Published September 16, 2015

Journal of Environmental Quality

Special Section Microbial Transport and Fate in the Subsurface

Swimming Motility Reduces Azotobacter vinelandii Deposition to Silica Surfaces Nanxi Lu,* Arash Massoudieh, Xiaomeng Liang, Dehong Hu, Tamir Kamai, Timothy R. Ginn, Julie L. Zilles, and Thanh H. Nguyen

B

acterial transport and fate in the soil and subsurface impacts soil microbial metabolism (Hunter et al., 1998; Murphy and Ginn, 2000), adaptation and evolution (Trevors et al., 1987), and nutrient cycling and contaminant degradation (Ginn et al., 2002; Hunter et al., 1998; Murphy and Ginn, 2000; Tufenkji, 2007), as well as risks associated with groundwater pathogens (e.g., Bradford et al., 2013). The roles of physicochemical factors on bacterial cells in groundwater streams regarding advection, dispersion, filtration, and straining have received significant attention (Sen et al., 2005; Tufenkji, 2007; Wang et al., 2013). After decades of studies on bacterial transport and deposition, it is evident that biological properties and activities need to be considered (Castro and Tufenkji, 2008; Tufenkji, 2007; Walker et al., 2005; Wang et al., 2008). Considering specifically the relationship between bacterial swimming motility and bacterial transport in porous media, previous research compared bacterial transport of motile and nonmotile strains under a range of physical and chemical parameters (Camesano and Logan, 1998; de Kerchove and Elimelech, 2008a, 2008b; Haznedaroglu et al., 2010). Different responses to ionic strength and divalent cation concentration have been observed for motile and nonmotile Pseudomonas strains, suggesting differences in the mechanisms controlling their attachments to silica and alginate-coated surfaces (de Kerchove and Elimelech, 2008a, 2008b). The presence of flagella resulted in greater sensitivity to ionic strength for Azotobacter vinelandii deposition (Lu et al., 2013). At a low fluid velocity (0.56 m/d), swimming Pseudomonas cells were able to avoid attachment in packed-bed columns (Camesano and Logan, 1998). For Escherichia coli, flagellar rotation also increased detachment of cells at a low fluid velocity (0.0044 cm/s) in a parallel flow setup, but under a higher fluid velocity (0.44 cm/s), enhanced attachment was observed for motile cells (McClaine and Ford, 2002a).

Abstract The transport and fate of bacteria in porous media is influenced by physicochemical and biological properties. This study investigated the effect of swimming motility on the attachment of Azotobacter vinelandii cells to silica surfaces through comprehensive analysis of cell deposition in model porous media. Distinct motilities were quantified for different strains using global and cluster-based statistical analyses of microscopic images taken under no-flow condition. The wild-type, flagellated strain DJ showed strong swimming as a result of the actively swimming subpopulation whose average speed was 25.6 mm/s; the impaired swimming of nifH− strain DJ77 was attributed to the lower average speed of 17.4 mm/s in its actively swimming subpopulation; and both the nonflagellated JZ52 and chemically treated DJ cells were nonmotile. The approach and deposition of these bacterial cells were analyzed in porous media setups, including single-collector radial stagnation point flow cells (RSPF) and two-dimensional multiple-collector micromodels under well-defined hydrodynamic conditions. In RSPF experiments, both swimming and nonmotile cells moved with the flow when at a distance ≥20 mm above the collector surface. Closer to the surface, DJ cells showed both horizontal and vertical movement, limiting their contact with the surface, while chemically treated DJ cells moved with the flow to reach the surface. These results explain how wild-type swimming reduces attachment. In agreement, the deposition in micromodels was also lowest for DJ compared with those for DJ77 and JZ52. Wild-type swimming specifically reduced deposition on the upstream surfaces of the micromodel collectors. Conducted under environmentally relevant hydrodynamic conditions, the results suggest that swimming motility is an important characteristic for bacterial deposition and transport in the environment.

Core Ideas • Statistical cluster analyses of cell trajectories separated actively swimming cells from nonswimming ones. • Motility was evaluated in multiple porous media model systems with increasing complexity. • Strong motility changed trajectories near the surface and reduced attachment.

N. Lu, J.L. Zilles, and T.H. Nguyen, Dep. of Civil and Environmental Engineering, Univ. of Illinois at Urbana-Champaign, Urbana, IL 61801, USA; A. Massoudieh and X. Liang, Civil Engineering Dep., The Catholic Univ. of America, 620 Michigan Ave. NE, Washington, DC 20064, USA; D. Hu, Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, 902 Battelle Blvd., Richland, WA 99352, USA; T. Kamai, Institute of Soil, Water and Environmental Sciences, Agricultural Research Organization, Bet Dagan, Israel; T.R. Ginn, Dep. of Civil and Environmental Engineering, Univ. of California, Davis, Davis, CA, USA. Assigned to Associate Editor Thomas Harter.

Copyright © 2015 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America. 5585 Guilford Rd., Madison, WI 53711 USA. All rights reserved. J. Environ. Qual. 44:1366–1375 (2015) doi:10.2134/jeq2015.03.0141 Supplemental material is available online for this article. Received 15 Mar. 2015. Accepted 6 July 2015. *Corresponding author ([email protected]).

Abbreviations: FCCP, Carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone; MOPS, 3-morpholinopropane-1-sulfonic acid; MSD, mean square displacement; RSPF, radial stagnation point flow cell.

1366

Several studies have investigated the role of chemoattractants on microbial distribution in both column (e.g., Wang and Ford, 2010) and micromodel (Wang et al., 2012) settings. Studies such as these indicate the need for more quantitative connection between single cell-scale motility and population-level microbial fate and transport in porous media. Microscopic techniques have been used to evaluate possible mechanisms for the effect of motility on bacterial transport. When either smooth or tumbling flagellar movement was eliminated, the cells showed lower attachment in parallel flow chambers at the ionic strength of 0.2 M (McClaine and Ford, 2002b). Qualitative microscope observations suggested that for smoothly swimming cells, the decreased attachment was linked to a decrease in time on the surface, while for continually tumbling cells the ability to approach the surface seemed to be impaired (McClaine and Ford, 2002b). A three-dimensional microscopic tracking technique revealed a tendency for individual E. coli cells to swim in circles parallel to a glass surface (Frymier and Ford, 1997; Frymier et al., 1995; Vigeant and Ford, 1997). Complementary approaches allowed three-dimensional trajectories to be reconstructed for multiple cells from the same time-series images, for example, by relying on image analysis to recognize cells that were outside the focal plane (Wu et al., 2006). With three-dimensional trajectory recording techniques, individual cell activities were identified and grouped to represent types of bacterial movement (Frymier and Ford, 1997; Frymier et al., 1995; Vigeant and Ford, 1997; Wu et al., 2006). These techniques cannot study the whole bacterial population, however, and could miss rare cell activities that may have important roles in transport. Recently, an innovative image analysis technique was developed that identifies not only bacterial trajectories but also the orientation of the individual rod-shaped cells of Pseudomonas within a population (Conrad et al., 2011). Using this approach, flagella-mediated, surfaceanchored spinning was identified, and vertically oriented cells were observed to have higher probabilities of detachment (Conrad et al., 2011). These trajectory-based approaches can quantitatively characterize bacterial motility at the population level and have the potential to provide new insight into its role in bacterial transport, but to our knowledge they have not yet been applied in transport and deposition studies in porous media. To evaluate the role of swimming motility on bacterial surface deposition, we conducted quantitative, trajectory-based characterization of swimming motility under no-flow condition and analyzed the corresponding cell movement under well-defined flow conditions using model porous media with increasing complexity. Specifically, we developed global and cluster-based statistical trajectory analyses of microscopic images to separate the actively swimming cells from nonswimming ones and quantified the motilities of actively swimming cells for the bacterial strains selected for deposition experiments. The selected bacterial strains and conditions were wildtype, flagellated A. vinelandii strain DJ; flagellated, impaired swimming DJ77; nonflagellated, nonmotile JZ52; and chemically treated DJ cells. We resolved bacterial movement on and above simplified collector surfaces using radial stagnation point flow cells (RSPF) and used two-dimensional, multiplecollector micromodels to investigate the impact of different

levels of swimming motility on bacterial deposition in porous media. The well-defined hydrodynamic conditions were maintained at an environmentally relevant range in all deposition experiments. Azotobacter vinelandii was selected as the model organism because it is a native soil organism and also naturally competent, that is, able to receive extracellular DNA through a common mechanism of horizontal gene transfer (Lorenz and Wackernagel, 1994, Page and von Tigerstrom, 1979). Its use here thus allows future research examining the role of motility in horizontal gene transfer in the environment. Silica surfaces were used to represent the most common soil and subsurface constituent.

Materials and Methods Bacterial Strains and Growth Conditions Three A. vinelandii strains were used: DJ (wild-type, flagellated, and motile), DJ77 (nif H-, flagellated but with impaired motility [ Jacobson et al., 1989; Lu et al., 2013]), and JZ52 (nif H- flhC-, nonflagellated and nonmotile [Lu et al., 2013]). Azotobacter vinelandii strains were grown on modified (no molybdenum) Burk’s medium plates (Strandberg and Wilson, 1968) with the addition of 0.013 M ammonium acetate at 30°C for 2 d before inoculation into liquid media of modified (no molybdenum, no iron) Burk’s medium with addition of 0.013 M ammonium acetate shaking at 170 rpm and 30°C for 18 to 20 h. This growth procedure was designed to induce competence for natural transformation assays and followed here to allow comparisons between transport and natural transformation behavior in A. vinelandii (Lu et al., 2013). The cells were then centrifuged at 1000 g for 10 min, decanted to remove the culture media, and resuspended in the same volume of 3-morpholinopropane-1-sulfonic acid (MOPS) buffer solution with 100 mM KCl at pH 7.2. No further washing was performed to avoid flagellar damage. Cells were then diluted five times in MOPS buffer with 100 mM KCl for use in motility characterization, micromodel, and RSPF experiments, resulting in a typical range of cell concentrations of 2.7 × 107 to 4.2 × 107 cells/mL as measured by hemocytometer counts. In addition to comparing these strains, where specified, the swimming motility in DJ was de-energized by the addition of the uncoupler Carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP) at a final concentration of 2.0 × 10-4 M (FCCPtreated DJ) (Turner et al., 2000).

Characterization of Cell Motility For characterization of motility under no-flow condition, 400 mL of the diluted cell suspension in MOPS buffer with 100 mM KCl was spread in a No. 0 glass bottom culture dish (MatTeck Corporation). The solution condition of 100 mM KCl was selected to investigate the role of flagellar motility and to facilitate surface attachment based on two previous studies. First, 100 mM KCl was found to be optimal for flagellar motion (de Kerchove and Elimelech, 2007). Second, enhanced column deposition was observed for A. vinelandii cells under 100 mM KCl in our previous work (Lu et al., 2013). A video composed of 1000 images with duration of 31 s was recorded for each strain using an inverted Axio Observer microscope (Carl Zeiss, Oberkochen) and a camera (Andor Technology iXon 897)

Journal of Environmental Quality 1367

controlled by Solis software (Andor Technology). The magnification was 40×, which gave an image with 0.4 mm per pixel size. The movement of individual cells was analyzed by particle tracking analysis, described below.

RSPF Setup and Experiments Radial stagnation point flow cell experiments were conducted as previously described (Liu et al., 2009). In brief, the surface was quartz (Quartz cover slip, Cat. No. #26016, Ted Pella Inc.), and the setup contained a 1-mm-radius injection capillary at 0.7 mm above the quartz surface as shown in Fig. 1A. The design enables real-time microscopic imaging of bacterial cells approaching and depositing on the front surface of porous media. Injection was controlled at a steady flow of 0.093 mL/min by a syringe pump (KD Scientific) to mimic groundwater flow conditions. Images and videos were recorded using the same setup described in the Characterization of Cell Motility section. Radial stagnation point flow cell experiments were conducted with the data collection designed to monitor cell movement at different distances above the surface. Videos were taken at the quartz surface and 20 and 60 mm above the quartz surface during separate RSPF experiments with each of the four strains/ conditions. The position of the focal plane was controlled using a three-dimensional automated microscope translation stage (Model: MS-2000XYZ, Applied Scientific Instrumentation). Each video was recorded for 31 s to contain sufficient cells for trajectory analyses at the fastest camera frame rate of 31 Hz. For a given strain/condition, sequential videos were collected at each of the three distances within 5 min.

Particle Tracking Analysis To investigate patterns of cell movement, particle tracking analysis was applied to the microscopic videos from the motility characterization. Cells were first identified by setting up a threshold for all 1000 images in a video. Each image was normalized based on the local average intensity to eliminate the effect of light intensity gradients (by subtracting the light intensity from a kernel smoothed version of the image with a large radius). This resulted in a uniform light intensity all over the image. The images were then transformed to binary images containing only the cells and the background. The thresholds related to microbe detection included minimum size, cutoff threshold, and a percentile parameter. The threshold parameters for bacterial detection were determined on the basis of the intensity distributions of background images and were adjusted to ensure that all of the cells were appropriately detected. The thresholds were chosen to include most of the bacterial cells on the focal plane according to comparisons between randomly selected original and the corresponding transformed images. An algorithm (available from the authors) was used to extract the locations of cells in each image. A C++ code was developed to infer the trajectory of each cell by considering their locations in consecutive images using a maximum likelihood approach. The code was verified extensively through visual comparison of the extracted trajectories with the videos. The image processing and trajectory analysis code is available on request. The mean square displacement (MSD) under no-flow (no advection) condition is calculated as (Qian et al., 1991) follows: 1368

Fig. 1. Characterization of cell trajectories from radial stagnation point flow cell (RSPF) experiments. The RSPF design and the observation distances above collector surface are marked in (A). DJ is motile; FCCP-treated DJ cells are nonmotile. Twenty trajectories for each type of cell were selected randomly from RSPF time-series images and analyzed for (B) slope and (C) linearity distribution of these cell trajectories moving outward from the center of the images radially. This analysis was conducted at distances of 20 and 60 mm above the surface. The average slopes of all conditions centered around 0, with approximately equal expansion into negative and positive regions, indicating movement in the radial direction. The r2 of the trajectories averaged close to 1, also suggesting that the cells were moving with the flow field at these distances above the surface.

MSD(t ) =

2 2ù 1 m é ê( x t , i - x0, i ) + ( yt , i - y0, i ) ú å û m i=1 ë

where m is the number of trajectories identified, and xt,i and yt,i are the location of cell i (i.e., trajectory i) at time t. The MSD functions of time are fitted by a power function in the form Journal of Environmental Quality

of MSD(t) » atb. The values of a and b indicate the motion behavior of the cells. A value of b = 1 indicates purely Brownian diffusion, whereas b < 1 shows subdiffusion and b > 1 shows superdiffusion. A b = 2 indicates ballistic motion: almost straight trajectory with little or no change in cell velocities during the movement (Conrad et al., 2011). In the analysis, some of the cells showed no significant movement, possibly due to attachment to the surface. These cells were excluded from the analysis. The instantaneous velocity magnitude distributions and turn angles were also calculated from all the trajectories from the same strain using C++ code. The instantaneous velocity magnitude was obtained using cell displacements over consecutive intervals. The turn angle for each trajectory at each time step was calculated using the position of the bacteria at that time and the position at the previous and the following time-steps. To compare cell trajectories in the RSPF videos from different distances, 20 random trajectories with more than six moving steps were selected for each distance. This analysis was conducted for DJ and FCCP-treated DJ and at distances of 20 and 60 mm above the surface. The total number of trajectories recognized in DJ at 20 mm was 20, which was the smallest number in all conditions. Therefore, the number was selected as the sample size to maintain the same sample depth in all the conditions compared. Linearity and direction were analyzed based on the slopes and the coefficients of determination r2 of these trajectories, as calculated using general linear regressions (Nelder and Baker, 1972). Cell trajectories at the RSPF collector surface were illustrated in superimposed images from the videos.

Cluster Analysis From the trajectory analysis of the motile strains DJ and DJ77, at any given time, some of the cells showed passive Brownian motion, while other cells were actively swimming. To extract the statistical parameters of actively moving cells, a cluster-based statistical analysis was used to separate the trajectories of actively and passively moving cells, based on the overall turn angle standard deviation and mean velocity magnitude obtained from each trajectory. The analysis was done using statistical measures extracted from the instantaneous turn angle (i.e., change of direction at each trajectory segment) and instantaneous velocity. In particular, the mean and standard deviation of instantaneous turn angles and velocities for each trajectory was used. K-means method (MacQueen, 1967) was further applied to group different motility behaviors for the strains wild-type DJ and DJ77. It is based on the following minimization expression: k

arg min å å x j -m i s

2



i =1 xi Î si

where k is the number of cluster centroids, x is a two dimensional real vector containing the standard deviations of turn angles and the mean velocity for the trajectory of each cell; s indicates the subset of swimming, nonswimming, and mixed-behavior clusters, and m is the centroid for each cluster. The goal is to find the optimal clusters in a way that the sum of squared distances of all points (each point representing one trajectory) from their corresponding clusters’ centroids is minimized and thereby to identify the members in each cluster group. A two-phase

K-means algorithm was used to perform the clustering. Using the corresponding cluster centroids, the clusters were separated and the instantaneous velocity magnitude distributions and turn angles from cluster groups of both DJ and DJ77 were computed.

Two-Dimensional Micromodel Fabrication and Experiments Silicon micromodels of porous media were used to visualize and quantify bacterial deposition in two dimensions under flow conditions. Fabrication, inoculation, and operation of the micromodels were described previously (Lu et al., 2013). A regular pattern of 1440 cylindrical collectors (180 mm diameter, 114 mm in pore-body size, 28 mm in pore-throat size, porosity 0.4) was etched to 22 mm depth in a Si wafer as depicted in Fig. 2A. The design was to represent the porous media structure and facilitate microscopic imaging of bacterial deposition. Influent was injected at a Darcy’s velocity of 0.0002 m/s to represent groundwater flow conditions. Images were collected using a Leica microscope connected to a charge-coupled device (CCD) camera (Qimaging Retiga 2000R Fast 1394) and Image Pro 7.0 plus software. To quantify the distribution of bacterial deposition along the flowpath, the micromodel was divided into 18 equal portions (each portion contains 16 images across the width) from the inlet to the outlet. We collected 18 (longitudinal) × 16 (latitudinal) gray-scale images of the entire micromodel collector field at 2 mm resolution to capture one whole collector in one image and recognize bacterial cells with diameters of 3.5 ± 0 mm as reported previously (Lu et al., 2013). The deposited bacterial cells on each collector were manually counted using these images. The single collector removal efficiency (h) in each of the 18 portions (longitudinal h) was calculated as the ratio of the average rate of attachment per single collector divided by the rate of cells approaching one collector. The average rate of attachment per single collector was calculated as the ratio of total attached cells in the region of interest divided by the product of the number of collectors in the same region and the experimental time (Liu et al., 2012). The rate of cells approaching one collector was defined as the product of the approaching flow projected area on one collector (the product of collector height and collector diameter), Darcy velocity, and influent cell concentration (Liu et al., 2012). To further evaluate the location of the deposition, an original approach was developed in this work to determine individual values of h for the front (upstream surfaces) and back (downstream surfaces) of the collectors, using a line that bisected each collector perpendicular to the flow direction. The longitudinal h and the h for the front and back of the collectors provided critical information about bacterial deposition behavior on the collectors as compared to reporting the time series of the overall h, which revealed the deposition dynamics (Lu et al., 2013). The results shown in Fig. 2 come from a single micromodel experiment for each of the three strains. To improve the statistical resolution in the analysis of longitudinal and front/back deposition, duplicate micromodel experiments were analyzed for all three strains. For DJ77 and JZ52, the second set of data came from additional analysis conducted on existing micromodel images (Lu et al., 2013). The previous analysis of these experiments

Journal of Environmental Quality 1369

included only changes in overall h from 5 to 15 min (Lu et al., 2013). In the current work, the same images were analyzed for longitudinal h and deposition locations. Because strain DJ was not included in the prior work, an additional micromodel experiment was conducted to provide a second replicate for this strain. The dataset for longitudinal h of 108 data points, combining two replicates for each strain, was used in two-sample t test (p = 0.05) to compare among the strains. A t test (p = 0.05) was also used to compare the h values of up to 108 data points obtained for the front and the back of the collectors among the strains, respectively.

Results and Discussion Motility Characterization under No-Flow Conditions

Fig. 2. Effects of motility on deposition in porous media micromodel experiments. (A) Micromodel top view and design of the collectors. (B) Single collector removal efficiency (h) for the three bacterial strains as averaged across each column (perpendicular to flow direction) along the flowpath of the micromodel (referred to as longitudinal h). (C) The distribution of deposition between the front or upstream (solid symbols) and back or downstream (open symbols) sides of the collectors. 1370

To study the effect of swimming motility on transport, it was necessary to have a detailed understanding of the motility of the strains and conditions used here. We therefore characterized the trajectories of individual cells from each strain or condition: flagellated wild-type DJ, flagellated DJ77 cells with impaired motility, nonflagellated and nonmotile JZ52 cells, and FCCP de-energized, nonmotile DJ cells. First, a global analysis of trajectories was conducted, allowing comparison with previous results for two of the strains, DJ77 and JZ52. Similar to the previous analyses (Lu et al., 2013), we found two types of movement: active swimming, as indicated by relatively smooth paths, and random Brownian motion, which exhibited more erratic movement with a lower overall velocity. Representative trajectories of all four strains/conditions are shown in Fig. 3 to provide a visual comparison of their motilities. DJ and DJ77 trajectories exhibited both actively swimming motion and Brownian motion characteristics. All trajectories of JZ52 and FCCPtreated DJ cells showed only Brownian motion. The distributions of the velocity magnitude for DJ and DJ77 had longer tails due to the higher velocities, indicative of active swimming (Fig. 4). DJ had an average instantaneous swimming speed of 13.1 mm/s. Flagellated DJ77 displayed a lower average speed of 8.7 mm/s, providing a quantitative measure of its impaired swimming ability. JZ52 and FCCP-treated DJ showed narrower velocity ranges centering around 0 to 5.0 mm/s. The MSD calculated from trajectories is a common metric for quantifying particle movement and evaluating the differences in the nature of the movement and is defined as the average of the squared displacements of a number of Journal of Environmental Quality

Fig. 3. Trajectories of individual cells from each strain/condition. Four strains/conditions showed distinctive motilities from the trajectories. DJ is a wild-type strain, while DJ77 has impaired motility. JZ52 lacks motility and flagella, and FCCP treatment de-energizes DJ cells, rendering them nonmotile. DJ and DJ77 trajectories exhibited both actively swimming motion and Brownian motion characteristics. All trajectories of JZ52 and FCCP-treated DJ cells showed only Brownian motion. The motility tests were done in MOPS buffer with 100 mM KCl, pH 7.2. Videos were taken at a frequency of 31 Hz.

bacteria as a function of time (Qian et al., 1991). The MSD is plotted versus time in log-log scale in Supplemental Fig. S1 for all four strains/conditions (DJ, DJ77, JZ52, and FCCP-treated DJ). The slopes of these fitted lines (b values) can be used to distinguish between random diffusive motion with slope of 1.0 and ballistic motion with slope of 2.0 (Conrad et al., 2011). The slopes for all strains were larger than 1, indicating movement that deviated from true Brownian motion. However, for DJ and DJ77, the movement was relatively ballistic, with slopes of 1.5 and 1.6, respectively. The movements of JZ52 and FCCP-treated DJ cells were closer to Brownian motion, with slopes of 1.2 and 1.3, respectively. The similar and low values for JZ52 and FCCP-treated DJ confirmed that both types of cells were not actively motile. The MSDs for these three strains showed the same trend as previously reported (Lu et al., 2013), although the values obtained here were slightly higher. Mean square displacement alone does not provide a clear description of the differences between strains DJ and DJ77. To characterize and quantify the differences in swimming motility at the level of individual cells between the two strains, cluster-based statistical analysis was applied to separate swimming and nonswimming cell trajectories for further analysis (Fig. 5). This analysis enabled the separation of the actively swimming motility from passive modes of motion and allowed us to determine the effect of actively swimming motility on the bacterial surface deposition. The turn angle was defined as the

change in the direction of the bacterial cells in each image. Based on the turn angle standard deviation and the mean velocity magnitude of each cell, each trajectory was classified into one of three clusters: swimming, nonswimming, and mixed-behavior groups (Fig. 5). In the swimming clusters, the velocity distribution was skewed toward higher values for DJ (average 25.6 mm/s) compared with DJ77 (average 17.4 mm/s). The discrete velocity distribution of the DJ77 swimming cluster and similar relative frequency of the swimming cluster, as compared to those for DJ cells, suggest that DJ77’s impaired motility was due to a lower swimming speed. The velocity distributions for nonswimming clusters of both DJ and DJ77 (Fig. 5A) were very similar to the overall velocity distributions for nonmotile strains/conditions (Fig. 4A, JZ52 and FCCP-treated DJ); these low velocities were due to Brownian motion. The motility analyses confirmed that DJ77 has impaired motility and provided new insight into the type of defect it exhibits. DJ77 cells showed slower movement than wild-type DJ cells, as observed visually in Fig. 3 and quantified by the cluster-based analysis. The turn angle distribution for swimming DJ was also the narrowest (i.e., had the smallest variance) compared with the JZ52 and FCCP-treated DJ cells undergoing solely random Brownian motion (Fig. 4B). Taken together, the four strains/conditions provide a motility gradient from strong swimming motility in DJ, to impaired motility in DJ77, and to the nonmotile cells JZ52 and FCCP-treated DJ.

Movement of Bacterial Cells in the RSPF Setup under Flow Conditions The four strains/conditions (DJ, DJ77, JZ52, and FCCPtreated DJ) were studied in an RSPF setup. In contrast to previous work quantifying cell deposition on the RSPF collector surface (Lu et al., 2013), our objective here was to identify the swimming patterns that affected deposition; so we collected data at different distances above the RSPF collector surface and analyzed the movement of individual cells at each distance. For swimming DJ and nonswimming, FCCP-treated DJ cells at 20 and 60 mm above the surface, 20 trajectories were randomly selected to examine the significance of swimming under flow conditions. The FCCP-treated DJ was selected for comparison with the swimming DJ strain because both types of cells have flagella. The trajectories of cells from both strains were all spreading out from the centers of the images in radial directions, as illustrated by representative trajectories at 20 mm above the surface in Supplemental Fig. S2. The slopes and coefficients of determination r2 for the general linear regressions of these trajectories were compared to analyze their linearity and directions (Fig. 1). The slope values for both strains at both focal planes ranged from –5.5 to +5.9 and centered at 0. A positive slope means that the cell traveled away from the center at angles between 0° and 90° or between 180° and 270° (Supplemental

Journal of Environmental Quality 1371

followed the straight radial streamlines in the RSPF as shown in Supplemental Fig. S2. On the surface, however, there were distinct differences between DJ and FCCP-treated DJ. The movement of DJ cells on the surface (Fig. 6A and B, numbered arrows) differed from the flow direction (Fig. 6, plain arrows in radial directions) and included both horizontal and vertical movement, as illustrated in Fig. 6A and B, respectively. Cells 1 and 3 swam in straight lines, deviating from the flow at average speeds of 8.8 and 18.5 mm/s, respectively. Cell 2 swam at an average speed of 16.3 mm/s, first in a straight line following the flow and then turning and deviating from the flow path. Cell 4 displayed another type of movement with compact trajectories. Vertical movement is indicated by the light intensity of the trajectories; since the camera is underneath the surface, brighter cells are closer to the surface and camera, while weaker intensity shows cells that are farther away from the surface. Both Cells 1 and 2 came in and out of focus but remained at a close proximity to the surface. Compared with swimming DJ, the FCCP-treated DJ cells consistently moved radially from the center following the flow and did not display vertical motion, as illustrated in Fig. 6C. Thus, we propose that the movement against the flow could reduce the attachment of swimming DJ cells to the collector surface. Although not directly examined here, these patterns are likely to require tumbling, which is also thought to contribute to E. coli cells leaving the horizontal plane after circling near a surface (Frymier et al., 1995). In these RSPF experiments, we found that motile cells showed different patterns of movement from the nonmotile cells. The movement on the surfaces suggested that swimming motility could reduce cell deposition on the front stagnation point. This finding encouraged us to pursue the investigation of how swimming motility affects bacterial deposition and transport in twodimensional porous media micromodels.

Bacterial Deposition in the Micromodel Setup

Fig. 4. Distributions of motility parameters for Azotobacter vinelandii strains. Motility parameters include the relative frequency distributions of (A) instantaneous velocities and (B) turn angles. The distributions of the velocity magnitude for DJ and DJ77 had longer tails due to the higher velocity indicative of active swimming. JZ52 and FCCP-treated DJ showed narrower velocity ranges centering around 0–5.0 mm/s. The motility tests were done in the same buffer solution as transport experiments: MOPS buffer with 100 mM KCl, pH 7.2. Videos were taken at a frequency of 31 Hz.

Fig. S2). A negative slope means that the cell traveled away from the center at angles between 90° and 180° or between 270° and 360° (Supplemental Fig. S2). The equal extension of slope values into both negative and positive regions in Fig. 1B indicated that the trajectories spread out evenly in radial directions (arrows marked in Supplemental Fig. S2). The distributions of the slope values obtained for trajectories at the 20 and 60 mm focal planes were not statistically different between DJ and FCCP-treated DJ cells. This observation suggested that the flow overcame swimming motility and the direction of movement did not deviate from the flow at these locations. The values of r2 in the general linear regression were close to 1 for all cases (Fig. 1C), indicating that at distances ≥20 mm above the surface, bacterial cells 1372

A two-dimensional micromodel setup was used to observe the role of swimming motility on deposition in porous media under flow conditions. Differences in h were observed along the longitudinal direction for all three strains (Fig. 2B); the longitudinal h for DJ were significantly smaller than those of DJ77 and JZ52 (t70 [degree of freedom = 70] = -5.26, p < 0.05, and t70 = -5.58, p < 0.05, respectively). For all strains, more deposition occurred on the collectors near the inlets and outlets compared with those in the middle of micromodels (Fig. 2B), which in agreement with previous studies with Cryptosporidium parvum oocysts (Liu et al., 2012). The deposition was also quantified separately for the front and back of the collectors, relative to the flow in the micromodel experiments (Fig. 2C). The longitudinal h values determined for the back of the collectors were overlapping and in the range from 1.4 × 10-5 to 2.7 × 10-4 for motile DJ, DJ77 with impaired motility, and nonswimming JZ52 cells. In contrast, on the front of the collectors, the h for DJ were significantly smaller than those for DJ77 and JZ52 (t70 = -5.86, p < 0.05, and t70 = -5.82, p < 0.05, respectively). JZ52 gave hforward of 5.4 × 10-4 and 1.1 × 10-3, at least five times larger than those for DJ. This micromodel trend of lower hforward for the motile strain agreed with the lower deposition suggested by surface movement patterns in RSPF experiments. This is also consistent with some Journal of Environmental Quality

Fig. 5. Cluster analysis-derived motility parameters for cells exhibiting a specific type of movement. Cluster analysis was used to divide DJ and DJ77 cell trajectories into three clusters: swimming, nonswimming, and mixed behavior. Motility parameters were calculated separately for each group, with the relative frequency distributions of (A) instantaneous velocities and (B) turn angles shown here. The velocity distribution for the swimming cluster was skewed toward higher values for DJ (average 25.6 mm/s) compared with DJ77 (average 17.4 mm/s). The low velocities for nonswimming clusters of both DJ and DJ77 were due to Brownian motion. To provide a visual representation of the effectiveness of the clustering, (C) representative trajectories from each cluster are also shown for strain DJ.

previous studies showing lower deposition with motile bacterial cells under certain experimental conditions (Camesano and Logan, 1998; McClaine and Ford, 2002a). However, it contradicts two previous RSPF studies (de Kerchove and Elimelech, 2008a; Haznedaroglu et al., 2010). In one case (de Kerchove and Elimelech, 2008a), the discrepancy is likely due to the use of different surfaces (silica dioxide in the current study vs. positively charged poly-L-lysine coated surface) and/or different bacteria (A. vinelandii here vs. Pseudomonas aeruginosa). As compared to

Haznedaroglu et al. (Haznedaroglu et al., 2010), the discrepancy could be attributed to their higher flow rate (1.5 mL/min vs. 0.093 mL/min in this study) or to differences among bacteria, as Salmonella strains were used in their work. The micromodel experiments further indicate that the difference in overall deposition is due to differences in front-side deposition. Because all three strains shared similar electrophoretic mobilities at ionic strengths ranging from 1 to 200 mM KCl (Supplemental Fig. S3), electrostatic interactions alone cannot

Journal of Environmental Quality 1373

explain the lower attachment of DJ. Several previous studies suggested that motile cells are mainly involved in reversible surface association rather than irreversible attachment (Kusy and Ford, 2009; Liu and Ford, 2009; Liu et al., 2011; Narayanaswamy et al., 2009). Attachment at secondary minimum was suggested to be the cause of backward deposition by colloids and bacterial cells, reversibly or weakly attaching to the front of a collector surface and rolling along the surface to deposit on the back of the collector (Elimelech and O’Melia, 1990; Hermansson, 1999, Kuznar and Elimelech, 2007). The secondary minima for all three A. vinelandii strains at the experimental ionic strength of 100 mM KCl were calculated in this study using Derjaguin–Landau– Verwey–Overbeek (DLVO) theory as developed and applied for bacterial adhesion (Redman et al., 2004). For all strains, the secondary minima had a depth of 24.9 kT with a corresponding separation distance of 4.5 nm. As a result, it is likely that all three strains overcame secondary minima at similar rates. However, we did not observe higher deposition on the back side of the collectors for all strains. Therefore, permanent primary attachment was expected, rather than the reversible or weak attachment to the front of a collector surface, which led to backward deposition. In addition, we did not observe a significant difference in deposition between DJ77 and JZ52 in micromodel experiments, suggesting that under the experimental conditions used here, the presence of the flagella without motility was not sufficient to change bacterial deposition and transport. Therefore, we propose that swimming motility controlled the surface attachment of the motile cells of A. vinelandii under well-controlled environmentally relevant hydrodynamics based on the direct microscopic evidence obtained in this study.

Conclusions This work used microscopically resolved, population-level quantification of individual cell movement to study bacterial transport and deposition in porous media. This approach could also be applied to study other bacterial behaviors such as chemotaxis and gene transfer. Swimming motility reduced surface attachment of A. vinelandii under environmentally relevant hydrodynamic conditions in this study. This swimming motility– surface attachment relationship, also suggested by other studies (Conrad et al., 2011; McClaine and Ford, 2002a, 2002b), may help these bacterial cells travel farther and spread more broadly. Bioremediation processes could benefit from broader transport of bacterial cells to reach a larger area for contaminant treatment. On the other hand, the complexity of pathogen monitoring and control increases with broader transport in soil and groundwater environments.

Acknowledgments Fig. 6. Representative cell trajectories observed on the radial stagnation point flow cell (RSPF) collector surface, depicting (A) horizontal and (B) vertical movement of swimming DJ cells compared with (C) movement in the flow direction by FCCP-treated DJ cells on the RSPF collector surface. White arrows indicate the flow direction within RSPF flow chambers; the yellow arrows mark the direction of movement for individual cells. Cells 1 and 3 in Panel A swam horizontally in straight lines, while cells 2 and 4 also moved horizontally but changed directions. As shown in Panel B, both Cells 1 and 2 moved vertically but remained at a close proximity to the surface. Compared with swimming DJ, the FCCP-treated DJ cells in Panel C consistently moved radially from the center following the flow and did not display horizontal or vertical motion. 1374

This work was supported by NSF Grants No. 1114385, 1215756, and 1114257, and part of the research was conducted in the William R. Wiley Environmental Molecular Sciences Laboratory, a scientific user facility of the US Department of Energy’s Office of Biological and Environmental Research and operated by the Pacific Northwest National Laboratory. This publication was made possible by research supported by grant R834870 from the US Environmental Protection Agency (USEPA). Its contents are solely the responsibility of the grantee and do not necessarily represent the official views of the USEPA. Further, the USEPA does not endorse the purchase of any commercial products or services mentioned in the publication. Journal of Environmental Quality

References Bradford, S.A., V.L. Morales, W. Zhang, R.W. Harvey, A.I. Packman, A. Mohanram, et al. 2013. Transport and fate of microbial pathogens in agricultural settings. Crit. Rev. Environ. Sci. Technol. 43:775–893. doi:10.1080/1064 3389.2012.710449 Camesano, T.A., and B.E. Logan. 1998. Influence of fluid velocity and cell concentration on the transport of motile and nonmotile bacteria in porous media. Environ. Sci. Technol. 32:1699–1708. doi:10.1021/es970996m Castro, F.D., and N. Tufenkji. 2008. Role of oxygen tension on the transport and retention of two pathogenic bacteria in saturated porous media. Environ. Sci. Technol. 42:9178–9183. doi:10.1021/es801677f Conrad, J.C., M.L. Gibiansky, F. Jin, V.D. Gordon, D.A. Motto, M.A. Mathewson, et al. 2011. Flagella and pili-mediated near-surface single-cell motility mechanisms. Biophys. J. 100:1608–1616. doi:10.1016/j.bpj.2011.02.020 de Kerchove, A.J., and M. Elimelech. 2007. Impact of alginate conditioning film on deposition kinetics of motile and nonmotile Pseudomonas aeruginosa strains. Appl. Environ. Microbiol. 73:5227–5234. doi:10.1128/ AEM.00678-07 de Kerchove, A.J., and M. Elimelech. 2008a. Bacterial swimming motility enhances cell deposition and surface coverage. Environ. Sci. Technol. 42:4371–4377. doi:10.1021/es703028u de Kerchove, A.J., and M. Elimelech. 2008b. Calcium and magnesium cations enhance the adhesion of motile and nonmotile Pseudomonas aeruginosa on alginate films. Langmuir 24:3392–3399. doi:10.1021/la7036229 Elimelech, M., and C.R. O’Melia. 1990. Kinetics of deposition of colloidal particles in porous media. Environ. Sci. Technol. 24:1528–1536. doi:10.1021/ es00080a012 Frymier, P.D., and R.M. Ford. 1997. Analysis of bacterial swimming speed approaching a solid-liquid interface. AIChE J. 43:1341–1347. doi:10.1002/ aic.690430523 Frymier, P.D., R.M. Ford, H.C. Berg, and P.T. Cummings. 1995. Three-dimensional tracking of motile bacteria near a solid planar surface. Proc. Natl. Acad. Sci. USA 92:6195–6199. doi:10.1073/pnas.92.13.6195 Ginn, T.R., B.D. Wood, K.E. Nelson, T.D. Scheibe, E.M. Murphy, and T.P. Clement. 2002. Processes in microbial transport in the natural subsurface. Adv. Water Resour. 25:1017–1042. doi:10.1016/S0309-1708(02)00046-5 Haznedaroglu, B.Z., O. Zorlu, J.E. Hill, and S.L. Walker. 2010. Identifying the role of flagella in the transport of motile and nonmotile Salmonella enterica serovars. Environ. Sci. Technol. 44:4184–4190. doi:10.1021/es100136m Hermansson, M. 1999. The DLVO theory in microbial adhesion. Colloids Surf. B Biointerfaces 14:105–119. doi:10.1016/S0927-7765(99)00029-6 Hunter, K.S., Y. Wang, and P. Van Cappellen. 1998. Kinetic modeling of microbially-driven redox chemistry of subsurface environments: Coupling transport, microbial metabolism and geochemistry. J. Hydrol. 209:53–80. doi:10.1016/S0022-1694(98)00157-7 Jacobson, M.R., K.E. Brigle, L.T. Bennett, R.A. Setterquist, M.S. Wilson, V.L. Cash, et al. 1989. Physical and genetic map of the major Nif gene cluster from Azotobacter vinelandii. J. Bacteriol. 171:1017–1027. Kusy, K., and R.M. Ford. 2009. Surface association of motile bacteria at granular porous media interfaces. Environ. Sci. Technol. 43:3712–3719. doi:10.1021/es8033632 Kuznar, Z.A., and M. Elimelech. 2007. Direct microscopic observation of particle deposition in porous media: Role of the secondary energy minimum. Colloids Surf. A Physicochem. Eng. Asp. 294:156–162. doi:10.1016/j. colsurfa.2006.08.007 Liu, J., and R.M. Ford. 2009. Idling time of swimming bacteria near particulate surfaces contributes to apparent adsorption coefficients at the macroscopic scale under static conditions. Environ. Sci. Technol. 43:8874–8880. doi:10.1021/es901865p Liu, J., R.M. Ford, and J.A. Smith. 2011. Idling time of motile bacteria contributes to retardation and dispersion in sand porous medium. Environ. Sci. Technol. 45:3945–3951. doi:10.1021/es104041t Liu, Y., D. Janjaroen, M.S. Kuhlenschmidt, T.B. Kuhlenschmidt, and T.H. Nguyen. 2009. Deposition of Cryptosporidium parvum oocysts on natural organic matter surfaces: Microscopic evidence for secondary minimum deposition in a radial stagnation point flow cell. Langmuir 25:1594–1605. doi:10.1021/la803202h Liu, Y., C. Zhang, M. Hilpert, M.S. Kuhlenschmidt, T.B. Kuhlenschmidt, and T.H. Nguyen. 2012. Transport of Cryptosporidium parvum oocysts in a silicon micromodel. Environ. Sci. Technol. 46:1471–1479. doi:10.1021/ es202567t Lorenz, M.G., and W. Wackernagel. 1994. Bacterial gene transfer by natural genetic transformation in the environment. Microbiol. Rev. 58:563–602.

Lu, N., T. Bevard, A. Massoudieh, C. Zhang, A.C. Dohnalkova, J.L. Zilles, et al. 2013. Flagella-mediated differences in deposition dynamics for Azotobacter vinelandii in porous media. Environ. Sci. Technol. 47:5162–5170. doi:10.1021/es3053398 MacQueen, J. 1967. Some methods for classification and analysis of multivariate observations. In: L.M. Le Cam and J. Neyman, editors, Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. University of California Press, Berkeley, CA. p. 281–297. McClaine, J.W., and R.M. Ford. 2002a. Characterizing the adhesion of motile and nonmotile Escherichia coli to a glass surface using a parallel-plate flow chamber. Biotechnol. Bioeng. 78:179–189. doi:10.1002/bit.10192 McClaine, J.W., and R.M. Ford. 2002b. Reversal of flagellar rotation is important in initial attachment of Escherichia coli to glass in a dynamic system with high- and low-ionic-strength buffers. Appl. Environ. Microbiol. 68:1280– 1289. doi:10.1128/AEM.68.3.1280-1289.2002 Murphy, E.M., and T.R. Ginn. 2000. Modeling microbial processes in porous media. Hydrogeol. J. 8:142–158. doi:10.1007/s100409900043 Narayanaswamy, K., R.M. Ford, J.A. Smith, and E.J. Fernandez. 2009. Surface association of motile bacteria and apparent tortuosity values in packed column experiments. Water Resour. Res. 45:W07411. doi:10.1029/2008WR006851 Nelder, J.A., and R. Baker. 1972. Generalized linear models. Wiley Online Library. Page, W.J., and M. von Tigerstrom. 1979. Optimal conditions for transformation of Azotobacter vinelandii. J. Bacteriol. 139:1058–1061. Qian, H., M.P. Sheetz, and E.L. Elson. 1991. Single particle tracking: Analysis of diffusion and flow in two-dimensional systems. Biophys. J. 60:910–921. doi:10.1016/S0006-3495(91)82125-7 Redman, J.A., S.L. Walker, and M. Elimelech. 2004. Bacterial adhesion and transport in porous media: Role of the secondary energy minimum. Environ. Sci. Technol. 38:1777–1785. doi:10.1021/es034887l Sen, T.K., D. Das, K.C. Khilar, and G.K. Suraishkumar. 2005. Bacterial transport in porous media: New aspects of the mathematical model. Colloids Surf. A Physicochem. Eng. Asp. 260:53–62. doi:10.1016/j.colsurfa.2005.02.033 Strandberg, G., and P. Wilson. 1968. Formation of the nitrogen-fixing enzyme system in Azotobacter vinelandii. Can. J. Microbiol. 14:25–31. doi:10.1139/m68-005 Trevors, J., T. Barkay, and A. Bourquin. 1987. Gene transfer among bacteria in soil and aquatic environments: A review. Can. J. Microbiol. 33:191–198. doi:10.1139/m87-033 Tufenkji, N. 2007. Modeling microbial transport in porous media: Traditional approaches and recent developments. Adv. Water Resour. 30:1455–1469. doi:10.1016/j.advwatres.2006.05.014 Turner, L., W.S. Ryu, and H.C. Berg. 2000. Real-time imaging of fluorescent flagellar filaments. J. Bacteriol. 182:2793–2801. doi:10.1128/ JB.182.10.2793-2801.2000 Vigeant, M., and R. Ford. 1997. Interactions between motile Escherichia coli and glass in media with various ionic strengths, as observed with a three-dimensional-tracking microscope. Appl. Environ. Microbiol. 63:3474–3479. Walker, S.L., J.A. Redman, and M. Elimelech. 2005. Influence of growth phase on bacterial deposition: Interaction mechanisms in packed-bed column and radial stagnation point flow systems. Environ. Sci. Technol. 39:6405– 6411. doi:10.1021/es050077t Wang, M., and R.M. Ford. 2010. Quantitative analysis of transverse bacterial migration induced by chemotaxis in a packed column with structured physical heterogeneity. Environ. Sci. Technol. 44:780–786. doi:10.1021/ es902496v Wang, M., R.M. Ford, and R.W. Harvey. 2008. Coupled effect of chemotaxis and growth on microbial distributions in organic-amended aquifer sediments: Observations from laboratory and field studies. Environ. Sci. Technol. 42:3556–3562. doi:10.1021/es702392h Wang, X., T. Long, and R.M. Ford. 2012. Bacterial chemotaxis toward a NAPL source within a pore-scale microfluidic chamber. Biotechnol. Bioeng. 109:1622–1628. doi:10.1002/bit.24437 Wang, Y., S.A. Bradford, and J. Šimůnek. 2013. Transport and fate of microorganisms in soils with preferential flow under different solution chemistry conditions. Water Resour. Res. 49:2424–2436. doi:10.1002/wrcr.20174 Wu, M., J.W. Roberts, S. Kim, D.L. Koch, and M.P. DeLisa. 2006. Collective bacterial dynamics revealed using a three-dimensional population-scale defocused particle tracking technique. Appl. Environ. Microbiol. 72:4987– 4994. doi:10.1128/AEM.00158-06

Journal of Environmental Quality 1375