Hydration-Controlled Bacterial Motility and Dispersal on Surfaces

Hydration-controlled bacterial motility and dispersal on surfaces

Arnaud Dechesnea, Gang Wangb, Gamze Güleza, Dani Orb, and Barth F. Smetsa,1

aDepartment of Environmental Engineering, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark; and bDepartment of Environmental Sciences, Swiss Federal Institute of Technology, 8092 Zurich, Switzerland

Communicated by James M. Tiedje, Center for Microbial Ecology, East Lansing, MI, June 14, 2010 (received for review October 5, 2009) Flagellar motility, a mode of active motion shared by many pro- from direct observation of bacterial dispersal at both individual karyotic species, is recognized as a key mechanism enabling pop- and population scales under conditions of controlled hydration, ulation dispersal and resource acquisition in microbial communities we used the porous-surface model (PSM) (12). In this experi- living in marine, freshwater, and other liquid-replete habitats. By mental system, bacteria are grown on a porous ceramic surface in contrast, its role in variably hydrated habitats, where water dynamics thin aqueous films, whose effective thickness is controlled by result in fragmented aquatic habitats connected by micrometric films, applying a prescribed suction similar to how matric potential is debated. Here, we quantify the spatial dynamics of Pseudomonas controls hydration in soils. The system allowed a quantitative putida KT2440 and its nonflagellated isogenic mutant as affected assessment of the dispersal rate and competition of Pseudomo- by the hydration status of a rough porous surface using an experi- nas putida KT2240 and a nonflagellated ΔfliM isogenic mutant as mental system that mimics aquatic habitats found in unsaturated influenced by water potential at the colony and individual scales. soils. The flagellar motility of the model soil bacterium decreased sharply within a small range of water potential (0 to −2 kPa) and nearly Results and Discussion ceased in liquid films of effective thickness smaller than 1.5 μm. How- At the colony scale, we observed constant front expansion rates, in ever, bacteria could rapidly resume motility in response to periodic accordance with the model proposed by Skellam (13) for a pop- increases in hydration. We propose a biophysical model that captures ulation dispersing by random walk and exponential growth. The key effects of hydration and liquid-film thickness on individual cell average rate of colony expansion for both the wild type and non- velocity and use a simple roughness network model to simulate colony flagellated mutant decreased with decreasing matric potential MICROBIOLOGY expansion. Model predictions match experimental results reasonably (Fig. 1). The most significant differences in expansion rates be- well, highlighting the role of viscous and capillary pinning forces in tween the strains were apparent at −0.5 and −1.2 kPa, where the hindering flagellar motility. Although flagellar motility seems to be wild type, capable of flagellar motility, dispersed more than 15 restricted to a narrow range of very wet conditions, fitness gains con- times faster than the mutant, which dispersed by cell shoving and ferred by fast surface colonization during transient favorable periods Brownian motion only. This clearly shows the potential of flagellar might offset the costs associated with flagella synthesis and explain motility for fast population dispersal on wet surfaces. A role of the sustained presence of flagellated prokaryotes in partially satu- swarming motility, which, in P. putida KT2440, relies on short pili rated habitats such as soil surfaces. rather than flagella, is unlikely in our experiments, because it is expressed only under specific conditions (14) and manifests itself flagella | biophysics | liquid film | fitness | motility by en masse cell movements, which we did not observe. The colony-expansion rate of the wild type decreased exponen- ispersal is recognized as a key ecological process enabling tially from an average of 521 μm/h for the wettest conditions (−0.5 Dpopulations’ access to new sites and pools of resources (1), kPa) to 14 μm/h at −2.0 kPa (Fig. 1). After this sharp decrease, thereby affecting structure and productivity of ecosystems (2, 3). the colonization rate leveled, suggesting that, on the PSM, −2.0 Active bacterial motion (motility) takes on many forms that require kPa marks a threshold, below which the contribution of flagellar various appendages (4). If surface-associated modes of motility motility to population dispersal becomes insignificant. Under drier such as twitching, gliding, or swarming seem restricted to some conditions, we expect both types of bacteria to disperse by cell species (5), the ability to swim by rotating one or more flagella is shoving only and thus, their colonies to expand at similar rates. The shared by a large diversity of prokaryotes. This swimming motility slightly faster expansion of the wild-type colonies observed at −3.6 has attracted considerable attention, primarily aimed at resolving kPa (Fig. 1) is attributed to the higher intrinsic-growth rate of the biophysical functioning of flagella and to a lesser degree, ex- this organism. Indeed, although the only difference between mu- ploring its adaptive value. In marine environments, a large fraction tant and wild type resides in the fliM knock-out, the former presents of bacterial populations are flagellated (6), and swimming motility asignificantly reduced growth rate. This was evidenced in compe- is often coupled with chemotaxis, conferring a clear benefit to these tition experiments in stirred liquid medium where the fitness of cells by allowing them to outswim diffusion and exploit transient the mutant relative to that of the wild type was significantly smaller substrate gradients (7, 8). than 1 (0.82, SD = 0.04, n =3,P = 0.016, two-tailed t test). In contrast to water-replete environments where flagellar mo- The mild matric potential that we prescribed does not per se tility is essentially unrestricted, there exists strong physical limi- restrict bacterial motion (12), but it acts through its control of tations to flagellar motility in partially saturated media where the effective liquid-film thickness on the ceramic-plate surface. The aquatic microhabitats are often fragmented and connected only by thin liquid films of bacterial size or smaller (9). The limitations to bacterial motility in thin liquid films have, thus, long been posited Author contributions: A.D., D.O., and B.F.S. designed research; A.D., G.W., and G.G. per- but never directly quantified or described biophysically beyond the formed research; A.D. made and analyzed single-cell observations; G.W. implemented the fl motility model, ran the simulations, and analyzed the results; G.G. made and analyzed general notion that agellar motility requires hydrated pathways. population-scale observations and competition experiments; and A.D., D.O., and B.F.S. In addition, the fitness benefit associated with flagellar motility in wrote the paper. partially saturated soils has been debated because of conflicting The authors declare no conflict of interest. experimental data (10, 11). 1To whom correspondence should be addressed. E-mail: [email protected]. Here, to avoid the complexity inherent to natural partially This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. saturated microbial habitats such as soil matrixes and benefit 1073/pnas.1008392107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1008392107 PNAS Early Edition | 1of4 Downloaded by guest on September 27, 2021 onset of normal capillary pressure, and introduction of a capillary pinning force (FC) (19). We combine hydration-dependent resistive forces into a simple model where attainable bacterial cell velocity is proportional to the residual force available for flagellar FM − Fλ − FC V ¼ V V FM − Fλ − FC < propulsion: 0 FM , with = 0 for 0. This model component provides estimates of the potential cell velocity for local hydration conditions (which may vary spatially over a natural surface) that may then be directed by local chemotactic gradient and a random component (tumble-like) to define the actual direction and extent of displacement during a single run (or time step). These modifications are implemented Fig. 1. Matric potential affects colony-expansion rates of P. putida KT2440 wild type (flagellated) and its ΔfliM isogenic mutant (nonflagellated) mea- by combining the random component (tumble-like change of run sured as radial expansion of colonies initiated from single cells. More neg- direction) and local chemotactic gradients by weighing these by ative matric potentials correspond to thinner surface liquid films. Error bars a factor of (1 − ξ)andξ, respectively, depending on normalized mark 1 SD (n varies from 6 to 14). Simulated colony-expansion rates are local chemoattractant gradient, ξ (defined as the ratio of local to depicted by the line and shaded area (representing 1 SD). maximal chemotactic gradient and calculated as the concentration difference across the local bond divided by the boundary concentration). The cellular motion is evaluated every 1.1 s (an fi relationship between matric potential and liquid- lm thickness assumed duration of a run and tumble cycle) (20), and the actual depends solely on surface wettability and roughness (15). At −2.0 * * * * run velocity is obtained as: V ¼ðRð1 − ξÞþNξÞV,withR de- kPa, which we identified as the limit beyond which the contribution * of flagellar motility to colony expansion is negligible on the ceramic scribing a random direction of cell motion and Ndescribing the plates, the predicted effective liquid film is thinner than 1.5 μm (12). displacement component along the chemoattractant gradient. To explain the strong reduction of colony-expansion rate ob- To validate the proposed model, we experimentally quantified served over a relatively small range of matric potentials and pro- individual cell trajectories at the surface of the PSM for different vide a predictive tool for bacterial-dispersal rates on partially matric potential values (Movies S1 and S2 show typical exam- saturated rough surfaces, a simple mechanistic model is proposed ples). The descriptors of individual trajectories varied consider- linking bacterial velocity to viscous and capillary forces acting on ably because of the specific microtopography experienced by motile cells. The key biophysical elements of the proposed model each cell; however, the average velocity and dispersal distance are summarized in Fig. 2. The model considers the effects of liquid- clearly decreased with decreasing matric potential (Fig. 3). Sin- film thinning on the propulsive force, FM ¼ 6πRηV0 (16), for fla- gle-cell motility was simulated by applying the model presented gellar motion at maximum velocity in bulk liquid, V0. On partially above to cells swimming in an idealized 2D rough surface con- saturated surfaces, hydrodynamic interactions between bacterial sisting of a network of angular channels of various geometries cells and solid surface hinder motion, preventing cells from that accommodate variable liquid configuration (21). The simu- attaining their maximum velocity. Considering, for simplicity, an lated descriptors of cell motility are in good agreement with the average 45° angle between the solid surface and cell trajectory, we experimental results, even without specific parameter fitting or obtainqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the following hydrodynamic interactions coefficient: other adjustments (Fig. 3). These results highlight the dominant λ ¼ λ2 þ λ2 , given in terms of cell-surface interactions for mo- role of capillary pinning forces in constraining bacterial motility. P N Subsequently, the cell-motility model was implemented to λ λ tion parallel, P (17), and normal, N (18), to surface, respectively. simulate population dispersal and reproduce experimental-col- These interactions affect bacterial cell velocity according to V ony expansion rates. Cell growth and motility were modeled at V ¼ 0 λ λ and are associated with a corresponding resistive force: the individual cell scale (22), and chemotaxis was included by 1 Fλ ¼ð1 − λÞFM . The most significant hindering force on partially biasing cell movement to neighboring channels with high nutri- hydrated surfaces sets in when the liquid film becomes thinner than ent content. An example of colony expansion is shown in Fig. S1. the cell diameter, resulting in interactions with liquid–air inter- The model predictions were consistent with the experimental faces, including formation of a contact line on the cell surface, dispersal rates (Fig. 1), showing that population dispersal cor-

Fig. 2. Cell velocity within idealized surface-roughness elements under different hydration conditions. (Left) A roughness element can be abstracted as a channel of triangular section, with depth H and spanning angle α. R is the cell radius, and r(μ) is the radius of curvature of the liquid-air meniscus determined by the ambient matric potential, μ. Depending on the channel geometry, cells can either be fully or partly immersed for a given matric potential. The forces exerted on swimming cells are different in these two situations, as depicted on Right. The cell velocity is modeled for cells swimming in two channels of the

same depth (H = 100 μm) but contrasting spanning angles (α = 120° or 150°). The maximum average velocity V0 was ﬁxed to 18 μm/s, a value typically observed in saturated systems (32). F0, Fλ, and FC are the viscous drag force opposing motion in bulk liquid, the viscous force associated with cell-wall interactions, and the capillary pinning force, respectively. The horizontal dashed line marks the onset of capillarity.

2of4 | www.pnas.org/cgi/doi/10.1073/pnas.1008392107 Dechesne et al. Downloaded by guest on September 27, 2021 Fig. 4. The competition between P. putida KT2440 wild type (flagellated; green) and its ΔfliM isogenic mutant (nonflagellated; red) at the surface of a porous ceramic plate is affected by matric potential and inoculation density. On average, 20 and 1,500 cells were inoculated at a 1:1 strain ratio for low and high inoculation densities, respectively. The images were ac- quiredafter2-or5-dgrowthat22°Cfor−0.5 and −3.6 kPa, respectively. The plate, 4.2 cm in diameter, is limited by a silicone o-ring, which appears bright red in the pictures. Each mosaic image is composed of about 45 fields of view. The images are representative of observations made on 5– 12 independent replicate plates. The contrast of the images has been digitally enhanced.

inoculation density, respectively). These values are significantly smaller than those observed in liquid culture (one-tailed t test, P = − MICROBIOLOGY 6.7 × 10 8 and P = 0.047 for low and high density, respectively), Fig. 3. Comparison of experimentally observed and simulated descriptors because the wild type quickly colonized the surface and thus, fl fi fl of swimming motility on partially saturated rough surfaces. Experimental val- intercepted substrate uxes more ef ciently than the non agel- ues were obtained by analyzing the trajectories of individual cells at the surface lated mutant. of the PSM set at various matric potentials. Simulated values are obtained by Natural partially saturated habitats such as soils and the surface of simulating the motion of 200 cells within an idealized roughness network. All plant leaves are subject to dynamic variations in hydration con- of the data are expressed as mean ± SEM, with n > 248 for experimental tra- ditions, the extent and frequency of which depend on climate and/or fi jectories and n = 600 for simulations. (A) Fraction of time cells display signi cant irrigation practices. To address the question of whether the occur- motility (significant motility is defined as velocities larger than 3 μm/s, the ex- fi rence of short periods of favorable wetness conditions might modify perimental detection limit). (B) Mean cell velocity during phases of signi cant fl motility. (C) Scaled maximal cellular travel distance. The maximal travel dis- agellar-based dispersal behavior, we experimentally evaluated tance is the distance between the starting point of a cell and the most distant the expression of flagellar motility under dry–wet cycles. Inoculated point of its trajectory. The trajectories were recorded over 28 s for experimental PSMs, maintained at dry conditions (−3.6 kPa), were subjected measurements and 33 s for simulations. The distance data were scaled by the to two short daily increases in hydration status (2 × 5min;−0.5 mean maximal travel distance observed in the wettest conditions (28 and 59 μm kPa). Despite the very short duration of the wet periods, the motile for experiments and simulations, respectively). strain was able to disperse, yielding larger colonies than on the control surfaces that were continuously maintained at −3.6 kPa (Fig. S2). Accelerated surface colonization associated with dry–wet rectly emerges from the individual-scale behavior in agreement cycles was not observed for the nonflagellated mutant strain. We, with previous theoretical derivations (13, 23, 24). SI Results and therefore, conclude that P. putida KT2440 can take advantage of Discussion has further discussion on the comparison between the short and infrequent wet events to disperse by flagellar motility. model and experiments. Our data show that, under conditions of partial hydration com- As evidenced by our observations and model predictions, thin fi mon in many terrestrial microbial habitats, the thickness and ge- liquid lms such as those found under most circumstances in ometry of liquid films control active bacterial motion and dispersal. soils strongly limit the dispersal rate of bacterial populations. fl Viscous and capillary pinning forces reduce the swimming velocity Does agellar motility, then, confer any selective advantage in of individual cells, resulting in low surface-colonization rates at this physically constrained environment? Competition experi- fl the population scale. More than bacterial intrinsic-growth kinetic ments between the wild type and non agellated mutant, coino- parameters, surface microtopography, hydration status, and bac- culated as randomly distributed single cells at the surface of fl fi terial agellation are essential parameters in surface colonization. the PSM, showed hydration-conditional tness effects (Fig. 4). Because the fitness gain associated with fast dispersal can be large − At 3.6 kPa, the two strains developed colonies of similar size, and when nutrient-rich microsites are available (Fig. 4) and because fi n the relative tness of the mutant was close to 1 (0.91, SD = 0.04, bacteria are able to take advantage of very short wet events to n = 9 and 0.96, SD = 0.03, = 12 for low and high inoculation disperse by flagellar motility (Fig. S2), even rare wet events could density, respectively). In these relatively dry conditions, the po- offset the cost associated with flagellar synthesis and explain the fl tentially motile wild type cannot disperse by agellar motility. sustained presence of flagellated cells in soil habitats. This shows Therefore, the sites colonized by each strain remained spatially the very tight couplings between microbial and physical processes separated, preventing direct competition, partly alleviating the in- in soil, which mediate the emergent properties of soil systems (26). trinsic inferiority of the mutant (25). The situation was different under wet conditions (−0.5 kPa), permissive for efficient flagellar Materials and Methods motility. In this condition, the mutant had a very low relative fitness Bacterial Strains. P. putida KT2440, a flagellated bacterium initially isolated (0.54, SD = 0.03, n = 9 and 0.74, SD = 0.06, n =5forlowandhigh from the rhizosphere (27), was used as the model strain. A nonflagellated

Dechesne et al. PNAS Early Edition | 3of4 Downloaded by guest on September 27, 2021 ΔfliM mutant was obtained by allelic exchange with a truncated version of fliM carrying the Gm-resistance gene aacC1 framed by lox sequences. The Wðm; wtÞ¼ln ðmF=m0Þ ln ðwtF=wt0Þ [1] aacC1 gene was then excised to yield an antibiotic resistance-free mutant (28). Both strains were tagged by inserting a constitutively expressed fluo-

rescent protein-encoding gene at a neutral position of their genome (29). To where m0 and mF are initial and final abundances of the mutant and wt0 and increase the fluorescence signal for single-cell observations, pJBA128, wtF are those of the wild type. a multicopy plasmid carrying a gfp gene (30), was additionally introduced Relative fitnesses were estimated in stirred liquid medium (17 mL FAB, into the gfp-expressing wild-type strain. The bacteria were routinely culti- benzoate 5 mM) and at the surface of the PSM (same medium). In the latter vated on FAB medium (31) supplemented with 5 mM benzoate. case, the areas colonized by each strain were used as proxy for their abundance to compute relative-fitness values. Experiments on the PSM. The PSM allows growing and observing fluorescent cells at the surface of a porous ceramic plate (diameter = 4.2 cm; maximum Modeling. We modeled a population of motile cells within a surface- × × pore size = 1.7 μm) under prescribed suction, which controls surface liquid- roughness network of size 17.2 17.2 mm (100 87 sites). The initial substrate concentration in aqueous phase was set to 0.2 mg/L and was main- film thickness (12). Liquid FAB medium with 5 mM benzoate as the sole tained constant at the top and right boundaries of the domain. The bottom carbon source was used to wet the plate and sustain microbial growth. The and left boundaries of the simulation domain were no-flux boundaries. inoculation of the surface of the PSM was performed as in ref. 12. Simulations were initiated by inoculating 200 bacterial cells in four sites at Observations of the PSM surface were realized at different scales using the bottom left corner of the domain. Triplicate simulations were conducted a Leica MZ16FA stereomicroscope. Time-lapse videos (27 s, 100 images) were for each value of matric potential: −0.0001, −0.5, −1, −1.5, and −2 kPa for 2 acquired at high magnification (field of view = 0.15 mm ) to document simulating single-cells motion and −0.5, −1, −1.5, −2, and −3.6 kPa for col- swimming motility at the cell scale. Individual trajectories were detected and ony-expansion analyses. Details of the model and its parameters (Table S1 analyzed with Image Pro Plus (MediaCybernetics). Population-scale dispersal and S2) are presented in SI Materials and Methods. was quantified as previously described (12). ACKNOWLEDGMENTS. We thank Alexis Bazire (Université de Bretagne Sud, Competition Experiments. To measure the relative fitness of the strains, we Lorient, France) for the construction of the ΔfliM mutant and Søren Molin (Tech- performed competition experiments where the abundance of the strains nical University of Denmark, Kgs. Lyngby, Denmark) for the kind gift of pJBA128. This work was supported by the Villum Kann Rasmussen Foundation Center of was measured at coinoculation time and at the end of incubation time. The Excellence Center for Environmental and Agricultural Microbiology (CREAM) fitness (W) of the mutant relative to that of the wild type was calculated as and the Swiss National Science Foundation (Project 200021-113442). B.F.S. was (2) (Eq. 1): supported by Marie Curie Excellence Grant MEXT-CT-2005-024004.

1. Tilman D (2004) Niche tradeoffs, neutrality, and community structure: A stochastic 16. Darnton NC, Turner L, Rojevsky S, Berg HC (2007) On torque and tumbling in theory of resource competition, invasion, and community assembly. Proc Natl Acad Sci swimming Escherichia coli. J Bacteriol 189:1756–1764. USA 101:10854–10861. 17. Lin BH, Yu J, Rice SA (2000) Direct measurements of constrained Brownian motion of 2. Kerr B, Riley MA, Feldman MW, Bohannan BJM (2002) Local dispersal promotes an isolated sphere between two walls. Phys Rev E 62:3909–3919. biodiversity in a real-life game of rock-paper-scissors. Nature 418:171–174. 18. Maude AD (1963) Movement of a sphere in front of a plane at low Reynolds number. 3. Venail PA, et al. (2008) Diversity and productivity peak at intermediate dispersal rate J Appl Phys 14:894–898. in evolving metacommunities. Nature 452:210–214. 19. Kralchevsky PA, Nagayama K (2001) Particles at Fluid Interfaces and Membranes 4. Jarrell KF, McBride MJ (2008) The surprisingly diverse ways that prokaryotes move. (Elsevier, Amsterdam). Nat Rev Microbiol 6:466–476. 20. Berg HC (2004) E. coli in Motion (Springer, New York). 5. Harshey RM (2003) Bacterial motility on a surface: Many ways to a common goal. 21. Long T, Or D (2007) Microbial growth on partially saturated rough surfaces: Annu Rev Microbiol 57:249–273. Simulations in idealized roughness networks. Water Resour Res 43:W02409. 6. Grossart HP, Riemann L, Azam F (2001) Bacterial motility in the sea and its ecological 22. Hellweger FL, Bucci V (2009) A bunch of tiny individuals—Individual-based modeling implications. Aquat Microb Ecol 25:247–258. for microbes. Ecol Modell 220:8–22. 7. Fenchel T (2002) Microbial behavior in a heterogeneous world. Science 296: 23. Phillips BR, Quinn JA, Goldfine H (1994) Random motility of swimming bacteria: single 1068–1071. cells compared to cell populations. AIChE J 40:334–348. 8. Stocker R, Seymour JR, Samadani A, Hunt DE, Polz MF (2008) Rapid chemotactic 24. Rivero MA, Tranquillo RT, Buettner HM, Lauffenburger DA (1989) Transport models response enables marine bacteria to exploit ephemeral microscale nutrient patches. for chemotactic cell populations based on individual cell behavior. Chem Eng Sci 44: Proc Natl Acad Sci USA 105:4209–4214. 2881–2897. 9. Or D, Smets BF, Wraith JM, Dechesne A, Friedman SP (2007) Physical constraints 25. Dechesne A, Or D, Smets BF (2008) Limited diffusive fluxes of substrate facilitate affecting bacterial habitats and activity in unsaturated porous media —a review. Adv coexistence of two competing bacterial strains. FEMS Microbiol Ecol 64:1–8. Water Resour 30:1505–1527. 26. Young IM, Crawford JW (2004) Interactions and self-organization in the soil-microbe 10. Turnbull GA, Morgan JAW, Whipps JM, Saunders JR (2001) The role of bacterial complex. Science 304:1634–1637. motility in the survival and spread of Pseudomonas fluorescens in soil and in the 27. Nakazawa T (2002) Travels of a Pseudomonas, from Japan around the world. Environ attachment and colonisation of wheat roots. FEMS Microbiol Ecol 36:21–31. Microbiol 4:782–786. 11. Boelens J, Vande Woestyne M, Verstraete W (1994) Ecological importance of motility 28. Quénée L, Lamotte D, Polack B (2005) Combined sacB-based negative selection and for the plant growth-promoting rhizopseudomonas strain-ANP15. Soil Biol Biochem cre-lox antibiotic marker recycling for efficient gene deletion in pseudomonas 26:269–277. aeruginosa. Biotechniques 38:63–67. 12. Dechesne A, Or D, Gülez G, Smets BF (2008) The porous surface model, a novel 29. Lambertsen L, Sternberg C, Molin S (2004) Mini-Tn7 transposons for site-specific experimental system for online quantitative observation of microbial processes under tagging of bacteria with fluorescent proteins. Environ Microbiol 6:726–732. unsaturated conditions. Appl Environ Microbiol 74:5195–5200. 30. Lee B, et al. (2005) Heterogeneity of biofilms formed by nonmucoid Pseudomonas 13. Skellam JG (1951) Random dispersal in theoretical populations. Biometrika 38: aeruginosa isolates from patients with cystic fibrosis. J Clin Microbiol 43:5247–5255. 196–218. 31. Hansen SK, et al. (2007) Characterization of a Pseudomonas putida rough variant 14. Matilla MA, et al. (2007) Temperature and pyoverdine-mediated iron acquisition evolved in a mixed-species biofilm with Acinetobacter sp. strain C6. J Bacteriol 189: control surface motility of Pseudomonas putida. Environ Microbiol 9:1842–1850. 4932–4943. 15. Or D, Tuller M (2000) Flow in unsaturated fractured porous media: Hydraulic 32. Berg HC, Brown DA (1972) Chemotaxis in Escherichia coli analysed by three- conductivity of rough surfaces. Water Resour Res 36:1165–1177. dimensional tracking. Nature 239:500–504.

4of4 | www.pnas.org/cgi/doi/10.1073/pnas.1008392107 Dechesne et al. Downloaded by guest on September 27, 2021