Molecular Adsorption Steers Bacterial Swimming at the Air/Water Interface

Abstract

Microbes inhabiting Earth have adapted to diverse environments of water, air, soil, and often at the interfaces of multiple media. In this study, we focus on the behavior of Caulobacter crescentus, a singly flagellated bacterium, at the air/water interface. Forward swimming C. crescentus swarmer cells tend to get physically trapped at the surface when swimming in nutrient-rich growth medium but not in minimal salt motility medium. Trapped cells move in tight, clockwise circles when viewed from the air with slightly reduced speed. Trace amounts of Triton X100, a nonionic surfactant, release the trapped cells from these circular trajectories. We show, by tracing the motion of positively charged colloidal beads near the interface that organic molecules in the growth medium adsorb at the interface, creating a high viscosity film. Consequently, the air/water interface no longer acts as a free surface and forward swimming cells become hydrodynamically trapped. Added surfactants efficiently partition to the surface, replacing the viscous layer of molecules and reestablishing free surface behavior. These findings help explain recent similar studies on Escherichia coli, showing trajectories of variable handedness depending on media chemistry. The consistent behavior of these two distinct microbial species provides insights on how microbes have evolved to cope with challenging interfacial environments.

Introduction

Swimming plays an integral role in the life of many microorganisms, aiding functions such as nutrient acquisition and efficient colonization. Swimming behavior near fluid boundaries is of special interest, bearing biological activities including the formation of biofilms. Boundary conditions create fluid flows distinct from the bulk, leading to new swimming behavior. It has been previously observed that the peritrichous bacterium Escherichia coli tend to swim in clockwise (CW) circles directly above a solid surface (1). The circular motion has a simple hydrodynamic origin: increased fluid drag near the boundary creates a net torque on the cell as the body and flagella rotate in opposite directions. Caulobacter crescentus is monotrichous during its swarmer phase, with a single flagellum attached to one pole of the crescent-shaped cell body. During the cell’s backward swimming phase, the cell tends to swim in tight CW circles above a solid surface (2); they readily leave the surface while swimming forward. Similar behavior has also been seen in monotrichous Vibrio Alginolyticus (3,4). The same turning handedness for the different swimming motions of E. coli and C. crescentus can both be accounted for by stronger viscous drag occurring near the solid boundary, noting that the flagellar filaments of the two species are of opposite helicity (5). Additionally, both bacteria have been observed accumulating near solid surfaces as cells spend increased amounts of time swimming near the boundary (6–8). Much of this behavior can be explained by the consequences of an elongated microswimmer colliding and then aligning with a flat solid surface, as shown by our recent work (7,9). These collision dynamics may also result in an accumulation of swimmers near the air/water interface, which remains flat due to high surface tension. This expected behavior is not addressed in this report; however, our experiments are performed using dilute samples of swimmers to focus on individual trajectories.

The interface between liquid and air can also create new swimming patterns. Recent observations of E. coli have shown that cells may swim in either CW or counter-clockwise (CCW) circles near such an interface (10,11). Modeling flows by treating the air/water interface as a free surface can explain only the later behavior. However, observations in these published studies of E. coli and experiments detailed in this report prove that different swimming patterns occur, including circular trajectories of opposite handedness, when the physical properties of the interface are altered by changes in surface chemistry.

In this study, we image and characterize the swimming trajectories of C. crescentus cells at the air/water interface using dark field microscopy. Experiments are performed using a mutant strain that only swims in the forward direction (cell body leading, like E. coli) and lacks pili, which wild-type strains use for adhering to surfaces. Cells are observed either approaching then leaving the surface in straight trajectories or continually retracing CW circular trajectories at the interface when swimming in growth medium containing proteins, phospholipids, and other biological material. The latter swimming pattern, however, is not observed when the cells are suspended in a motility medium that contains only simple salts. Additionally, we find that swimming behavior is very sensitive to small amounts of surfactant present in the medium, which reduces the CW swimming population by an order of magnitude. As a control, we also observe the diffusion of micron-sized beads near the interface when suspended in the media used for the swimming experiments to characterize the resistive forces acting on near-surface particles.

We propose, based on experimental findings presented in this work, that molecular adsorption alters the rheological properties of the air/water interface, which in turn determines the resultant swimming patterns of nearby cells. Though an ideal air/water interface acts as a free surface, the adsorption of proteinacious material creates a thin viscous or viscoelastic sheet, which increases drag acting on nearby cells much as a solid surface does. As a result, the turning handedness of swimming is reversed and cells can become hydrodynamically trapped at the interface. Competitive adsorption due to even small concentrations of surfactant, however, can displace the film and restores free surface features of the interface. The phenomena observed could potentially affect a wide variety of motile microswimmers using various different motility strategies (12) by modifying the fluid flows that these movements generate. In particular, near-surface effects are very pronounced in confined geometries such as swimming in microchannels or swarming in thin fluid layers (13–15). Changes in swimming patterns and accumulation will likely impact many important biological functions, including nutrient acquisition, predation, efficient colonization, and biofilm formation, which often rely on motility at surfaces (16).

Materials and Methods

Experiments in this report were performed using C. crescentus strain SB 3860, a forward swimming only mutant that lacks pili normally protruding from the cell body near the base of the flagellum. 10 mL of PYE (peptone yeast extract) growth medium (2 g/L Bacto peptone, 1 g/L Bacto yeast extract, 1.2 mM MgSO₄, and 0.5 mM CaCl₂) was inoculated with the strain and incubated at 30°C on a shaker plate at 100 rpm for 8 h. The sample was then transferred to a plastic petri dish along with additional fresh growth medium. The sample was incubated at 30°C while undergoing 40 rpm gyration over night, allowing swarmer cells to adhere to the bottom of the dish and differentiate into reproductive stalked cells. The growth medium was removed and the petri dish washed with DI water before being placed back into the incubator with fresh growth medium for several hours to ensure the health of attached stalked cells. To prepare a sample for observation, the dish was thoroughly washed and rinsed with 1 mL of medium added back following the last rinse. The dish was allowed to sit for 5 min as daughter swarmer cells separate from the attached predivisional cells before harvesting the fresh swarmer cells in the suspension. This effectively synchronized the sample (2), because all cells present in the medium are swarmer cells that divided off from their parents within the 5 min time frame. This synchronization step reduced variation in swimming behavior due to cells being in different phases of their life cycle, ensuring that all observed swimmers were <20 min old during the entire course of observation. For growth medium with surfactant observations, Triton X100 was added to the samples directly after synchronization. Samples were created with seven different Triton concentrations ranging from 1.7 mM (0.1% by volume) to 0.0017 mM (0.0001%). Samples for observations of swimming in motility medium were made by adding back 1 mL of motility medium instead of growth medium during the synchronization stage. The motility medium used for this study is a standard buffer used in biological studies of C. crescentus commonly referred to as M2 salts (20 mM Na₂HPO₄, 20 mM KH₂PO₄, 9.3 mM NH₄Cl, 0.5 mM MgSO₄, 0.5 mM CaCl₂, 0.01 mM FeSO₄, and 0.008 mM EDTA). Once prepared, 80 μL of the sample was pipetted into a specially constructed observation well composed of a rubber O-ring 12.7 mm in diameter attached to a glass slide. This geometry created a fluid layer with a solid glass boundary at the bottom and an air/water interface at the top. The high surface tension of the sample ensured that the interface remains flat and in focus during observation. A glass coverslip was placed on top of the O-ring, although still leaving a pocket of air, to prevent drafts from perturbing the top liquid surface.

Samples were observed using a Nikon (Tokyo, Japan) Eclipse e800 upright microscope with a 40× dark field objective. The stage was then slowly raised until the cells swimming closest to the air/water interface came into focus. Due to the depth of field of the microscope, swimmers within ∼5 microns of the surface were observed. Images of the sample were taken repeatedly at 20 frames/s over 10 s using a Photometrics (Tucson, AZ) Coolsnap HQ CCD camera attached to the microscope. The resulting images were cleaned using Metamorph (Molecular Devices, Sunnyvale, CA) software and a threshold was applied to obtain the positions of visible cells in each frame. The centroid position of these objects were fed into custom written MATLAB (MathWorks, Natick, MA) scripts, which tracked the movement of individual cells and analyzed the properties of the resulting trajectories. The speed and curvature of each swimmer were calculated along its trajectory using positions in consecutive frames.

For bead diffusion measurements, the same apparatus and media were used. Movies of the beads’ motion were captured at 10 frames/s over 4 s. Images were thresholded. Using positions of individual beads in consecutive frames, net displacements of the beads over a fixed time interval were calculated. This procedure measures diffusion in two dimensions coplanar to the interface over the same depth of field as our C. crescentus observations. The hydrodynamic mobility (μ) and drag coefficient (ζ = 1/μ) of an individual bead were solved using D = μk_BT. Here, D is the diffusion coefficient, which is related to the average of the square of the net displacement in one dimension 〈x²〉 = 2 Dt, where t is the time interval between consecutive observations. Thus, by tracking one bead in two dimensions, two values of the drag coefficient were obtained based on mean-squared displacement in two orthogonal directions. For reference, we calculated the drag coefficient for a 0.5 micron diameter bead in pure water using ζ = 6πηr, where η is the fluid viscosity and r is the bead radius. For (a typical pure) water viscosity of 0.001 Pa·s, the drag coefficient is ∼4.7 × 10⁻⁹ Pa·m·s. Amine group coated beads were purchased from Polysciences (Warrington, PA).

Results

By changing a microswimmer’s chemical environment, we are potentially altering the fluid properties and boundary conditions governing the hydrodynamic flows generated during swimming. Because dynamic interactions between the swimmer and the surrounding medium are responsible for motility, we also expect to observe new swimming patterns. Two key properties of cell motility that we measure are speed and direction of motion over time. It has been previously observed that in the bulk (away from interfaces/boundaries), C. crescentus swims in straight trajectories at ∼50 microns/s (17). Due to the fast rotation rate of the cell body (∼50 revolutions/s with the flagellar motor rotating on the order of a 100 revolutions/s), its crescent shape does not cause the swimming trajectory to markedly deviate from a straight path. At an air/water interface, however, the presence of an imposed boundary breaks axial symmetry and allows for curved swimming patterns. Furthermore, physical properties of a particular interface (such as surface tension and viscosity) alter swimming dynamics. From observed cell trajectories (Fig. 1) we derived the instantaneous swimming speed and curvature for all cells, which are then averaged for each individual swimmer over the entire trajectory. Fig. 2 shows in comparison the distribution of average speeds and curvatures for cells swimming in both motility medium and growth medium. In the motility medium, cells move in relatively straight trajectories with a very small bias toward negative (CCW) average curvature and with a wide spread of swimming speeds. Visual inspection shows that these trajectories consist of relatively straight sections interspersed with negative curvature regions (Fig. 1 A). The curvature values of swimmers in growth medium, however, do not conform to a single peak bell curve. Instead, two clearly separated populations are noted. One population is centered at near zero curvature, whereas the other peaks at 0.2 μm⁻¹. A large spread of swimming speeds is present in both populations, although the high curvature population has a lower average speed. We additionally noted that cells swimming in CW circular trajectories tend to continue swimming in the same pattern at the surface, whereas straight swimmers tend to approach the surface then leave within a short period of time. The local minimum between the two peak values seen in the average curvature histogram (Fig. 2 C) confirms our observation that individual cells do not switch between these two behaviors during the timescale of our recordings. Frequent switching between straight and CW circular swimming would yield a larger number of measured average curvature values between the two peaks.

Figure 1.

C. crescentus trajectories at the air/water interface swimming in motility medium (A) and growth medium (B). Displayed images are overlays of every other frame from recorded movies, thereby depicting sample trajectories of swimming cells. The added arrows indicate the directions of swimming.

Figure 2.

Scatter plots of average swimming speeds and curvatures for individual cell trajectories. The data compares cells swimming at the interface of air and motility medium (A) and that of air and growth medium (B). The graphs contain 1466 and 2255 data points, respectively. Each data point represents a distinct swarmer cell. Positive curvature is defined as CW movement as viewed from above the fluid surface. (C) Histograms showing the fractions of observed cells swimming with average curvature values within 0.01 μm⁻¹ bins.

The characteristic swimming patterns observed are indicative of the physical properties of the air/water interface. We therefore expect changing these properties to alter swimming dynamics. Surfactants are amphiphilic molecules that adsorb at the interface between polar and nonpolar media such as water and air. The presence of adsorbed surfactant lowers the surface tension at the interface and also alters other surface properties. To test whether surface chemistry was the dominant factor separating swimming patterns observed in growth medium versus motility medium, we repeated our observation procedure with low concentrations of nonionic surfactant Triton X100 added to the growth medium. At the low concentrations used in our study (0.1% or less by volume), changes to bulk viscosity and density of the swimming medium were negligible. Additionally, the mutant strain of C. crescentus used in this study is unable to perform chemotaxis, which would require changing motor rotation and consequently swimming direction. Therefore, we expect no changes in biological or chemical signaling within the cell to alter swimming trajectories. As predicted, we found that at high surfactant concentrations the second population of high CW curvature swimmers disappeared. All cells swam in relatively straight trajectories, much like in motility medium. Fig. 3 displays a histogram of average trajectory curvature for various concentrations of added Triton. Note that for the highest concentrations of surfactant, the distribution consists of only one peak at near-zero curvature. Free Triton molecules in the bulk are known to correspondingly increase the surface density of adsorbed Triton at the interface until, at a sufficiently high concentration, surface adsorption reaches a saturation value of 2.6 μmol/m². The exact relationship between bulk concentration and surface concentration can be determined by minimizing the free energy of a system of adsorbing molecules (18,19). Using this information, we calculated the true bulk concentration for each of our samples by requiring that the total amount of Triton (dissolved in bulk plus adsorbed at surface) is conserved and equal to the amount added to the sample. It should be noted that when very small amounts of Triton are added to our thin sample, their strong adsorption may reduce the bulk concentration by an order of magnitude. By defining CW circular swimmers as cells with average curvature >0.05 μm⁻¹ (the local minimum value between the two observed populations in growth medium), we determined the fraction of observed cells swimming in circular trajectories under all conditions (Fig. 4). We find that the CW circular population follows a very similar pattern to Triton adsorption where the percentage of observed cells following circular trajectories falls quickly with increased surfactant at low concentration before reaching a constant background value of ∼5% at high concentrations. The most compelling feature is that these transitions occur over the same range of bulk surfactant concentration values, suggesting that surface adsorption is directly responsible for whether swarmer cells are trapped at the interface or able to leave freely. Because our growth medium contains both yeast extract and peptone, which both contain proteinacious and other biological material; we performed control experiments by adding each ingredient individually to our motility medium before resuspending swarmer cells. Circular swimming was present in both conditions with only one nutrient source added (Table 1), suggesting a variety of nutrient sources can alter swimming patterns. The greatest number of CW circular swimmers was seen in growth medium, suggesting that swimming dynamics are sensitive to the total concentration of biological material in the swimming medium.

Figure 3.

Distribution of average trajectory curvature for swimming in growth medium with various concentrations of Triton X100. Concentrations of added surfactant range from 0.00171 mM (0.0001% by volume) to 1.71 mM (0.1% by volume). Over 1000 trajectories were measured for each condition.

Figure 4.

Fraction of observed cells moving in CW circular trajectories (circles) and surface density of adsorbed Triton (solid line) (18) as a function of bulk Triton concentration. The dashed line is a power law fit that serves as a guide for the eye. CW circular swimmers are defined as having average curvature >0.05 μm⁻¹. Error bars indicate the standard error of measured values over multiple observations.

Table 1.

Percentage of observed swimmers in CW circular trajectories when suspended in various media

Swimming medium	Percent trapped in CW circles
Motility medium	4.0 ± 2.4
Motility medium + yeast extract	15.0 ± 1.6
Motility medium + peptone	25.6 ± 2.3
Growth medium	59.7 ± 5.9
Growth medium + surfactant	6.8 ± 1.3

2 g/l yeast extract and 4 g/l peptone (the concentrations used in the growth medium) were added to the motility medium to test effects of the two ingredients individually. Error shown is standard error of the mean based on multiple preparations and observations for each condition.

Hydrodynamic flows near an interface are highly dependent on the physical properties of the interface itself. We therefore expect that surface rheology greatly impacts cell swimming under the conditions of our experiments. In particular, some interfaces display significant surface viscosity, which resists flow parallel to the interface. A wide variety of methods have been developed to measure shear and dilation surface viscosity of a fluid interface (20,21). How these properties affect particles moving near the interface, however, is a related but significantly different problem. In our experiments, C. crescentus cells move directly underneath and parallel to the air-water interface when trapped at the surface. To model this system and assess the drag forces acting on the swimmer, we observed the Brownian motion of 0.5 micron diameter beads under similar constraints; the rate of diffusion is dependent on the resistive forces the bead experiences. The beads were suspended in the same motility media used in our swimming experiments and imaged using the same apparatus and methods. Additionally, because we observed hydrodynamic trapping of cells near the surface, we used amine-coated beads such that a weak attractive force was present between the beads and negatively charged proteins adsorbed at the surface. This attraction simulates the component of cell propulsion into the interface when the cell’s swimming orientation is angled toward the surface. In both cases, this force acts to minimize the distance between the cell or bead and the air-water interface, which may enhance drag due to a viscous surface. Additionally, we found that uncoated beads diffused away from the surface before a statistically significant number of frames could be captured. The amine-coated beads, however, are not permanently trapped at the surface. At any moment, they can be seen over the entire depth of field we observed.

By observing the diffusive movement of the beads in two dimensions, we can calculate two independent drag coefficients for each bead. Under all three conditions, the most frequently measured value for the drag coefficient was ∼5 × 10⁻⁹ Pa·m·s, This is also the rate we expect to observe for 0.5 micron diameter beads diffusing in bulk pure water. When suspended in growth medium, however, there was also a substantial population of beads diffusing much slower than expected. Reduced mobility was not prevalent in the motility medium or the growth medium with added surfactant (see Fig. 5). This result is consistent with our swimming observations in that cells swimming in CW circles are common only in growth medium without Triton. We conclude that objects sufficiently near the air/water interface in growth medium experience increased resistive forces when moving parallel to the surface. The large variation among the high effective drag measurements is likely due to the beads’ variable distances from the interface over time. Swimmers angled even slightly into the interface tend to accumulate near the surface. This explains the larger number of cells experiencing increased drag as compared to beads under the same conditions as many observed beads are too far away from the surface to display near-surface effects. Comparing our measurements with previous measurements of surface viscosity also yields good agreement. Many low molecular weight surfactants have been shown to create negligible or low shear surface viscosities of the order 1 μPa·m·s, whereas a wide variety of proteins adsorbed at the interface can create shear surface viscosities of the order 1 mPa·m·s or larger due to interactions between adjacent molecules (22–24). These results also agree with recent experimental work showing that particles suspended on the air-water interface above an E. coli swarm have decreased mobility (14). Although our setup differs in that the beads are suspended slightly below the surface instead of being trapped at the interface, both experiments highlight increased surface viscosity.

Figure 5.

Measured drag coefficients for micron-sized beads near the air/water interface. Data points in scatter plots show independent values for motion in the x and y directions for individual beads. Histograms use bins of width 10⁻⁹ Pa·m·s and display measurements from motion in the x and y direction together. Beads are suspended in motility medium (A and D), growth medium (B and E), and growth medium with 0.01% Triton X100 by volume (C and F). Insets show beads diffusing beneath interfaces with different surface chemistry. Beads close to a viscous surface (illustrated by a shaded bead) have decreased mobility.

Discussion

Free surface hydrodynamics in motility medium

Hydrodynamically, an ideal air/water interface is considered to be a free surface such that shear stress parallel to the interface is effectively zero (25). Such a system allows for flow parallel to the surface, as opposed to at a solid/water interface, which requires that the flow decays to zero, known as the no-slip condition. Additionally, high surface tension at the interface between two dissimilar fluids such as air and water requires minimizing the surface area shared by the two fluids. The hydrodynamic flow generated by a swimmer near a free air/water interface must meet the boundary condition that the flow normal to the surface must go to zero at the surface, whereas the flow parallel to the surface can be nonzero. A simple way to satisfy this requirement is to invoke an image swimmer on the other side of the interface with the cell body and flagellum or flagellar bundle rotating in directions opposite that of the real swimmer (10). The flow created by the image swimmer exerts lateral forces and a net torque on the real swimmer, resulting in circular trajectories. The direction of turning is opposite to that of the circular swimming near a solid interface (1). Given our experimental setup using an upright microscope, where we view a swimmer underneath an air/water interface from above, we would expect to see E. coli swim in CW circles and C. crescentus swim in CCW circles due to the opposite handedness in the helical structure of their flagellar filaments. Our results observing cells suspended in the motility medium within the finite depth of field of observation agree with this model. Most cells have slightly negative average curvature as trajectories are composed of sharp CCW turns interspersed between straight sections associated with approaching and leaving the interface.

One notable difference between E. coli and C. crescentus is that the latter does not continually trace out circular trajectories near the free surface but instead it quickly reorients and swims away from the air/water interface. We have previously reported that, unlike E. coli, C. crescentus does not become trapped near a solid surface during forward swimming (2). Whether or not a cell can escape the surface is dictated by the strength of Brownian rotation of the cell’s central axis after orienting along the surface (6). The tendency of forward swimming C. crescentus to leave both the air/solid and air/water interfaces may be attributable to a higher coefficient of rotational diffusion of its smaller cell body and single flagellum or misalignment between the axes of the cell body and flagellar filament due to the flexibility of the hook attaching the two. In fact, recent work has implicated that the flexibility of this hook plays a vital role in swimming reorientation during chemotaxis in similar monotrichous bacteria (26).

Viscous surface effects in growth medium

A free surface model, however, does not explain C. crescentus swimming trajectories in growth medium where ∼60% of cells swim in CW circles or a similar E. coli experiment in which both CW and CCW trajectories were observed (11). The sign of curvature for C. crescentus swimming in CW circles and E. coli swimming in CCW circles is the same as when they swim near a solid interface. Considering the opposing directions of rotation of the cell body and the flagellum as the main cause of circular swimming, the fluid drag on the cell must be greater on the side of the cell facing the interface to obtain the correct sign of torque. This interpretation is reinforced by our measurements of decreased mobility of beads diffusing near an air/growth medium interface. Both the motility medium and the growth medium used in our experiments have bulk viscosity and density nearly identical to pure water, leading to similar bulk swimming behavior. However, adsorption of molecules dissolved in the medium at the air/water interface can drastically alter surface chemistry. In particular, proteins and peptides contained in growth medium tend to adsorb at the air/water interface due to their moderate amphiphilic behavior (23,27). A thin layer of proteinacious material creates a viscous or viscoelastic surface due to lateral interactions between molecules (20,22). Because both peptone and yeast extract contain a wide variety of biological material, it may be possible that the adsorption of a variety of other molecules, such as phospholipids and polysaccharides, contributes to this effect. However, the physical implication is clear. A viscous surface, unlike a free surface, resists deformation through either shear or dilation (28). The energy cost of deforming a viscous surface imposes new boundary conditions on our swimming system. For a near surface swimmer, drag acting on both the cell body and the flagellum as they rotate in opposing directions is enhanced. This drag is most pronounced on the side of the cell facing the interface much like the case of a nearby solid surface imposing a no slip condition. Thus, when applied to our C. crescentus swimmers, the unbalanced forces on the cell create a net torque resulting in CW circular motion, opposite to what is expected at the free surface (see Fig. 6). The increased total drag acting on the swimmer would also explain why CW circular swimming cells have reduced average speed. Cells swimming near a viscous surface display many properties associated with swimming near a solid surface. In fact, in the limit of very large surface viscosity, the interface will become effectively stationary and impose the same boundary condition as a solid surface.

Figure 6.

Illustration of a cell swimming directly beneath a viscous surface as viewed from the side (A) and from above (B). Due to increased resistive drag on the side of the cell facing the surface, a net torque angles the cell into the interface, preventing the cell from leaving it. The turning of the cell body and flagellum also generates a torque that results in CW circular swimming. Near a free surface, the turning handedness is inverted as drag is decreased rather than augmented at the interface.

A highly viscous or viscoelastic surface also allows for hydrodynamic trapping of near surface swimmers. As a cell moves through a fluid, there is also a linear drag force opposing its forward motion that, in the bulk, can be described as applied directly along the axis of the cell due to symmetry. While moving parallel to a viscous surface, however, the cell experiences unbalanced drag forces on its sides facing and facing away from the surface. For simplicity, we model the cell body as a propelled sphere moving through a viscous fluid while in contact with the air/water interface. The sphere will experience a net torque that results in a rolling motion in the presence of large surface drag (29). In the case of a swarmer cell pushed from behind by a helical flagellum, this effect results in net torque acting on the cell body angling the cell's swimming orientation toward the plane of the surface (see Fig. 6). If this torque is strong enough, Brownian rotational diffusion would not be sufficient to orient the cell away from the surface and the cell becomes hydrodynamically entrapped. The component of swimming propulsion normal to the surface is counterbalanced by strong surface tension, preventing deformation of the interface. Additionally, nonspecific binding between the cell and the adsorbed protein layer may create a weak attractive potential that could enhance trapping. However, such a force alone might not be sufficiently strong to overpower the propulsive force generated by a swimmer if it were able to angle itself away from the surface. Previous work has also shown that pusher cells swimming parallel to a nearby solid surface are hydrodynamically attracted to it (6,8). However, this model does not apply to puller cells such as C. crescentus during its backward swimming phase or swimming next to free surfaces due to different boundary constraints.

Competitive adsorption and restoration of free surface dynamics

Surfactants can displace a thin protein layer at the air/water interface through competitive adsorption (30), which also removes the viscous surface boundary condition from our system. When plotted as a function of bulk Triton concentration, the proportion of CW swimmers undergoes a steep drop over the same range as the surface concentration of adsorbed Triton increases steeply (Fig. 4). In the transition region, the proteinacious layer is only partially removed and surface viscosity is reduced but still significant. The variation in the CW circular swimmer population could be a consequence of a reduction in entrapping forces as surface viscosity continuously decreases until Triton adsorption reaches saturation. Alternatively, Triton adsorption could be locally variable at subsaturation concentrations, leaving patches of high viscosity protein films. In either case, the probability of an individual cell becoming hydrodynamically trapped is very sensitive to changes in the chemical composition of the interface, as evidenced by the steep reduction in the population of CW circular swimmers. Although this work only studies the effects of a single chemical surfactant, it is reasonable to expect that any molecule that modulates surface properties through adsorption may alter swimming behavior. This includes other surfactants and phospholipids, which may have different characteristics such as lower solubility and decreased critical micelle concentrations. Regardless, we expect that resultant swimming behavior is directly related to the viscous or viscoelastic properties of the interface. An interesting consequence of this phenomenon is that by simply observing the swimming patterns of a large collection of cells, it may be possible to infer certain physical characteristics of a nearby surface.

Concluding Remarks

Our experiments demonstrate that altering one physical parameter in a swimming system can result in several different, independently measurable phenomena. The adsorption of proteinacious material at the air/water interface changes a free surface into a viscous surface, giving rise to increased resistive drag on nearby cells, decreased swimming speed, net torque that results in CW circular swimming, and the potential for hydrodynamic trapping. The contradictory swimming behaviors of E. Coli in previous studies can be accounted for by this same surface chemistry-dependent mechanism. Both C. crescentus and E. coli adopt new swimming patterns when the air/water free surface is replaced with a viscous surface. Because the movement of any microswimmer through its environment is a function of the hydrodynamic flow it produces, we expect that this viscous surface effect is also relevant to other swimming strategies, including the propagation of transverse waves along the flagella of spermatozoa and synchronization of cilia along the body of certain eukaryotic cells. From a biological perspective, maneuverability of an organism through its environment can affect survivability by impacting its ability to acquire nutrients and escape predators. With this in mind, we expect some common swimming strategies to remain effective under the surface constraints examined in this study. It is even possible that near-surface swimming patterns could enhance an organism’s ability to climb chemical gradients or form biofilms, resulting in an evolutionary advantage for certain swimming strategies.