Flagellar rotation comes full circle

Hosu et al. (1) provide experimental confirmation that rotation of the bacterial flagellum, proposed 50 y ago by Berg (2), is driven by the rotation of multiple small wheels. Flagellar anatomy is shown in figure 1A (1). A basal body made up of a membrane-embedded MS-ring connects to a rod that passes through the periplasmic space, the peptidoglycan cell wall, and the outer membrane. The tip of the rod is attached to a flexible hook, which connects the rod to the helical flagellar filament. The hook is a universal joint that converts rotation of the rod perpendicular to the cell wall into rotation of a bundle of helical flagellar filaments parallel to the long axis of the cell body that propel the cell.

The cytoplasmic C-ring attaches to the MS-ring via the FliG protein. The C-ring also contains the FliM and FliN proteins. Default rotation is counterclockwise (CCW). The phosphorylated form of the CheY protein (CheY-P) interacts with FliMN to generate clockwise (CW) rotation. CheY-P couples input from the polar chemoreceptor patch to mediate chemotaxis. CCW rotation propels the cell in a straight swim. CW rotation disrupts the flagellar bundle to cause a chaotic tumbling motion that reorients the swimming direction (3).

Flagellar rotation requires two peripheral proteins, MotA and MotB (4). These proteins form a complex located at the periphery of the MS-ring, which is called a stator because it is firmly attached to the cell wall via MotB. The MotA subunit spans the cytoplasmic membrane four times and extends into the cytoplasm to interact with FliG (5).

The protonmotive force (PMF) across the cytoplasmic membrane is coupled to flagellar rotation. While a postdoc in Berg’s lab, I learned that there is a flagellated Streptococcus (strain V405; 6) that can be rendered immotile by a brief period of nutrient starvation. It lacks an outer membrane and is thus sensitive to the K⁺ ionophore valinomycin.

I tested whether nonmotile deenergized cells swam after imposing a valinomycin-induced K⁺ diffusion potential or a pH shift to generate a PMF. Either could initiate swimming (7). The rotational speed of cells tethered to a slide via a single short flagellum was directly proportional to the magnitude of the protonmotive force (PMF), whether the PMF depended on a membrane potential or a pH gradient (8). A PMF of opposite sign (inside positive relative the external medium) also drove flagellar rotation. Thus, protons moving into the cell, the normal case, or out of the cell could be coupled to flagellar rotation.

It was puzzling that the flagella rotated CCW regardless of the direction of proton flow. Thinking chemotaxis might confound the issue, I isolated nonchemotactic mutants of V405 (9). One mutant rotated its flagella CCW even in the presence of repellents, which caused wild-type cells to rotate their flagella CW. When a PMF of opposite sign was generated in these mutant cells, their flagella turned exclusively CW. Thus, reversing the direction of proton flow in a cell with its flagellum locked in CCW rotation caused that flagellum to rotate CW. At the time, I could not explain that phenomenon. Now I can (see below).

The protons that drive flagellar rotation flow through the stator units (5). Biochemical analysis predicted that the MotA/MotB stoichiometry in the stator unit was 4:2 (5). Both MotA and MotB contribute to the ion channel, and a highly conserved aspartate residue in the middle of the single membrane-spanning helix of MotB is critical for proton conduction (5). The 4:2 stoichiometry suggested that each MotB subunit interacts with two MotA subunits.

The cytoplasmic domain of MotA interacts with FliG via an interaction of conserved charged residues (10). Various models predicted how protonation of the conserved aspartate could change the conformation of the cytoplasmic domain of MotA such that it pushes against FliG to apply the torque that drives rotation of the C-ring. These models were speculative and had no experimental support.

Stator units float freely in the cytoplasmic membrane and associate and dissociate with the C-ring (11). Free stator units normally do not conduct protons because their ion-conducting channels are plugged by a short amphipathic helix in MotB.

Overproduction of wild-type stators has no detrimental effect. However, when stators in which this helix is deleted are overproduced, the cells quickly stop growing because their cytoplasm is acidified by an influx of hydrogen ions at a rate that exceeds the ability of the cells to pump them out again (12). This result suggests that, when a stator unit interacts with the C-ring, its MotB subunits undergo a conformational change that removes the plug from the ion-conducting channel. This conformation allows the periplasmic domains of the MotB dimer to extend and attach to the cell wall. This prediction was confirmed experimentally (13).

The critical insight that got everyone thinking about stator rotation came from two papers published in 2020 (14, 15). Single-particle cryoelectron tomography of isolated stator units demonstrated that the true stoichiometry was five MotA proteins to two MotB proteins. This insight immediately suggested that the peripheral ring of five MotA protomers might rotate around a stationary axle of the MotB dimer.

This discovery came at a propitious time for me. I was approached by Frontiers in Microbiology to organize an issue devoted to rotary motors. I accepted the invitation. The result is a collection of 20 reviews of the enormous variety of biological rotary nanomotors (16). These include different bacterial 5:2 motors, the independently evolved adenosine triphosphate (ATP)-driven rotary motor of the archaellum, and the ubiquitous ATP synthase that generates ATP in all of the different domains of life.

“In PNAS, Hosu et al. provide experimental confirmation (1) that the rotation of the bacterial flagellum proposed by Howard Berg more than 50 y ago (2) is driven by the rotation of multiple small wheels within the cell.”

The most relevant of these reviews to the present work is a comprehensive discussion of how a 5:2 rotary motor might operate (17). The authors suggest that the proton channel is actually not continuous but rather consists of two half channels. One conducts protons from the outside of the cell (the periplasm in Escherichia coli) to the essential aspartate residue in MotB. The second conducts protons that dissociate from the aspartate residue into the cytoplasm. Providing access of the second half channel to the protonated aspartate residue requires a 36° displacement of the MotA ring.

There are two active proton channels per stator unit, one with its half channel open to the outside of the cell and the other with its half channel open to the cytoplasm. Thus, ten proton translocations through the membrane generate one 360° rotation of the pentameric MotA ring. This process is shown in figure 1B (1).

One can model the MotA ring as a small gear and the C-ring as a large gear, with the oppositely charged residues of MotA and FliG serving as gear teeth. If these charged stator residues are positioned outside the C-ring, then CW rotation of the MotA ring drives CCW rotation of the C-ring. If the C-ring expands in response to binding of CheY-P, then CW rotation of the MotA ring inside the C-ring drives its CW rotation. A schematic image of this mechanism is shown in figure 1C (1).

The theory was there, but where was the experimental proof? The stator units are tiny. The diameter of the MotA ring is 7.5 nm, compared to the 44 nm diameter of the FliG ring (18). Direct visualization of MotA ring rotation using conventional fluorescence microscopy is impossible. The current study recognized that fluctuations in the polarized fluorescence of MotA or MotB could detect rotation.

The fluorescent tag they chose was the Venus variant of yellow fluorescent protein. Venus was attached to the cytoplasmic C-terminus of MotA or the cytoplasmic N terminus of the transmembrane helix of MotB. The linkers were designed to be rigid enough to prevent excessive movement of Venus relative to the tagged Mot protein and to be positioned so as not to interfere with the function of the labeled Mot protein. This was straightforward with MotB, but with MotA, it required considerable protein engineering to position Venus to interfere minimally with the portion of MotA that interacts with FliG. The labeled proteins were only slightly less functional than their wild-type equivalents.

There should be no fluctuation of fluorescence of motors containing uniformly labeled Mot proteins. The trick used in this study was to photobleach some MotA or MotB proteins with a brief high-intensity pulse of polarized light. Fluorophores in the plane of S-polarized light should bleach preferentially, leading to uneven bleaching among the subunits of multimers. This uneven bleaching gives fluctuations in the intensity of fluorescence from a rotating structure excited with low intensity S-polarized light. The basic idea is illustrated in figure 2A (1).

To image the labeled Mot protons from a flagellum fixed in space, E. coli cells lacking a filament and producing a sticky FlgE hook protein were employed. The cells tethered spontaneously through interaction of the hook with the slide. Note that this attachment keeps the C-ring of the tethered flagellum fixed with in the laboratory frame of reference; the body of the tethered cell rotates around the stationary C-ring. There was little wobble during rotation as the tether was short.

Figure 2C (1) shows that a cell with labeled MotB produced a very regular pattern of fluorescence, with four peaks over two rotations (720°) of the cell body. Thus, there were two peaks of fluorescence per rotation of the cell body. Figure 2E (1) shows that the same pattern was seen in the majority of the 21 cells analyzed. The result was the one expected if the two MotB proteins in a stator rotate with the same frequency as the cell body because they are tethered to the cell wall of the rotating cell. The C-ring remains fixed in space, rigidly connected to the tethered flagellar hook.

Figure 2D (1) shows that a cell with labeled MotA produced a very different pattern of fluorescence fluctuation. In a recording made over one-sixth of a rotation (60°), there were two peaks of rotation at intervals of about 20°. The average peak separation for the 36 cells analyzed was ~25°. Using that figure, there were 360/25 = 14.4 rotations of the MotA ring per rotation of the cell body. Subtracting the contribution of the rotation of the cell body to the MotA signal, there were 12.4 rotations of the MotA ring per cell body rotation. Thus, in a cell in which the flagellum is not tethered, the MotA ring rotates 12.4/2 = 6.2 times faster than the cell body.

These numbers are enlightening. The C-ring contains 34 FliG subunits (19). If each 72° rotation of the MotA ring brings a new MotA subunit into contact with a FliG subunit, one rotation of the MotA ring leads to the displacement of the C-ring by five FliG subunits if rotation of the MotA ring and rotation of the C-ring are tightly coupled (20). One rotation of the MotA ring should move the C-ring by 5/34 of a rotation, or ~53°. Thus, there should be 360/53 = 6.8 rotations of the MotA ring per rotation of the C-ring. The experimentally determined ratio of 6.2 rotations of the MotA ring per rotation of the cell body is remarkably close to this value.

It is also suggestive that the ratio of the diameters of the FliG ring and MotA ring is 44 nm/7.5 nm = 5.86. The exact overlap of the MotA and FliG rings is unknown. Therefore, this value is remarkably close to the values for the measured and calculated ratios for the rotation rates of the tightly coupled MotA and FliG rings.

This work establishes two important features of the flagellar motor. First, it provides direct evidence that the pentameric MotA ring rotates around the MotB dimer of a stator unit. This same feature has not been demonstrated for the 5:2 stators found in a variety of other motors predicted to be rotary (1, 16), but it strongly suggests that such may be the case. Second, it demonstrates that the movement of protons through the stator units of the flagellar motor is tightly coupled to the rotation of the C-ring (20).

I chose the title for this commentary because this study (1) represents closure of an adventure that began with Berg’s original proposal that bacterial flagella rotate (2). He published that idea long before we had any idea about the structure and operational properties of the flagellar motor. I recall my excitement as a graduate student in 1973 upon reading that paper. It set a new course for my scientific career in a direction for which I was totally unprepared. The rotation of the MotA ring in the CCW direction in response to an outward proton flow explains the mystery behind my observation that a reversed proton flow causes a normally CCW-locked mutant motor to rotate CW. With an outward proton flow, the MotA ring should rotate CCW and the C-ring CW. The publication of this paper fulfills the quest to which I devoted much of my scientific life: To understand how rotation of the flagellar nanomachine is generated.