Bivalent Cations Increase Both the Subpopulation of Adhering

U.K.) with polyethylene frits (25 µm pore diameter) as described previously (18). .... collision efficiency Rsm can be calculated from the Maxwell di...
0 downloads 0 Views 119KB Size
Environ. Sci. Technol. 2000, 34, 1011-1017

Bivalent Cations Increase Both the Subpopulation of Adhering Bacteria and Their Adhesion Efficiency in Sand Columns STEFANO F. SIMONI,† TOM N. P. BOSMA,‡ H A U K E H A R M S , * ,§ A N D ALEXANDER J. B. ZEHNDER Swiss Federal Institute for Environmental Science and Technology (EAWAG), CH-8600 Du ¨ bendorf, Switzerland, and Swiss Federal Institute of Technology Zu ¨ rich (ETHZ), CH-8000 Zu ¨ rich, Switzerland

The need to understand important factors affecting the spread of bacteria in groundwater aquifers is evident for fields as diverse as drinking water safety or environmental engineering concerned with bioremediation of polluted sites. For example, increasing concentrations of dissolved minerals tend to increase the deposition efficiency of bacteria in porous media. As bacteria and mineral surfaces are mostly negatively charged, this is generally assumed to be a consequence of the higher ionic strength, which leads to stronger shielding of the surface charges by the counterion-cloud in solution. We found Mg2+ to enhance deposition of Pseudomonas sp. strain B13 in sand columns with respect to a solution of identical ionic strength containing Na+. Hence bivalent cations are likely to affect microbial deposition specifically, for example due to specific binding to the cell surface. Moreover, low concentrations of Pb2+ or Cu2+ reverted the surface potential of strain B13, thus providing additional evidence for this hypothesis. Recently, we showed strain B13 to split up in a well-adhering and in a nonadhering subpopulation. In experiments conducted with Mg2+ and Na+ at various ionic strength, bivalent cations seemed to increase the welladhering subpopulation as well as its adhesion efficiency.

Introduction Microbial adhesion is affected by the composition of the suspending medium. This is a consequence of the importance of electrostatic forces for adhesion, and of their sensitivity to modification of the actual surface charge by bound ions, or to shielding by the counterion-cloud in solution (1, 2). If the adhesion substratum is negatively charged like most bacteria are at circumneutral pH, an increase in electrolyte concentration reduces electrostatic repulsion and adhesion * Corresponding author phone: ++ 41 21 693 37 73; fax: ++ 41 21 693 56 70; e-mail: [email protected]. † Present address: CSD AG Environmental & Geotechnical Consulting, Schachenallee 29, CH-5000 Aarau, Switzerland. ‡ Present address: TNO Institute of Environmental Sciences, P.O. Box 342, NL-7300 AH Apeldoorn, The Netherlands. § Present address: Swiss Federal Institute of Technology Lausanne (EPFL), IATE-Pe´dologie, GR Ecublens, CH-1015 Lausanne, Switzerland. 10.1021/es990476m CCC: $19.00 Published on Web 02/04/2000

 2000 American Chemical Society

increases (3-7). It is therefore not surprising that several research groups found pronounced effects of electrolyte concentration on the removal efficiency of bacteria in porous media (8-12). Groundwater composition is thus expected to influence the spread of bacteria in the subsurface, which is of concern to diverse fields ranging from drinking water safety to engineered bioremediation of contaminated aquifers. The important role of electrolyte chemistry for microbial adhesion is also confirmed by the stimulation of adhesion by polyvalent cations (3, 8 ,13). As for the concentration effects described above, this might be due to more effective shielding of the surface charge by compression of the electric doublelayer, because both higher concentrations of ions or higher charge per ion lead to an increased charge density in solution. The contribution of individual ion-types to the total charge density of an electrolyte is often summed up in the ionic strength I (1)

I)

∑C z

1 2

2

(1)

i i

where Ci and zi are the concentrations and the valencies of every ion species i in solution, respectively. When we converted published data which showed that MgSO4 increased the reversible sorption of Achromobacter strain R8 with respect to NaCl (3) from salt concentrations to solution ionic strengths, we indeed discovered the two series to superimpose. On the other hand, cell envelopes of gramnegative and gram-positive bacteria are known to bind cations (14), and lipopolysaccharides (LPS) on the surface of gram-negative bacteria were shown to contain high-affinity binding sites for bivalent cations (15-17). It therefore seems likely that the influence of polyvalent cations on microbial adhesion is not restricted to the effects of ionic strength I. During our studies on transport and deposition of Pseudomonas sp. strain B13 in sand columns, we found Mg2+ to increase deposition of bacteria in comparison to Na+, even when the ionic strength was kept constant. We present here the result of our efforts to understand this finding and its consequences for the subsurface transport of bacteria. As described in an earlier paper (18), we found strain B13 to split up in an adhering and in a nonadhering subpopulation. Intrapopulation heterogeneity with respect to adhesion seems to be quite widely spread (19-23). In this report, we point out that this phenomenon prevents a straightforward discussion of the effects of different electrolytes on bacterial filtration because it obstructs an easy calculation of the common filtration parameters. Our interpretation of the data obtained with strain B13 is based on a mechanistic model of the deposition process, which relies on an extended DLVOtheory of colloid stability.

Materials and Methods Organism and Culture Conditions. Pseudomonas sp. strain B13 (24) is a gram-negative organism able to utilize 3-chlorobenzoic acid (3-cba) as the sole source of carbon and energy. Strain B13 is rather hydrophilic and negatively charged at circumneutral pH (25). Cells are covered by a dense layer of lipopolysaccharides (LPS), extending 20-50 nm from the outer membrane (18). We grew the organisms in phosphate buffer containing minerals and trace elements, which was additionally amended with 5 mM of 3-cba (26). Cell suspensions were prepared with the desired electrolytes after harvest of starved cells by centrifugation (18). VOL. 34, NO. 6, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1011

Electrolyte Chemistry. Electrolytes used for preparing cell suspensions and for equilibrating the columns were based on the growth medium, in which phosphate was replaced by MOPS/NaOH buffer at pH ) 7.2 and (NH4)2SO4 was used instead of NH4NO3. We then added MgSO4 or Na2SO4 to this solution in order to obtain electrolytes with a total ionic strength of I ) 100 mM. These main electrolytes accounted for approximately 70% of the final ionic strength. Electrolytes of lower I were obtained by dilution. In some cases, other salts replaced MgSO4 or Na2SO4 in solutions with I ) 10 mM. Although thermodynamic data suggest the formation of soluble Ca or MgSO40 complexes for electrolytes containing the respective salts and I > 10 mM (2), we found that solution conductivity was not reduced accordingly (not shown). In contrast to a previous report (18), we therefore decided to neglect the effect of complex formation on I. All electrolytes were prepared with deionized water (NANOpure Cartridge System, SKAN, Basel, Switzerland). Deposition Experiments. We conducted filtration experiments in glass columns (2.5 cm I.D., Omnifit, Cambridge, U.K.) with polyethylene frits (25 µm pore diameter) as described previously (18). Briefly, the columns were packed wet with washed cristobalite sand (Fluka, Buchs, Switzerland) to bed-heights of 3.2 or 6.4 cm with a porosity p ) 0.45. Ten, 50, and 90% of the grains (number based) were below 150, 238, and 345 µm in diameter, respectively. We operated the columns in downflow mode with a peristaltic pump (Ismatec, Glattbrugg, Switzerland) at Darcy-velocities 1.2 cm h-1 < U < 1.3 cm h-1. Connecting tubings consisted of polyethylene and tygon (Ismatec, Glattbrugg, Switzerland). We started filtration studies with bacterial suspensions (C0 ∼ 108 cells mL-1) after equilibrating the columns for at least 10 pore volumes (PV) with the respective solution without bacteria. Column outlets were sampled with fraction collectors. After a temporary steady-state was reached (normalized time t* > 2 PV), sampling was continued for another 2-3 PV. We deduced relative cell densities in the column outlet with respect to the column inlet (C/C0) from turbidity measurements at 280 nm after carefully shaking the sample vials. Quantification was possible down to 5 × 106 cells mL-1 (OD280 ∼ 0.02 AU) or less. The resolution was ∆OD280 < 0.01 AU. The breakthrough of 1.0 mM KBr was followed online at 220 nm with a Jasco 870-UV detector (Jasco, Tokyo, Japan) in tracer tests without cells. Filtration Model. About 10 years ago, clean bed filtration theory (27) was suggested to be applicable to irreversible first-order removal of bacteria in porous media at low surface coverages (10, 28):

(

C/C0 ) exp -

)

3 (1 - p) ηRL ) exp(-RλL) 4 rc

(2)

In eq 2, C0 and C are inflowing and outflowing cell concentrations, respectively, L is the length of the flow path, p is the porosity of the packed bed, rc is the radius of the collector sand grain, L is the length of the flow path, η is the collector efficiency, and R is the collision efficiency. Equation 2 is often simplified by defining a filtration coefficient λ. The collector efficiency η describes the transport of particles from the bulk solution to the collector surface and can be theoretically derived from approximate solutions to the convective-diffusion equation. We used a corrected version (10, 29) of an equation originally published by Rajagopalan and Tien and found η to be mainly determined by diffusive transport in the flow-velocity range we used. The collision efficiency R corresponds to the fraction of collisions which finally result in attachment of the particle and is expected to depend on the interaction energy ∆G DLVO-AB (h) and other factors. In the absence of repulsive forces, R approaches 1. 1012

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 6, 2000

A simple way to account for heterogeneous deposition within a population of bacteria is to postulate two subpopulations of cells with different collision efficiencies Rfast and Rslow (18)

C/C0 ) ffast exp(-RfastλL) + fslow exp(-RslowλL) ffast + fslow ) 1

(3)

where ffast and fslow denote the fractions of the population undergoing fast and slow removal, respectively. If the differences in the adhesion efficiencies between the two subpopulations are pronounced (Rslow , Rfast), eq 3 can be approximated by

C/C0 ≈ ffast exp(-RfastλL) + (1 - ffast)

(4)

in the first portion of the filtration path. It is clear from eq 4 that the breakthrough level C/C0 for given hydrodynamic conditions can be affected either by a change in the welladhering fraction ffast or by a change in Rfast. Surface Characterization. (i) Electrophoretic Mobility and ζ-Potential. Electrophoretic mobility uE of suspended bacteria and sand particles was measured by dynamic light scattering (Zetamaster, Malvern Instruments, Malvern, U.K.) in the media used for the column experiments. We calculated ζ-potentials at the electrokinetic shear plane from uE with the Henry correlation (1, eq 3.3.5), which is adequate except for very low I. To obtain estimates for the ζ-potential of sand, we ground sand in a mortar and determined uE of the fraction remaining in suspension for several hours. (ii) Contact Angles. Contact angles of liquid droplets on a bacterial lawn were determined as described by (30). Constant Capacitance Model of the Solid-Liquid Interface. Following a simplified Stern-layer concept (1, 2), the negative charge of the cell surface σ0 ( 0):

σd ) -σ0

(5)

The surface charge σ0 may consist of structural charge, proton charge, or charge arising from complexed ions. The two charged planes are separated by a small distance d, and they can be understood as a small capacitor with a constant potential at the surface ψ0, a potential ψd at distance d, and the capacitance of the Stern-layer KS:

σd ) -(ψ0 - ψd)KS ) -ψ0KS + ψdKS

(6)

As suggested by experimental evidence, the ζ-potential at the electrokinetic shear-plane corresponds quite well to ψd (1), which implies that d corresponds to the distance of the electrokinetic shear plane from the cell surface. An approximation for σd can be obtained from the ζ-potentials by integrating the Poisson-Boltzmann equation from infinity up to distance d. There are different expressions available for this integral, which we found to yield comparable results. We used the expression derived for the double layer around a sphere (1, eq 2.3.37), which is appropriate unless I is very low. After rewriting eq 6

σd ) -ψ0KS + ψdKS ≈ σmax + ζKS d

(7)

we find that a plot of σd vs ζ allows to deduce KS from the slope of the regression line and σmax from its intercept (ψd, d ψ0, and ζ < 0 for negatively charged cells). As the charge in the diffuse cloud of counterions σmax is counterbalanced by d the charge on the cell surface (eq 5), the intercept σmax also d provides a measure for the maximal charge density at the cell surface at high ionic strength σmax 0 .

TABLE 1. General Parameters for Calculation of Interaction Energy ∆G DLVO-AB parameter

rc rp µ k T Acwg r 0 λLW λAB

h0

source

sand grain/collector radius [m] cell/particle radius [m] water viscosity [kg m-1 s-1] Boltzmann constant [J K-1] temperature [K] Hamaker constant cell-water-glassa [J] dielectric constant for water [-] permittivity of the vacuum [C2 J-1 m-1] decay length for retarded Lifshitz-van der Waals interactions [m] decay length for acid-base interactions [m] distance of closest approach for acid-base interactions [m]

10-4

1.3 × 5.0 × 10-7 8.9 × 10-4 1.38 × 10-23 298 6.4 × 10-21

(18) (25)

(47)

78.4 8.85 × 10-12 1.0 × 10-7

(33)

1.0 × 10-9

(38)

1.6 × 10-10

(38)

a Values for glass were used to approximate interaction energies for sand. The result is almost identical if Ac derived from γ LW (Table 2) is used to calculate the compound Hamaker constant.

TABLE 2. Contact Angles and Surface Energy Components Needed for Calculation of ∆G0AB contact angle θ [°]a water bacteria glassc waterf formamidef diiodomethanef

32 11

fa 40 11

components [mJ m-2]

dm

γLW

52

32.4b

d

50.3e 21.8 39.0 50.8

γ+ 0.22b

0 and results in repulsive forces. In contrast, ∆G LW is usually attractive. A bacterium approaching a sand grain typically experiences a shallow secondary minimum (∆Gsm) of ∆G DLVO at cell-tosurface separations h > 10 nm. Upon further approaching the surface, an energy-barrier of up to several hundred kT often prevents access to the deep primary minimum for h < 1 nm where ∆G LW dominates. We calculated ∆G DLVO for sphere-plate geometry (rc . rp), using an expression for retarded ∆G LW (33) with a compound Hamaker-constant Acwg from literature (Table 1). Use of a retarded term for ∆G LW instead of an unretarded one results in less pronounced secondary minima. To calculate ∆G EL, we used an expression derived for constant-potentials (34), where we inserted ζ-potentials for the surface potentials (Table 4). Following an “extended” DLVO-approach like others before us (3537), we included an additional term (∆G AB) in order to account

FIGURE 1. Influence of cation type on relative cell density C/C0 in the outflow of sand columns. Representative data from duplicate columns (I ) 10 mM, L ) 3.2 cm) are shown together with a tracer curve (‚‚‚, 1 mM KBr-pulse). The shielding of the negative surface charges on cells and sand is expected to depend on ionic strength I. As I was identical for both electrolytes, MgSO4 must have reduced the electrostatic repulsion by reducing the surface charge directly in order to increase deposition as observed. for relatively short-ranging “acid-base”-type forces (38). Although the existence of such forces, often ascribed to “hydrophilic” or “hydrophobic” effects, is widely recognized, their physical basis as well as the adequate way to describe them are still a matter of debate (39). Due to its short range, ∆G AB affects mainly the primary minimum region of ∆G DLVO-AB (h). We calculated ∆G AB from estimates for polar surface energy components derived from contact angle measurements on bacterial lawns with three different liquids (water, formamide, and diiodomethane, Table 2). Calculation of Theoretical Collision Efficiency rsm. Taking up a proposition by Hahn et al. (40), we recently suggested Rfast to depend on the theoretical escape probability from the secondary minimum ∆Gsm in the interaction energy ∆G DLVO-AB (18). According to this approach, a theoretical collision efficiency Rsm can be calculated from the Maxwell distribution of the kinetic energies of the cells Ekin

f (Ekin) )

2

xπkT

Rsm ) 1 -

x



(

)

Ekin Ekin exp kT kT

(8)

f (Ekin)dEkin

(9)



-∆Gsm

Results and Discussion Effect of Cation Type and Ionic Strength on Deposition of Bacteria. Deposition of Pseudomonas sp. strain B13 in sand columns was shown to be irreversible unless the chemistry of the suspending medium was changed (18). As indicated by the height of the plateau reached after initial breakthrough, deposition of bacteria was more efficient in solutions containing Mg2+ as the dominant cation than in solutions of identical ionic strength I but containing Na+ (I ) 10 mM, L ) 3.2 cm; Figure 1). In nine independent pairs of column experiments conducted under these conditions, we found Mg2+ to increase cell deposition (1 - C/C0) by 54% on average in comparison to Na+ electrolytes (P ) 0.014 in a one-sided Student t-test). The change in cell deposition was accompanied by a change in the ζ-potential of suspended cells (Table 3). This points to electrostatic forces being involved. In support of this hypothesis, we found changes in I to affect both cell deposition (Figure 2) and ζ-potential (Table 4). In contrast to the replacement of bivalent cations by monovalent ones, substitution of the SO42- anions with NO3- did neither VOL. 34, NO. 6, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1013

TABLE 4. ζ-Potentials Used for Calculation of ∆G EL

TABLE 3. Electrophoretic Mobility uE and ζ-Potential of Pseudomonas sp. Strain B13 and Ground Quartz Sand Particles Pseudomonas sp. B13 electrolytea K2SO4 KNO3 Na2SO4 NaNO3 CaSO4 Ca(NO3)2 MgSO4 Mg(NO3)2

uEb [10-8 m2 V-1 s-1]

n

-2.6 ( 0.33 5 -2.4 ( 0.43 5 -2.4 ( 0.23 10 -2.3 ( 0.28 7 -1.7 ( 0.10 6 -1.6 ( 0.13 7 -1.9 ( 0.17 9 -1.8 ( 0.19 7

ζc

[mV]

-34 -31 -32 -31 -23 -21 -25 -23

electrolytea

ground quartz sandd

uEb [10-8 m2 V-1 s-1]

n

-4.0 -4.2 -3.6 -4.3 -2.6 -2.7 -2.9 ( 0.77 -2.9

1 1 1 1 1 1 3 1

ζc

FIGURE 2. Influence of solution ionic strength I on relative cell density C/C0 in the outflow of sand columns. Filled and open symbols are representative data from columns with L ) 3.2 cm and L ) 6.4 cm, respectively. For MgSO4 electrolytes and I ) 100 mM, formation of cell aggregates was observed (symbols with crosses), and these data points were omitted from discussion. Data for I ) 10 mM are means from several independent experiments and error bars indicate 95% confidence intervals (n ) 12 for short columns with NaSO4 electrolytes; for the MgSO4 electrolytes, 23 experiments were conducted with short columns, and six with long columns). The diamond in the upper left corner of the lower panel was obtained from an experiment with deionized water (I < 10-6 M). The dotted horizontal line corresponds to the fraction of nonadhering cells (64%) we found for MgSO4 electrolytes and I ) 10 mM (18). It indicates the minimal C/C0 to be expected for these conditions and is added to facilitate a comparison with the other experiments. alter deposition (not shown) nor affect ζ-potentials (Table 3). Similar to our findings for MgSO4 electrolytes (18), rinsing with deionized water resulted in a washout of 20-35% of the 9

I [mM]

bacteria ζb [mV]

sand ζ [mV]

Na2SO4

1 10 100 1 10 100

-42 -32 -19 -25 -25 -19

-80d -46b -25c -39b -38b -24b

[mV]

-52 -54 -46 -55 -33 -35 -38 -37

a MOPS buffers where the indicated salts accounted for ∼70% of the total ionic strength (I ∼10 mM). b Electrophoretic mobility with 95% confidence intervals from n independent experiments with multiple measurements each. c ζ-potentials were derived from uE. d Particles small enough to remain suspended for several hours were measured.

1014

salt

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 6, 2000

MgSO4

a MOPS buffers where the indicated salts accounted for ∼70% of the total ionic strength. b Derived from electrophoretic mobility. c Streaming potentials for glass (47). d Approximated from sources reporting streaming potentials for glass ((47), 78.8 mV) and for quartz ((50), ∼85 mV).

deposited cells for Na2SO4 electrolytes (not shown). This indicates that deposition was only partially reverted by the increased electrostatic repulsion between sand and bacteria resulting from the lowered ionic strength. Although a modest influence of ionic strength on cell surface hydrophobicity has been reported for other strains (41), water contact angles remained constant at 32 ° regardless of the electrolyte used to prepare the cell lawns (data not shown). The finding that the cation-type affects the electrostatics at the negatively charged surfaces of sand and bacteria, but the anion-type does not, is consistent with the well-known dominant role of counterions in colloid chemistry (1). In contrast to other studies (3, 8, 13), we may rule out the observed effect to be a consequence of a stronger shielding of the surface charge due to compression of the adjacent ion cloud alone. The thickness of this double layer is assumed to depend on I (1), which was identical for the respective experiments with MgSO4 or Na2SO4. The reduced ζ-potential can therefore only be explained by sorption of Mg2+ within the electrokinetic shear-plane, which results in reduced apparent surface charge σ0 and surface potential ψ0. Constant Capacitance Model of the Solid-Liquid Interface Applies to Monovalent Cations. Upon plotting calculated σd against measured ζ, we found a linear relationship for Na2SO4 electrolytes and I g 10-4 M (Figure 3). Only the point obtained at I ) 10-5 M to the left clearly deviates from the straight line obtained with the other four points and was therefore omitted from the regression analysis. For the MgSO4 electrolytes, this linear relationship could not be observed. As discussed in the following, the regression parameters obtained for Na2SO4 electrolytes (slope KS ) 47 µF cm-2, intercept σmax ) -σmax ) 1.4 × 1017 e m-2) lie well within the d 0 range to be expected: The Stern-layer capacitance KS can be approximated independently from information on the separation between the two capacitor plates d, the dielectric constant at the interface S, and the permittivity of the vacuum 0 are known (1):

KS )

S0 d

(10)

We may assume that d is bigger than the radii of unhydrated anions (1); for SO42-, which is the dominating anion in our case, d must be > 0.23 nm (42). Furthermore, experimental evidence suggests that d < 1 nm and 6 < S < 20 (1). With 0 taken from Table 1 we thus expect 5 µF cm-2 < KS < 77 µF cm-2. The negative charge on gram-negative bacteria can be attributed to the presence of carboxyl and phosphate groups in the core-region of the lipopolysaccharides (LPS) on their surface (15, 17, 41). Indeed, we found an LPS-layer on the surface of strain B13 (18), but the number of molecules was

FIGURE 3. Charge density σd in the diffuse double layer at the electrokinetic shear-plane as a function of ζ-potentials. ζ-potentials were derived from measured electrophoretic mobilities uE and σd was calculated by integrating the Poisson-Boltzmann equation from infinity up to the electrokinetic shear-plane (1, eq 2.3.37). Data are plotted for ionic strength increasing from I ) 10-5-10-1 M from left to right. The assumption of a constant surface potential ψ0 and a constant capacity KS seems justified for Na2SO4 electrolytes and I g 10-4 M (Four points to the right; dotted regression line according to eq 7 with slope KS ) 45 ( 0.7 µF cm-2, intercept σdmax ) -σ0max ) 1.4 × 1017 ( 0.01 × 1017 e m-2, r 2 ) 1.00). not quantified. But our estimate for the maximal charge density σmax ) -1.4 × 1017 e m-2 is in the same order of 0 magnitude as LPS-densities reported for other gram-negative organisms (43, 44). The assumption of constant capacitance and constant potential close to the cell surface seems thus to be adequate for the Na2SO4 electrolytes, where ion distribution apparently follows basic electrostatic rules. Evidence for Specific Binding of Bivalent Cations to the Cell Envelope. For cells of strain B13 suspended in MgSO4 electrolytes, the assumption of constant KS and constant ψ0 is obviously wrong (Figure 3). We suggest Mg2+ cations bind to specific ligands in the cell surface, thus reducing the net charge within the electrokinetic shear plane. As a consequence, fewer counterions are expected to accumulate in the diffuse layer, resulting in a lower σd. Bivalent cations are known to bind to phosphate and carboxylate groups in the LPS-layer of gram-negative bacteria (15-17). We found additional evidence for an involvement of binding mechanisms more specific than mere electrostatics in the divergent effects of different bivalent cations on uE of strain B13 (Figure 4). Cu2+ or Pb2+ even led to a charge reversal for CMe g 0.1 mM, a finding reported previously for bacterial cells (45). For the highest concentrations of these heavy-metals, we calculated an excess density of positive charge of σ0 ≈ 0.5 × 1017 e m-2, which corresponds to about a third of σmax 0 derived from the Na2SO4 data. The high affinity of Cu2+ or Pb2+ for organic ligands, especially carboxylic acids, is known, and is a general feature of metals readily forming hydroxoand carbonato-complexes (2). It remains unclear, whether the apparent lack of an effect for CMe < 0.1 mM was due to the large excess of Na+ in the background electrolyte. The relevance of the above findings is not limited to the colloid-mediated transport of the sorbed heavy metals. Although concentrations of dissolved heavy-metals are often quite low in natural waters (2), it is important to realize that microbial surfaces offer competing ligands and that metals sorbed to these ligands might alter the deposition behavior of the entire cells drastically. This applies especially to metal contaminated aquifers or soils, where microbial travel distances might increase (in case of positively charged surfaces, e.g. Fe-oxides) or decrease (in case of negatively charged surfaces, e.g. quartz sand).

FIGURE 4. Electrophoretic mobility uE of Pseudomonas sp. strain B13 suspended in solutions containing different bivalent cations (3 mM MOPS buffer at pH ) 7.2, KCl to adjust Itot to 11 mM). An increase in Mg2+ concentration had the smallest effect, whereas Cu2+ or Pb2+ led to a charge reversal at the cell surface for CMe g 10-4 M. This indicates that these metals were bound to organic ligands in the layer of lipopolysaccharides found on the cells. Influence of Solution Chemistry on Subpopulations of Bacteria. As mentioned in the Introduction, we discovered Pseudomonas sp. strain B13 to split up in a well-adhering and in a nonadhering subpopulation (18). This fact must be taken into account in any discussion of the results presented above. For the simplified case of a bimodal distribution of the collision efficiency R, the observed influence of the cationtype on cell filtration in the sand beds might be due to either a variation in the fraction of well-adhering cells ffast or to a variation in the collision efficiency Rfast (eq 4)ssuppose the collector efficiency η remains the same, which is a fair assumption for unchanged hydrodynamic conditions. With the limited data we have, it is not possible to clearly distinguish between these two possibilities. Such a distinction would need detailed breakthrough data along the filtration path for every electrolyte tested. Nevertheless, our data suggest that solution chemistry affects the amount of well adhering cells as well as their adhesion efficiency (Figure 2). On one hand, the relative cell density C/C0 in the outlet of the long columns was almost identical to that in the short columns for I g 10 mM; the well adhering subpopulation thus seems to be completely removed under these conditions. As C/C0 for Na2SO4 electrolytes was generally higher, we can conclude that the fraction of well adhering cells is reduced in comparison to MgSO4 solutions. On the other hand, not only decreasing I resulted in an increase in C/C0 for both column lengths and both electrolytes but also the gap between the two became bigger. We may conclude that the well adhering subpopulation is not entirely removed any more on the first 3.2 cm (length of the short columns) of the filtration path. This is in contrast to our findings for I g 10 mM and hints at decrease of Rfast with decreasing ionic strength. Discussion in Terms of an Extended DLVO-Theory of Colloid Stability. Although an interpretation of bacterial adhesion data in terms of ∆G DLVO-AB is debatable (31, 35), this approach might be helpful to develop a mechanistic model of the adhesion process. The analysis of ∆G DLVO-AB can usually be confined to the region of the secondary minimum ∆Gsm (18, 35, 46). For most bacteria, repulsive forces prevent close contact with the solid surface. For high I, this is due to strong “hydrophilic” repulsion for small separation distances (∆G AB > 0 for h < 10 nm; (35, 46)), whereas strong electrostatic repulsion at larger separation dominates for low I (Figure 5). The situation is almost identical for the “classical”DLVO approach considering ∆G LW and ∆G EL only, where electrostatic repulsion leads to energy barriers of at least 40 kT in our case (not shown). Moreover, surface polymers (LPS) of several tens of nanometers in length must be assumed to prevent many bacteria from closely approaching a surface VOL. 34, NO. 6, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1015

TABLE 5. Characteristics of Calculated Curves for ∆G DLVO-AB (h) electrolytea salt

I [mM]

∆Gsmb [kT]

hsmc [nm]

rsmd

Na2SO4

1 10 100 1 10 100

-0.05 -0.81 -4.01 -0.06 -0.91 -4.01

124 27 11 108 25 11