diffraction data recorded from EF₁-δ(∊S65D) crystals were slightly improved in resolution as compared to wild-type EF₁-δ crystals (3.15 versus 3.26 Å), the reproducibility of EF₁-δ(∊S65D) crystals was dramatically enhanced. More consistent growth of EF₁-δ(∊S65D) crystals, additional refinement of additives and dehydration procedures should help us achieve higher-resolution structures in the future.

Figure 2.

Effects of additives on crystallization of EF₁-δ. (a), (b) Representative crystals of EF₁-δ(∊S65D) obtained by the hanging-drop vapour diffusion method using microseeding. Crystals obtained (without seeding) in the presence of 4.0%(v/v) of 1-propanol (c), (d), acetone (e), (f), or 1-butanol (g), (h).