In 1973, Howard Berg and Robert Anderson argued that bacterial movement is powered by the rotation of their filaments, or flagella.1 The ultrastructure of the flagellum was soon visualized using electron microscopy.2 Negative-stain electron microscopy studies of the flagellar motor revealed that the machine is made up of a long filament, connected through a flexible hook to a basal body embedded in the bacterial cell envelope and comprising several rings. The cell envelope in Escherichia coli (E. coli) consists of an inner membrane and an outer membrane separated by a peptidoglycan layer. The energy for the rotation of the flagellum comes from dispersion of a proton motive force.3
E. coli has several flagella on its cell body. When these spin counterclockwise, they come together to form a bundle; the cell moves forward powered by the bundle’s in-unison rotation. If the cell encounters an unfavorable environment, chemical signals cause some of the flagella to rotate in the opposite direction. The bundle falls apart; the cell moves erratically or tumbles. Once the flagella recover and resume counterclockwise motion, the bundle forms again and the cell once more swims straight, but this time, on average, in a different direction. As tumbling is more likely under unfavorable conditions, E. coli can swim away from molecules it dislikes and toward those it likes.4
Classical genetics showed that the power to drive the rotation of the flagellum is derived from the motility proteins, MotA and MotB. Further studies indicated that MotA and MotB form a stator unit with proton channel activity. It is the passage of protons down an electrochemical proton gradient and through the MotAB complex that acts as an energy source. Freeze-fracture electron microscopy studies showed that stator units surround the basal bodies in the inner membrane.5 A model emerged in which stator units, which remain anchored to the rigid peptidoglycan wall of the cell, engage with the cytoplasmic portion of the basal body, or rotor, to power the rotation of the flagellum.
X-ray crystallography provided further insight into the structure of the flagellar motor.6 These advances, together with genetic and biophysical studies, allowed the piecing together of the motor’s organization.
One major finding came through crystallography: the C-ring, named for its position in the cytoplasmic portion of the basal body, can change its conformation upon binding the phosphorylated chemotaxis-signaling molecule CheY. One of the components of the C-ring, FliG, was shown to have two conformations involving an approximately 180° flip of a torque helix, which, in turn, was thought to interact with the MotAB complex.7
It was clear that this information was crucial to understanding how bacterial flagella could rotate and switch rotation direction.
The details remained out of reach in the absence of understanding the structure of the stator complexes, their interaction with FliG, or their working mechanism.
The Resolution Revolution
For many decades, x-ray crystallography—which, like electron microscopy, can provide near-atomic resolution—has been the method of choice in structural biology. The technique requires the macromolecule of interest to form large, ordered, three-dimensional arrays—crystals. Not all macromolecules can do this—especially not the proteins and complexes normally embedded in cell membranes.
Electron microscopy seemed to have many advantages over crystallography, but, given radiation damage to biological samples and the inability of negative-stain electron microscopy to provide high-resolution details, it was not an option for examining the flagellar motor. To overcome radiation damage, samples could, in theory, be frozen, since damage is less at lower temperatures. But freezing aqueous materials leads to ice formation, which again impedes imaging.
At the beginning of the 1980s, Jacques Dubochet and Alasdair McDowall managed to obtain vitreous ice by rapidly freezing aqueous solutions in liquid ethane. For the first time, researchers had a way to examine biological specimens at very low temperatures, approximately –173ºC.8
In the following years, image processing developments allowed researchers to obtain higher-resolution reconstructions of large molecules. Computer algorithms served to integrate a number of noisy micrographs into one three-dimensional image. This process led to several higher-resolution reconstructions of very large complexes such as viruses or ribosomes. The method was still no match for the mature crystallographic techniques, which routinely obtained far greater resolutions when complexes could be crystallized; and small proteins remained inaccessible to cryogenic electron microscopy (cryo-EM) techniques.
The situation changed with the advent of direct electron detectors, which, together with better image processing software, greatly improved the capabilities of cryo-EM. Before the advent of cryo-EM, researchers had collected data on photographic media, a process that was tedious and suffered from the drifting of the sample in the microscope. Or they used charge-coupled devices that converted electrons to light and back to electrons, effectively blurring the signal. New electron detectors were able directly to detect electrons, and, furthermore, allowed to correct for drift. A much better image was the result. These advances heralded a resolution revolution in cryo-EM. The 2017 Nobel Prize in Chemistry was awarded to the pioneers of this method: Dubochet, Joachim Frank, and Richard Henderson.9
What Cryo-EM Revealed
Studies of the bacterial flagellar motor were some of the first to take advantage of cryo-EM. David DeRosier and Keiichi Namba used cryo-EM to study the basal body and the flagellar filament, respectively.10 Structural insight into the stator units around the basal body was restricted to negative-stain reconstructions, again with limited resolution.11 Tomography gave additional insight into their organization,12 but the limited resolution of these reconstructions prevented detailed molecular understanding of torque generation.
Spurred by the resolution revolution, my research group set out to understand how torque is generated at the molecular level. To this end, we aimed to obtain a single-particle cryo-EM image of the stator unit.13 A similar strategy was taken in parallel by Susan Lea and colleagues,14 whereas Jun Liu and coworkers tried to answer the same question using cryo-electron tomography.15 In this review, I mainly describe the findings of my own group in collaboration with the groups of Berg and Marc Erhardt, but the three studies came to similar conclusions and almost identical working models for flagellar rotation.
The first step in single-particle cryo-EM is to isolate high quantities of the complex of interest. Membrane proteins, before being imaged, must be extracted from the membrane using detergents. In the case of the stator unit, extraction turned out to be tricky. After screening stator units from many different bacteria, we obtained well-behaving complexes from three species. Cryo-EM revealed secondary structural elements in two of them. Studies of Campylobacter jejuni MotAB resulted in reconstructions of sufficient quality, at 3.1Å resolution, to allow us to build an atomic model of the stator unit.
To our great surprise, we found that the stator unit is a complex with a stoichiometry of 5:2 MotA:MotB, and not 4:2 as previously believed.16 This ratio introduces an asymmetry, as 4:2 stoichiometry would most likely have been symmetric. MotA has four helices spanning the inner membrane and a cytoplasmic domain. The part of this domain most distal from the membrane is well conserved and binds the rotor. The five MotA molecules make a nearly symmetric, ringlike structure surrounding the N-terminal helices of two MotB molecules, which contain a universally conserved aspartate residue. Each of the MotB N-terminal helices is followed by a plug helix that lies between two MotA molecules. The two plugs cross over and are thought, based on prior experiments, to block activity of the stator unit. As expected, the channel appears in closed conformation.
To investigate what would happen when the channel opens, we deleted these plug helices, resulting in unplugged channels. The cryo-EM structure of unplugged MotAB was similar to that of the plugged version, except for two residues, namely the conserved aspartate residue of one of the MotB chains and a conserved hydrophobic residue at the top of a putative channel in one of the MotA chains. The residue is in one position in the plugged version, but in two conformations in the unplugged channel, suggesting a possibility for regulating proton access.
We next wanted to find out what happens upon protonation of the conserved aspartate, which is thought to be an essential step in the rotation mechanism. To this end, we mutated this aspartate in both chains to asparagine, which mimics protonation. Another study showed this protonation affects the proteolysis pattern of MotA. This led the authors to suggest that a protonation- (or deprotonation-) induced conformational change in MotA powers the flagellar motor.17 To our surprise, the structure of the protonation-mimic unplugged stator unit was similar to that of the unplugged stator unit. The only difference was the local conformation around one of the two conserved aspartates in MotB.
This puzzling result can be explained if it is the rotation of MotA around MotB that drives the rotation of the motor. We proposed that one of the two MotB aspartates—call it MotB1—is protonated and anchored to MotA. This is a high-energy state for MotB1, which would drive MotA rotation were it not for the aspartate of MotB2. This aspartate is both negatively charged and unprotonated: the neutral surface of MotA cannot rotate across it. But when MotB2 accepts a proton from the periplasm, it is neutralized, and rotation of the hydrophobic MotA surface across the neutralized MotB2 aspartate can now occur. After rotation, MotB2 grabs on to MotA in this new position, and MotB1 releases its proton. MotB1 then waits at the cytoplasmic side to pick up a new proton. After these steps, a 36º rotation of the MotA ring around the MotB dimer has occurred. The whole structure is in the same state as before, except that MotB1 and MotB2 have switched roles, so the cycle can begin again.
Through independent work, Lea and colleagues arrive at a similar hypothesis based on the structure of the plugged stator unit from another bacterium. Further biophysical experiments will be needed to prove or disprove these models.
How does this miniature MotAB rotary motor power the rotation of a large flagellum? Upon incorporation in the motor, the C-terminal domains of MotB dimerize and bind to the peptidoglycan layer that forms part of the cell envelope. Peptidoglycan binding is accompanied by the unplugging of the MotAB ion channel, which allows ions to flow from the periplasm to the cytoplasm. Upon ion flow, MotA rotates clockwise around MotB. Given normal swimming conditions, the rotor is engaged at the proximal side of the MotA ring, and the clockwise rotation of MotA rotates the rotor counterclockwise. When all flagella spin counterclockwise, they form a bundle and swim in a straight line.
Upon chemotactic signaling, a whole signaling cascade takes place and results in the phosphorylation of CheY. Phosphorylated CheY binds to the rotor and switches its conformation. The rotor now interacts with the distal side of the MotA ring. The same clockwise rotation of MotA around MotB induces a clockwise rotation in the motor. One or several flagella change their rotation direction, breaking up the flagellar bundle. The bacterium starts tumbling, until CheY is dephosphorylated and gets released from the rotor and all flagella again spin in the same counterclockwise direction and the bundle reforms.
Researchers have come a long way since Antonie van Leeuwenhoek observed moving bacteria several centuries ago. They now know bacteria swim using long filaments powered by a bidirectional, rotary proteinaceous motor. They know the molecular makeup of the rotary motor, and that this bidirectional motor is itself driven by unidirectional miniature rotary motors. Yet several unanswered questions still remain. What is the exact energy consumption of the motor—how many ions are used per rotation of MotA around MotB? How can the motor use several stator units? Do they need to act cooperatively, or can single stator units drive rotation even in the presence of other bound but non-active stator units? A combination of single-molecule light microscopy, biophysical experiments, and cryo-electron tomography is likely necessary to answer these questions.
This essay is dedicated to the memory of Howard Berg, a pioneer in understanding the molecular basis of bacterial motility. It was a pleasure and honor to collaborate with him on our study of the stator unit.
