Results

In order to unambiguously identify individual filaments on a microscope slide over multiple hours, we required the majority of cells to possess on average a single filament. A serendi-pitous discovery was made that a strain deleted for the fliO gene, harboring a PflhD P1 and P4 promoter up mutation[49] and in addition missing the anti- factor FlgM (termed fliO*) preferentially assembled only a single filament (Figure 2.2). The FliO component of the flagellar-specific type-III secretion apparatus is essential for export apparatus function and a fliO strain is non-flagellated under normal export substrate conditions[50, 51]. Ho-wever, it was recently found that the requirement for fliO could be bypassed by mutations in fliP[52]. We found that the fliO* strain retained slight motility in soft-agar plates (Figure 2.2A) and that more than half of the cells of the fliO* strain produced at least one flagellum (Figure 2.2 B and C).
We next introduced a single cysteine amino acid substitution (T237C) in the flagellin fliC into the fliO* strain to allow observation of flagellar filaments by fluorescent microscopy. Residue T237 in the variable loop of the FliC flagellin of S. enterica was chosen for cysteine substitution Figure 2.2 – (A) Enhanced motility of strain TH16123 that is deleted for fliO and that has increased flagellar gene expression resulting from deletion of the negative-regulator FlgM and a promoter-up mutation in the flhDC operon (ΔfliO*). Motility plates were incubated overnight for 18 hours before imaging. The parental strain TH10548 deleted for fliO displays a non-motile phenotype. (B) Fluorescent microscopy analysis revealed the preferential forma-tion of a single flagellum in the fliO* strain TH16123. Exemplary fluorescent microscopy image of the fliO* strain. Flagellin FliC was immunostained as described in Materials and Methods. Membranes were stained using FM-64 and DNA using DAPI. Scale bar 2 µm. (C) Quantification of numbers of flagella per cell of the fliO* strain by anti-FliC immunostai-ning. (D) Graphical visualization of the surface localization of the cysteine-substituted residue T237 (shown in red) using PDB no. 1IO1 of flagellin. (E) Relative swimming motility of strain TH9671 harboring a T237C substitution in the flagellin FliC compared to the wildtype control TH6232. Swimming motility was assayed using soft-agar swimming plates containing 0.3 % agar.
(Figure 2.2D). As shown in Figure 2.2E, the motility of otherwise wildtype Salmonella cells harboring the fliC(T237C) mutation was approximately 64% of the wildtype fliC allele.
As described in the Materials and Methods section, cells of the fliO* fliC(T237C) strain were immobilized in a custom-made flow chamber and labeled with Alexa-Fluor maleimide 546 dye after incubation. Only cells that were firmly attached to the coverslip and that had one single flagellum were selected for shearing of the filament by laser ablation. In order to ensure that the observed cell was alive and healthy, we considered only the rotating filaments. In addition, we selected filaments that were not only rotating on their axis, but also slowly gyrating (i.e. the filament axis itself was rotating around slowly). Indeed, initial trials showed that if the filament was not gyrating, the laser pulses cutting the filament often stopped the rotation of the motor. It is not exactly clear why that was the case, but presumably non gyrating filaments have a much stronger tendency to stick to the cell body or the coverslip upon laser ablation. The ideal candidate was therefore a rotating filament that was also gyrating in a somewhat uniform circular trajectory. The laser beam was then positioned in the vicinity of the bacterium so that its filament would cut itself on the train of ultrafast laser pulses (at 250 kHz repetition rate). A successful cutting operation was clearly identified by the acceleration of the filament gyration, and the cut portion was often seen diffusing away. To be sure that the laser did not damage the flagellar motor or compromised the cell membrane, we made sure that the filament was still rotating after ablation. Figure 2.3 shows the same bacterium before (panel A) and after (panel B) its filament was cut. The length of the cut filament was reduced from ∼3.5 µm to ∼2 µm.
A total of 82 individual bacterial filaments were cut using the femtosecond laser and observed after a two-hour incubation period. We never observed any filament re-growth on any of the filaments that had been cut. Statistically this observation allows us to conclude (with 95% confidence) that the proportion of filaments that can regrow after being cut by the laser is less than 4% (by the “rule of three”, 3/n = 3/82= 4%)[53]. Table 2.1 breaks down the number of observed filaments by strain and by their rotation status when we revisited them after the incubation period. The fact that a filament was still rotating after incubation demonstrates that this particular bacterium was alive and healthy (and therefore potentially able to re-grow filaments). However, if a filament was stopped, it was most likely because it simply stuck to the poly-L-lysine-coated coverslip surface. We considered it valid data because dead bacteria could easily be identified due to their cell bodies filling up with fluorophore. As a control, we observed many filaments that were left intact on the same coverslip. As shown in Figure 2.3C, the portion of the filament that grew during incubation is clearly visible as a green extension at the end of the orange filament (when images in both channels are combined digitally). To acquire such image, the filament’s rotation had to be stopped either by exposing the bacterium to a large amount of blue-green light, or by punching a hole in the cell body with the laser. Overall, between 90% and 95% of the uncut filaments that were still turning Figure 2.3 – Flagellar filament of strain EM800 (A) before and (B) after being cut by an ultrafast laser beam. The cell body is barely visible (highlighted with white dotted line) and the filament shows up large and fuzzy because it is rotating much faster than the image acquisition rate. The white arrow points to the cut portion of the filament drifting away and out of focus. Scale bars are 2 µm. (C) Control cells of strain EM800 whose filaments were left intact. The filaments were first labeled with an orange fluorophore and then, after a 2-hour incubation at 37°C in TB, labeled again with a green fluorophore. The portions of filaments that grew during incubation are clearly distinguishable. (D) Example of a bacterium (EM800) that grew a new flagellum during incubation. The top arrow points to the new filament that grew after the first labeling (in the 2 hour incubation). The filament is blurry since it was rotating during the exposition. The bottom arrow points to the cut filament (orange) that did not regrow. The continued rotation of the flagellar filament demonstrates that the cell was still alive and potentially able to re-synthesize a new filament. after the incubation period showed a green “regrowth” portion.
On a few occasions, we observed cells that grew a second filament during the incubation period. Figure 2.3D shows a cell on which the new filament (green and fuzzy due to rotation) is seen besides the old cut filament (orange) that clearly did not regrow. Such cases with “built-in” control further support our conclusion that filaments do not regrow after being cut by laser ablation.
Table 2.1 – Number of filaments that were cut and observed after a 2-hour incubation for each strain used. The rotation status of the filaments when we revisited them is also detailed. None of these 82 filaments continued to grow after being cut.
The fliO* strain EM800 harbored a deletion of the anti- factor FlgM that ensures constant 28-dependent gene expression from Class 3 promoters. Accordingly, both flagellin subunits and the filament cap FliD should be available for filament regrowth. In fact, as shown in Figure 2.3D, the cells were able to grow a new flagellum when the first filament was damaged by laser ablation. However, to provide an excess of cap protein, we additionally performed the laser shearing experiments with a strain that overexpressed fliD from an inducible arabi-nose promoter (EM1283, see Figure 2.5 in Supplementary Material). This enabled us to test the possibility that excess FliD in the cytoplasm could accelerate the formation of a new cap structure, and thereby allow filament growth after damage. We performed the same laser abla-tion manipulations with strain EM1283 and, as shown in Table 2.1, none of the 18 filaments that were cut grew back (44% were still turning after incubation), while undamaged two-color filaments were frequently observed. Supplementary Figure 2.5A shows complementation of a fliD strain by overexpressed fliD in a motility plate assay. The same arabinose-inducible fliD construct did not affect motility of the fliO* strain EM1283.

Shearing

A surprising aspect of the results described above is that they differ from the conclusions of Turner et al. who found that mechanically sheared E. coli filaments can re-grow[48]. The principal difference between the experiments described in that paper and the ones reported here is the method used for breaking the filaments : mechanical shearing (with small syringe needle) in [48] versus ultrafast laser ablation here. That aspect will be discussed in detail below, but to test whether the different results could simply be explained by the different bacterial species (e.g. Salmonella versus E. coli), we mechanically sheared the filaments of the Salmonella strain used here. To this end, we constructed strain EM2046 that harbored the flagellar master operon flhDC under control of the tetracycline-inducible P promoter[44]. Induction of flagellar synthesis by addition of tetracycline allowed us to synchronize production of basal bodies within the population of strain EM2046. In addition, this strain allowed us to stop the production of new basal bodies before the end of the incubation period by removal of the inducer tetracycline and this ensured that most filaments (∼90%) were lon-ger than 2 µm before shearing. As described in the Methods section, the filaments of strain EM2046 were sheared using a 22-gauge needle. After comparing the distributions of filament lengths with and without shearing (shown in Figure 2.4), we concluded that the shearing is effective at shortening filaments (4.0 ± 1.6 µm (s.d.) on average for normal population vs 2.0 ± 1.1 µm on average for sheared population). The distribution of filaments that failed to produce a green segment (i.e. did not regrow) is shown by the white bars Figure 2.4B. To answer the question whether the mechanically sheared filaments regrow, filaments below 2 µm length were examined in detail.
The drawback of the two-color labeling approach after mechanical shearing is that we can only make statistical arguments. Indeed we can never be sure that a given filament observed to be two-colored (i.e., did grow after the shearing) was in fact broken by the initial shearing process.
Figure 2.4 – Results of the shearing experiment. The two top panels (A) show the data without mechanical shearing between the two labeling of EM2046 bacteria and the two bottom panels show the results with mechanical shearing. The left panels are sample images of two-color fluorescent labeling of EM2046 bacteria (scale bar is 5 µm). The length of the first portion of the filament (near the cell body, labeled in orange) is generally shorter when the bacteria are sheared (in B). This is easily observable on the right panels, which show histograms of the filament length measurements in the two situations (200 and 239 filaments in total in the non-sheared and sheared population respectively). One also notices that most orange filaments have a green tip, which indicates that the filament grew back after shearing. The white bars (in front of the orange bars on the bottom right panel) indicate the number of filaments that did not regrow after the first labeling.
However, the use of the EM2046 strain increased the odds as it allowed us to synchronize filament assembly. As displayed in Figure 2.4A, only 8% of the filaments (16 out of a total of 200) are less than 2 µm long when the culture was not sheared, as opposed to 49% (116/239) after shearing. Therefore, we concluded that the shearing process was efficient. In the worst case scenario, if no filament below 2 µm could ever be broken by mechanical shearing, we would expect a maximum of about 8% of the sheared filaments to be intact and therefore continue to grow. Among the 116 short filaments that had been sheared, 22 did not show a green portion, which leaves 94 filaments that continued to grow (39% of the total number). That 39% is significantly higher than 8% (p<0.0001, or the difference is 31 ± 8% with 95% confidence – see details in Supplementary section 2.8.1), and we conclude that mechanically sheared filaments from Salmonella are able to re-grow, in contrast to the filaments broken by mechanical laser ablation. The observed filament growth after shearing also demonstrated that cells were alive and healthy after the mechanical shearing process.

Discussion

This work studied whether it is possible for a flagellar filament of S. enterica to continue to grow after being damaged by laser shearing. The combination of femtosecond laser ablation and the use of a single-filament bacterial strain enabled us to achieve the technical challenge of inducing specific damage to identified filaments and revisit each one individually after an incubation period. Our conclusion is unequivocal : the bacterial filaments do not re-grow under these conditions. In contrast, we observed re-growth of filaments after applying breaking forces by viscous shearing.