Analysis of Bacterial Random Motility in a Porous Medium Using

Jan 8, 2003 - Department of Chemical Engineering, Bucknell University,. Lewisburg, Pennsylvania 17837, and Department of. Civil Engineering, Universit...
1 downloads 0 Views 144KB Size
Environ. Sci. Technol. 2003, 37, 781-785

Analysis of Bacterial Random Motility in a Porous Medium Using Magnetic Resonance Imaging and Immunomagnetic Labeling JULI L. SHERWOOD,‡ JAMES C. SUNG,‡ R O S E A N N E M . F O R D , * ,† ERIK J. FERNANDEZ,† JAMES E. MANEVAL,§ AND JAMES A. SMITH# Program of Interdisciplinary Research in Contaminant Hydrogeology, Department of Chemical Engineering, University of Virginia, Charlottesville, Virginia 22904-4741, U.S. Department of Defense, Charlottesville, Virginia 22911, Department of Chemical Engineering, Bucknell University, Lewisburg, Pennsylvania 17837, and Department of Civil Engineering, University of Virginia, Charlottesville, Virginia 22904-4741

In this study, we demonstrate the application of immunomagnetic labeling and magnetic resonance imaging (MRI) for the noninvasive visualization of changes in bacterial density distributions as a function of time in a water-saturated porous medium. Magnetite particles (50-60 nm diameter) were attached via a monoclonal antibody to the surface of Escherichia coli K12 NR50 cells. The cells maintained their motility after labeling, and the presence of the magnetite did not significantly alter cell swimming speed. Diffusive migration for both motile and nonmotile E. coli through a porous medium with a particle-diameter distribution of 250300 µm was compared. The movement of the nonmotile cells was described by an effective random motility coefficient consistent with Brownian diffusion of a nonmotile colloid. An effective coefficient determined a priori from bacterial motility in an aqueous medium and properties of the porous medium adequately described the movement of the motile cells. The ability to noninvasively visualize bacterial concentrations within an opaque porous medium in real time provides researchers with a powerful tool for studying bacterial transport in porous media. This is important for understanding the impact of bacterial transport on remediation strategies for environmental cleanup of polluted groundwater.

Introduction The distribution and transport of bacteria in the subsurface environment are critical factors in the application of engineered technologies in situ. For example, successful implementation of both in situ bioremediation and microbially * Corresponding author phone: (434)924-6283; fax: (434)982-2658; e-mail: [email protected]. Mailing address: Department of Chemical Engineering, University of Virginia, 102 Engineers’ Way, P.O. Box 400471, Charlottesville, VA 22904-4741. † Department of Chemical Engineering, University of Virginia. ‡ U.S. Department of Defense. § Bucknell University. # Department of Civil Engineering, University of Virginia. 10.1021/es011210u CCC: $25.00 Published on Web 01/08/2003

 2003 American Chemical Society

enhanced oil recovery is dependent on the movement and accumulation of bacteria in the small interstices of the surrounding porous matrix (1, 2). In advection-dominated systems, it is usually assumed that the effects of random motility will be negligible compared to advection. Yet, many bacteria can swim through aqueous media at speeds (20-40 µm/s) (3-5) greater than typical groundwater flow rates (0.10.5 m/d or about 1-5 µm/s) in aquifers (6, 7). Motility may allow microorganisms to penetrate and degrade contaminants effectively in regions of low subsurface permeability. Because the run-and-tumble motion of motile bacteria resembles the random diffusive behavior of gas molecules (8), it is often characterized by the aqueous phase random motility coefficient, µo, which is analogous to a molecular diffusion coefficient in the gas phase. However, the restricted geometry of a porous medium disrupts the natural swimming behavior of bacteria, so that µo is no longer valid and must be replaced by an effective random motility coefficient, µeff. While analogy to gas diffusion provides a theoretical basis for analysis of bacterial random motility, the predictions do not always agree with experimental results (9). Analyses of bacterial migration in porous media are most often based on packed column studies where breakthrough curves of the effluent (10-12) are the only data available. Profiles which include attached bacteria also are frequently obtained by sectioning the column and counting the bacteria within each section (10, 13-15), but this is a time-consuming technique that provides spatial information for only a single point in time. Several researchers have shown that magnetic resonance imaging (MRI) can be used to quantify bacterial populations, both in suspension (16, 17) and in porous media (18), but only at relatively high (1010 cells/mL) concentrations. However, by labeling the cells with magnetite particles prior to imaging, we can detect concentrations on the order of 108109 cells/mL, dilute enough so that bacterial cells migrating independently of one another can be studied. Magnetically susceptible nuclei placed in a static magnetic field align with the field (19). Application of a radio frequency (RF) pulse can excite the nuclei and force them out of alignment with the static field. Relaxation of the nuclei from the excited state proceeds via first-order processes characterized by the spin-lattice (T1) and spin-spin (T2) relaxation times (19). As the nuclei return to their equilibrium state, the net magnetization remaining out of alignment can be measured using the same antenna that applied the initial RF pulse. The strength of the signal at a particular location is a function of the concentration of the nuclei, the relaxation times, and the timing of the signal acquisition (TE time). The presence of a superparamagnetic contrast agent, such as magnetite, causes decreases in both T1 and T2 values in a manner inversely proportional to its concentration; however, the effect on T2 is usually much greater (20). Thus, changes in T2 values can be related to changes in the concentration of magnetite and, in the case where the magnetite is attached to motile bacterial cells, to changes in bacterial cell concentration as well. In this study, we demonstrate the application of this technique for the nondestructive visualization of motile bacteria as a function of time in a watersaturated porous medium.

Experimental Section Bacteria. Escherichia coli K12 NR50, originally obtained from the culture collection at the University of Pennsylvania, was selected for study because of the availability of commercially produced antibodies to E. coli and because its swimming VOL. 37, NO. 4, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

781

properties have been well studied (3, 5, 21, 22). Cultures were grown aerobically at 27 °C in minimal medium (23) amended with 5.4 mM galactose and 0.2 mM thiamine. Cells were harvested in the middle of exponential growth phase as measured by optical density (Abs590 ) 1.0) to obtain maximum cell homogeneity and motility. Random motility buffer, a phosphate buffer (6.43 mM K2HPO4, 3.53 mM KH2PO4, and 0.014 mM EDTA) that does not support growth, was used for all experiments. Immunomagnetic Labeling. Magnetite was provided as 50-60 nm ferrofluid nanoparticles (Immunicon Corporation, Huntingdon Valley, PA), composed of a quasi-spherical crystalline magnetite core coated with a monolayer of derivitized bovine serum albumin (BSA), at a concentration of 4.8 mg Fe/mL (4.3 × 1012 particles/mg Fe). A purified antiE. coli monoclonal antibody specific to an epitope on the cell surface (CR1224M, Cortex Biochem, San Leandro, CA) was obtained at a concentration of 0.1 mg/mL. The procedure for attaching the magnetite nanoparticles to the bacteria was adapted from that of Nakamura et al. (24). In the first step, 100 µL of the antibody was reduced with 4 mg of dithiothreitol (DTT, Pierce Chemical, Rockford, IL), and 80 µL of ferrofluid was derivatized with 0.5 mg of the heterobifunctional cross-linker N-succinimidyl-3(2-pyridyldithio) propionate (SPDP, Pierce Chemical, Rockford, IL) in 420 µL of buffer and 25 µL of dimethyl sulfoxide (DMSO). The antibody and the ferrofluid reacted separately for 1 h on a Model 151 rotating mixer (Scientific Industries, Inc., Bohemia, NY) set at 60% of maximum rotation speed and then were washed three times with 500 µL of buffer. For each wash, the samples were spun for 15 min at 8163xg in Ultrafree 0.5-mL centrifugal filtration devices (10 000 MWCU, Millipore Corporation, Bedford, MA). After resuspension in buffer, the antibody (50 µL) and magnetite (800 µL) were combined and returned to the rotating mixer for 18 h to allow the thiol group on the reduced antibody to react with the pyridyl disulfide end of the SPDP molecule on the surface of the nanoparticles. E.coli K12 NR50 cells harvested during midexponential growth were labeled by mixing them with the magnetite-antibody conjugate in a volumetric ratio of 10:1 on the rotating mixer for 90 min. Free conjugated antibody was separated from the labeled bacteria by vacuum filtration through a 0.22-µm membrane filter (Micro Separations, Inc., Westboro, MA), and the labeled bacteria retained on the filter were gently resuspended in buffer. Motility Observations. Qualitative observation of bacterial motility before and after labeling was performed at 400X using a Zeiss Std. 16 microscope equipped with phase contrast filters. Quantitative measurements of bacterial motility were performed using a 3-D tracking microscope designed and built by Berg (23, 25). This specially designed microscope was able to follow the movement of an individual bacterium in three dimensions (26). Each bacterial trace contained swimming speed, turn angle, and run time data. Nonmotile Control. To verify that any observed changes in bacterial cell concentrations were the result of cell motility alone, controls were performed in which the bacteria were rendered nonmotile before labeling. The technique involved exposing bacteria to high shear rates to remove the flagella from the organisms (27), rendering them incapable of selfpropulsion. After harvesting, aliquots of the bacterial suspension were collected in a 3-mL disposable plastic syringe and forcibly expressed through a 1-in. 26-gauge disposable stainless steel needle (Becton-Dickinson, Franklin Lakes, NJ). This process was repeated 15 times for each aliquot, and the aliquots were combined into a single suspension prior to labeling. Qualitative observations of bacterial suspensions before and after shearing provided evidence that cells processed in this manner exhibited minimal motility while retaining cell structural integrity. 782

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 4, 2003

FIGURE 1. Schematic of the packed column apparatus used in diffusion experiments. Impinging flow of buffer and E. coli bacteria creates a quasi-step-change in bacterial concentration at the center of the column. The fluid exits the center of the column through four identical ports located symmetrically around the circumference. For clarity of illustration, only two ports are shown here. Cell Counting. Concentrations of labeled bacterial suspensions used for these experiments were determined by the acridine orange direct count (AODC) method (28).

Experimental Protocol Experiments were conducted in an adjustable-bed borosilicate glass chromatography column (Biorad, Hercules, CA) with an inside diameter of 1.5 cm. The column was modified to permit impinging flow similar to that established in the Stopped-Flow Diffusion Chamber (SFDC) (29) by drilling four evenly spaced holes around the circumference of the column, as illustrated in Figure 1. The holes were located 4 cm from the base of the column so as to be situated at the midpoint of the bed when it was packed to a final height of 8 cm. A small circle of nylon fabric, secured in each hole by a plastic tubing connector sealed in place with silicone adhesive, acted as a frit to retain the porous medium. The column was packed wet by filling it with buffer and adding the prewetted porous medium using a transfer pipet. A well-mixed suspension of glass-coated polystyrene microbeads (SoloHill Engineering, Inc., Ann Arbor, MI) with a size distribution of 250 µm to 300 µm were added in 1-2 mL increments and allowed to gravity settle, while the column was gently tapped to pack the porous medium into its final configuration. This technique provided a more uniform T2 profile for us than dry packing which was prone to trapping air bubbles within the porous matrix. A flow adapter with a 40 µm glass frit was used to keep the bed in place and to distribute the incoming flow evenly across the entire cross-section of the bed. The column was connected to a syringe pump and waste reservoir with Tygon tubing. After assembly, the column was pretreated with unlabeled E. coli (109 cells/mL) at a flow rate of 2 mL/min for approximately 25 min, until the optical density of the effluent was equal to that of the incoming bacterial suspension. Without pretreatment, it was found that the BSA-coated magnetite particles bound essentially irreversibly to the packing material, presumably by adsorbing to exposed polystyrene patches thought to exist on the surfaces of the microbeads (30). The column was flushed with 15 mL of buffer at 2 mL/min to remove unsorbed bacteria from the pore fluid prior to initiation of the experiment. With buffer flowing into the column from one side, labeled bacteria flowing in from the other side, and both streams exiting at the midpoint, impinging flow was used to establish an initial step change in the bacterial concentration, as shown in Figure 1. While impinging flow was maintained (2 mL/ min for 6 min), no mixing occurred between the two halves of the column (30), and a sharp interface was established at the center. When the flow was stopped, the bacteria were free to migrate into the other half of the column. Magnetic resonance images were collected under no-flow conditions in the column. Magnetic Resonance Imaging. All experiments were performed in a 1.75 T, 12-cm bore horizontal magnet (Nalorac Cryogenics Corporation, Martinez, CA) and a spectrometer (TecMag, Inc., Houston, TX) equipped with shielded gradient coils (Magnex Scientific, UK) providing a maximum gradient

strength of 20 G/cm. Power was supplied to the coils by three power amplifiers (Techron, Elkhart, IN). A bird-cage resonator radio frequency coil 2.54 cm in diameter and 8 cm in length was mounted around the flow cell. The coil was tuned to the 1H resonating frequency (74.57 Hz) using a 60 MHz oscilloscope (Tektronix, Inc., Beaverton, OR) and a sweep generator (Wavetek RF Products, Inc.). All pulse programming was conducted using MacNMR version 4.5.9 software (Tecmag, Inc.). The imaging protocol was a T2-weighted, one-dimensional, x-slice spin-echo sequence with 20 TE times, a last delay of 8 s, and 8 acquisitions. The slice thickness was 0.29 cm and the field of view (FOV) was 8.52 cm divided into 256 pixels, with a pixel size (spatial resolution) of 330 µm. T2 profiles were generated from analysis of the one-dimensional signal intensity data using a MATLAB program. At each spatial location (256 points), the natural logarithm of the average signal intensity (8 scans) versus the TE time (20 values per point) was fit to a linear function whose slope was the negative inverse of the T2 value for that spatial position. Mathematical Description of Diffusional Migration in Porous Media. A mathematical description of the experimental system is based on a differential mass balance on the diffusing species. By approximating the porous medium as a continuum, assuming radial symmetry, and neglecting adhesion, reaction, and advection, the model equation was reduced to simple one-dimensional diffusion

∂b ∂  ) - Jb ∂t ∂x

(1)

where  is the porosity of the packed bed, b is the concentration of the diffusing species, t is time, and x is the distance from the initial interface. The bacterial flux Jb is defined by

∂b Jb ) - Deff ∂x

(2)

where Deff is the effective diffusion coefficient. For molecular diffusion, the effective diffusion coefficient can be related to the bulk aqueous diffusion coefficient according to (31, 32)

 Deff ) D0 τ

(3)

where Do is the bulk aqueous diffusion coefficient, and τ is the tortuosity of the porous medium. Since bacterial motion in the absence of an attractant or repellant gradient can be characterized as a three-dimensional random walk analogous to molecular diffusion in gases, eqs 1-3 can be used to describe bacterial motility by substituting µeff for Deff and µo for Do (13). If the initial interface is a step change, and one assumes boundary conditions of constant concentration at the both ends of the column, eq 1 can be solved analytically. In the experiments described herein, however, the initial interface was not a perfect step change; impinging flow rates would have to be infinitely fast and exit ports perfectly aligned to achieve an ideal step change. It was clear that assuming a step-change for the initial condition was inappropriate when examining the profiles for the nonmotile bacteria which were unable to diffuse a significant distance from the initial distribution. Thus, eq 1 was solved numerically using a CrankNicolson finite difference method. The first data set (collected at t ≈ 23 min) was used as the initial condition b(0,x) ) b0(x). To generate a function b0 (x) describing the first data set we used an analytical solution of the diffusion equation with the diffusion coefficient as a fitting parameter. No flux boundary conditions were assigned to each end of the column. The numerical solution yielded concentration profiles having the same general shape as the analytical

FIGURE 2. Transmission electron microscopy (TEM) micrographs of E. coli K12 NR50 bacteria, (a) unlabeled and (b) labeled with 40 nm-diameter magnetite particles. The magnetite particles (dark spots) can be seen bound to the surface of the cells in image (b). The scale bar represents 200 nm. Samples were prepared using the embedding-and-slicing technique with 80-nm-thick slices. Debris in the images is pieces of flagella. solution, but the data were described by lower values of µeff than if an ideal step change had been assumed as the initial condition.

Results and Discussion Immunomagnetic Labeling. In applying an immunomagnetic labeling technique to observe bacterial motility, two key assumptions were made: (1) the magnetite particles were attached to the surface of the bacterial cells and (2) the presence of the magnetite particles did not significantly impede the natural swimming behavior of the bacteria. Evidence to support the validity of these assumptions is presented below. Magnetite Attachment. Transmission electron microscopy (TEM) was used to confirm that magnetite particles were attached to the cell surface. Figure 2a shows a TEM image of unlabeled E. coli K12 NR50 cells that served as a negative control to the magnetite-labeled cells seen in Figure 2b. The image on the right clearly shows the conjugated magnetite particles attached to the surface of the labeled cells. The antibody is reactive with a number of E. coli serotypes, including K12, and is presumed to bind to a common surface antigen (33). Estimates of magnetite loading were determined using T2 measurements and AODC analysis of bacterial suspensions labeled with 50-60 nm magnetite particles. The concentration c of a relaxing agent in solution is inversely proportional to its measured T2 value according to the relationship

c)

(

1 1 1 R T2 T2,0

)

(4)

where R is the relaxivity constant and T2,o is the T2 value for pure water. These values were determined experimentally to be 4.16 ms-1(mg Fe/mL)-1 and 1500 ms, respectively. Examination of seven samples of labeled bacteria yielded an average loading of (4.0 ( 2.0) × 10-12 mg Fe/cell. Motility Observations. Visual observations of labeled and unlabeled E. coli cells using a phase contrast microscope at 400× suggested that the labeling process did not alter motility. The speed of the cells and the overall fraction of motile cells were used as the basis for comparison. Clumping of cells, which would be indicative of bacterial cross-linking, was not VOL. 37, NO. 4, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

783

seen in any of the light or electron microscope observations, nor was it expected. Chalmers (34) did not observe cell aggregation or cross-linking when using magnetite particles with a low antibody loading (one antibody per particle); the antibody loading in this study also was estimated to be low at six antibodies per magnetite particle. Moreover, the number of magnetite particles on each bacterium was not sufficient to significantly affect the density of the cell. Cells labeled with (4.0 ( 2.0) × 10-12 mg Fe/cell would experience a 0.4% increase in bacterial weight, assuming an initial cellular mass of 9.5 × 10-13 g (35). Settling velocity calculations based on the Svedberg equation (8) predict a sedimentation velocity of 0.685 µm/s for a 2 µm × 1 µm elliptical cell of this mass; the sedimentation rate increases only to 0.688 µm/s if the mass of the cell is increased by 0.4%. Since the typical swimming speed of an E. coli bacterium is 20-40 µm/s (5, 23, 36), the extra mass added to the cell by the magnetite particles would not be expected to affect motility. The tracking microscope allowed a quantitative comparison of the average run time, average turn angle, and average run speed for E. coli with and without magnetite particles on the surface of the cell. A one-tailed Student’s t-test (0.05 significance level) performed on the run speed data revealed no statistically significant decrease in the average swimming speed for labeled bacteria (15). The same was true for the average run times. Statistical comparison of average turn angles for the unlabeled and labeled bacteria indicated that the average turn angle of the unlabeled bacteria (53° ( 33°) was less than that of the labeled bacteria (66° ( 42°). This difference in average turn angles would correspond to a 33% lower motility coefficient for the labeled bacteria. We note further, that the average turn angle observed for the labeled bacteria is actually closer to values previously reported in the literature (68 ( 36) which were for unlabeled bacteria (25). In our subsequent analysis of migration data we neglected any decrease in motility of the labeled bacteria. Diffusive Migration Studies. MnCl2 Tracer Experiments. The general applicability of the MRI technique for observing and quantifying diffusive behavior in a porous medium was demonstrated by duplicate experiments performed using 0.034 mM MnCl2 as the diffusing agent. The concentration of MnCl2 was calculated to provide an initial T2 value of approximately 350 ms, comparable to that expected for the labeled bacteria. For each experiment, T2 profiles were collected once an hour for 12 h. Equation 4 and a value of R for MnCl2 of 0.034 ms-1 (mM)-1 (37) were used to convert the T2 profiles to dimensionless concentration profiles, an example of which is shown in Figure 3. We obtained an average value for Deff/ from concentration profiles by curve fitting using visual inspection at several times points for each of two replicate experiments. Using eq 3 and Do ) 1.26 × 10-5 cm2/s for MnCl2 at infinite dilution (38), the tortuosity factor τ for the column was determined to be 5.2 ( 0.4. We checked for retardation of Cl- using a tracer test in the column and observed that breakthrough was consistent with the calculated interstitial velocity, thus confirming that MnCl2 could be treated as a conservative tracer. Bacterial Motility Experiments. Six bacterial motility experiments were performed, four with motile E. coli and two in which the bacteria were rendered nonmotile by removing their flagella. Each experiment ran for a total of 12 h, with T2 profiles collected at least once an hour. Dimensionless concentration profiles for the motile bacteria (shown in Figure 4 for one experiment) indicated a significant amount of penetration by the cells when compared to similar profiles for the nonmotile bacteria (Figure 5). In fact, transport of the nonmotile bacteria could be attributed entirely to Brownian diffusion. A nonmotile colloid having the same size as an E. coli bacterium would be expected to have an effective 784

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 4, 2003

FIGURE 3. Dimensionless concentration profiles for diffusion of MnCl2 in the packed column at 21 min (ο), 143 min (∆), and 692 min (×) after impinging flow was halted. Curves are best-fit solutions to model eqs 1-3 with Deff/E as the fitting parameter. Deff/E ) 2.8 × 10-6 cm2/s was used for 143 min and ) 2.4 × 10-6 cm2/s for 692 min. A fit of the data at 21 min was used as the initial condition in the model equations.

FIGURE 4. Dimensionless concentration profiles for motile E. coli K12 NR50 in the packed column at 23 min (ο), 367 min (∆), and 736 min (×) after impinging flow was halted. The curves are theoretical predictions with µeff/E of (5.0 ( 1.1) × 10-7 cm2/s. Three curves are plotted at each time point which show the average value and the range defined by the confidence intervals. A fit of the data at 23 min was used as the initial condition in the model equations. diffusion coefficient of 1.5 × 10-10 cm2/s (8), and as Figure 5 illustrates, this value matched the experimental profiles well. For the motile bacteria, expected µeff/ values were calculated from eq 3. Bulk aqueous phase random motility coefficient µo values determined previously for E. coli NR50 from stopped-flow diffusion chamber measurements were (2.6 ( 0.5) × 10-6 cm2/s (unpublished data). Assuming the same µo was valid for our experiments and using the same tortuosity value that described MnCl2 diffusion, µeff/ for the motile experiments was expected to be (5.0 ( 1.1) × 10-7 cm2/s. This value inserted into eqs 1-3 yields the theoretical predictions shown in Figure 4. Three theoretical curves are shown for each time point. They represent the average value and the limits of 95% confidence intervals for the parameter µeff/. The reasonable agreement of the predictions with the experimental data suggests that we were able to successfully monitor bacterial migration due to random motility within the packed columns. As demonstrated here, immunomagnetic labeling and MRI employed in the study of bacterial transport holds great potential. Although E. coli were used in this study, any number of different bacteria could be used with a suitable antibody.

FIGURE 5. Dimensionless concentration profiles for nonmotile E. coli K12 NR50 in the packed column at 23 min (ο), 367 min (∆), and 736 min (×) after impinging flow was halted. Curves are theoretical profiles for each of the plotted times (which essentially overlie each other) using a value for µeff of 1.5 × 10-10 cm2/s, as predicted for Brownian diffusion. In the future, this technique could be used to observe the movement of bacterial populations in the vicinity of contaminants such as NAPLs or to study the deposition of cells within heterogeneous porous media. Extension of this technique to natural aquifer materials will introduce some new challenges. Iron oxides that are naturally present in these materials will tend to reduce the contrast and thus a higher loading of magnetite per bacterium would be needed to compensate.

Acknowledgments The authors extend sincere thanks to Mr. Ronald N. Keener and Dr. Matthew L. Dickson for helpful assistance with MRI. Funding for this work was provided by the donors of the Petroleum Fund (ACS-PRF 33577-AC9) administered by the American Chemical Society. Additional support for Dr. Sherwood was provided by a Dreyfus Foundation Environmental Chemistry Fellowship granted to the Program of Interdisciplinary Research in Contaminant Hydrogeology at the University of Virginia. We also thank the anonymous reviewers for helpful comments that improved the manuscript.

Literature Cited (1) Jenneman, G. E.; McInerney, M. J.; Knapp, R. M. Appl. Environ. Microbiol. 1985, 50, 383-391. (2) Harvey, R. W.; Garabedian, S. P. Environ. Sci. Technol. 1991, 25, 178-185. (3) Frymier, P. D.; Ford, R. M.; Berg, H. C.; Cummings, P. T. Proc. Natl. Acad. Sci. 1995, 92, 6195-6199. (4) Harwood: C. S.; Fosnaugh, K.; Dispensa, M. J. Bacteriol. 1989, 171, 4063-4066. (5) Lowe, G.; Meister, M.; Berg, H. C. Nature 1987, 325, 637-640. (6) Baedecker, M. J.; Cozzarelli, I. M.; Bennett, P. C.; Eganhouse, R. P.; Hult, M. F. Water Resources Investigations Report 94-4015, Colorado Springs, CO, September 20-24 1993; U.S. Geological Survey.

(7) LeBlanc, D. R. Water Resources Investigations Report 94-4015, Colorado Springs, CO, September 20-24 1993; U.S. Geological Survey. (8) Berg, H. C. In Random Walks in Biology; Princeton University Press: Princeton, NJ, 1993; pp 75-94. (9) Chen, K. C.; Ford, R. M.; Cummings, P. T. J. Theor. Biol. 1998, 195, 481-504. (10) Clement, T. P.; Peyton, B. M.; Skeen, R. S.; Jennings, D. A.; Petersen, J. N. J. Contam. Hydrol. 1997, 24, 269-285. (11) Reynolds, P. J.; Sharma, P.; Jenneman, G. E.; McInerney, M. J. Appl. Environ. Microbiol. 1989, 55, 2280-2286. (12) Sharma, P. K.; McInerney, M. J. Appl. Environ. Microbiol. 1994, 60, 1481-1486. (13) Barton, J. W.; Ford, R. M. Appl. Environ. Microbiol. 1995, 61, 3329-3335. (14) Sharma, P. K.; McInerney, M. J.; Knapp, R. M. Appl. Environ. Microbiol. 1993, 59, 3686-3694. (15) Sung, J. C. M.S. Thesis, University of Virginia, Charlottesville, VA, 1996; p 74. (16) Van Zijl, P. C. M.; Moonen, C. T. W.; Faustino, P.; Pekar, J.; Kaplan, O.; Cohen, J. S. Proc. Natl. Acad. Sci. 1991, 88, 32283232. (17) Pilatus, U.; Shim, H.; Artemov, D.; Davis, D.; Zijl, P. C. M. v.; Glickson, J. D. Magn. Reson. Med. 1997, 37, 825-832. (18) Potter, K.; Kleinberg, R. L.; Brockman, F. J.; McFarland, E. W. J. Magn. Reson., Ser. B 1996, 113, 9-15. (19) Maudsley, A. A. In Medical Magnetic Resonance: A Primer - 1988; Budinger, T. F., Margulis, A. R., Eds.; Society of Magnetic Resonance in Medicine, Inc.: Berkeley, CA, 1988; p 385. (20) Gibby, W. A. In Imaging in Neuroradiology, Part II; Lee, S. H., Zimmerman, R. A., Eds.; W. B. Saunders Company: Philadelphia, 1988; Vol. 26 (5), pp 1047-1058. (21) Adler, J.; Templeton, B. J. General Microbiol. 1967, 46, 175-184. (22) Berg, H. C.; Brown, D. A. Nature 1972, 239, 500-504. (23) Adler, J. J. General Microbiol. 1973, 74, 77-91. (24) Nakamura, N.; Burgess, J. G.; Yagiuda, K.; Kudo, S.; Sakaguchi, T.; Matsunaga, T. Anal. Chem. 1993, 65, 2036-2039. (25) Berg, H. C.; Brown, D. A. Antibiot. Chemother. 1974, 19, 55-78. (26) Frymier, P. D. Ph.D. Dissertation, University of Virginia, Charlottesville, VA, 1995. (27) Kalmokoff, M. L.; Jarrell, K. F.; Koval, S. F. J. Bacteriol. 1988, 170, 1752-1758. (28) Hobbie, J. E.; Daley, R. J.; Jasper, S. Appl. Environ. Microbiol. 1977, 33, 1225-1228. (29) Ford, R. M.; Phillips, B. R.; Quinn, J. A.; Lauffenburger, D. A. Biotechnol. Bioeng. 1991, 37, 647-660. (30) Sung, J. C. Ph.D. Dissertation, University of Virginia, Charlottesville, VA, 2000; p 171. (31) Geankoplis, C. J. Transport Processes and Unit Operations, 2nd ed.; Allyn and Bacon, Inc.: Boston, 1983. (32) Epstein, N. Chem. Eng. Sci. 1989, 44, 777-779. (33) Lau, M. T.; Cortex Biochem Inc., personal communication, 1998. (34) Chalmers, J. Ohio State University, Columbus, OH, personal communication, 1997. (35) Neidhardt, F. C.; Ingraham, J. L.; Schaechter, M. Physiology of the Bacterial Cell: A Molecular Approach; Sinauer Associates, Inc.: Sunderland, MA, 1990. (36) Macnab, R. M.; Aizawa, S. W. Annu. Rev. Biophys. Bioeng. 1984, 13, 51-83. (37) Pykett, I. L.; Rosen, B. R.; Buonanno, F. S.; Brady, T. J. Phys. Med. Biol. 1983, 28, 723-729. (38) Cussler, E. L. In Diffusion: Mass Transfer in Fluid Systems; Cambridge University Press: Cambridge, 1997; pp 143-153.

Received for review August 20, 2001. Revised manuscript received August 26, 2002. Accepted November 18, 2002. ES011210U

VOL. 37, NO. 4, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

785