Oxygen Depletion Triggers Switching between Discrete Speed Modes of Gonococcal Type IV Pili

Abstract

Type IV pili are polymeric bacterial appendages that affect host cell interaction, motility, biofilm formation, and horizontal gene transfer. These force-generating motors work in at least three distinct velocity modes—elongation, and retraction at two distinct speeds, high and low. Yet it is unclear which regulatory inputs control their speeds. Here, we addressed this question for the human pathogen Neisseria gonorrhoeae. Using a combination of image analysis and surface analytics, we simultaneously monitored the speed of twitching motility and the concentration of oxygen. While oxygen was detectable, bacteria moved in the high-speed mode (1.5 μm/s). Upon full depletion of oxygen, gonococci simultaneously switched into the low-speed mode (0.5 μm/s). Speed switching was complete within seconds, independent of transcription, and reversible upon oxygen restoration. Using laser tweezers, we found that oxygen depletion triggered speed switching of the pilus motor at the single-molecule level. In the transition regime, single pili switched between both modes, indicating bistability. Switching is well described by a two-state model whereby the oxygen level controls the occupancy of the states.

Introduction

Motility can enhance the fitness of bacteria. Various bacterial species have evolved diverse molecular motors for movement (1). In response to environmental changes, the direction (2) or the speed (3–5) of the molecular motor can be regulated. In particular, the speed of the flagella motor changes in response to limitation of metabolite (e.g., with decreasing concentration of metabolites, the speed decreased continuously (3)). We have recently shown that individual type IV pilus motors of the human pathogen Neisseria gonorrhoeae (gonococcus) and of Myxococcus xanthus show two discrete speed modes (6,7). Bimodality of speed is an unusual property of molecular motors, and it is unclear which external triggers control their occupancy.

Type-IV pili (T4P) are among the most ubiquitous bacterial cell appendages and they are generated by a large number of bacterial species (8). Their functions are diverse including adhesion to host cells and inert surfaces, twitching motility, biofilm formation, and transformation. Most likely, T4P can support these diverse functions because the pilus itself serves as a polymeric scaffold. Functionality is then defined through integration of minor pilins (9,10) and through posttranslational modification of the major pilin subunit (11).

For the human pathogens N. gonorrhoeae (gonococcus) and Pseudomonas aeruginosa it has been shown that the length of the T4P is highly dynamic, as it can increase through polymerization and decrease through depolymerization (12,13). During retraction, single T4P generate mechanical force exceeding 100 pN (14). Dynamics and force generation by gonococcal T4P trigger cytoskeletal rearrangements during infection (15,16). Furthermore, T4P retraction controls the architecture of biofilms formed by Neisseria meningitidis (17), which is closely related to N. gonorrhoeae. Dynamic T4P mediate twitching motility by cycles of pilus polymerization, adhesion to surfaces, and subsequent retraction by depolymerization (12) (Fig. 1 a). They are polymerized from the major pilin subunit PilE and polymerization is supported by the ATPase PilF (18). The hexameric ATPase PilT supports depolymerization of the pilus (19,20). Gonococcal pili are, on average, ≈1-μm long (21) and 6-nm thick (22).

Figure 1.

Global speed switching. (a) Molecular mechanism of twitching motility. Multiple adhering pili retract and pull the bacterium by a distance δ toward the points of attachment. (b) Bright-field image of multiple motile bacteria on a glass surface 17.0 min and (c) 17.8 min after sealing the sample. (Solid lines) Path of the respective bacterium over a time period of 10 s. Scale bar 10 μm. (d) Overlay of speeds v of 25 bacterial tracks as a function of time. The value t_gs roughly denotes time-point of global switching. (e) Relative frequency of speeds for t ≪ t_gs (solid, N = 825) and for t ≫ t_gs (shading, N = 265). N, number of bacterial tracks. (f) 〈δ²(τ)〉 for t ≪ t_gs (open circles, N = 138) and for t ≫ t_gs (shaded triangles, N = 65). (Solid lines) Fit to Eq. 1. N, number of bacterial tracks longer than 1 min.

We have recently demonstrated that the type-IV pilus motor is a multistate system (21). The pilus motor can reverse its direction, switching from retraction to elongation, and back (6). The probability of elongation increases with external force and with decreasing concentration of the retraction protein PilT. Retraction at a clamped force of 8 pN can occur either in the high-speed mode ≈ 2 μm/s or in the low-speed mode at ≈1 μm/s. Bimodal speed distribution has been shown to be conserved in the phylogenetically distant bacterial species M. xanthus (6). Thus, the pilus motor has at least three different states: elongation, retraction at high speed, and retraction at low speed. This multistate system is likely to enable the bacterium to adapt very rapidly to changing environment through nongenetic switching. It is therefore important to find external inputs that trigger switching between the different states.

In this work, we address the question how the speed of gonococci is regulated in response to environmental changes. We address this question both at the level of a single pilus motor and at the level of bacterial motility that involves simultaneous action of multiple pili.

Materials and Methods

Growth conditions and media

N. gonorrhoeae wild-type strain MS11 was grown overnight at 37°C and 5% CO₂ on gonococcal base agar plus supplements as described previously (6). Before each experiment, gonococcal colonies were resuspended in retraction assay medium consisting of phenol red-free Dulbecco's modified Eagle's medium (GIBCO, Grand Island, NY), 4.5 g/L Glucose (GIBCO), 2 mM L-glutamine (Roth, Darmstadt, Germany), 8 mM sodium pyruvate (GIBCO), and 30 mM HEPES (Roth). A quantity of 5 mM ascorbic acid (Roth) was added unless stated otherwise. To generate anaerobic conditions, an oxygen scavenger system based on 2.5 mM protocatechuic acid (Sigma-Aldrich, St. Louis, MO) and 50 nM protecatechuate-3,4-dioxygenase (Sigma-Aldrich) was added to the media (23). To investigate the role of transcription in speed switching, RNA polymerase was inhibited by adding 120 μM rifampicin (Sigma-Aldrich) to cell suspensions 60 min before sealing the sample chamber.

A two-dimensional oxygen sensor

An oxygen sensor based on the oxygen-sensitive dye Pt(II) meso-tetra(pentafluorophenyl)porphine (PtTFPP; Frontier Scientific, Logan, UT) was fabricated to monitor oxygen consumption and bacterial motility simultaneously (Fig. 3 a) (24). Stock solution of PtTFPP (20 mM in toluene) was stored at room temperature. PtTFPP is embedded in a Sylgard 184 polydimethylsiloxane network (PDMS; Dow Corning, Midland, MI). Therefore, PDMS was mixed with Sylgard 184 curing agent (Ratio 10:1) and 1 mM PtTFPP and directly spin-coated on cover slides to result in ∼30-μm thin layers. In the end, oxygen sensors were cured at 60°C for at least 3 h. Calibration and oxygen measurements are described in the Supporting Material.

Figure 3.

Global switching occurs upon complete consumption of oxygen. (a) Setup for simultaneous measurement of oxygen concentration and speed. (b) Oxygen concentration as a function of time (solid circles) and average speed v of 30–40 bacteria over a period of 30 s for each time-point (shaded triangles). (c) Oxygen concentration as a function of time (solid circles) and average speed v of 10–20 bacteria per time-point (shaded triangles) after addition of oxygen scavenger. (b and c) Without ascorbic acid.

Twitching motility assays

Twitching motility assays of N. gonorrhoeae were performed on bovine serum albumin (Sigma-Aldrich) coated glass in a commercial microscope equipped with a heated thermal insulation box equilibrated at 37°C. Bacterial motility was monitored via standard video microscopy (10 frames/s) and subsequent cell tracking in MATLAB R2009b (The MathWorks, Natick, MA). Tracking is based on the algorithms of J. C. Crocker and D. Grier originally written in the programming language IDL, and transferred to MATLAB code by D. Blair and E. Dufresne available on http://physics.georgetown.edu/matlab/ (25) (further details are available in the Supporting Material).

Single pilus retraction assays

Single pilus retraction events were measured with an optical tweezers in force-clamp mode (6) (details in the Supporting Material). Experiments were conducted at 37°C as for twitching motility assays. Bacteria were fixed to polystyrene spin-coated glass slides. Carboxylated polystyrene microspheres (Molecular Probes, Eugene, OR), 2 μm in diameter, were added to the cell suspension before sealing the chamber.

Results

Speed of bacterial motility switches rapidly and globally between two discrete speed modes

For characterizing the temporal development of speed of twitching motility, a suspension of bacteria was added to a bovine-serum-albumin-coated cover slide and sealed. Immediately after attachment, bacteria started to move over the surface (Fig. 1 b). Bacterial traces were obtained by tracking the movement of individual bacteria (≈30) within one field of view (≈6700 μm²). The speed distribution was broad with an average speed v_H = (1.47 ± 0.02) μm/s (Fig. 1 e). Around the maximum, the distribution was Gaussian-shaped, but at low speeds, the distribution showed a pronounced tail. The distribution was in agreement with previous experiments (26). Unexpectedly, we observed that several minutes after sealing the sample, the speed of all bacteria (≈ 30) in one field of view continuously decreased for several seconds (see Movie S1 in the Supporting Material). After this switching period, the average speed assumed an average value of v_L = (0.51 ± 0.02) μm/s (Fig. 1, c–e) and was in good agreement with a Gaussian. This speed did not decrease further for at least 30 min, indicating that bacterial speed switched from a high-speed mode to a low-speed mode. Before global switching, the mean-squared displacement of movement was in agreement with a correlated random walk, as demonstrated previously (21),

$$\langle {\delta }^2\left(\tau \right)\rangle =2{\tau }_c{v}^2\left(\tau -{\tau }_c\left(1-e^{-\frac{\tau }{{\tau }_c}}\right)\right)+A.$$ (1)

Here, we fitted the data for 0.1 s < τ < 15 s and obtained a characteristic speed of v = (1.6 ± 0.1) μm s⁻¹, a correlation time τ_c = (2.3 ± 0.5) s, and an offset A = (0.02 ± 0.04) μm² (Fig. 1 f). After global switching, movement was still persistent with v = (0.52 ± 0.02) μms⁻¹, a correlation time τ_c = (5.9 ± 0.9) s, and A = (0.01 ± 0.01) μm² (fit between 0.1 s < τ < 30 s). Please note that the speed of twitching motility v is not to be confused with the speed of single pilus retraction v_s. These results demonstrate that the correlation length, l_c = v · τ_c with l_c = (3.7 ± 0.8) μm, before global switching decreased only slightly to l_c = (3.1 ± 0.5) μm after global switching, and that persistent movement is not affected after oxygen depletion. In this set of experiments, the mean twitching speed and the correlation time was higher than in previous experiments (21), which is most likely due to slight variations between different bacterial stocks and an improved evaluation of bacterial tracks.

Oxygen depletion reversibly triggers speed switching

N. gonorrhoeae consumes oxygen by respiration. To assess the hypothesis that oxygen depletion caused speed switching, we added an oxygen scavenger system consisting of the enzyme protocatechuate-3,4-dioxygenase and the substrate protocatechuic acid in a flow chamber. We verified that the addition of the individual components did not affect the time-point of global switching t_gs (see Fig. S1 in the Supporting Material). We found that global switching from the high-speed mode to the low-speed mode occurred after ∼2 min (Fig. 2 a), consistent with the time reported for consumption of oxygen by this scavenger system (23). Subsequently, we exchanged the medium with fresh (oxygen-rich) medium and found that all bacteria within the field of view immediately switched back to the high-speed mode with a switching period of ≈1 s (Fig. 2 b). These experiments clearly demonstrate that oxygen depletion triggers the switch from the high-speed mode to the low-speed mode and that switching is fully reversible.

Figure 2.

Change in oxygen concentration triggers global speed switching. (a) Overlay of speeds v of 81 bacterial tracks as a function of time. The oxygen-scavenger injection period started at 0.5 min and ended 2.0 min after twitching was monitored. Time-point of global switching t_gs = 2.2 min after injection. (Solid line) Fit to sigmoidal function. (b) Single bacterial track during influx with fresh medium saturated with oxygen. Medium was added ∼10 min after global switching shown in panel a. (Solid line) Fit to sigmoidal function. (c) Relative frequency of speeds under anaerobic conditions (shading, N = 32) and after reenergizing with fresh medium (solid, N = 42). N, number of bacterial tracks.

We next sought to correlate the speed of twitching motility with oxygen levels. Therefore we simultaneously measured the oxygen concentration and the speed of individual bacteria as a function of time via a two-dimensional oxygen sensor (Fig. 3 a). The phosphorescence of the embedded porphyrin dye PtTFPP is quantitatively quenched in the presence of oxygen following the Stern-Volmer equation (see Fig. S2). Two-point calibration was performed at saturated oxygen [O₂] = 180 μM (37°C) and zero oxygen conditions that led to a Stern-Volmer constant of K_SV = (0.70 ± 0.09) L μmol⁻¹. The average speed was slightly lower on the PDMS surface than on the glass surface (Fig. 3 b). We found that bacteria switched into the low-speed mode when the oxygen was decreased to a level close to the sensitivity of our sensor, indicating that mode switching occurred upon depletion of oxygen (Fig. 3 b). To further support this result, we simultaneously measured speed and oxygen concentration after addition of the oxygen scavenger, and found that again switching into the low-speed mode occurred upon depletion of oxygen (Fig. 3 c). Please note that global switching on the oxygen sensor occurs at later time-points, because the PDMS layer is an additional oxygen reservoir. The oxygen depletion rate in the medium was nearly constant up to full depletion, with an average rate of (0.015 ± 0.001) fmol/(min^∗cell) (see Fig. S3). Altogether, the results clearly show that N. gonorrhoeae switch into a motility mode with low speed once the oxygen in the medium is near depletion.

The switching kinetics is too fast to be explained by regulatory mechanisms driven by changes in gene expression. To further support the result that gene expression is not necessary for speed switching, we measured the global switching time t_gs with and without the RNA polymerase inhibitor rifampicin and found no significant difference, confirming that modulation of gene expression induced by anaerobiosis (27) was not involved in mode switching (see Fig. S4).

Single pilus retraction speed switched from a high-speed mode to a low-speed mode upon oxygen depletion

Next, we investigated whether speed switching was due to speed switching of single pilus retraction. Using laser tweezers, we characterized retraction speed of individual type-IV pili before and after global switching for three different forces clamped at 8 pN, 30 pN, and 60 pN, respectively (Fig. 4 a). Before and after global switching, the speed distributions were monomodal, e.g., at 8 pN with an average high-speed mode of v_s = (2053 ± 33) nm/s and an average low-speed mode of v_s = (1121 ± 50) nm/s (Fig. 4 a). This behavior was independent of external force, while both absolute values shifted slightly to lower values with increasing force. Oxygen scavenger treatment at early time-points resulted in retraction only in the low-speed mode. Fig. 4 b shows an overlay of three single pilus retraction events measured consecutively before, during, and after global switching at a clamping force of 30 pN. Their corresponding speed distributions are displayed in Fig. 4 c. During the switching period (t ≈ t_gs), the speed distribution was bimodal and interestingly, switching between both modes was observed within a single retraction event within the switching period. This result shows that the pilus motor functions in two states that yield different speeds and that the availability of oxygen controls their occupancy in N. gonorrhoeae.

Figure 4.

Mode switching of single pilus speed v_s is triggered by oxygen depletion. (a) Overlay of three single pilus retraction events for t < t_gs (light shading), t ≈ t_gs (solid), and t > t_gs (dark shading) at 30 pN. Pilus length change d is plotted versus time t. (b) Corresponding speed histograms of single pilus retraction for t < t_gs (light shaded), t ≈ t_gs (solid), and t > t_gs (dark shading). (c) Dependence of average speeds v_s of single pilus retraction on force for t < t_gs ((light shading) and t > t_gs (dark shading), after addition of oxygen scavenger (open). Data are averaged over 20–170 individual pilus retractions for each condition.

Dynamics of switching depends on the rate of oxygen consumption

Next, we investigated the dynamics of switching at varying cell densities and found that the time-point of global switching t_gs decreased with increasing cell density equivalent to increasing oxygen consumption rate (Fig. 5, a–e). Here, an optical density OD₆₀₀ = 1 at 600-nm wavelength corresponds to 1.1 × 10⁹ cells/ml. When adding ascorbic acid as an additional oxygen consumer, switching occurred earlier. The effect of ascorbic acid was negligible for cell concentrations exceeding OD₆₀₀ = 0.24.

Figure 5.

Dynamics of switching. (a) Speed v of 59 bacterial tracks as a function of time (shaded circles) at OD₆₀₀ = 0.06. (Solid line) Fit to Eq. 4. (b) Speed v of 55 tracks as a function of time (shaded circles) at OD₆₀₀ = 0.24. (Solid line) Fit to Eq. 4. (c) Speed v of a single bacterial track as a function of time (shaded circles) at OD₆₀₀ = 0.06. (Solid line) Fit to Eq. 4. (d) Speed v of a single bacterial track as a function of time (shaded circles) at OD₆₀₀ = 0.24. (Solid line) Fit to Eq. 5. (e) The value t_gs at varying cell densities without (solid circles) and with 5 mM ascorbic acid (open squares). (Solid line) Fit to b + a/OD.

Interestingly, the time window of global switching was also dependent on cell density (Fig. 5, a and b). To investigate whether individual cells switched at different rates or whether the starting point for switching showed a stronger scatter at lower cell densities, we investigated the switching dynamics of individual cells (Fig. 5, c and d). We found that individual cells also showed faster switching with increasing cell density, corresponding to a higher oxygen consumption rate.

Discussion

Dynamics of switching is consistent with a two-state model

The single molecule experiments strongly suggest that the gradual decrease of speed of gonococcal twitching motility during the switching period (Fig. 5) results from frequent switching between the two speed states at the single pilus level. Because motility is driven by multiple pili simultaneously, and because we had averaged over the speeds of movement of multiple bacteria, we conclude that the gradual decrease of speed during the switching period results from averaging over the two speed modes of individual pili. Pausing and switching from retraction to elongation occurred infrequently (see Fig. S5). Therefore, we will not consider these two states of pilus dynamics in the following. It is interesting to note that another bacterial species, P. aeruginosa, switches back and forth between different speeds of T4P-mediated surface motility at constant environmental conditions (28,29). The underlying mechanism in this case is different from our mechanism as it is associated with its rodlike shape.

To understand the switching behavior quantitatively, we consider the pilus system as a two-state system, whereby one state corresponds to the high speed and the other to the low speed (Fig. 6 a). The average speed of the ensemble of motors v(t) at time t is given by

$$v\left(t\right)=p_H\left(t\right)v_H+p_L\left(t\right)v_L,$$ (2)

where p_H and p_L are the probabilities of finding the motor in the high-speed mode or the low-speed mode, and v_H and v_L are the speeds in the high and low mode, respectively. We assume that the occupancy of the states was near equilibrium during the switching period and that the decrease of speed during the switching period was a result of continuous decrease of oxygen concentration. If we further assume that the occupancy of the states follows a Boltzmann distribution (see the Supporting Material), then

$$v\left(t\right)=v_H-\frac{v_H-v_L}{e^{-\beta \Delta \epsilon \left(t\right)}+1},$$ (3)

with β = (k_BT)⁻¹, and Δε describes the difference of free energies of both states dependent as a function of time (Fig. 6 a). For simplicity, we assume that Δε(t) ∼ t near the time-point of global switching and we fit our data with the function

$$v\left(t\right)=v_H-\frac{v_H-v_L}{\text{exp}\left(k\left(t_{gs}-t\right)\right)+1},$$ (4)

where Δε(t) = k(t − t_gs)/β, t_gs is the time-point of global switching, and k is the rate at which the free energy difference between the states changes. At different cell densities (associated with different rates of oxygen consumption), this model can describe the data averaged over multiple bacteria well (Fig. 5, a and b). Nevertheless, we point out that the assumption that the relative energies of the states change linearly with oxygen concentration, is likely to be an oversimplification.

Figure 6.

Two-state model for speed switching. (a) Two-state model. (b) Rate at which the free energy difference between the state changes k as a function of the oxygen consumption rate r_ox without (solid circles) and with 5 mM ascorbic acid (open squares).

Because we have found that the oxygen consumption rate r is constant for t < t_gs (see Fig. S3), it was estimated as the ratio between the saturating oxygen concentration at room temperature c₀ = 254 μM and t_gs. We found that the rate k was dependent on r for r > 17 μM/min (Fig. 6 b), as expected if the system is near equilibrium. In the presence of ascorbic acid, k showed a similar behavior but the values were slightly lower, potentially due to an unidentified interaction of bacteria with ascorbic acid. When the rate of change of free energy difference, k, is lower than the rate at which the transition from the high-speed state to the low-speed state occurs, k_HL, the system is near equilibrium, consistent with our initial assumption. For r > 17 μM/min, the rate k did not change when further increasing the oxygen consumption rate. Therefore, we propose that in this regime, we see the dynamics of the transition from the high-speed state to the low-speed state. Assuming an Arrhenius-like behavior and k_HL ≫ k_LH, the relaxation should have the form

$$v\left(t\right)=v_L+\left(v_H-v_L\right)e^{-k_{HL}t},$$ (5)

whereby k_HL is the rate constant of the transition from the high-speed state to the low-speed state. Fitting the data yields a rate of k_HL = (0.19 ± 0.03)s⁻¹ at r = 20 μM/min and k_HL = (0.17 ± 0.04)s⁻¹ at r = 48 μM/min. We conclude that the two-state model describes the speed switching well.

Speed switching occurs at forces up to 60 pN

We have previously reported that the high-speed mode does not occur at forces exceeding 30 pN (6), whereas in the experiments reported here the high-speed mode was observed at 60 pN. We attribute this discrepancy to the fact that at high forces, successful retraction events are much less frequent and therefore experiments were extended to longer time periods. In addition, the average unbinding force of a pilus from the polystyrene bead is only 30 pN (21) or lower, depending on the batch of beads used. Thus, it is very likely that for most retraction events that we have characterized in previous experiments at high force, oxygen had already been depleted.

Putative molecular mechanisms coupling oxygen depletion to speed switching

How does oxygen depletion trigger speed switching?

It is conceivable that the oxygen levels are directly sensed by a sensor protein such as FNR (30,31). FNR is known as a transcriptional regulator; however, FNR or a different oxygen-binding protein may have a second function in protein-interactions. The oxygen concentration may then be translated into motor velocity through tuning the functionality of the motor, reminiscent of flagella rotation in E. coli (3) and Bacillus subtilis (5). Oxygen is the final electron acceptor of the respiratory chain that helps in maintaining the proton gradient between the cell's interior and its exterior space. Thus, a different coupling mechanism may involve proton-motive force. Finally, when respiration is impaired, the function of the ATP synthase is inhibited and consequently the intracellular ATP pool depletes. Structural studies strongly suggest that the six PilT retraction ATPases hydrolyze ATP in a coordinated manner (20,32). Thus, the enzymatic activity and the velocity may depend on the ATP concentration in a nonlinear way, defining a cutoff concentration of ATP for coordinated fast and uncoordinated slow modes. Future studies will have to assess these hypotheses.

Possible biological functions of speed switching

N. gonorrhoeae has been coisolated with obligate anaerobic bacteria and recently it has been shown that it can grow under anaerobic conditions using a truncated denitrification pathway (33,34). When N. gonorrhoeae is grown anaerobically, the level of piliation is reduced because pilC2 expression is downregulated (35). pilC is expressed at a low level even under aerobic conditions and therefore regulation of pilC may be less costly than degrading and expressing major components of the pilus under fluctuating oxygen concentrations. In Neisseria species, gene expression in response to oxygen is regulated by the transcriptional activator FNR, which senses oxygen by direct binding and activates the truncated denitrification pathway when oxygen becomes limited. FNR induces the expression of the nitrite reductase AniA, which reduces nitrite (NO⁻₂) to nitric oxide (NO). Subsequently, the presence of NO activates the nitric oxide reductase NorB, which finally reduces NO to nitrous oxide (N₂O) (30,36). Here, we have shown that gonococci can respond to oxygen depletion considerably faster than by activating the FNR pathway.

What are potential biological functions of speed switching?

One explanation for speed reduction upon oxygen depletion may be that reduction of pilus retraction velocity correlates with a decrease in energy consumption. Thus, the low velocity can be considered a power-saving mode when oxidative phosphorylation is inhibited through oxygen depletion. Furthermore, it is conceivable that velocity switching may be one of the initial steps toward biofilm formation. Twitching motility influences the architecture of gonococcal biofilms (17). It is likely that already during the initial stage of biofilm formation oxygen gradients build up, and that oxygen is depleted faster in the center of the microcolonies, thus speed switching may influence the architecture of biofilms. Another function may be aerotaxis as reported for other bacterial species (37). As we found speed-switching exclusively at very low oxygen concentrations, either gradients at very low concentrations or very steep concentrations would be necessary to test whether speed switching is involved in aerotaxis.

Conclusion

In conclusion, we have shown that the type IV pilus retraction motor has two discrete speed modes whose occupancy can be controlled by the oxygen concentration. Oxygen is essential for ATP synthesis by oxidative phosphorylation. Thus, oxygen depletion corresponds to the loss of an energy source. An advantage of speed reduction upon oxygen depletion may be, therefore, that reduction of pilus retraction speed correlates with a decrease in energy consumption. Thus, we propose that low speed may be considered a power-saving mode. In future studies, it will be interesting to assess the molecular mechanism of speed reduction.