Motile ghosts of the halophilic archaeon, Haloferax volcanii

Significance

Molecular motors are natural molecular machines that convert chemical energy to directional mechanical motion (e.g., ATP [adenosine-5′-triphosphate]-driven linear motors myosin, kinesin, and dynein in Eukaryotes; the ion-coupled rotary bacterial flagellar motor in Bacteria). Reconstituted systems such as in vitro motility assays are powerful tools for understanding the mechanisms of these motors. In this study, we established a membrane-permeabilized ghost system that controls the ATP-driven archaellar rotary motor in halophilic archaea. With this assay, we demonstrated high nucleotide selectivity and negative binding cooperativity and measured the energetic requirements for rotation. Our ghost model is also a platform for further studies of the archaellum. More generally, it is also a powerful tool for the investigation of other intracellular molecular functions in archaea.

Abstract

Archaea swim using the archaellum (archaeal flagellum), a reversible rotary motor consisting of a torque-generating motor and a helical filament, which acts as a propeller. Unlike the bacterial flagellar motor (BFM), ATP (adenosine-5′-triphosphate) hydrolysis probably drives both motor rotation and filamentous assembly in the archaellum. However, direct evidence is still lacking due to the lack of a versatile model system. Here, we present a membrane-permeabilized ghost system that enables the manipulation of intracellular contents, analogous to the triton model in eukaryotic flagella and gliding Mycoplasma. We observed high nucleotide selectivity for ATP driving motor rotation, negative cooperativity in ATP hydrolysis, and the energetic requirement for at least 12 ATP molecules to be hydrolyzed per revolution of the motor. The response regulator CheY increased motor switching from counterclockwise (CCW) to clockwise (CW) rotation. Finally, we constructed the torque–speed curve at various [ATP]s and discuss rotary models in which the archaellum has characteristics of both the BFM and F₁-ATPase. Because archaea share similar cell division and chemotaxis machinery with other domains of life, our ghost model will be an important tool for the exploration of the universality, diversity, and evolution of biomolecular machinery.

Motility is seen across all domains of life (1). Prokaryotes exhibit various types of motilities, such as gliding, swimming, and twitching, driven by supramolecular motility machinery composed of multiple different proteins (2). In archaea only swimming motility is reported, driven by the archaellum (archaeal flagellum), which consists of a reversible rotary motor and a helical filament acting as a propeller (3, 4). The archaellar motor has no homology with the bacterial flagellar motor (BFM) but is evolutionarily and structurally related to bacterial type IV pili (T4P) for surface motility (3). In Euryarchaeota, the filament is encoded by two genes, arlA and arlB [recently, the nomenclature of archaellum genes was changed from fla to arl (5)], and the motor is composed of eight, arlC to -J (ref. 3 has details in Crenarchaeota). Euryarchaeota encode the same chemotaxis system as flagellated bacteria, cheA, B, C, D, R, W, and Y, which might have been acquired by horizontal gene transfer from Bacillus/Thermotoga groups (6).

Fig. 1A shows the current association of functions with motor genes, based on analysis of mutants and biochemical data: ArlC/D/E as mediators of directional switching of archaellar rotation, coupled to chemotaxis signals (7); an ArlF/ArlG complex interacting with the surface layer (S-layer), with ArlF regulating ArlG filament assembly (8, 9); ArlH as a regulator of the switch between assembly of the archaella and rotation (10); ArlI as an adenosine triphosphatase (ATPase) essential for both assembly and rotation (11); and ArlJ as the membrane-spanning component. An inhibitor of proton translocating ATP (adenosine-5′-triphosphate) synthases reduced both intracellular [ATP] and swimming speed in Halobacterium salinarum (12), suggesting that archaellar rotation is driven by ATP hydrolysis at ArlI. However, direct evidence is lacking due to the lack of a reconstituted system.

Fig. 1.

Swimming ghosts. (A) The current model of the archaeallar motor in Euryarchaeota (details in the Introduction). (B, Left) Phase-contrast image of live cells. (B, Center) Fluorescent image of the same cells labeled with streptavidin-Dylight488, which binds to biotinylated filaments. (B, Right) Merged phase-contrast (magenta) and fluorescent (green) images. (Scale bars: 5 μm.) (C) Procedures to observe swimming ghosts. Live cells were permeabilized in a tube and then added to a flow chamber to observe swimming motility. (D) The same as B but for ghosts. (Scale bars: 5 μm.) (E) Sequential images of a change from a live cell to ghost (0.6-s intervals) (Movie S3). From 2.4 s, the ghost remains in focus and continues to swim. (Scale bars: 3 μm.) (F) Time course of swimming speed (triangles) and cell optical density (circles) from E. The arrow indicates the time of permeabilization. (G) Histograms of swimming speed in detergent before (Upper; 1.51 ± 0.34 μm s⁻¹) and after (Lower; 1.30 ± 0.21 μm s⁻¹) transformation into ghosts (mean ± SD, n = 24). B–G show data for Hfx. volcanii strain FYK3 (w.t. for chemotaxis).

To test the above hypothesis of ATP-driven motility in archaea, we established a permeabilized ghost model that enables the manipulation of intracellular contents. Ghosts in ATP solution swim like intact cells. Their motor rotation speed increases with [ATP] in a way consistent with negative binding cooperativity. Other nucleotide triphosphates (NTPs) support slow rotation but with ∼200× reduced binding affinity compared with ATP. Measured motor torque indicates the energetic requirement for at least 12 ATP molecules to be hydrolyzed per revolution. We also demonstrated that CheY increases motor switching from counterclockwise (CCW) to clockwise (CW) rotation, and we discuss different rotary models.

Results

Swimming Motility of Haloferax volcanii Live Cells.

In this study, we used strains of the halophilic archaeon Haloferax volcanii (FYK3, FYK5) (SI Appendix, Result S1 and Table S1) with genetically introduced cysteine substitutions on the surface of the archaeal filament, ArlA1(A124C). Hfx. volcanii is rod shaped, with a diameter of 0.9 μm and length of 2 to 10 μm, and swims using multiple polar archaella (Fig. 1B and Movie S1). We increased the fraction of swimming Hfx. volcanii cells from 20 to 30% (13) to 80% by adding 20 mM CaCl₂ (SI Appendix, Fig. S2A). The swimming-speed distribution of cells wild type (w.t.) for chemotaxis (FYK3) had two peaks, at 1.7 and 2.7 μm s⁻¹ (SI Appendix, Fig. S1D). We observed only the slower peak in a CheY deleted strain (FYK5) (SI Appendix, Fig. S3A).

Construction of Swimming Ghosts.

We present an in vitro experimental system for the archaellum, similar to the triton model for the eukaryotic flagellum (14) and the permeabilized ghost model for gliding Mycoplasma mobile (15, 16). In those experiments, samples were immobilized, allowing rapid addition of high concentrations of detergent and rapid removal after permeabilization. Because rapid buffer exchange is not possible with swimming Hfx. volcanii cells, we instead suspended motile cells in buffers containing lower concentrations of detergent (0.015% [wt/vol] sodium cholate) and 2.5 mM ATP (Fig. 1C). The detergent reduced the refractive index of cells, indicating permeabilization of the cell membrane and corresponding loss of cytoplasm. Fluorescent imaging revealed that permeabilized cells still possessed archaellar filaments, the cell membrane, and S-layer (Fig. 1D and SI Appendix, Fig. S4). Remarkably, the permeabilized cells still swam (Movie S2 and SI Appendix, Fig. S5A). We named them “ghosts,” as in similar experiments on M. mobile (15).

Fig. 1E shows a typical example of a live swimming cell changing to a ghost, marked by a sudden change of image density at 2.4 s (Movie S3). The solution contained 2.5 mM ATP, and the swimming speed did not change dramatically when this cell became a ghost (Fig. 1F; SI Appendix, Fig. S5B shows more examples). Fig. 1G shows histograms of swimming speeds of cells in detergent before and after permeabilization, indicating that ghosts swim at the same speed as live cells in this saturating ATP concentration (P = 0.421834 > 0.05 by t test, ratio 0.93 ± 0.24, n = 24) (SI Appendix, Fig. S5C). The absence of the faster swimming peak (compare Fig. 1 G, Upper with SI Appendix, Fig. S1 D, Lower for the same strain with no detergent) indicates suppression of the faster swimming mode by detergent, before permeabilization.

Characterization of ATP-Coupled Motor Rotation by a Ghost-Bead Assay.

Although we succeeded in producing swimming ghosts, we found them difficult to track due to their low contrast, and we were not able to determine the direction of archaellar filament rotation. We therefore established a ghost-bead assay for measuring ATP-coupled motor rotation (Fig. 2A). We attached cells with sheared, biotinylated archaellar filaments nonspecifically to the cover glass surface and then introduced 500-nm streptavidin beads, which attached to the filaments, allowing direct observation of the speed and direction of archaellar rotation (Materials and Methods and SI Appendix, Fig. S6). Addition of 0.1 mg mL⁻¹ streptavidin (which would cross-link adjacent filaments in a rotating bundle) did not stop bead rotation, indicating that shearing removed most filaments and rotating beads are attached to a single archaellum (Movie S4) (17).

Fig. 2.

Visualization of motor rotation in ghosts via beads attached to archaellar filaments. (A) Schematic of experimental setup (Upper) and phase-contrast images of a live cell (Lower Left) becoming a ghost (Lower Right). (Scale bars: 5 μm.) (B) An example of ghost formation, with reduced flow rates to allow tracking of the bead without loss of focus. (Upper) The y coordinate of the rotating bead during ghost preparation (markers at 500-nm intervals). (Lower) Shaded sections (Upper) are expanded (Left); with corresponding speed distributions by Fourier transform analysis (Right) using the same colors as in Upper. Blue shows the live cell, orange shows the motor stopped after treatment with detergent, and red shows the motor reactivated after addition of ATP. Rotation cannot reliably be measured during media exchange time, ∼10 s (data from Movie S5). (C, Left) The x–y plots of locations of two different beads attached to archaeallar filaments. Green and blue represent CW and CCW rotation, respectively. (C, Right) Angle vs. time for several similar beads. The slopes increase with [ATP] in both rotation directions, indicating an ATP-coupled motor rotation. (D) Rotation rate for different NTPs at 10 mM. The means ± SD were 6.14 ± 1.48 Hz for ATP (n = 43), 0.69 ± 0.24 Hz for GTP (n = 30), 1.18 ± 0.36 Hz for CTP (n = 32), and 0.99 ± 0.32 Hz for UTP (n = 29). (E) Lineweaver–Burk plot of rotation rate and inhibitors. Blue, ocher, green, and red represent data with 0.5 mM ATP-γ-S (n = 118), 2 mM ADP (n = 140), 2 mM ADP + Pi (n = 114), and without inhibitors (n = 345), respectively. Data are representative of three independent experiments.

For the preparation of ghosts, live cells labeled with rotating beads were treated in a flow chamber with detergent (0.03% sodium cholate) for less than 10 s to permeabilize their cell membrane. This higher concentration of detergent increased the yield of ghosts compared with the swimming cell assay. ATP was absent from the detergent buffer, allowing ghost formation to be detected as cessation of rotation. Immediate removal of detergent, by replacement with a buffer containing ATP, minimized further damage to ghosts (observed with prolonged exposure to detergent) (SI Appendix, Result S2 and Fig. S7) and reactivated rotation (Fig. 2B and Movie S5). To assess the effects of detergent, we measured rotation of beads with chemotaxis wild-type ghosts (FYK3) in the presence of detergent (SI Appendix, Fig. S7). The rotation speed fluctuated, often to zero, and beads rotated exclusively CCW (n = 11). This result indicates that detergent inhibited CW rotation in ghosts, which led us to remove detergent in all other bead experiments.

We investigated the effect of different NTPs in the chemotaxis w.t. strain (FYK3). We observed both directions of rotation in this assay (Fig. 2C). We did not see any differences between CW and CCW rotation rates (SI Appendix, Fig. S8) and therefore, analyzed speeds collectively. Previous in vitro experiments showed that purified ArlI hydrolyzes different NTPs at similar rates (18). However, the archaellar rotational rates in ghosts in 10 mM GTP (guanosine-5′-triphosphate), CTP (cytidine-5′-triphosphate), and UTP (uridine-5′-triphosphate) were 5 to 10 times slower than in ATP (Fig. 2D). Other NTPs supported rotation at speeds similar to those in ∼200× lower ATP concentrations (SI Appendix, Fig. S9). This result suggests that ArlI might specifically recognize the adenine base of ATP, rather than distinguishing only between purines and pyrimidines as in F₁-ATPase (19), and/or prevent extra energy consumption in vivo like the endopeptidase Clp (figure 1B of ref. 20).

We also tested the inhibitory effects on rotation of ADP (adenosine-5′-diphosphate), ADP + Pi (inorganic phosphate), and the nonhydrolyzable ATP analog ATP-γ-S (adenosine-5′-[γ-thio]triphosphate). We saw no rotation with ATP-γ-S alone. We measured the rotation rates of 500-nm beads attached to archaella in ghosts over a range of [ATP] between 63 μM and 10 mM, with and without each of ATP-γ-S (0.5 mM), ADP (2 mM), and ADP + Pi (each 2 mM). Fig. 2E shows the results as a Lineweaver–Burk plot. All three caused large reduction of rotation rates at lower [ATP] but much smaller reductions of the speed at saturating [ATP] (f_max), indicating competitive inhibition. The inhibitor constants, K_i, were estimated to be 0.11 mM for ATP-γ-S (blue), 1.94 mM for ADP (ocher), and 1.22 mM for ADP.Pi (green). We also observed modest effects of pH and ion concentration on rotation (SI Appendix, Result S3 and Fig. S10).

Archaeal CheY-Mediated CW Rotation.

Bidirectional rotation of the flagellar motor in bacteria is mediated by the response regulator CheY, which induces CW rotation (21). Live Hfx. volcanii cells with CheY deleted (FYK5) were observed to rotate in either direction, without switching during our typical recording time of 30 s (n = 5 for CW rotation, n = 76 for CCW rotation). To observe the role of archaeal CheY in motor switching, we extended our recording time to 300 s (Fig. 3). Switching from CCW to CW rotation was frequent in chemotaxis wild-type live cells (FYK3) (Fig. 3A and SI Appendix, Fig. S11) but very rare in ΔCheY live cells (FYK5) even during 5-min recordings (Fig. 3B). FYK3 ghosts also switched very rarely (Fig. 3C), but their probability of CW rotation (P = 0.27) was similar to that of live cells with CheY (P = 0.25); both were greater than live cells with CheY deleted (P = 0.07) (Fig. 3 and SI Appendix, Table S3). We observed CheY-GFP (green fluorescent protein) in Hfx. volcanii FYK7 using fluorescence microscopy (SI Appendix, Fig. S12). Live cells (SI Appendix, Fig. S12 A, Upper) showed CheY-GFP distributed throughout, with additional polar clusters, as in our previous report (13). By contrast, we could detect no CheY-GFP in ghosts (SI Appendix, Fig. S12 A, Lower). These results indicate that CheY increases the probability of, but is not absolutely required for, CW rotation and switching in live cells. CheY is lost during ghost formation, but the CW probability remains at the live cell level: motors stop switching, and one in three remains CW even though CheY is missing. We discuss this further in SI Appendix, Result S4.

Fig. 3.

Archaeal CheY-mediated motor switching. (Left) Representative 5-min recordings of single motors using 970-nm beads in wild-type (FYK3; A) and ΔCheY (FYK5; B) live cells and 500-nm beads in wild-type ghosts (C). Positive and negative speeds represent CW and CCW rotation, respectively. (Insets) The y–x plots of bead rotation. Grids represent 500 nm. (Right) Histograms of speeds for each cell.

ATP and Load Dependence of Motor Rotation.

Fig. 4A shows the dependence of rotation speed (f; revolutions per second) of 200-, 500-, and 970-nm beads upon [ATP] in the range 8 μM to 10 mM. Michaelis–Menten fits to the data (solid lines) are poor below 30 μM ATP (Movies S6 and S7; SI Appendix, Fig. S9 shows more examples). Fig. 4B shows the relationship between log([ATP]) and log(f/(f_max − f)), where f_max is estimated from the Michaelis–Menten fit (Fig. 4A), and the slope represents Hill coefficients of 0.63, 0.82, and 0.89 for 200-, 500-, and 970-nm beads, respectively. This result indicates negative cooperativity in ATP-driven archaellar rotation (see below for discussion). Fig. 4C shows the relationships between torque, speed, and [ATP] for archaellar rotation. The maximum motor torque was estimated to be ∼200 pN nm for live cells and ∼170 pN nm for ghosts, comparable with Hbt. salinarum live cell experiments (160 pN nm) (22).

Fig. 4.

ATP- and load-dependent archaeal motor rotation. (A) Rotation rates of 200-nm (green), 500-nm (pink) and 970-nm beads (purple) attached to archaellar filaments vs. [ATP]. The solid lines show fits to the Michaelis–Menten equation $\frac{f_{\text{max}}\times \left[\text{ATP}\right]}{K_{\text{m}}+\left[\text{ATP}\right]}$, where f_max and K_m are 10.3 Hz and 188 μM for 200-nm beads (n = 287), 6.8 Hz and 249 μM for 500-nm beads (n = 438), and 2.6 Hz and 132 μM for 1,000-nm beads (n = 303). (Right) Corresponding rotation rates of live cells: 18.47 ± 6.86 Hz for 200-nm beads (n = 19), 10.17 ± 2.39 Hz for 500-nm beads (n = 32), and 3.22 ± 1.27 Hz for 970-nm beads (n = 72). (B) A Hill plot of the same data. The Hill coefficient, determined from the slope of the plots, was 0.63 for 200-nm beads, 0.82 for 500-nm beads, and 0.89 for 970-nm beads. (C) Torque vs. speed (mean ± SD) of live cells and ghosts at various [ATP]. Torque was estimated as T = 2πfξ, where f is rotation speed and ξ = 8πηa³ + 6πηar² is the viscous drag coefficient of the bead. ξ was varied by using bead size (d = 2a = 200, 500, and 970 nm) and viscosity (η = 2.5, 3.9, and 7.2 mPa·s in buffer, 5% Ficoll, and 10% Ficoll, respectively). r is the major axis of the ellipse describing the orbit of the bead center. [ATP] and number of cells are indicated.

Discussion

Our estimated maximum torque (T) of ∼170 pN nm corresponds to the motor doing work for a single rotation (2πT, ∼1,000 pN nm) equivalent to the free energy of hydrolysis of ∼12 ATP molecules per revolution, assuming a free energy of 80 to 90 pN nm per ATP molecule (23). Conservation of energy therefore sets a lower limit of 12 per revolution per motor on the ATP hydrolysis rate, ∼15 times higher than that measured in vitro for ArlI (24). This indicates that motor assembly enhances ATPase activity in the archaellum, as observed in other systems: for example, the PilC–PilT interaction in T4P (25) and β- and γ-subunit interaction in F₁-ATPase (26). Hydrolysis of 12 ATP molecules per revolution is consistent with previous reports (22) and with models of a twofold ArlJ rotor rotating within a sixfold ArlI ATPase (22, 27).

Hill coefficients <1 in fits to the data of Fig. 4B indicate negative cooperativity in archaellar rotation at low [ATP]: subsequent ATP molecules bind the multimeric motor with lower rate constants than previous ones (28). Negative cooperativity in the archaellar motor might be explained by a mechanism similar to that proposed for F₁-ATPase (29), where each subsequent nucleotide that binds changes the conformational state of the entire complex, and the binding affinity at each site is lowered when the other two sites are occupied. In this scenario, negative cooperativity would be the result of the same interactions within the ArlI hexamer that power rotation (22, 27). Negative cooperativity could also arise from communication between ArlI and ArlH rings, similar to interring effects in chaperonins (30). Although ArlH has only the Walker A motif, ATP binding is known to modulate the interaction between hexametric rings of ArlI and ArlH (10).

Our finding that the time-averaged motor torque decreases with increasing speed at low loads (Fig. 4C) differs from a previous report (22), which assumed constant torque irrespective of viscous load and speed and explained the observed speed variations by assuming an extra contribution to the viscous drag from an unseen remnant of the filament. The required length of these remnants (ξ; 0.8 pN nm s) would be about 4 μm (22), which seems unlikely given our observation that most filaments are removed by shearing. The apparent trend of increasing torque with speed, at low speeds in live cells, is within the error bars and may be no more than experimental noise (SI Appendix, Fig. S13). Negative cooperativity (Fig. 4B) suggests that the hexametric ATPase ArlI, in which two ATP molecules can be bound (25, 27), may have two distinct mechanisms for torque generation, depending on [ATP]. At low [ATP], only one catalytic site would be occupied, the second site having too-low binding affinity. At high [ATP], two ATP molecules could be bound and simultaneously hydrolyzed. In this scenario, the maximum torque would increase with [ATP], consistent with Fig. 4C. The curves in Fig. 4C are qualitatively similar to those reported for the BFM with varying ion-motive force (31). By contrast, the equivalent data for F₁-ATPase correspond to Michaelis–Menten kinetics and torque that decreases linearly with increasing speed (figure 2 of ref. 32). Simple models for the torque–speed relationships, similar to those applied to the BFM (31, 33) and high-resolution detection of steps in rotation (34) using gold nanoparticles (32, 35), may reveal the details of the rotation mechanism of the archaellum in future.

With due care to account for reflections in our microscope (SI Appendix, Fig. S6), the rotational direction of the motor was directly determined by our bead assay. Together, our results (Fig. 3 and SI Appendix, Fig. S6) demonstrate that in our experiments, at temperatures 20 °C below those of the organism’s natural habitat, CheY greatly increases the probability of, but is not absolutely required for, CW rotation and switching in the archaellum. The same is true in the flagellar motor, which requires CheY for CW rotation at room temperature, but not below ∼6 °C (36). This similarity seems remarkable given the lack of evolutionary relation between the two motors but perhaps not so given that they use the same chemotaxis proteins to perform the same function. Deletion of CheY also removed the fast-swimming mode (SI Appendix, Fig. S3), consistent with previous reports that linked fast swimming to CW rotation in other archaeal species—with right-handed helical filaments pushing cells (“forwards”) when rotating CW and pulling when CCW (SI Appendix, Fig. S3B) (37, 38). The absence of the fast-swimming (CW) peak in cells in detergent before permeabilization (Fig. 1G) indicates a possible direct or chemotactic effect of detergent on the switch. Our findings appear to contradict a previous study reporting CW-only rotation in a CheY deletion strain of Hbt. salinarium (39). However, that study did not measure rotation direction directly, instead inferring it from forward swimming under the assumption that filaments were right handed in their strain, Hbt. salinarium M175, as in ref. 37. Subsequent work revealed that filaments are in fact left handed in Hbt. salinarium M175 (40), and thus, that rotation in the CheY deletion was CCW—in agreement with our direct observations. When CheY is lost during ghost formation, cells stop switching, many remaining “locked” in CW rotation. This indicates that other factors also control switching of the archaellum.

There is, to date, no direct evidence published that answers the fundamental question of which components of the archaellum are fixed relative to the cell (“stator”) and which rotate with the filament (“rotor”). ArlF and G are most likely part of the stator, anchored to the S-layer. Previous reports indicate interaction between ArlF and the S-layer and a deficiency in swimming motility of S-layer deleted cells (8, 9). Our observations of increased motility with [CaCl₂] and the speed fluctuations at low [CaCl₂] (SI Appendix, Figs. S2 and S8), given that calcium stabilizes S-layer (41), support this hypothesis. Homology to F₁-ATPase and T4P is generally taken to favor a model where rotation of an ArlJ dimer within the central core of the ArlI hexamer is driven by cyclic changes in the conformation of the ArlI hexameric ATPase, coupled to ATP hydrolysis (22, 27). In this model, ArlJ is the rotor, and all other motor components are the stator (SI Appendix, Fig. S14, Left). For switching, changes caused by CheY binding, presumably somewhere on ArlC/D/E, would have to propagate all of the way to the core of ArlI, which would need separate mechanochemical cycles for CW and CCW rotation. SI Appendix, Fig. S14, Right illustrates the other extreme possibility, most similar to the BFM. In this model, conformational changes in ArlI would push on ArlF/G, either directly or via ArlC/D/E. In the latter case, the switch mechanism could reside entirely within ArlC/D/E, and ArlI need not have separate modes for CW and CCW rotation. Intermediate models are also possible (SI Appendix, Fig. S14, Center). Our ghost model may allow labeling of archaellar components to observe directly which are part of the rotor (42, 43), analysis of rotational steps as in isolated F₁ and other molecular motors (32, 44, 45), and direct investigations of the role of CheY.

Our ghost assay represents an experimental system that allows manipulation of the thermodynamic driving force for an archaeal molecular motor, following previous examples including eukaryotic linear motors (46), the PomAB-type BFM (45), and the Mycoplasma gliding motor (15). We anticipate that this assay will be helpful for other biological systems. Archaea display chemotactic and cell division machinery acquired by horizontal gene transfer from bacteria (6, 47). Although the archaellum and bacterial flagellum are completely different motility systems, they share common chemotactic proteins. Theoretically, only our ghost technique allows monitoring of the effect of purified CheY isolated from different hosts on motor switching. Similarly, our ghost cells offer the potential to manipulate and study the archaeal cell division machinery as with in vitro ghost models of Schizosaccharomyces pombe (48). Ghost archaea offer the advantages of both in vivo and vitro experimental methods and will allow the exploration of the universality, diversity, and evolution of biomolecules in microorganisms.

Materials and Methods

Strains and Cultivation.

Strains, plasmids, and primers are summarized in SI Appendix, Tables S1 and S2. Hfx. volcanii cells were grown at 42 °C on a modified 1.5% Ca agar plate (2.0 M NaCl, 0.17 M Na₂SO_4, 0.18 M MgCl₂, 0.06 M KCl, 0.5% [wt/vol] casamino acid, 0.002% [wt/vol] biotin, 0.005% [wt/vol] thiamine hydrochloride, 0.01% [wt/vol] l-tryptophane, 0.01% [wt/vol] uracil, 10 mM [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid]-NaOH [Hepes-NaOH] [pH 6.8], 1.5% [wt/vol] Agar). Note that 20 mM CaCl₂ should be added (SI Appendix, Fig. S2A). Colonies were scratched by the tip of a micropipette and subsequently suspended in 5 mL of Ca liquid medium. After a 3-h incubation at 37 °C, the culture was centrifuged at 5,000 rpm and concentrated to 100 times volume. The 20-μL culture was poured into 25 mL fresh Ca medium and again grown for 21 h with shaking of 200 rpm at 40 °C. The final optical density would be around 0.07.

Gene manipulation based on selection with uracil in ΔpyrE2 strains was carried out with polyethylene glycol 600 (PEG 600), as described previously (49). For the creation of knockout strains, plasmids based on pTA131 were used carrying a pyrE2 cassette in addition to ∼1,000-bp flanking regions of the targeted gene. arlA1(A124C) was expressed by tryptophane promotor (SI Appendix, Fig. S1).

Preparation of Biotinylated Cells.

The culture of Hfx. volcanii FYK3 cells was centrifuged and suspended into buffer A (1.5 M KCl, 1 M MgCl₂, 10 mM Hepes-NaOH, pH 7.0). Cells were chemically modified with 1 mg mL⁻¹ biotin-PEG2-maleimide (Thermo Fischer) for 1 h at room temperature, and excess biotin was removed with 5,000 × g centrifugation at room temperature (R.T.) for 4 min.

Motility Assay on Soft-Agar Plates.

A single colony was inoculated on a 0.25% (wt/vol) Ca-agar plate and incubated at 37 °C for 3 to 5 d. Images were taken with a digital camera (EOS kiss ×7; Canon).

Microscopy.

All experiments were carried under an upright microscope (Eclipse Ci; Nikon) equipped with a complementary metal oxide semiconductor (CMOS) camera (LRH1540; Digimo) and a 40× objective (EC Plan-Neofluar 40 with Ph and 0.75 numerical aperture [N.A.]; Nikon) or 100× objective (CFI Plan Apo 100 with Ph and 1.45 N.A.; Nikon). Images were recorded at 100 frames s⁻¹ (fps) for 10 to 30 s. For a motility experiment at 45 °C, a phase-contrast microscope (Axio Observer; Zeiss) equipped with a 40× objective (EC Plan-Neofluar 40 with Ph and 0.75 N.A.; Zeiss), a CMOS camera (H1540; Digimo), and an optical table (Vision Isolation; Newport) were used.

For fluorescence observations, a fluorescence microscope (Nikon Eclipse Ti; Nikon) equipped with a 100× objective (CFI Plan Apo 100 with Ph and 1.45 N.A.; Nikon), a laser (Nikon D = eclipse C1), an electron-multiplying charge-coupled device (EMCCD) camera (ixon⁺ DU897; Andor), and an optical table (Newport) were used. The dichroic mirror and emitter were Z532RDC (C104891; Chroma) and 89006-ET-ECFP/EYFP/mcherry (Chroma) for an FM4-64 experiment, Z442RDC (C104887; Chroma) and 89006-ET-ECFP/EYFP/mcherry for an S-layer experiment, and Z442RDC and ET525/50m (Chroma) for a Dylight488 experiment.

Construction of Swimming Ghosts.

The flow chamber was composed of a 22 × 22-mm coverslip and a standard microscope slide. Two pieces of double-sided sticky tape, cut to a length of ∼30 mm and aligned on either side of an ∼5-mm-wide gap, were used as spacers between coverslips, forming a tunnel chamber of ∼15-μL volume (38). The glass surface was modified with a Ca medium containing 5 mg mL⁻¹ bovine serum albumin (BSA) (C1254; Sigma Aldrich) to avoid cells attaching.

To construct swimming ghosts, 10 μL each of cell culture in Ca medium and buffer B (2.4 M KCl, 0.5 M NaCl, 0.2 M MgCl₂, 0.1 M CaCl₂, 10 mM Hepes-NaOH, pH 7.2) containing 1 mg mL⁻¹ DNase, 5 mM ATP (A2383; Sigma Aldrich), and 0.03% sodium cholate hydrate (Sigma Aldrich) were mixed in an Eppendorf tube. Subsequently, the 20-μL mixture was infused into the flow chamber.

Phase-contrast images were captured at 20 fps for 15 s. Swimming trajectories were determined by the centroid positions of cells and subjected to analysis using Igor pro 6.30. Given the trajectory of cells, r(t) = [x(t), y(t)], the swimming velocity v(t) was defined as v(t) = $\frac{\mathit{r}\left(\mathit{t}+\mathrm{\Delta }\mathit{t}\right)-\mathit{r}\left(\mathit{t}\right)}{\mathrm{\Delta }t}$.

Bead Assay.

For the observation of a rotational bead attached to an archaellar filament, archaellar filaments were sheared by 30 times pipetting with a 200-μL pipette (F123601; Gilson), infused into a flow chamber, and kept for 10 min. Streptavidin-conjugated fluorescent beads (200 nm [F6774; Molecular Probes], 500 nm [18720; Polysciences], or 970 nm [PMC 1N; Bangs Laboratory, Inc.]) in buffer B (2.4 M KCl, 0.5 M NaCl, 0.2 M MgCl₂, 0.1 M CaCl₂ 10 mM Hepes-NaOH, pH 7.2, 0.5 mg mL⁻¹ BSA) were added into the flow chamber, incubated for 15 min, and then rinsed with buffer to remove unbound beads. The solution was replaced with buffer B plus 0.03% sodium cholate hydrate and 1 mg mL⁻¹ DNase. Immediately after the optical density of cells decreased (typically ∼5 s after buffer exchange), buffer B plus 0.03% sodium cholate hydrate was replaced with buffer B plus ATP. This process for preparing ghosts typically ended ∼10 s after the addition of detergent solution. For Fig. 2B and Movie S5, which aimed to record bead rotation for demonstration purposes, we exchanged buffers slowly and carefully to avoid the cell going out of focus. This increased the duration of the process to ∼30 s.

For pH measurements, the following buffer was used: bis(2-hydroxyethyl)iminotris(hydroxymethyl)methane⋅HCl (Bis-Tris⋅HCl) for pH 5.7 and 6.1 experiments, Hepes-NaOH for a pH 8.0 experiment, and tris(hydroxymethyl)aminomethane⋅HCl (Tris⋅HCl) for pH 8.6 and 9.3 experiments (SI Appendix, Fig. S10). For nucleotides experiments, ADP (A2754; Sigma Aldrich), ATP-γ-S (A1388; Sigma Aldrich), GTP (ab146528; Abcam and R0461; Thermo Fischer Scientific), CTP (R0451; Thermo Fischer Scientific), and UTP (R0471; Thermo Fischer Scientific) were used. Most data were collected at 100 fps for 10 s.

Bead position was determined by centroid fitting, giving cell trajectories r(t) = [x(t), y(t)]. The rotation rate was determined from either Fourier transform analysis (Fig. 2B) or linear fits to bead angle vs. time. (Fig. 2C). The rotational torque against viscous drag was estimated as T = 2πfξ, where f is rotational speed and ξ = 8πηa³ + 6πηar² the viscous drag coefficient, with r the radius of rotation of the bead center (estimated as the major axis of the observed elliptical projections of circular bead orbits), a the bead radius, and η the viscosity. We neglected the viscous drag of filaments (22), which is expected to be negligible compared with these beads (50).

To measure the viscosity of the medium, we tracked fluorescent beads diffusing near the coverslip surface for 30 s at 100 fps and performed an analysis of their mean-squared displacement vs. time. From this analysis, the viscosities are estimated to be 0.0025 Pa·s in buffer, 0.0039 Pa·s in buffer + Ficoll 5%, and 0.0072 Pa·s in buffer + Ficoll 10% at 25 °C, which are slightly higher than previous estimates (22). We inferred that this discrepancy might be due to the proximity of the glass surface (51).

Fluorescence Experiments.

For visualization of archaellar filaments, biotinylated cells were subsequently incubated with 0.1 mg mL⁻¹ Dylight488-streptavidin (21832; Invitrogen) for 3 min, washed by centrifugation, and resuspended.

FM4-64 (F34653; Life Sciences) was used to stain the archaeal cell membrane. The powder was dissolved in buffer B (1.5 M KCl, 1 M MgCl₂, 10 mM Hepes-NaOH, pH 7.0), and the cells were incubated for 30 min. The extra dye was removed by centrifugation. For microscopic measurements, the glass surface was cleaned using a plasma cleaner (PDC-002; Harrick plasm).

Quantum dots 605 (QD605) (Q10101MP; Invitrogen) was used to stain the archaeal cell surface, S-layer (38). Cells were biotinylated with biotin-NHS-ester (21330; Thermo Fischer) and incubated for 15 min at R.T. Extra biotin was washed by centrifugation. Biotinylated cells were subsequently incubated with the buffer containing QD605 at a molar ratio of 400:1 for 3 min, washed by centrifugation, and resuspended.

For visualization of CheY-GFP, Hfx. volcanii FYK7 cells were infused into a flow chamber and immobilized on a 18 × 18-mm glass surface (Matsunami glass). Unbound cells were washed away with buffer A. GFP signal was captured at 5 fps using a fluorescence microscope (Nikon Eclipse Ts2; Nikon) equipped with a 40× objective (EC Plan-Neofluar 40 with Ph and 0.75 N.A.; Nikon) and a CMOS camera (LRH1540; Digimo).

Evaluation of the Purity of NTP Solutions.

NTPs (ATP, CTP, GTP, and UTP) were purchased from abcam, Thermo Fisher Scientific, and Sigma Aldrich. Their purity was evaluated using a reverse-phase chromatography column (02485-81; Cosmosil). The column was connected to a high-performance liquid chromatography (HPLC) (PU-4180; JASCO) and preequilibrated with 20 mM potassium phosphate buffer (KPi), pH 7.0. Nucleotides were dissolved into 20 mM KPi buffer and eluted with a flow rate of 2 mL/min.

Data Availability.

All study data are included in the article and SI Appendix.