Environ. Sci. Technol. 2010, 44, 2406–2411
Ultrafiltration Membranes Incorporating Amphiphilic Comb Copolymer Additives Prevent Irreversible Adhesion of Bacteria †
‡
ATAR ADOUT, SEOKTAE KANG, AYSE ASATEKIN,§ ANNE M. MAYES,| AND M E N A C H E M E L I M E L E C H * ,† Department of Chemical Engineering, Environmental Engineering Program, Yale University, New Haven, Connecticut 06520-8286, Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada T6G 2W2, and Department of Chemical Engineering and Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
Received September 24, 2009. Revised manuscript received January 19, 2010. Accepted February 10, 2010.
We examined the resistance to bacterial adhesion of a novel polyacrylonitrile (PAN) ultrafiltration membrane incorporating the amphiphilic comb copolymer additive, polyacrylonitrile-graftpolyethylene oxide (PAN-g-PEO). The adhesion of bacteria (E. coli K12) and the reversibility of adhered bacteria were tested with the novel membrane, and the behavior was compared to a commercial PAN ultrafiltration membrane. Under static (no flow) bacterial adhesion tests, we observed no bacterial adhesion to the PAN/PAN-g-PEO membrane at all ionic strengths tested, even with the addition of calcium ions. In contrast, significant adhesion of bacterial cells was observed on the commercial PAN membrane, with increased cell adhesion at higher ionic strengths and in the presence of calcium ions. Under crossflow filtration conditions, initial bacterial deposition rate increased with ionic strength and with addition of calcium ions for both membranes, with generally lower bacterial deposition rate with the PAN/PAN-g-PEO membrane. However, deposited bacteria were readily removed (between 97 and 100%) from the surface of the PAN/PAN-gPEO membrane upon increasing the crossflow and eliminating the permeate flow (i.e., no applied transmembrane pressure), suggesting reversible adhesion of bacteria. In contrast, bacterial adhesion on the commercial PAN membrane was irreversible, with approximately 50% removal of adhered bacteria at moderate ionic strengths (10 and 30 mM) and less than 25% removal at high ionic strength (100 mM). The resistance to bacterial adhesion of the PAN/PAN-g-PEO membrane was further analyzed via measurement of interaction forces with atomic force microscopy (AFM). No adhesion forces were detected between a carboxylated colloid probe and the PAN/PAN-g-PEO membrane, while the probe exhibited strong adhesion to the * Corresponding author phone: (203)432-2789; e-mail:
[email protected]. † Yale University. ‡ University of Alberta. § Department of Chemical Engineering, MIT. | Department of Materials Science and Engineering, MIT. 2406
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 7, 2010
commercial PAN membrane, consistent with the bacterial adhesion tests. The exceptional resistance of the PAN/PAN-gPEO membrane to bacterial adhesion is attributable to steric repulsion imparted by the dense brush layer of polyethylene oxide (PEO) chains.
Introduction A major challenge in long-term operation of membrane filtration systems is the growth of biofilm on membrane surfaces, or biofouling (1). Biofouling leads to the use of higher operating pressure, more frequent chemical cleaning, and shorter membrane life. Designing membranes that resist biofouling is, therefore, of utmost importance for sustainable use of membrane technology in water and wastewater treatment as well as desalination and wastewater reuse. The key stages in biofilm formation are bacterial deposition and irreversible adhesion, formation of microcolonies, and biofilm maturation (2, 3). Most of biofouling control techniques have been developed to prevent or retard one or more of these stages. For example, membrane surfaces have been modified to be hydrophilic, negatively charged, and/or smooth to minimize the initial adhesion of bacteria (4-7). To inactivate irreversibly adhered microorganisms and prevent the formation of microcolonies, various antimicrobial strategies have been investigated, including coatings that incorporate Ag or photobactericidal metal oxide nanoparticles (8-11), self-assembling peptides that disrupt bacterial membranes (12, 13), and quorum quenching enzymes that obstruct microcolony signaling linked to biofilm growth (14). Bacterial deposition in membrane systems is inevitable due to the ubiquitous convective permeate flow and resulting permeation drag force during membrane filtration. While coatings that employ natural and engineered nanomaterials with antimicrobial activity show promise for reducing bacterial adhesion (9, 15), such agents cannot maintain their antimicrobial activity after adsorption of organic foulants such as natural organic matter and soluble microbial products (9). An ideal antibiofouling membrane should not only provide effective hindrance of bacterial adhesion and cell inactivation, but also ensure that the deposition of microorganisms on the membrane surface is reversible. To minimize the strength of microbial adhesion and to ensure that microbial adhesion to the membrane surface is reversible, various types of polymers have been grafted on membrane surfaces, including poly(ethylene oxide) (PEO), poly(2-hydroxyethyl methacrylate), poly(2-hydroxyethyl acrylate), poly(acrylic acid), poly(acrylamide), poly(N,N-dimethylacrylamide), phosphorylcholine (PC), poly(2-acrylamido2-methyl-1-propanesulfonic acid), and poly(2(dimethylamino)ethyl methacrylate) (7, 16, 17). Among these polymers, PEO has been shown to be a very effective material to prevent adhesion of biomacromolecules due to its hydrophilicity, large excluded volume, and unique ability to coordinate surrounding water molecules in an aqueous medium (18-20). For example, we have investigated the use of amphiphilic graft/comb copolymers having hydrophobic backbones (polyvinylidene fluoride, PVDF) and hydrophilic PEO side chains (poly(oxyethylene) methacrylate) as surfacemodifying membrane coatings (21). More recently, the antifouling property of a novel polyacrylonitrile (PAN) ultrafiltration (UF) membrane incorporating the amphiphilic comb copolymer additive, polyacrylonitrile-graft-polyethylene oxide (PAN-g-PEO), has been successfully demonstrated with bovine serum albumin, sodium alginate, and humic acid as model organic foulants (19, 22). 10.1021/es902908g
2010 American Chemical Society
Published on Web 03/01/2010
In this study, we examined the resistance of the PAN/ PAN-g-PEO ultrafiltration membrane to bacterial adhesion and, more importantly, the reversibility of bacterial adhesion to the membrane. A Gram negative bacterial strain, Escherichia coli K12 MG1655, chromosomally tagged with green fluorescence protein, was used as a model bacterium. Static adhesion tests and crossflow membrane filtration experiments with a system that allows direct observation of cell deposition on the membrane surface demonstrated the superb resistance of the membrane to bacterial adhesion. Measurements of interaction forces with atomic force microscopy (AFM) were used to elucidate the mechanisms governing the resistance of the PAN/PAN-g-PEO membrane to bacterial adhesion.
Materials and Methods Ultrafiltration Membrane. A novel fouling resistant ultrafiltration membrane, denoted PAN/PAN-g-PEO, was fabricated following two steps. First, a polyacrylonitrile-graftpoly(ethylene oxide) (PAN-g-PEO) comb copolymer was synthesized by free radical polymerization of acrylonitrile and poly (ethylene glycol) methyl ether acrylate (PEGA, Mn)454 g/mol). Next, the PAN/PAN-g-PEO blend membrane was prepared from a mixture of polyacrylonitrile (PAN) and PAN-g-PEO solutions by precipitation followed by annealing in a water bath. Details on the preparation and characterization of the PAN/PAN-g-PEO membrane are given elsewhere (19, 22). A commercial PAN ultrafiltration membrane (Osmonics, Minnetonka, MN), with a similar hydraulic permeability as that of the PAN/PAN-g-PEO blend membrane, was selected for comparison. The commercial PAN membrane is made from polyacrylonitrile that has been modified to possess a hydrophilic surface with minimal surface roughness (23). The manufacturer rates the commercial PAN membrane as having a nominal molecular weight cutoff of approximately 45 kDa at 90% retention of dextran. Measurement of Membrane Contact Angle and Zeta Potential. Contact angle measurements were performed using a VCA-2500 system (AST Products, Bellerica, MA). Prior to measurement, the membrane coupons were air-dried overnight. The dried membrane coupon was then adhered to a glass slide by a double-sided tape and was brought near the tip of the syringe. DI water was injected with the syringe until a water droplet formed on the membrane surface (∼2.5 µL per droplet). About 10 contact angle measurements were performed across the membrane coupon at equally spaced intervals. Each measurement provided left and right contact angles from a cross-sectional view of a water droplet placed on the membrane surface. There was no difference between the left and right contact angles, and the average of 10 measurements (20 values) was calculated. An asymmetric clamping cell was employed to measure the streaming potential of the membranes (EKA, Brookhaven Instruments, Holtsville, NY). Streaming potential measurements were taken 8 times with alternating flow direction of the 10 mM NaCl solution (pH∼5.5, unadjusted). Detailed experimental procedure and the method to calculate the zeta potential from the measured streaming potential are given elsewhere (24). Model Bacteria. Escherichia coli K12 MG1655 was used as a model bacterial cell (25). The strain was kindly received from S. Molin, the Technical University of Denmark. To allow live cell detection under fluorescent microscopy, bacterial strains were tagged with a plasmid coding for green fluorescent protein (MG1655 was introduced with the suicide plasmid pSM1696). The E. coli cells were incubated and harvested at midexponential growth phase in Luria Bertani (LB) broth with 50 mg/L kanamycin at 37 °C. The bacterial suspension was centrifuged (Sorvall SS-34) at 15,000 rpm for
3 min and then washed/resuspended with 154 mM NaCl solution. The centrifugation and resuspension were repeated three times to remove organic and inorganic impurities. After the last centrifugation, the bacterial cell pellet was resuspended in the electrolyte solution which was used in the experiment and was vortexed shortly. The electrophoretic mobility of the bacterial cells was measured as a function of ionic strength at ambient pH (∼5.5) and room temperature (22 °C) using ZetaPALS (Brookhaven Instruments, Holtsville, NY). Electrophoretic mobilities were then converted to zeta potentials using the Smoluchowski equation because of the relatively large size of the bacterial cells and the ionic strength used. Solution Chemistries. Four different electrolyte solutions were used for the bacterial adhesion and deposition experiments: 10, 30, and 100 mM NaCl, and a combination of 27 mM NaCl and 1 mM CaCl2 (total ionic strength of 30 mM). Chemicals were ACS grade (Fisher Scientific, Pittsburgh, PA). The electrolyte solutions were prepared with deionized water (Millipore, MA). Solutions were used without pH adjustment, with the ambient pH ranging from 5.5 to 5.7. Static Bacterial Adhesion Tests. Two membrane coupons (each with dimensions of 1 cm by 1 cm) were placed in 20 mL cell suspension (4 × 107 cells/mL) at the desired electrolyte solution. The cells and the membranes were incubated in a shaker (Lab-line 4631 Maxi Rotator) at 20 rpm and room temperature (22 °C) for 1 h. The membrane coupons were then rinsed gently with a bacteria-free electrolyte solution having the same concentration as that used for the adhesion test to remove weakly bound cells. Membrane coupons were then observed under a fluorescent microscope (Olympus BX41, Japan), and at least 10 images were taken across the membrane surface. The average number of cells on the membrane was normalized by the observed membrane area (0.145 mm2). Direct Microscopic Observation of Bacterial Deposition and Reversibility. The crossflow membrane filtration unit for direct observation of bacterial deposition is depicted in Figure S1. We have used such a system previously for studying protein fouling of ultrafiltration membranes (19). Briefly, the feed solution (2 L) was pressurized to 150 kPa (1.5 bar) and circulated by a gear pump (Cole-Parmer, Vernon Hills, IL) at a fixed crossflow velocity of 10 cm/s in a crossflow channel (9 cm long, 3 cm wide, and 0.1 cm high). The permeate flux was kept constant at 3 × 10-5 m/s (108 L/m2 · h) during the runs by an 8-roller digital peristaltic pump (Cole-Parmer, Vernon Hills, IL) mounted on the permeate line. The membrane was first compacted and equilibrated for 50 min with bacteria-free electrolyte solution that had solution chemistry identical to that used in the subsequent fouling runs (10, 30, and 100 mM NaCl, and 27 mM NaCl plus 1 mM CaCl2). An appropriate amount of bacterial cells was then added to the feed tank to achieve cell concentration of ∼1.3 × 105 cells/mL. The bacterial deposition experiments were carried out for 30 min at a constant permeate flux, with an image of the membrane taken every 5 min. The deposition rate coefficient of bacterial cells was calculated from the slope of the curve plotting the number of deposited cells per unit membrane area versus time divided by the number concentration of cells in the influent (26). Bacterial deposition runs were carried out at least three times for each electrolyte condition at an ambient pH (unadjusted, pH of 5.5-5.7) and a temperature of 22 °C. At the conclusion of the fouling runs, physical cleaning experiments were preformed by increasing the crossflow velocity from 10 to 150 cm/s while turning off the permeate pump (relaxing the transmembrane pressure) for 10 min, without changing the feed solution or the pressure (150 kPa) in the system. After 10 min of physical cleaning, images of the membrane were taken. For cases where we had less than VOL. 44, NO. 7, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2407
FIGURE 1. Electrophoretic mobility (EPM) and calculated zeta potential of E. coli cells at different ionic strengths and composition. Experimental conditions: unadjusted (ambient) pH of 5.5-5.7, temperature of 22 °C. ∼20 cells in the image, another 7 images at different locations of the membrane were taken. Cleaning efficiency was calculated as the percentage of the number of cells remaining on the membrane after the 10 min of cleaning compared to the number of cells on the membrane after the 30 min deposition run (just before starting the cleaning). All reversibility tests were performed at least three times. Interaction Force Measurements. A Nanoscope IIIa Multimode atomic force microscope (AFM) (Digital Instruments, Santa Barbara, CA) was used to quantify the interaction forces between foulant and membrane surfaces. A carboxylate-modified latex (CML) particle (Interfacial Dynamics Corp., Portland, OR), 3.9 µm in diameter, was glued to the end of a commercial 0.06 N/m SiN tipless cantilever (Veeco Metrology Group, Santa Barbara, CA). The particle was used as a surrogate for the bacteria cells because it carries negative carboxylic functional groups similar to biomacromolecules (19, 22, 27) and bacterial cells (28). The AFM was operated in force mode, with an approach/retraction speed of 1 µm/s and 1 µm of piezo-movement. The experiments were performed in a liquid cell, and the system was allowed to equilibrate for 60 min for each experimental condition. Raw data were converted from cantilever deflection and z-piezo position into force-distance curves. The force was then normalized to the radius of the CML particle, R. The detailed experimental procedures for using the AFM colloidprobe technique to quantify interaction forces relevant to organic fouling are given in our previous publications (29-31). Force measurements were performed at four different locations on the membrane surface, with 20 measurements at each location to minimize inherent variability in the force data. The solution chemistries of the test solutions in the liquid cell were the same as those used in the adhesion and deposition experiments.
Results and Discussion Bacterial Electrokinetic Properties. The electrophoretic mobility (EPM) and corresponding calculated zeta potential for the E. coli cells as a function of solution ionic strength and the presence of divalent (calcium) cations are presented in Figure 1. Measured electrophoretic mobilities (EPM) followed expected behavior of decreasing in magnitude with increasing ionic strength due to compression of the electric double layer. Addition of divalent Ca2+ ions further decreased the magnitude of bacterial cell EPM, likely due to specific interaction with bacterial surface functional groups and charge neutralization. The relatively high residual electrophoretic mobility at high ionic strength is indicative of the presence of the “soft” polyelectrolyte layer at the bacterial surface (32). 2408
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 7, 2010
FIGURE 2. Number of E. coli cells attached on the unit membrane surface (per mm2) at various solution ionic strengths. Membrane coupons were incubated for 1 h with ∼4 × 107 cells/mL. Experimental conditions: unadjusted (ambient) pH of 5.5-5.7, temperature of 22 °C. No cell attachment to the PAN/ PAN-g-PEO membrane was observed. Bacterial Adhesion and Reversibility. Under static, batch conditions (no pressure, no flow), there is a significant difference in the affinity of the bacterial cells to the PAN/ PAN-g-PEO and the commercial PAN membranes (Figure 2). The number of attached bacterial cells on the commercial PAN membrane increased as the ionic strength of the solution increased. Similarly, the addition of Ca2+ ions resulted in an increase of the number of attached cells to the commercial membrane, compared to the case with no Ca2+ but a similar total ionic strength (30 mM). The observed behavior with the commercial PAN membrane agrees with the trends observed with the EPM of E. coli cells (Figure 1), thereby suggesting that electrostatic interactions govern the cell adhesion behavior. In a stark contrast, no bacteria attached to the PAN/PAN-g-PEO membrane at any of the solution conditions investigated. This observation suggests that the PAN/PAN-g-PEO membrane has strong antifouling properties that prevent irreversible adhesion of bacteria at the surface. Further discussion on the bacterial adhesion mechanism, supported by AFM measurements, is given later in the paper. Bacterial Deposition Rate and Reversibility in Crossflow Filtration. The bacterial deposition rate coefficients on both membranes, determined from the crossflow filtration experiments, are presented in Figure 3a. For both the commercial and the PAN/PAN-g-PEO membranes, the deposition rate increases continuously with increasing ionic strength and following the addition of 1 mM of calcium ions at a fixed total ionic strength of 30 mM. The deposition rate on the commercial PAN membrane is higher than that on the PAN/ PAN-g-PEO membrane at ionic strengths of 10 and 30 mM, but attains similar values at high ionic strength (100 mM). The deposition of bacterial cells on the membrane surface is determined by the interplay between permeation drag resulting from the convective permeate flow and the repulsive interactions between the bacteria and the membrane surface at close separations. The latter interactions include electrostatic interactions as well as steric forces for the PAN/ PAN-g-PEO membrane as discussed in detail later. As the ionic strength increases, the range and magnitude of the electrostatic repulsion is decreased, and bacterial cells can either attach to the membrane or be held very close to the membrane surface by the permeate drag force that acts continuously on the cells. Note that the permeation flow in ultrafiltration is relatively high compared to reverse osmosis or nanofiltration membranes (3 × 10-5 m/s or 108 L m2- h-1 in our case), resulting in a significant permeation drag force. The deposition or accumulation of bacteria on the UF membrane surface is inevitable because of the ubiquity of the permeation drag force during membrane filtration. A key aspect to investigate is the reversibility of bacterial deposition on the membrane surface. Because UF membrane
FIGURE 4. Distributions of adhesion forces (at contact or zero separation) between the AFM colloid probe and the PAN/ PAN-g-PEO membrane. Solution conditions are the same as those used in the corresponding cross-flow cell deposition and static adhesion experiments. Solid lines represent the fit of Gaussian distribution.
FIGURE 3. (a) Observed cell deposition rate coefficients of E. coli during 30 min of crossflow filtration. Deposition experiments were carried out at a crossflow velocity of 10 cm/s, permeation velocity of 30 µm/s, and cell concentration of ∼1.3 × 105 cells/mL at a constant pressure of 150 kPa (1.5 bar). (b) Percent removal of E. coli cells from the membrane after 10 min of rinsing with the same solution at 150 cm/s cross-flow velocity and with no permeate flow at 150 kPa (1.5 bar). systems involve frequent physical cleaning of the membrane (with no chemicals), bacteria that reversibly deposit on the membranes will be readily washed away, thus preventing bacterial growth and subsequent biofilm formation. Figure 3b presents the percent of deposited cells that were removed from the surface when subjected to “physical cleaning” at elevated crossflow (150 cm/s) with no permeate flow (or applied transmembrane pressure). We observe relatively low removal of deposited bacteria from the commercial membrane (at most ∼50%) compared to above 97% and often close to 100% for the PAN/PAN-g-PEO membrane. Furthermore, while the removal efficiency from the commercial PAN membrane is dependent on ionic strength, dropping to less than 25% at 100 mM, the removal of deposited bacteria from PAN/PAN-g-PEO membrane is independent of ionic strength. As we discuss later, bacteria attach irreversibly on the commercial PAN membrane in deep energy wells because no repulsive forces exist at very close separations between the bacteria and the membrane surface. On the other hand, strong repulsive forces between the bacterial cells and the PAN/PAN-g-PEO membrane exist very close to the membrane surface, and once the permeation drag force is eliminated (when permeate flow vanishes), the bacterial cells near the membrane surface are washed away by the crossflow. Why Is Bacterial Adhesion to the PAN/PAN-g-PEO Membrane Negligible? As our previous studies demonstrated (19, 29, 31), atomic force microscopy (AFM) allows both quantitative measurement of adhesion forces and mechanistic understanding of foulant-membrane interaction. We have also shown that F/R, that is the adhesion (pull-off) force (F) normalized by the radius of the colloid/foulant probe (R), serves as an indicator for the fouling propensity of polymeric membranes. Bacterial adhesion behavior observed
in the static adhesion tests (Figure 2) and the measured adhesion forces (Figure 4) are in general agreement with our previous findings. The results imply that the presence of adhesive forces (F/R < 0) between the model colloid probe and the membranes is an indicator for bacterial adhesion potential, as we showed earlier for adhesion of biomacromolecules (19, 31, 33). For the commercial PAN membrane, relatively strong adhesion forces (i.e., negative values of F/R) are measured for the entire range of ionic strength investigated (Figures S2 and S3), thereby explaining the high number of attached cells on the membrane surface. In contrast, no adhesion forces (i.e., no negative values of F/R) were detected between the CML probe and the PAN/PANg-PEO membrane (Figure 4, Figures S2 and S3). As shown in Figure 4, F/R values (or adhesion events) were distributed between zero to positive values for all solution chemistries tested, with or without calcium, which explains the absence of cell attachment on the PAN/PAN-g-PEO membrane surface. Note that the positive adhesion force values in Figure 4 correspond to the repulsive force at zero separation, as no negative (adhesion) events were detected. The observed cell deposition kinetics of E. coli cells under crossflow membrane filtration conditions were similar for the commercial PAN membrane and PAN/PAN-g-PEO blend membrane (Figure 3a). This observation is attributable to the permeation drag force which acts on the cells when the filtration system is under pressure. Cells are transported from the bulk suspension toward the membrane by the convective permeate flow and subsequently deposit on the membrane surface. On the commercial PAN membrane surface, the deposition of cells resulted in irreversible attachment (Figure 3b), supported by the strong adhesion forces measured with AFM. On the other hand, on the PAN/PAN-g-PEO membrane surfaces, cells are only held on the membrane surface by the permeation drag force. Consequently, when the permeation flow is stopped (by relaxing the applied transmembrane pressure), the deposited E. coli cells diffuse back to the bulk solution because only repulsive forces exist at contact with the membrane surface (Figure 4). Figure 3b demonstrates the near complete removal (detachment) of E. coli cells from the membrane when subjected to elevated crossflow velocity (150 cm/s) for 10 min with no permeation flow (or applied transmembrane pressure). For the commercial membrane, removal of deposited cells at higher crossflow velocity was not effective (
