Dynamics in an unusual acyl carrier protein from a ladderane lipid‐synthesizing organism

1 INTRODUCTION

Bacteria and archaea use a wide array of acyl chains in their lipid membranes, ranging from saturated- and partially unsaturated linear chains to highly branched and even cyclopropyl group-containing moieties.1 Arguably, the most surprising and interesting acyl chains encountered in microbial lipids, however, are the ladderanes,2 which consist of linearly fused cyclobutane rings forming a ladder-like structure (Figure 1A). Ladderanes are exclusively found in the lipids of anaerobic ammonium oxidizing (anammox) bacteria, which use the energy released by the oxidation of ammonium with nitrite to generate a proton gradient across an internal membrane, which in turn drives ATP synthesis.3, 4 As recent measurements have confirmed that membranes containing ladderane-based lipids are less permeable to protons and hydroxide ions, it is currently believed that anammox bacteria employ ladderanes in their membranes to help maintain their metabolic proton gradient.5 How ladderane lipids are produced, however, remains enigmatic; some mechanistic proposals have been put forward,6, 7 but no direct information on ladderane biosynthesis is available to date. However, bioinformatics has identified a gene cluster unique to anammox bacteria that is both conserved and expressed at high levels,8 and which encodes homologs of canonical fatty acid biosynthesis enzymes as well as other proteins, such as a putative radical SAM enzyme.9 Currently, the proteins encoded by this gene cluster are the most likely candidates for enzymes involved in ladderane biosynthesis.9

image (A) Pentacycloanammoxic acid, a typical 20[5]-ladderane fatty acid encountered in the lipid membranes of anammox bacteria. (B) Schematic view of an acyl carrier protein with its attached 4′-phosphopantetheine linker. (C) Acyl chains are bound as thioesters to the terminal thiol group of the linker. (D) The sequence of Kuenenia stuttgartiensis amxACP Kuste3603 (CAJ74366.1) was aligned to its orthologues from the corresponding gene cluster in Brocadia sinica JPN1 (BROSI_A3844, GAN35295.1), Brocadia fulgida (BROFUL_02173, KKO19130.1), Jettenia caeni (Planctomycete KSU-1, KSU1_C1689, WP_051006204.1), and Scalindua brodae (SCABRO_02231, KHE92060.1) using ClustalO.63 The alignment also contains ACP sequences from Pseudomonas aeruginosa PAO1 (NP_251656.1), Shewanella oneidensis MR-1 (NP_718356.1), Escherichia coli (strain MG1655, NP_415612.1), Neisseria meningitidis MC58 (NP_273277.1), Thermotoga maritima MSB8 (NP_228471.1), Spinacia oleracea (spinach chloroplast ACP 1 without leader sequence, XP_021844779.1), Bacillus subtilis (NP_389474.1), and Nostoc sp. PCC 7120 (Asl1650, WP_010995819.1). The sequence identities of K. stuttgartiensis amxACP to its anammox orthologues range from 79% to 94% and are much lower to other bacteria (16%–33%). Interestingly, the sequence identities to the amxACP paralogue in Kuenenia KsACPII (Kustd1387, CAJ72132.1) are only about 20%, too. The conserved serine residue binding the 4′-phosphopantetheine linker is highlighted in red and marked with a red star. The double phenylalanine (FF) motif in amxACP orthologues is highlighted in green. Secondary structures are derived from the Kuenenia stuttgartiensis amxACP structure

One of the unusual proteins on this gene cluster is amxACP, a unique version of an acyl carrier protein (ACP), specific to anammox bacteria. ACPs10, 11 are small, four-helical-bundle proteins that shuttle nascent fatty acid molecules from enzyme to enzyme during fatty acid biosynthesis, or nascent polyketide chains during polyketide synthesis. ACPs either occur as a domain of a large, multidomain synthetase consisting of a single polypeptide (Type I ACPs), or as discrete proteins (Type II ACPs), as is the case for amxACP. ACPs bind the nascent acyl chain via a thioester linkage to a 4′-phosphopantetheine moiety, which in turn is bound to a conserved serine on the surface of the protein (Figure 1B). How the acyl chain interacts further with the ACP varies,12 but typically Type II ACPs sequester their acyl cargo in an internal cavity, which expands as needed to accommodate nascent acyl chains at various stages of maturation.13, 14 This expansion is achieved without any major changes to the fold of the protein13, 14 but does appear to involve a repositioning of helix III.12, 15-17 A displacement of helix III was also implicated in opening the side of the protein to allow the acyl chain to move into and out of its binding cavity in a switchblade-like manner, to be presented to various enzymes for modification.18, 19

Biosynthetic pathways typically use a dedicated ACP that is tailor-made for its particular cargo and for the required interactions with the pathway's enzymes. Interestingly, the amino acid sequence of amxACP shows unique features conserved in anammox organisms. A dual phenylalanine (FF) motif, located between helices I and II, was found to be typical for amxACPs.9 Moreover, the sequence of helix III is distinct from that in other ACPs but highly conserved among amxACPs. To investigate how these unique features might influence the properties of the protein, we determined the structure of the amxACP from the anammox model organism Kuenenia stuttgartiensis. This protein can be loaded with various acyl chains, including ladderane-containing acyl chains (Dietl et al., in preparation). We find that in K. stuttgartiensis amxACP, the functionally important helix III forms a rare, six residue-long 310-helix, and we study the protein's conformational flexibility in crystallo by ensemble refinement and in silico by molecular dynamics. The results point to a hitherto unknown, distinct manner in which amxACP achieves the remarkable flexibility that all Type II ACPs require to carry out their functions.

2 MATERIALS AND METHODS 2.1 Protein expression and purification

The gene coding for the anammox acyl carrier protein from Kuenenia stuttgartiensis (amxACP, kuste3603) was amplified from genomic DNA by polymerase chain reaction (PCR), using primers KsE3603_F1 (5′-GGAGGACTACCATGGACGACGAAGAG-3′) and KsE3603_R1 (5′-CGAATAACCCTCGAGCTTCCCACTCTCTGCC-3′).

The PCR product was cloned into the pET-24d vector with a C-terminal hexa-histidine tag using the NcoI and XhoI restriction sites. The protein was heterologously expressed in Escherichia coli BL21 (DE3) cells. 2 × 2 L lysogeny broth (LB) medium supplemented with kanamycin (50 μg/ml) were inoculated with 1% (v/v) overnight pre-culture and grown at 37°C while shaking at 100 rpm in baffled 5 L Erlenmeyer flasks. Protein expression was induced at an OD600 of 0.6–0.8 by the addition of 500 μM isopropyl-β-d-thiogalactopyranoside (IPTG), after which incubation was continued at 20°C, 90 rpm for 16–18 hr. The bacterial cells were harvested by centrifugation at 8000×g, 15 min, 4°C. The cell pellet was resuspended in wash buffer (WB, 50 mM Tris-HCl pH 8.0, 300 mM NaCl and 10 mM imidazole) with the addition of one spatula tip of each lysozyme and DNaseI (Roche, Mannheim, Germany). The cells were disrupted using a microfluidizer (Microfluidics, Westwood) operated at a pressure of 0.7 MPa, and the lysate was clarified by centrifugation at 50 000×g for 1 hr at 4°C. The cleared lysate was filtered through a 0.45 μm syringe filter and loaded onto a gravity-flow column containing 8 ml of Ni-NTA (nickel-nitrilotriacetic acid) agarose (Qiagen, Hilden, Germany) pre-equilibrated with buffer WB. The column was further washed with 30 ml of buffer WB. The C-terminally His-tagged amxACP was eluted using elution buffer EB [50 mM Tris-HCl pH 8.0, 300 mM NaCl and 250 mM imidazole] and collected in 2 ml fractions. ACP-containing fractions were pooled and concentrated using a 3-kDa cut-off Amicon concentrator (Millipore Bioscience, Schwalbach, Germany). The concentrated protein was further purified using gel filtration chromatography on a Superdex 75 (16/60) column (GE Healthcare, Uppsala, Sweden) in gel filtration buffer (GFB, 50 mM HEPES-NaOH pH 7.5 and 150 mM NaCl) at a flow rate of 1 ml/min at 8°C. Finally, ACP-containing fractions were pooled, concentrated, and buffer-exchanged to a buffer containing 25 mM HEPES-KOH pH 7.5 and 25 mM KCl using a 3-kDa cut-off Amicon concentrator up to a concentration of about 40 mg/ml. Protein purity was assessed by 15% SDS-PAGE and protein concentration was determined based on the absorbance at 280 nm using an extinction coefficient of ε280 = 5500 M−1 cm−1 on a NanoDrop spectrophotometer (Peqlab, Erlangen, Germany). The protein solution was flash-frozen in small aliquots in liquid nitrogen and stored at −80°C until further use.

2.2 Protein crystallization

C-terminally His-tagged apoACP from Kuenenia stuttgartiensis (amxACP) was crystallized using the vapor-diffusion hanging-drop method by mixing 1.5 μl protein solution (10 mg/ml in 25 mM HEPES-KOH pH 7.5 and 25 mM KCl) with 1.5 μl precipitant on glass cover slips that were placed above a reservoir containing 800 μl of precipitant solution in pre-greased 24-well crystallization plates (Crystalgen Inc, Commack, New York) at 20°C. Hexagonally shaped crystals with dimensions of 30 × 10 × 10 μm3 were obtained after 2–3 days from a precipitant containing 28% (v/v) PEG 400, 200 mM calcium acetate, 100 mM sodium acetate pH 4.5, and 10 mM zinc chloride. The crystals were flashed-cooled in liquid nitrogen after brief soaking in a cryoprotectant containing 35% (v/v) PEG 400, 150 mM calcium acetate, 100 mM sodium acetate, pH 4.5, and 10 mM zinc chloride.

2.3 Data collection and structure determination

A highly redundant, 2.5 Å resolution Zn-SAD dataset was recorded from an apo-amxACP crystal at 1000 Å wavelength (at which f” for Zn is 2.6 e−) at the PXII beamline of the Swiss Light Source (Paul Scherrer Institute, Villigen, CH). Data were processed using XDS20 and showed appreciable anomalous signal up to 2.8 Å resolution, as judged by the significance of the anomalous correlation at the 0.1% confidence level. These data were phased by phenix.autosol21, 22 using the anomalous signal from the zinc ions. Six sites were found automatically, and after density modification, a figure-of-merit of 0.388 was obtained. Automatic building resulted in a partial model that contained 67 residues (out of 91 in the final model) of which >50% had the correct sequence assigned. The structure was completed manually using Coot.23, 24 To obtain a high-quality dataset with minimal radiation damage at higher resolution for refinement, the first 100° of the data were reprocessed taking data to higher resolution into account. This dataset showed considerable anisotropy, extending to ~1.8 Å resolution along the k axis but only to ~2.5 Å along the l axis. After integration with XDS,20 the data were therefore merged with STARANISO, while applying an ellipsoidal resolution cut-off.25 The final single-structure model was obtained by refinement against this data set using phenix.refine.26 Data and model statistics are reported in Tables 1 and 2.

TABLE 1. Data collection and processing statistics amxACP, Zn-SAD amxACP, high res. processing Diffraction source SLS PX-II SLS PX-II Wavelength (Å) 1.0000 1.0000 Temperature (K) 100 100 Detector Pilatus 6M Pilatus 6M Crystal-detector distance (mm) 390 390 Rotation range per image (°) 0.1 0.1 Total rotation range (°) 315 100 Exposure time per image (s) 0.1 0.1 Space group P64 P64 a, b, c (Å) 76.8, 76.8, 30.9 76.8, 76.8, 30.9 α, β, γ (°) 90.0, 90.0, 120.0 90.0, 90.0, 120.0 Mosaicity (°) 0.2 0.2 Resolution range (Å) 38.4–2.5 (2.6–2.5) 38.4–1.8 (1.9–1.8) Total no. of reflections 61 600 (4700) 39 342 (1123) No. of unique reflections 3741 (280)a 13 897 (935)a Completeness (%) 99.9 (100.0)a 91.6 (57.8)b Redundancy 16.5 (16.8)a 5.3 (2.8)b ⟨I/σ(I)⟩ 33.5 (13.6)a 16.4 (1.5) Rr.i.m. 0.072 (0.426)a 0.025 (0.652) CCano 0.28a Not determined Overall B factor from Wilson plot (Å2) 48.6 32.3 Note: Values for the outer shell are given in parentheses. a Treating Friedel mates as individual reflections. b An ellipsoidal cut-off was applied using STARANISO. TABLE 2. Structure solution and refinement amxACP apo stand-alonePDB# 7AUF amxACP apo ensemble (75 models)PDB# 7AX5 Resolution range (Å) 38.4–1.8 (1.9–1.8) Completeness (ellipsoidal, %) 91.6 (57.8) σ cut-off F > 1.34σ(F) No. of reflections, working set 13 208 (715) No. of reflections, test set 689 (36) Final Rcryst 0.2301 0.1747 Final Rfree 0.2862 0.2343 ML-Based DPI 0.23 0.16 No. of non-H atoms Protein 1451 1451 Ligands 7 Zn ions 7 Zn ions Water 36 36 R.m.s. deviations Bonds (Å) 0.012 0.008 Angles (°) 1.394 0.664 Average B factors (Å2) Protein 57.7 45.7 Ligands 84.0 48.7 Water 46.6 52.6 Ramachandran plot Most favored (%) 95.35 87.23 Allowed (%) 4.65 7.54 Outliers 0.00 5.24a Rotamer outliers (%) 1.27 16.95a Note: Values for the outer shell are given in parentheses. a A large number of unfavorable (φ, ψ) combinations and rotamers are expected, as ensemble refinement samples high-energy conformers.27 2.4 Ensemble refinement

Ensemble refinement was performed using phenix.ensemble_refinement27 according to the protocol described previously.28 Explicit hydrogens were added to the final, refined single-model structures using phenix.refine, after which a three-dimensional grid search was set up scanning different values for pTLS (0.6, 0.8, 0.9, and 1.0), for Tbath (2.5, 5.0, and 10.0) and tx (0.3, 0.6, and 1.2). The individual refinements were carried out in parallel on 36 cores of a cluster of Intel Xeon CPU X7560 processors running at 2.3 GHz. The lowest Rfree (0.2343) was obtained at pTLS = 0.8, Tbath = 2.5 and tx = 0.6. The final ensemble was analyzed and the results displayed using custom-written python scripts employing NumPy29, 30 and Matplotlib.31, 32

2.5 Molecular dynamics simulations

MD calculations were performed with GROMACS 202033, 34 using the AMBER99SB-ILDB force field35 and TIP3P water36 in a simulation box of the size of the molecule plus 10 Å on all sides. After energy minimization, 100 ps each of NVT and NpT equilibrations were performed with positional restraints on the protein atoms. Then, a production MD run of 1 μs with a time step of 2 fs was performed at 300 K using a modified Berendsen thermostat and Parrinello-Rahman pressure coupling, while employing periodic boundary conditions. Simulations were carried out on a cluster of 4 Intel Xeon CPU E7-4890 v2 processors with 15 cores each, running at 3.4 GHz.

2.6 Bioinformatics

A custom-written python script using the PyPDB module37 was used to conduct a search of the Protein Data Bank38 for entries mentioning “acyl carrier protein” in any text record, revealing 1006 entries. These were retrieved individually and passed to the DSSP program (CMBI version 3.0),39, 40 and the results searched automatically for stretches of 4 or more residues in a 310-helical conformation in a chain shorter than 150 residues. NMR structure ensembles were searched model by model. Entries flagged as possibly containing an ACP with a 310 helix were manually inspected in PyMol (Schrödinger, Inc.).

To investigate how often 310-helices of 6 residues (2 turns) or longer occur in general, we performed automated DSSP 3.0 analysis of 10 000 randomly sampled structures of any size in the PDB, searching for 5, 6, 7, 8, or more consecutive residues in a 310-helical conformation in any chain, and in any model of a multi-model structure, without visual inspection.

3 RESULTS

Kuenenia stuttgartiensis amxACP was heterologously expressed, crystals were grown and diffraction data collected to a resolution of 1.8 Å. These were phased by single-wavelength anomalous diffraction, using the anomalous signal from zinc ions present during crystallization. Although phasing and initial model building were easily achieved by automatic means, refinement was far from trivial; various refinement strategies were evaluated but none resulted in satisfactory R-factors. Pathologies that could explain these difficulties, such as twinning or translational NCS, were not detected, but the presence of pronounced anisotropy in the data pointed to a degree of disorder in the structure, as did the fact that in some regions the electron density appeared smeared out. We therefore employed ensemble refinement27, 28, 41-43 to account for any such flexibility. This resulted in an ensemble model for amxACP containing 75 separate structures which has acceptable R-factors. Data collection statistics are given in Table 1, and statistics for both the stand-alone and ensemble models are given in Table 2.

The results show that amxACP displays the typical four-helical bundle fold of a Type II acyl carrier protein (Figure 2A), as demonstrated by the relative orientations of helices II, III, and IV.44 In addition to these helices, a short, two-turn α-helix between helices I and II is present that also occurs in several other ACPs, and which we call helix Ia. The conserved Ser41, to which the 4′-phosphopantetheine linker binds, is located at the N-terminus of helix II. Strikingly, helix III, which consists of residues 60–67, adopts a 310-helical conformation as determined by DSSP 3.039, 40 and STRIDE,45, 46 rather than a regular α-helical conformation as the other helices do. This 310-helix stretches over an unusual length of six residues (two complete turns, residues 61–66). Importantly, ensemble refinement identifies this helix III as well as the loop consisting of residues 30–40 which contains the FF motif specific to amxACPs (Phe33–Phe34), as highly flexible (Figure 2B). Within the structures obtained by ensemble refinement, these two structural elements displayed positional RMSDs of up to 0.6 Å for the Cα atoms. The rest of the protein, with the exception of the N- and C-termini, showed much lower RMSDs (Figure 2C). This apparently higher structural variability is reflected in the electron density maps (Figure 2D) and geometry statistics. However, the electron density for helix III is of sufficient quality to confidently assign a 310-helical conformation to this structural element.

image

(A) Structure of amxACP from Kuenenia stuttgartiensis, as determined using conventional (single model) refinement. Ser41 is shown as sticks, and helix III, which is a 310-helix, is shown in orange. (B) Result of ensemble refinement, showing considerable flexibility in the 30–40 loop and in helix III. (C) RMSD of Cα positions plotted as a function of residue number for the ensemble refinement (blue) and MD (red) results. The various structural elements are indicated below the horizontal axis. In both cases, helix III (hIII, orange) shows the largest structural variability. (D) 2mFo-DFc electron density map from ensemble refinement (blue, 1.0 σ) overlayed on helices III and IV. The electron density is more well-defined for helix IV than it is for helix III

The difference in flexibility between helix III and the other helices could be due to crystal packing; in the crystal, helix III has room to move whereas helices Ia, II and IV are involved in crystal contacts. We therefore performed a 1 μs molecular dynamics simulation of amxACP in water and compared the results with those of the ensemble refinement. Inspection of the trajectory shows that the overall fold of the molecule is maintained over the time course of the simulation. However, plotting the positional RMSD of Cα atoms reveals that the overall flexibility of the structure in the MD simulation is indeed higher than in the X-ray structure with helix III being considerably more flexible during the simulation than the rest of the protein.

During large parts of the MD simulation's time course, the 310-helical structure in helix III is stable, with periods of unfolding in between: residues 61–65 adopt a 310-helical conformation more than 50% of the time with residue 66 doing so 28%, and 6% of the time three or more of these residues assume an α-helical conformation, which is in line with simulation results reported by Narwani and coworkers.47 In the ensemble refinement results, the 310-helix is present in about half of the individual models (Figure 3). We therefore investigated whether the high RMSD values observed are solely due to an unfolding of the helix or whether they, to a first approximation, represent motions of an intact helix III. We therefore split the ensemble refinement result and the MD trajectory into frames with an intact 310-helix in helix III (defining five consecutive residues in the correct conformation as “intact”) on the one hand and frames with a partially unfolded helix III on the other. We then recalculated the RMSD values for each set of frames (not shown), which revealed that the comparatively high RMSD values for helix III are due to both effects: a partial unfolding of the 310-helix as well as motions of the folded helix III as a whole (Figure 4).

image (A) Variation of the secondary structure in the K. stuttgartiensis apo-amxACP ensemble refinement results. (B) Variation of the secondary structure during a 1.0 μs MD simulation of K. stuttgartiensis apo-amxACP, showing the result for 100 equally spaced time points. In both cases, helices I, II, and IV are stable, whereas considerable conformational variability is observed in the 30–40 loop and in helix III, which varies between a 310-helical- and turn/bend/coil conformation. Secondary structures were determined using DSSP39 image

(A) Distance between the centres of gravity of helices II and III during the MD simulation. (B) Stereo figure, showing two structures from the MD simulations showing extremes of the distance between helices II and III (0.535 μs, short distance and 0.399 μs, large distance). In the long-distance structure, helix III is both displaced and partially unfolded

Our crystal structure shows that in amxACP, helix III forms exclusively hydrophobic interactions with the rest of the protein (Figure 5A), over a surface area of 370 Å2. Of this buried surface, more than 40% is accounted for by Ile67 alone, which is not part of the 310-helix. The sidechain of Ile67 contacts Ala45 and Ile47 on helix II, as well as Thr73 on helix IV. In crystal structures of acyl-loaded E. coli ACP, this conserved residue at the C-terminus of helix III was found to be the only residue undergoing a large change in backbone conformation upon ligand binding.14 Other interactions with helix III are made by Leu45 and Leu48 on helix II and Leu76 on helix IV.

image

(A) Stereo figure of the interactions between helix III (orange) and the hydrophobic core of the protein in the crystal structure. (B) Stereo figure showing the packing of helix III (orange sticks) with the protein's surface in the crystal structure. In projection, the helix has approximate 3-fold rotational symmetry

To investigate how often 310-helices occur in helix III of ACPs, we mined the Protein Data Bank for other ACP structures with a 310-helix in helix III. Apart from a few NMR structures in which short 310-helices occur in some of the models in the ensemble, the ACP Asl1650 from Nostoc sp. PCC 7120 (formerly called Anabaena sp. PCC 7120)48 and spinach chloroplast ACP13 are the only ACPs we found in which a part of helix III predominantly adopts a 310-helix conformation, albeit for only three or four consecutive residues, respectively. Thus, amxACP appears to be the only ACP structurally characterized to date with a helix III that mainly consists of two full turns of 310-helix, that is, six consecutive residues. By contrast, investigation of 10 000 randomly sampled PDB structures of any size or description revealed that around 11% contain a stretch of six residues that adopt a 310-helical conformation as determined by DSSP 3.0.

4 DISCUSSION

ACPs require a considerable degree of conformational flexibility to perform their function. They need to be able to bind acyl chains of various lengths, and which are in various states of functionalization as encountered during their synthesis. Moreover, they need to signal these states to binding partners, and need sufficient flexibility to allow access of even very large acyl chains to and from their binding cavity.13, 14, 49 As shown by our crystallographic and simulation results, a large degree of conformational flexibility in amxACP resides in helix III, which largely consists of a two-turn (six residue) 310-helix. This length is comparatively rare for a 310-helix; the vast majority (96%) of 310-helices are only four residues or less in length,50 in fact, their average length is just one turn.51 Most 310-helices are found at the termini of α-helices and it has been proposed that they represent an intermediate state between the α-helical and random coil conformations. Indeed, they appear to be less conformationally stable than, for example, an α-helix,47 and as such have been associated with structural transitions in both peptides and proteins. A switch between 310- and α-helix conformation was observed for a synthetic peptide in response to a change in solvent polarity52 and another peptide was found to be in equilibrium between the two states.53 A transition between the two states was suggested to explain the behavior of another peptide in response to an applied electric field.54 In aspartate aminotransferase, a switch of one helical turn from α- to 310-helical conformation assists in closing the enzyme's active site55 and a switch between an α- and 310-helix is implicated in the functioning of certain voltage-gated ion channels.56, 57

Interestingly, the sequence of helix III in amxACP is strongly conserved among anammox genera yet distinct from that in other ACPs, and appears to have evolved specifically to favor the formation of a 310-helix. For instance, the N-terminus of helix III consists of two negatively charged residues in sequence (Glu61 and Glu62). The presence of N-terminal glutamate residues was found to have a stabilizing influence on 310-helices.58 Moreover, the presence of a glutamate at position 62 as well as an aspartate at position 66 results in an electrostatically unfavorable urn:x-wiley:08873585:media:prot26187:prot26187-math-0001 interaction, which would destabilize an α-helical conformation, favoring a 310-helix formation.59

Apart from being inherently less stable than an α-helix, the properties of 310-helices also dictate specific ways in which they pack with other structural elements. As pointed out before,

留言 (0)

沒有登入
gif