M.
Fauvart
a,
P.
Phillips
b,
D.
Bachaspatimayum
a,
N.
Verstraeten
a,
J.
Fransaer
c,
J.
Michiels
a and
J.
Vermant
*b
aCentre of Microbial and Plant Genetics, Katholieke Universiteit Leuven, B-3001, Leuven, Belgium
bDepartment of Chemical Engineering, Katholieke Universiteit Leuven, B-3001, Leuven, Belgium. E-mail: jan.vermant@cit.kuleuven.be
cDepartment of Metallurgy and Materials Engineering, Katholieke Universiteit Leuven, B-3001, Leuven, Belgium
First published on 20th October 2011
Bacterial swarming is one of the most efficient methods by which bacteria colonize nutrient-rich environments and host tissues. Several mechanisms have been proposed to explain the phenomenon and the associated intricate macroscopic pattern formation, but so far no conclusive evidence has been presented that identifies the factors that control swarming. Vice versa, little is known about how swarming can be controlled. Here, by using a series of complementary genetic and physicochemical experiments and a simple mathematical analysis, we show how the bacterial swarming can be caused by a surface tension driven flow. The opportunistic pathogen, Pseudomonas aeruginosa, is studied, as it is relevant for such bacteria to control and arrest swarming. Moreover, P. aeruginosa bacteria secrete strong surface active components as part of their quorum sensing system. Our results demonstrate that surface tension gradient control can even be the dominant mechanism that drives swarming. It can be quantitatively predicted and can be expected to play a role in a wide variety of bacterial systems. The modeling reveals subtle dependencies on both the wetting conditions and the physical properties of the slime. Based on these dependencies, strategies can be devised to arrest swarming under certain conditions by simple physicochemical means.
Three main mechanisms have been put forward to explain swarming. Some authors point to the role of the quorum sensing system, which is suggested to trigger a change of the cells from an undifferentiated vegetative state to multi-nucleated and hyper-flagellated swarmer cells, which form rafts of bacteria as their flagellar filaments become entangled.7,8 Swarming is seen as an example of dynamic self-assembly in microbiology.9 A repeated switching is suggested to give rise to the appearance of typical bulls eye patterns, for example in Proteus mirabilis.10 However, the types of patterns formed by colonies living on nutrient rich agar plates differ strongly, ranging from fractal-like in Serratia marcescens11 and Rhizobium etli12 to striking dendritic patterns in e.g. P. aeruginosa.13 To rationalize the rich variety of patterns formed by swarming colonies and the sensitivity of the pattern formation to the conditions in the substrate, the second explanation has been that swarming can be described by models which couple nutrient diffusion to bacterial density. Reaction-diffusion models, and extensions taking into account phenomena such as chemotaxis and lubrication fluid layers,14 are highly non-linear and can produce a wide range of morphologies. However, the modeling provides limited insight into the biological and physical mechanisms underlying swarming dynamics. So far, there have been no attempts to rationalize, predict or control the spreading velocities.
Conceptually adhering to a continuum modeling approach, the role of flows driven by surface tension, which are known as Marangoni flows, has been proposed as a third explanation. For the bacterium R. etli both the pattern formation and the spreading speeds were observed to be consistent with those expected for a Marangoni flow for surface tension gradients, film thicknesses and viscosities observed experimentally.12N-acylhomoserine lactone (AHL) molecules, the signaling molecules of quorum sensing in most Gram negative bacteria,15 were identified mainly by genetic knock out experiments as playing a dominant role in the swarming. This class of AHL molecules was proven to be surface active for biologically relevant concentrations and surface tension gradients can develop and cause the resulting Marangoni flows.12 More recently, experiments on Bacillus subtilis confirmed the generic character of the role of Marangoni flows in this specific case for biofilm dynamics. B. subtilis produces the molecule surfactin, a strong biosurfactant. Model experiments were used to show how the differential accumulation rates induced by the geometry of the bacterial film give rise to surfactant waves causing bacterial biofilms to climb up a wall.16 Within the continuum description, the bacteria act as little factories of surfactant production, and their spatial distribution or differentiation can lead to gradients of concentration. Although the role of these surface tension gradients has not yet received much attention, it is widely recognized that surface active components play an important role in motility (see e.g. Wilking et al. for a recent review17).
In the present work, we demonstrate that surfactant concentration gradient driven flows are a factor to be reckoned with and can even be the dominant mechanism controlling swarming of P. aeruginosa colonies. The quorum sensing system controls the production of rhamnolipids and swarming.13,18 Swarming is also accompanied by striking dendritic patterns. In this respect it is important that rhamnolipids are known to be strong biosurfactants and the present bacterial model system was selected and placed in conditions where the effects of surfactant production are maximized.19 Using a combination of genetic knock out experiments and green fluorescent protein (GFP) reporters for both the cell density and rhamnolipid production, the effect of surfactant production and quorum sensing on the swarming patterns and spreading velocities are investigated. The presence of surfactant waves under swarming conditions is shown using white light interferometry, in situ and for different instants during the swarming process. A simple mathematical model is used to explain the height increase and predict the spreading velocity. The physicochemical parameters that influence the phenomena are determined experimentally and can be used to predict or control the spreading velocities, even in such a way that swarming can be brought to a stop using simple counter-gradients of surfactant concentrations.
Timelapse movies were recorded using a Nikon D80 camera (Nikon, NY, USA). Fluorescent images were obtained using a LT-9500 imaging system (Lightools Research, CA, USA) or an Olympus BX51W equipped with a Hamamatsu C8800 CCD. Image analysis was carried out using ImageJ (National Institutes of Health, MD, USA). White light interferometry was performed using a Wyko NT3300 interferometer (Veeco, AZ, USA). Samples were prepared by gel trapping, as described previously,22 to lock in the height profile at a give moment in time such that high resolution height profiles could be measured.
Rheological experiments were carried out on a rotational rheometer (MCR501 Paar Physica), equipped with cone and plate geometries of varying cone angles. Colonies were harvested from the agar plates, the bacteria were killed using UV and the samples were measured immediately to yield an effective viscosity of the colony and its medium. A double solvent lock was used and a Peltier plate and hood were used to ensure constant temperature. At low stresses significant wall slip was present, whereas at stresses above 0.2 Pa the results became independent of measurement geometry. Only data in this regime will be reported. The surface pressure measurements were carried out using a Wilhelmy plate and an electromagnetic balance and a Langmuir trough (from KSV instruments, Finland).
Fig. 1 Swarming colonys edges are characterized by a height increase. A–H: Visible light images of a swarming colony with signal intensity plots below; images are in chronological order (0.5 h, 2.5 h, 5 h, 9 h, 12.5 h, 16 h, 18.5 h). Left half: wild-type strain P. aeruginosa PA14, right half: rhlA mutant deficient in rhamnolipid production. Gray rectangles indicate regions for which signal intensity was calculated after background subtraction. White square boxes in panels D and H indicate regions for which white light interferometry data is shown in panels K–N. See also supplementary movie 1.† I–L: height profiles of swarming colony recorded with white light interferometry; sampling time points correspond to panels A, B, D and H, respectively. M–N: 3D images generated from white light interferometry data shown in panels K and L, respectively. |
To prove more clearly that the higher intensity of the microscopy images near the edges of the colony is caused by an increase in height, the thickness of the bacterial films has been measured in situ using white light interferometry. Profiles I–L give the measured height profiles for the wild-type strain P. aeruginosaPA14 for the panels A, B, D and H, respectively. Whereas the initial spot inoculation and drying leads to a thin film of a few micrometres thick (panel I), the height of the colony gradually increases and a pronounced rim of roughly twice the height of the colony develops as is shown in Fig. 1J for a full cross section. A more detailed graph for half the cross section is given in panel K (corresponding to a situation as in panel D). Panel K was taken just before the branching instability takes place. A 3D height profile generated from white light interferometry data shown in panels K and L corresponds to the situation right at the onset of branching and within a tendril, respectively. The wave front near the edges of the colony is also present within a branch as is shown in the profile in Fig. 1 K and even more clearly in the 3D reconstructions. It should be pointed out that the mutant which does not produce the biosurfactant shows no increase of the height. Image analysis of the spreading colonies of the wild-type strain P. aeruginosa PA14 was used to measure the spreading velocities. The initial spreading velocities of the circular shaped colonies were on the order of 5–10 μm min−1. In the branches the spreading speeds up to values of 40 μm min−1. In the valleys between the branches, the spreading velocities of the colonies slow down. Similar to the case of R. etli,12 we could not detect a correlation between the velocities of the individual bacteria moving in the slime and the overall spreading velocity. More intricate local scale mechanisms such as coordinated flagellar motion25 were also not observed.
To identify the role of the bacteria in driving these collective mechanisms, we performed experiments with modified P. aeruginosa strains, allowing to detect the cell density or to provide information on the rhamnolipid production. Fig. 2A shows bright field images of a swarming colony. The bacteria in this experiment were genetically modified to produce a green fluorescent protein (GFP) and the corresponding fluorescence image is shown in Fig. 2B. The fluorescence is present throughout the colony and is most intense in the dendritic arms. It can be concluded that the fluorescence intensity mimics the intensity observed in the bright field images, and hence the difference in light (2A) and fluorescence (2B) intensity can be traced back to the differences in film thickness shown in Fig. 1. Hence it can be concluded that the bacteria are distributed uniformly throughout the colony. For a colony of bacteria expressing rhlA-gfp fusion to monitor rhamnolipid production, the fluorescence microscopy image in Fig. 2C also shows a fluorescence throughout the colony with higher intensity in the arms, clarifying that rhamnolipids are being produced everywhere in the colony. Fig. 2D shows a detailed image of the edge of the colony just before the onset of swarming. Although no direct quantitative comparison can be made it can be concluded that there are only limited or localized spatio-temporal variations in production of rhamnolipids. A recent microarray study has shown that rhlA production is down regulated in the dendrite tips, compared to the swarm center.26 This down-regulation, or the mere fact that there is a region of lower bacterial density near the very edges of the colony (where the bacteria cannot penetrate the film) can lead to local gradients in rhamnolipid concentration.
Fig. 2 Swarming by Pseudomonas aeruginosa. A: swarmed state, visible light, B: swarmed state, GFP filter shows autofluorescence = cell density, C: swarmed state, GFP filter shows PrhlA activity = rhamnolipid production, D: zoom of swarmed state, GFP filter shows PrhlA activity = rhamnolipid production. |
In order to identify the possible role of nutrient diffusion in the subphase to bacterial density, which is a key element in some of the reaction diffusion approaches, we investigated if the colonies could swarm over a barrier of about 1 mm thickness. The setup is shown in Fig. 3A. A petri dish with an insert was filled with agar in such a manner that the insert was just covered with a very thin layer of agar. A planar interface was then obtained. Spot inoculation was carried out only on one side of the barrier. If nutrient diffusion would be an important issue, the barriers would have been ‘felt’ by the swarming colony and the swarming should have stopped. However, both 3B and C show that in the final swarmed state, the colony has swarmed over a barrier. In the experiments reported in Fig. 3C a dye molecule was added to the right hand side to demonstrate that there is no significant diffusion over the barrier. These experiments suggest that nutrient diffusion does not play a critical role in the swarming of P. aeruginosa.
Fig. 3 Swarming over a barrier. A: schematic cartoon of the dishes with an insert in the agar (side view). The colony was spot inoculated on the right. B: swarmed state, bright field image showing dendrils going over the barrier. C: swarmed state, the right hand element was colored with a food dye. |
(1) |
Fig. 4 Physical characteristics of the biofilm and calculated height profiles. A: viscosity versus shear rate of the biofilm and (insert) surface pressure of the rhamnolipids versus the surface concentration (Cs in mol cm−2) for 2 consecutive compression expansion cycles. B: renormalized height profile of a spreading film. |
The viscosity of the bacterial colony has been measured and reveals a non-Newtonian behavior with a strong shear thinning, i.e. the viscosity decreases as the deformation gradient or stress is increased, as is shown in Fig. 4A. In the range of deformation gradients (0.1–1 s−1 given by the range of U/H) that are relevant for the spreading, the viscosity is of the order of 0.1 to 1 Pa s. As the orders of magnitude of all parameters of eqn (1) are known, the spreading velocity expected for a Marangoni driven flow is expected to be 0.5 to 5 μm s−1. The experimentally observed velocities before the patterns become unstable agree very well with the order of magnitude predicted from this simple analysis.
A slightly more elaborate analysis, still using a simplified continuum view of the bacterial colony, can be used to evaluate not only the velocities, but also the characteristic height increase near the edges. As the liquid film is thin (ε = H0/L0 ≪ 1), the lubrication approximation can be used, and only a 1D Stokes flow equation needs to be solved.27 The concentration gradients that drive the flow appear in the boundary conditions of the flow equations. The model is based on the Marangoni driven spreading of a monolayer of surfactant on the surface of a Newtonian liquid layer.28 Writing down the momentum and mass balances, a set of coupled PDE is obtained, to predict the spatial temporal variables, the height of liquid layer, H(x,t), and the surfactant concentration, Γ(x,t). The resulting equations are
(2) |
(3) |
These equations reflect the balance between the Marangoni stresses that drive the motion and the viscous terms and the capillary stresses that oppose the creation of curvature, as well as a surface diffusive term, with the surface diffusivity Ds. To complete the formulation of the problem suitable initial conditions for both H(x,t) and Γ(x,t) have been selected. For H(x,0) a pancake like profile was assumed, which mimics the height profile that was measured experimentally using white light interferometry at the moment the colony starts to swarm and moves radially outwards. A similar profile was chosen for the concentration profile at the start of the calculations, based on the observations in the GFP experiments that the surfactant concentration follows the bacterial density which is assumed to be more or less uniform in the bulk but lower near the edges17 and the observations in literature that the expression of biosurfactant may be down regulated near the edges of the colony.26 The set of PDEs is rendered non-dimensional and solved numerically using a finite element method for the spatial discretisation. The resulting non-dimensional groups are the Péclet number (Pe = UL0/Ds), characterizing the relative importance of convection over diffusion. For the sizes and molecular weight of the rhamnolipids, the surface diffusion was estimated and Pe was estimated to be of the order of 5000. The capillary number (Ca = ηU/σ) can be estimated to be of order one, but a higher value (Ca = 500) was used to speed up the computations (hence the time scales are faster compared to the experimental time frame). The solution of the problem was found to be very sensitive to the boundary condition used near the edge of the colony. To mimic the experimental conditions as much as possible, we allow slip at certain length below the boundary, denoted as l, the slip length which was taken to be of order 0.05. We assume this to be a realistic boundary condition as there is water everywhere on the agar film. The boundary condition becomes:
(4) |
and
u(x,−l,t) = 0 | (5) |
Fig. 4B shows the characteristic features for the initial Marangoni driven spreading for conditions which are assumed to be relevant for the present case. The height and length were normalized relative to the initial stage, the time scale will be faster than in the experiments due to choice of the dimensionless parameters. The calculations serve to demonstrate how some of the experimentally observed features during the initial stages of spreading are effectively reproduced. Specifically, a characteristic ‘bulge’ develops near the edge of the droplet, which is reminiscent of the features observed during the white light interferometry experiments in Fig. 2, especially in Fig. 2M. Such height increases have also been observed in experiments and calculations of the spreading of surfactant droplets atop of viscous liquids,28,29 and indirectly through bright field microscopy images for a bacterial biofilm of B. subtilis climbing up an inclined slope.16 The occurrence of the bulge near the edge of the colonies, both during the initial stages as within the fingers (see Fig. 1M and 1N and Fig. 4b), can be viewed as a signature of the start of Marangoni driven flow. Very similar to the formation of ‘tears of wine’,30 the process of Marangoni driven flows will speed up as a fingering instability develops along the front of the film. The new area which is being created can be expected to have a lower surface concentration of the surfactant, hence a higher surface tension. As a consequence, the liquid is pulled out into the branches and this leads a self-sustained propagation.
Although the pattern formation is not discussed in detail here, it can be pointed out that like the reaction-diffusion models, the set of equations describing the Marangoni flows is highly nonlinear and can be expected to give rise to a rich variety of morphologies. The details will depend on the link between the various physical parameters that control the Marangoni flow (concentration gradients, surface tension, surface diffusion, dynamic surface tension, viscosity) and the bacterial characteristics, which enter the picture predominantly as producers of biosurfactants. How the motility of the individual bacteria intervenes is as yet unclear, but the spatial heterogeneity of surfactant concentration may be affected by differences in motility. Further work with motility mutants is clearly required to clarify this link.
To see if the present model can also predict the patterns which arise when the flow becomes unstable requires a full linear and non-linear stability analysis. Whereas a full stability analysis of eqn (2)–(3) lies beyond the scope of the present work, some important remarks can already be made. First, the mathematical structure of the height-concentration evolution equations is very similar to the reaction diffusion models, and hence a similar richness in possible patterns can be expected. Second, the nonlinearity of the equations will be enhanced even when the shear thinning of the viscosity is taken into account. Finally, the number of arms observed in the dendritic structures is in line with predictions for the dominant wavelength using a linear stability analysis,31 although the variability in the experiments is considerable.
Fig. 5 Inhibition of swarming. A–D: swarming can be inhibited by a exogeneously added biosurfactant. A: control; B–D: increasing concentrations of rhamnolipids were added in a circular pattern (concentration relative to the in vivo one: B:1/100, C:1/10; D:1/1). |
Footnote |
† Electronic supplementary information (ESI) available: suppmovie1a.mov and suppmovieRL.mov. See DOI: 10.1039/c1sm06002c |
This journal is © The Royal Society of Chemistry 2012 |