The 3D architecture of a bacterial swarm has implications for antibiotic tolerance

Abstract

Swarming bacteria are an example of a complex, active biological system, where high cell density and super-diffusive cell mobility confer survival advantages to the group as a whole. Previous studies on the dynamics of the swarm have been limited to easily observable regions at the advancing edge of the swarm where cells are restricted to a plane. In this study, using defocused epifluorescence video imaging, we have tracked the motion of fluorescently labeled individuals within the interior of a densely packed three-dimensional (3D) region of a swarm. Our analysis reveals a novel 3D architecture, where bacteria are constrained by inter-particle interactions, sandwiched between two distinct boundary conditions. We find that secreted biosurfactants keep bacteria away from the swarm-air upper boundary, and added antibiotics at the lower swarm-surface boundary lead to their migration away from this boundary. Formation of the antibiotic-avoidance zone is dependent on a functional chemotaxis signaling system, in the absence of which the swarm loses its high tolerance to the antibiotics.

Figure 1

Tracking 3D trajectories of fluorescent bacteria in a multilayered swarm using defocused imaging (DI) methodology. (A) A macroscopic view of a S. marcescens swarming colony inoculated in the center of the agar plate and allowed to swarm out. The region observed for tracking was ~100–400 μm behind the advancing edge (downward pointing arrow). (B) A sketch of the cross section of the colony near the edge, illustrating the approximate shape and dimensions in this region. The height of the colony at the point of observation was estimated to be ~40 μm (see Methods). This height was arbitrarily divided into 15 levels or bins. (C) A snapshot of off-focus fluorescent images viewed from the top. The diameters of the diffraction rings report on bacterial location at different heights in the colony. Arrow points to an example of a relatively large diffraction ring corresponding to location of a cell 35 μm above the agar. There are, on average, 1000 times more unlabeled non-fluorescent cells in any view of the fluorescence image as determined by viewing the same field in phase contrast (see Movie S4). (D) The calibration curve correlating the diameter of the diffraction ring as a function of the depth of the cells, as described in Methods. (E) A cartoon showing how height is estimated from the size of the diffraction ring for bacteria in C. (F) The x-y trajectory of a typical cell moving inside the 3D colony. (G) The height of the same cell plotted in F as a function of time. (H) The 3D trajectory of the same cell plotted in (F,G). Small red arrows indicate instantaneous vectors of the torsion and curvature of the trajectory (see Methods).

Figure 2

Inhabitation, speeds and turning rates of WT and Che- bacteria in a 3D swarm. The swarms contain a mixture of fluorescently-labeled and unlabeled cells. The fluorescent cells were tracked by DI as described in Fig. 1C–H. (A,D) The occupancy of WT (JP1020) and Che^- mutants (JP2529 and JP2531) across 15 levels in a region whose maximal colony height is ~40 μm (see Fig. 1B). PDF stands for probability density function, where the Y-axis is normalized to a sum of all the data points (>200,000). (B,E) The average speed in the x-y plane at different heights for cells tracked in A and D, respectively. (C,F) The average speed along the z-direction at different heights for cells tracked in A and D, respectively. Speeds were calculated using a Matlab program as described under Methods. Error bars indicate standard deviations from the mean. (G,H) PDF of curvatures and torsions for WT (>250,000 data points) and Che^- mutants (>200,000 data points). See Movie S5. (I) The graph plotting the steps between significant curvatures and torsions events for WT yields a power law distribution. See the section on ‘Curvature and torsion of trajectories’ under Methods for a fuller discussion of the results in G-I.

Figure 3

Inhabitation of Srw- bacteria in a 3D swarm. The swarms contain a mixture of fluorescently-labeled and unlabeled cells. The fluorescent cells were tracked by DI as described in Fig. 1C–H. (A) A Srw^- mutant of S. marcescens (RH1041) is distributed uniformly across the height of the colony. In (B), a commercial source of surfactin from B. subtilis was added to the Srw^- mutant, recovering the WT cell occupancy pattern (black trace). (C) A glass cover slip (1 × 1 cm, 100 μm in thickness) was placed gently on the top of the WT colony without disturbing the swirling motion underneath. Cell distribution changed to that seen in the Srw^- strain in A. PDF values represent >100,000 data points in each figure. (D,E) The average speed in the x-y and z planes at different heights for cells tracked in C, respectively, calculated as described in Fig. 2. PDF, probability density function (see Fig. 2 legend).

Figure 4

Inhabitation of WT and Che- bacteria within the swarm in the presence of antibiotics. Cells were inoculated directly on the swarm agar containing antibiotics and tracked by DI as in Fig. 2. (A) In the presence of indicated concentrations of Kan and Cipro, the WT strain showed absence of bacteria in levels 1 to 4 (~8–11 μm). PDF values represent >100,000 data points for each antibiotic. (B,C) The experiment in (A) was repeated with a range of antibiotic concentrations but with a non-fluorescent WT strain, and cells were observed using Live-Dead staining, where live cells show green fluorescence and dead cells show red fluorescence (see Methods). The color chart on the right indicates % of live cells. The MIC (μg/ml) for Kan and Cipro for S. marcescens in liquid vs. swarm was reported to be <5 vs. >20 and <0.025 vs. >0.5, respectively. (D) Cell occupancy of the fluorescent Che^- strain JP2529 was monitored in a swarm with the indicated antibiotics present. The habitation pattern resembled that of WT on antibiotics-free agar (black trace). (E,F) Live-dead staining of non-fluorescent Che^- JP2529 cells under antibiotic conditions as in D. (G) A fluorescent Kan^R strain (JP2703_GFP) shows an inhabitation pattern indistinguishable from that of WT on antibiotics-free agar (black trace). (H) Live-dead staining of non-fluorescent Kan^R JP2703 cells under Kan concentrations identical to that shown in (D).