Introduction

Bioaugmentation is an soil bioremediation process in which cultured bacteria are introduced into contaminated soil to degrade pollutants [1]. Among numerous factors that determine the process efficacy, a critical prerequisite that is often challenging to achieve is to disperse the cells to the pollutant source, particularly to low permeability zones that are abundant in heterogeneous, unsaturated soil [2]. A successful bioaugmentation thus requires directing decomposer bacteria to the target soil micropores deep in the subsurface where the contaminants are present [3, 4]. While bacteria can passively advect across permeable regions of the subsurface via hydrodynamic dispersion, impervious areas, which are prevalent in heterogeneous soil matrix, can only be accessed by their active motility or Brownian motion. These areas often carry a significant amount of contaminants since they cannot be easily displaced hydrodynamically, thus limiting the effectiveness of remediation [5, 6].

The fact that decomposer bacteria can thrive on toxic contaminants makes them naturally chemotactic toward these contaminants [7–9]. In fact, it is well known that chemotaxis directs cells to pollutant-rich regions and improves bacterial retention during inoculation, implying active migration toward low-permeability zones where contaminants tend to persist [10–13]. Nevertheless, the inherently stochastic nature of flagellar motility (e.g., run-and-tumble) may limit the efficiency of this navigation [14, 15]; thus, rather deterministic transport mechanisms may offer more reliable targeting. Here, we hypothesize that diffusiophoresis could provide this deterministic guidance, enabling more directed bacterial migration toward contaminant-rich regions.

Diffusiophoresis describes the directed motion of colloidal particles along solute gradients, where the particle velocity can be precisely controlled by manipulating solute gradients [16]. A number of recent studies have shown that diffusiophoresis can be used to control colloidal transport in confined spaces, e.g., porous media [17–23] and impervious dead-ends [24–30], as such regions typically display diffusion-limited environments. Also, the native surface charge of cell membranes makes bacteria susceptible to experiencing diffusiophoresis [31–33], although it remains elusive how diffusiophoresis affects flagellar motility. Motivated by these prior studies, here we investigate the impact of diffusiophoresis on the motility of soil bacterium, Pseudomonas putida F1, and demonstrate the use of diffusiophoresis in enhancing chemotactic dispersion toward non-aqueous phase liquid (NAPL) in confined geometries.

Results

Salt Gradients Induce Directional Cell Migration

To observe the motility of P. putida F1 under the influence of salt gradients, we use a transparent microfluidic device consisting of two parallel flow channels connected by a narrow pore (Figure 1a). One end of the pore is sealed with a thin hydrogel membrane (polyethylene glycol diacrylate) to suppress undesired convective flows [34]. The membrane allows the diffusion of solute molecules, enabling the formation of stable salt gradients. Additional details on the fabrication process are provided in the SI and also in our previous work [35].

FIG. 1

Initially, cells suspended in a 10 ×-diluted random motility buffer (11.2 g/LK₂HPO₄, 4.8 g/L KH₂PO₄, and 0.029 g/L EDTA) supplemented with 0.1 M NaCl are introduced into the open-sided flow channel (left channel in Figure 1a), while the same buffer containing 0.1 M NaCl is simultaneously injected into the gel-sided flow channel (right channel). Bacteria from the left flow channel swim into the pore, eventually achieving an even distribution across the pore after one hour (upper panel in Figure 1a). Then, a cell suspension at reduced salt concentration (1 mM NaCl with dilute motility buffer) is injected into the left flow channel to create salt gradients across the pore (lower panel in Figure 1a).

Upon the introduction of NaCl gradients, we observe that the overall cell population in the pore changes over time. As presented in Figure 1b, the total number of cells in the pore gradually increases in the presence of NaCl gradients, whereas the cell population remains nearly unchanged when the NaCl concentration is uniform throughout the channel. Moreover, not only does the total cell population increase, but also their spatial distribution changes, as shown in Figures 1c,d; the color code represents local number of cells normalized by the total cell number at t = 0. While the overall cell distribution remains more or less the same in the absence of NaCl gradients (Figure 1c, Movie S1), with NaCl gradients the cells tend to locate deeper into the channel toward higher salt concentration (Figure 1d, Movie S2). These results suggest that salinity gradients can attract the bacteria into the deep pores and promote their dispersion toward higher salinity.

Directional Alignment and Biased Motility under Salt Gradients

To better understand the observed salt-tactic behavior, we evaluate the motility of individual cells by analyzing their trajectories. Sample trajectories (unfiltered) with or without NaCl gradients at different time windows over a period of 25 s are presented in Figures 2a,b, which also display more cell trajectories found deep in the pore over time under NaCl gradients (Figure 2b).

FIG. 2

To quantitatively determine whether more cells move to the end of the pore, we analyzed the distribution of cell positions over time (Figure 2c,d). In the absence of NaCl gradients, the distribution remains spatially consistent, with no progressive shift in the position and distribution (Figure 2c). We did not observe any temporal change between early (0–25 s) and late (150–175 s) segments (Δx = -17.43 ± 18.9 μm, p = 0.36, n = 375 trajectories) [36]. In contrast, under NaCl gradients, the distribution shifts progressively deeper into the pore, as reflected by increases in the median and a broader distribution at later times extending toward larger x values (deeper into the pore), as shown in Figure 2d. The median position is increased from 216.9 ± 15.1 μm to 343.2 ± 19.6 μm (Δx = 126.30 ± 24.7 μm, p < 0.0001, n = 255 trajectories). More statistical details can be found in Table S1 (SI).

Upon analyzing the trajectories, we immediately identify that the cells tend to be more aligned along the salt gradient, where Figures 2e,f show the probability distribution of angle θ between the cell’s instantaneous velocity vector u and the pore axis (+x direction). The slightly reduced distribution in the lateral direction (θ → ±90°) in the absence of NaCl gradients (Figure 2e) is likely due to the lateral confinement imposed by the slender geometry of the pore (width ≈ 65 μm). In the presence of NaCl gradients, the noticeable alignment toward the pore entrance (θ → 180°), i.e., away from the salt, is due to the run-and-reverse motility of P. putida, which is one of the pronounced motilities of lophotrichous P. putida [37–39]. Regardless of whether the cells are headed toward or away from the salt, the cells are aligned more along the pore direction (Figure 2f).

Next, we analyze their mean run speed, i.e., the average speed of the cell between successive tumble events. Here, we define ‘tumble’ as any sudden change in the cell speed and direction between two successive runs (Figure S1, SI). After processing over 1,412 runs combined, we did not observe any noticeable difference in the average run speed between the control group (no gradient) and the NaCl gradient cases (Figure 3a). However, when we plot the run speed in their corresponding directions (Figures 3b,c), we notice that there is a significant improvement in the run speed toward the salt (θ → 0°), and at the same time, the run speed decreases in the direction away from the salt (θ → 180°). We also note that we did not observe any significant difference in the tumble rates between the control and NaCl gradient cases (Figure 3h; Figure S2, SI), suggesting that NaCl gradients do not interfere with chemoreceptors [40].

FIG. 3

While the observed bias in the run speed may be speculated as due to the additional diffusiophoretic drift toward salt, the speed change in fact cannot be explained by diffusiophoresis alone. Under current conditions, the diffusiophoretic velocity, u_d ≈ ℳ_d/ℓ where ℳ_d = 𝒪(100 μm²/s) is the diffusiophoretic mobility [32] and ℓ = 𝒪(100 μm) is the characteristic length scale of the salt gradient, is estimated to be no more than 1 μm/s, which is an order of magnitude smaller than the measured speed change. Another possibility of the observed bias may be due to chemokinesis, which is a non-directional response in which cells modulate their swimming speed in response to local chemical concentration [41]. This behavior has been reported in uniform attractant concentrations without gradients, e.g., in Rhodobacter sphaeroides Escherichia coli and Azospirillum brasilense [42–44]. We analyze the instantaneous run speed of the bacteria and found no significant differences between 1 mM and 100 mM NaCl concentrations. Nonetheless, the difference was markedly significant (p < 0.001) under the salt gradients, as illustrated in Figure 3d. More statistical analysis can be found in Table S2 (SI). Therefore, we conclude that the observed bias in the run speed cannot be fully explained by chemokinesis.

Interestingly, as we observe how cells run by analyzing the straightness of each run, we notice that cells tend to run straighter toward the salt (Figures 3e-g). The run straightness, , where x_j is the cell position at j-th frame within a single run, which is defined as the ratio of the end-to-end distance to the total distance the cell has traveled within a run, represents an effective tortuosity of runs [45, 46]. As shown in Figure 3g, the run straightness is noticeably improved when swimming toward the salt. This was unexpected since run straightness is mainly affected by the Brownian rotation or the inherent asymmetry of the cell, both of which are sensitive to the cell morphology [47].

P. putida is a lophotricous cell, i.e., multiple flagella (length f ~ 5 μm, thickness ~ 20 nm) are attached at one end of the cell body (body length a ~ 1.5 μm, body width b ~ 0.8 μm) [37, 48]. Therefore, when P. putida propels forward the flagella tuft bundles up and forms a long, slender helix that is attached to the rod-like body where the body and the flagellar bundle are spatially segregated [49]. Given the asymmetry of this geometry, we hypothesize that the improved straightness toward the salt is due to non-uniform diffusiophoresis acting distinctly on the cell body and flagellar bundle.

Non-uniform Diffusiophoresis Steers Cells

Under the influence of salt gradients, the surface charge of the cell membrane triggers bacterial diffusiophoresis [32]. However, because there is a significant difference in the size and surface charge between the body (mainly consists of lipopolysaccharides and phospholipids) and the flagella (consist of flagellin), we expect that diffusiophoresis will not be occurring uniformly over the entire cell owing to the size- and charge-dependent nature of diffusiophoresis [25, 50–54]. As the cell body is larger and more charged than flagella, the body is expected to experience stronger diffusiophoresis. We confirm that the surface charge of the flagella is much smaller than the cell body, where the measured zeta potential of flagella isolated from the cell body is -0.6 ± 1.0 mV whereas the entire body is -16.3 ± 0.6 mV (SI). Considering the size and surface charge, the diffusiophoretic mobility of the body and flagellar bundle are estimated to be and , respectively [25, 50].

Therefore, when a cell swims toward salt, such an asymmetric action of diffusiophoresis tilts the cell body to align with the gradient (Figure 4a). This biased steering restores orientation along the gradient, stabilizes the trajectory, and enhances run speed by producing straighter paths.

FIG. 4

To assess whether the rotation due to diffusiophoresis can overcome Brownian rotation, we evaluate the rotational Péclet number,

where and are the diffusiophoretic velocity and mobility difference between the body and the flagella bundle, respectively, and . is the rotational diffusivity of the body plus flagella bundle [55]. We estimate Pe_r ~ 10, suggesting that such a subtle diffusiophoretic motion is yet powerful enough to overcome Brownian rotation, leading to improved motility toward salt. As shown in Figures 4b,c (Movies S3,S4), the run trajectories repositioned to start at the origin (x₀ , y₀) = (0, 0) indeed show the steering of cells toward the salt with straighter runs, confirming the impact of salt gradients on cell motion.

Cell steering due to diffusiophoresis can be further characterized by correlating the change of body angle Δθ over its arc length s along individual run trajectories, i.e., 〈cos(Δθ)〉_s, where 〈⋅〉_s is the ensemble average over all segments pairs with equal arc length separation across entire run trajectories (SI). As shown in Figure 4f, in the absence of salt gradients, the angle correlation decays slowly along its displacement, indicating persistent swimming along its initial direction within a run. In contrast, under salt gradients, the angle correlation decays faster, indicating increased directional decorrelation over its run path. This long-range decay is not attributed to any change in the rotational Brownian noise; rather, it arises from continuous, deterministic reorientation of the swimming direction enabled by the diffusiophoretic torque. Such forced steering introduces long-range curvature into the run trajectories, leading to a faster decay of angle correlation as the motion becomes globally biased toward the salt gradient.

Further analyzing the mean square displacement (MSD) of the run trajectories reveals that the observed long-range steering arises mainly from the change of cell motion along the gradient direction (SI). The MSD anisotropy, which is defined as the ratio of MSD components along the gradient direction (MSD_||) and transverse to the gradient direction (MSD_⊥), i.e., MSD_||/MSD_⊥. In the absence of salt gradients, as shown in Figure 4h, the MSD anisotropy remains approximately unity across all lag times τ. However, under salt gradients, the MSD anisotropy is substantially increased, suggesting that the directional persistence is enhanced only along the gradient axis. Also, the MSD anisotropy is large even at very short lag times, indicating that the reorientation driven by diffusiophoretic torque occurs continuously throughout the entire run.

This result is fully consistent with the angle correlation analysis, where the correlation function decays much faster under salt gradients compared to the no-gradient conditions. Moreover, the elevated MSD anisotropy at the shortest lag times, together with the rapid decay of 〈cos(Δθ)〉, provides a clear signature of continuous orientation rather than biased run-time modulation characteristic of chemotaxis. The absence of any significant and noticeable change in the mean run time between the no-gradient and salt gradient conditions (Figure 3h and Figure S2) further rules out salt-taxis or chemotactic mechanisms, supporting non-uniform diffusiophoresis as the origin of the observed behavior here in this study. Comprehensive statistical analysis, including permutation tests for the MSD anisotropy, confirms the robustness of these findings (Table S3, SI).

The diffusiophoresis-driven steering can also be simulated by solving Langevin equations for cell translation and rotation (SI). The rotation of Brownian cell subject to diffusiophoresis can be modeled as , where is the rotation due to non-uniform diffusiophoresis, and η_r (t) is the random Brownian rotation. The simulation results are shown in Figures 4e where NaCl gradients effectively steer the cells toward salt and enable straighter runs, displaying a qualitative agreement with the experimental trajectories shown in Figure 4c. The non-uniform diffusiophoresis also yields the rapid decay in the angle correlation (Figure 4g) and the increased MSD anisotropy (Figure 4i). Furthermore, the simulations also predict the change in the global (Figure 1b) and local (Figure S3) cell population under the impact of nonuniform diffusiophoresis, which is consistent with the experimental results shown in Figures 1b and 2b.

Diffusiophoresis Enhances Bacterial Dispersion toward Pollutants

Finally, we evaluate the diffusiophoretic cell dispersion in the presence of NAPL; toluene. While toluene is highly toxic to most microorganisms due to its membrane-disrupting effects, P. putida is resistant and capable of metabolizing toluene, which drives its natural chemotaxis toward the compound [56–59]. Using the same microfluidic device, we inject P. putida suspended in low salinity water (1 mM NaCl) into a channel that is initially filled with high salinity water (0.1 M NaCl) while toluene is present in the gel-sided flow channel (Figures 5a,b), effectively simulating diffusiophoresis-enabled bioaugmentation of NAPL-contaminated subsurface. Toluene gradually dissolves and creates toluene gradients in addition to salt gradients, creating dual chemical gradients.

In the presence of toluene, we observe that cells initially chemotax toward toluene, but soon after they slow down due to the toxicity at high toluene concentration [60, 61], making it challenging to disperse cells closer to the contaminant source (Figure 5c, Movie S7). On the other hand, when NaCl gradients are additionally imposed, the cells continuously migrate into the pore via diffusiophoresis despite the lack of flagellar motility, as diffusiophoresis is forcing the cells to move toward the contaminant source (Figure 5d, Movie S8). This demonstration suggests the potential utility of diffusiophoresis for bioremediation of NAPL-contaminated soil using externally imposed salt gradients.

FIG. 5

Conclusion

Our results indicate that diffusiophoresis can improve the swimming motility of P. putida to reach targets in confined geometries, offering a promising strategy for bioremediation in contaminated soil environments. While this work only focuses on P. putida, we expect other similar lophotrichous cells to experience similar diffusiophoresis-enhanced motility improvement due to the asymmetric geometry of the cell and the native surface charge of bacterial membranes [32]. In the same reasoning, we expect that peritrichous cells (e.g., Escherichia coli) are likely to experience the diffusiophoresis-induced steering much less since their body and flagella are not spatially separated, unlike lophotrichous cells. This warrants further investigations to uncover how flagellar configuration determines cellular diffusiophoresis.

Data availability

All codes, data files, and documentation for tracking can be accessed via the GitHub repository https://github.com/alinikkh/-Pseudomonas-putida-F1-tracking.

Acknowledgements

This material is based upon work supported by the National Science Foundation under Grants No. 2223737 and 2237177.

We thank Drs. Roseanne Ford and Rhea Braun for providing P. putida.

Additional information

Author Contributions

Conceptualization: S.S.; Methodology: V.S.D., and A.N.; Investigation: V.S.D., and A.N.; Resources: S.S.; Writing–original draft: all authors; Writing–reviewing and editing: all authors; Funding acquisition: S.S.

Funding

National Science Foundation (NSF) (2223737)

• Sangwoo Shin

National Science Foundation (NSF) (2237177)

• Sangwoo Shin

Additional files

Supplementary material.

Movie S1.

Movie S2.

Movie S3.

Movie S4.

Movie S5.

Movie S6.

Movie S7.

Movie S8.