Environ. Sci. Technol. 2009, 43, 1546–1552
Enhanced Transverse Migration of Bacteria by Chemotaxis in a Porous T-Sensor TAO LONG AND ROSEANNE M. FORD* Department of Chemical Engineering, University of Virginia, Charlottesville, Virginia 22904
Received September 30, 2008. Revised manuscript received December 15, 2008. Accepted December 23, 2008.
Subsurface bioremediation is often hindered by the inability to achieve good mixing between injected bacteria and residual contaminants. Chemotaxis, which is the ability of bacteria to migrate preferentially toward higher concentrations of certain chemical attractants, could potentially increase bacterial transport into the contaminated zone. To observe and quantify this chemotactic enhancement to bacterial dispersion transverse to groundwater flow, a microfluidic devicesa porous T-sensorswas created. It allowed two streams of equal flow rate to enter side-by-side into a porous channel; the transverse mixing of the two streams was controlled primarily by dispersion. When a suspension of the chemotactic bacteria Escherichia coli HCB1 and a solution of chemical attractant R-methylaspartate were injected as the two incoming streams, enhanced bacterial migration into the attractant stream was observed relative to a control experiment with dispersion alone. Chemotaxis was observed under lower flow rates comparable to natural groundwater flow. The chemotactic response was greater than that predicted by an advectiondispersion equation model using a chemotactic coefficient derived under quiescent experimental conditions, which suggests that flow in porous media may further enhance transverse migration for chemotactic bacteria. This study provided direct evidence of the significance of bacterial chemotactic transverse migration at groundwater flow rates.
1. Introduction Groundwater pollution by chemical waste has been discovered at hundreds of thousands of sites throughout the United States (1). In situ bioremediation has demonstrated great potential to remediate organic contamination in groundwater and deep soil effectively and economically (2, 3). This technology uses engineered systems to deliver microorganisms directly into the contaminated zone (bioaugmentation) or to stimulate the microbial activities at the contaminate source (biostimulation), optimizing the subsurface hydrological and chemical conditions to achieve maximum rates of contaminant degradation. Enhancing the pore-scale mixing that brings microorganisms in contact with the contaminants remains a challenging problem. Chemotaxis is the ability of bacteria to swim toward higher concentration of potential substrates. It has been observed in many bacterial strains in response to various environmental contaminants (4). When the bacterial flagella * Corresponding author phone: (434)924-6283; fax: (434)982-2658; e-mail:
[email protected]. 1546
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 5, 2009
rotate in a counterclockwise direction, the cell swims forward (runs); when one or more flagella reverse their rotation, the cell reorients (tumbles) (5). When sensing an increase in attractant concentration in the ambient environment, the bacterium reduces its tumble frequency, thereby prolonging the course of straight swimming, which results in chemotaxis at population scale (5). If chemotactic transport transverse to groundwater flow is significant, migration of microorganisms into the low permeability zones or contaminant plumes may be greatly enhanced, resulting in faster degradation of the residual contaminants (6). However, bacterial motility and chemotaxis in porous media is still a field full of unknowns. Experimental data sets are limited, including many earlier works that reported only the “penetration rate” rather than effective diffusion coefficients or random motility, which are more suitable for quantitative analysis. In static saturated columns in which 1D temporal diffusion profiles of bacteria were measured, Barton and Ford (7) found chemotaxis was insignificant, and Olson et al. (8) reported a reduced extent of chemotaxis, which could be due to the obstructed bacterial motility in the pore space. In swarming plate assays packed with a section of porous media, Roush et al. (9) estimated a 100fold increase in bacterial chemotactic sensitivity, which might be a result of increased attractant gradient in pore water as a result of consumption. When pore water flow was present, chemotaxis was found to reduce bacterial deposition in a column study (10), but detailed transport modeling was not attempted. No previous studies directly quantified bacterial chemotactic transport in pore water flow. In bioremediation, transverse mixing becomes an important factor, and measuring bacterial transport in 2D is critical. Transport models in the format of advection dispersion equations (ADE) have been used to simulate chemotaxis in groundwater (11) and chemotaxis involved in bioremediation (12), but no existing experimental data were available to test these models. Microfluidic devices are designed to control fluid flow on a length-scale smaller than 1 mm (13). In hydrologic studies, microfluidic devices (also termed micromodels) have been widely applied to provide direct visualization and quantification of transport phenomena in porous media, including transport of colloids and bacteria (14-18). Optical techniques are often used to achieve nondestructive, online measurement of several different length scales in microfluidic devices. One such device called the T-sensor (19) (or similar designs) succeeded in creating a stable chemical gradient for the study of chemotaxis in bulk liquid (20-25). The primary objective of this study was to directly observe bacterial chemotactic transport in a model porous medium. A microfluidic devicesthe porous T-sensor (Figure 1)swas modified from the original T-sensor design (19). Two injectates were injected into the two arms of the porous T-sensor under the same flow rate. When they flowed side by side into the head of the main channel, mixing between the two streams started. The mixing was driven primarily by transverse dispersion, because the device was operating under low Reynolds numbers. In this study, the transverse transport properties of the solute and the bacteria were quantified. Then the attractant and bacteria were injected in the two input streams to capture the impact on bacterial transverse migration by chemotaxis. Finally, the experimental results were compared with the traditional ADE model incorporating chemotactic terms. 10.1021/es802558j CCC: $40.75
2009 American Chemical Society
Published on Web 01/27/2009
FIGURE 1. (a) Illustration of the porous T-sensor design. The main channel has a total length of 8.3 cm, width of 6 mm, depth of 13 µm, and porosity of 40%. (b) An image of the inner structure taken under a bright field microscope (Carl Zeiss, 20× objective). Each impermeable cylinder had a diameter of ∼200 µm, and the dimension of pore throats was ∼46 µm. (c) Example of an image taken under a wide field microscope (Olympus IX-70, 60× objective). The bright square shows the area chosen for bacteria counting. (d) Example of a transverse profile consists of 25 pore throats.
2. Materials and Methods Bacterial Strain and Chemicals. The bacterial strain used in the experiments was Escherichia coli HCB1, a wild type strain obtained from Dr. Howard Berg at Harvard University. The cells are rod-shaped, with a length of approximately 2 µm and cross sectional diameter of 1 µm (25). A bundle of flagella located at one end of the cell provides motility to E. coli HCB1. A 100 µm aliquot of -70 °C frozen stock (40% v/v glycerol) of E. coli HCB1 (25) was used to inoculate 50 mL of autoclaved Luria broth (Fisher) in a sterile 250 mL baffled shake flask. Cells were cultured for ∼9 h in a LabLine Environshaker (model 3528-5) at 150 rpm under 30 °C to midexponential phase, where the optical density of the culture reached 1.0 at 590 nm (Beckman DU-7 Spectrophotometer). A 5 mL quantity of the culture was filtered on a 0.22 µm filter (Millipore GSWP14250) and resuspended in 5 mL of 10% RMB (random motility buffer, which consisted 11.2 g of K2HPO4, 4.8 g of KH2PO4, and 0.029 g of EDTA per liter of distilled water) (25-27). The diluted RMB with an ionic strength of 0.02 M (28) was used in this study to reduce bacterial sorption to the microfluidic device. R-Methylaspartate (R-mASP, Sigma-Aldrich) was dissolved in 10% RMB to yield a 3 × 10-4 M solution, a concentration which elicited
observable chemotactic attraction of E. coli HCB1 in a previous study (25) and was also confirmed in this study using an agarose plug assay (29). Fluorescein (Fisher) at a concentration of 1.2 × 10-4 M was used as the tracer. All the solutions were prepared with 10% RMB as the solvent. The experiments were performed under room temperature (∼20 °C). Microfluidic Device Design and Operation. The design of the porous T-sensor is illustrated in Figure 1. Typical soft lithography methods were followed to produce the microfluidic device (30), and detailed procedures can be found in the Supporting Information. The main channel of the device has a total length of 8.3 cm, width of 6 mm, depth of 13 µm, and porosity of 40%. Each impermeable cylinder has a diameter of ∼200 µm, and the dimension of pore throats is ∼46 µm. The porosity and cylinder diameter were chosen to serve as a 2D simplification of a homogeneous sandy aquifer. Liquid was delivered into the porous T-sensor using 100 µL GC microsyringes (Becton-Dickinson) and 1/16 in. Teflon tubing (Upchurch Scientific), driven by a syringe pump (Harvard PHD 2000). Three tests were performed in the porous T-sensor with different groups of injectates. In test 1 (tracer test), the two VOL. 43, NO. 5, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1547
injectates were 1.2 × 10-4 M fluorescein solution and phosphate buffer. In test 2 (bacterial control), the injectates were bacterial suspension and phosphate buffer. In test 3 (chemotaxis), bacterial suspension and 3 × 10-4 M R-mASP were used. In each test, the linear velocity in the main channel was adjusted to 20, 10, and 5 m/day, representing typical groundwater flow rates. Images were captured after the channel had been flushed at the desired linear velocity for 0.5-1 h to ensure steady state was achieved. In each test for each velocity, images were recorded across three transverse profiles at 2, 4, and 6 cm from the head of the main channel. Each profile had 25 pores separated by 24 impermeable grains (Figure 1d). Snapshots were taken in these pore bodies to measure fluorescein and bacterial concentration. Measurements from the first and the last pore were dropped due to accumulation near the channel boundaries. In the tracer test, three snapshots of fluorescent light were taken at each pore with an epifluorescent microscope (Carl-Zeiss Standard 16, 20× objective, arc lamp; Hamamatsu 4742-95 digital camera) at 80 ms exposure time. The fluorescent intensity at the pore centers was converted to fluorescein concentration with previously obtained standard curves. In the bacterial control and chemotaxis tests, ten bright field snapshots were taken at each pore with a wide field microscope (Olympus IX-70, 60× objective) equipped with a digital camera (Hamamatsu C4742-98) at 1 s intervals. A ∼50 × 50 µm square region at the center of each pore was selected (Figure 1c), and the number of bacteria in the square was counted manually with ImageJ (NIH) and the point picker plug-in (by ´ cole Polytechnique Fe´de´rale de LauPhilippe The´venaz, E sanne). The sum of bacteria counted from the ten snapshots was used to get the normalized bacterial concentration at the specific pore body. In each group of ten snapshots, the standard deviation of counted cell number was less than 5% of the averaged value. Mathematical Modeling. The governing equation for bacteria transport including chemotaxis was provided by Olson et al. (12), which is Rb
∂b ∂b ∂2b ∂2b ∂(vChxb) ∂(vChyb) ) -vf + Dbx 2 + Dby 2 ∂t ∂x ∂x ∂y ∂x ∂y (1)
where Dbx ) Rbxvf + vChx )
µ0,eff µ0,eff , Dby ) Rbyvf + ε ε
(2)
KCh KCh 1 χ0,eff 1 χ0,eff ∂C ∂C , v ) (3) 3 ε (K + C)2 ∂x Chy 3 ε (K + C)2 ∂y Ch Ch
and Rb is the bacterial retardation factor, b is the bacterial concentration, vf is the averaged linear velocity of the flow, C is the concentration of the tracer, ε is the porosity, Dbx and Dby are the longitudinal and transverse dispersion coefficients for bacteria, vChx and vChy are the longitudinal and transverse chemotactic velocities describing the convective transport of bacteria caused by chemotaxis in shallow attractant gradients. In eq 2, µ0,eff is the effective random motility coefficient. The magnitude of the chemotactic velocities are described by eq 3, where χ0,eff is the chemotactic sensitivity and KCh is the constant describing the apparent binding rate of the chemotactic sensors (31). The coefficients µ0,eff and χ0,eff are also expected to satisfy (8, 32)
1548
9
µ0,eff µ0 ) ) µa ε τb
(4a)
χ0,eff χ0 ) ) χa ε τb
(4b)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 5, 2009
where µ0 and χ0 are random motility and chemotactic sensitivity coefficients in bulk aqueous phase under quiescent conditions, µa and χa are defined as the apparent coefficients in porous media, and τb is the effective tortuosity for bacteria to transport through the porous medium. Values of µ0, χ0, and KCh are available for several bacterial species (6), including E. coli HCB1. Olson et al. (33) observed that under quiescent nonchemotactic conditions, the bacterial τb was much greater than the tortuosity experienced by a solute, which may be because bacterial runs were continuously restricted by the pore space. The value of τb may depend on porous media properties, flow field characteristics, and bacterial swimming properties, although the specific form of the dependence was not quantified. In static pore water, accounting for only the restricted run lengths due to the solid phase, bacterial movement resembles Knudson diffusion and τb can be estimated by
(
τb ) τ 1 +
3µ0 vbdp
)
(5)
where vb is the bacterial swimming speed in bulk liquid, dp is the pore diameter (33), and τ is the tortuosity. The tortuosity for transverse dispersion in hexagonal cylinder arrays is approximately 2 based on numerical simulations by Acharya et al. (34). In static saturated soil columns, incorporating the Knudson diffusion still significantly underestimated the effective tortuosity experienced by bacteria (33). However, we note that this expression may not be valid when pore water flow exists, as will be discussed later. The tortuosity for bacteria was 3.7, as calculated from eq 5, given an averaged swimming velocity for E. coli HCB1 of 22.8 µm/s and pore diameter dp of 46 µm (6). The channel walls were impermeable boundaries (y ) 0, y ) 6 mm). A step change of bacterial or attractant concentration was assumed at the head of the channel [e.g., C(x)0,0ey
