Transport and viability of Escherichia coli cells in clean and iron oxide coated sand following coating with silver nanoparticles.
Bryne T. Ngwenya1*, Philip Curry1, Leon Kapetas2.
1School of Geosciences, University of Edinburgh, John Murray Building, James Hutton Road, Edinburgh EH9 3FE
2Department of Geoscience & Engineering, Technical University of Delft, Stevinweg 1/ PO-box 50482628 CN Delft/2600 GADelft, The Netherlands.
*Corresponding author ()
As accepted on
13th May 2015
Journal of Contaminant Hydrology
Abstract: A mechanistic understanding of processes controlling the transport and viability of bacteria in porous media is critical fordesigning in situ bioremediation and microbiological water decontaminationprograms. We investigated the combined influence of coating sand with iron oxide and silver nanoparticles on the transport and viability of Escherichia coli cells under saturated conditions. Results showed thatiron oxide coatings increase cell deposition which was generally reversed by silver nanoparticle coatings in the early stages of injection. These observations are consistent withshort-term, particle surface charge controls on bacteria transport, where a negatively charged surface induced by silver nanoparticles reverses the positive charge due to iron oxide coatings, but columns eventually recovered irreversible cell deposition. Silver nanoparticle coatings significantly increased cell inactivation during transit through the columns. However, when viability data is normalized to volume throughput, only a small improvement in cell inactivation is observed for silver nanoparticle coated sands relative to iron oxide coating alone. This counterintuitive result underscores the importance of net surface charge in controlling cell transport and inactivation, and implies that the extra cost for implementing silver nanoparticle coatings on porous beds coated with iron oxides may not be justified in designing point of use water filters in low income countries.
Keywords: Bacteria transport, silver nanoparticles, iron oxide coated sand, retention, zeta potential.
1. Introduction
A thorough understanding of the processes affecting the transport and viability of bacteria in porous media is critical in solving practical problems ranging from in situ bioremediation to water supply and health (Sayler et al., 2000; Weiss et al., 2005; Mthombeni et al., 2012). Research efforts to date have shown that the transport of bacteria in porous media is affected by a variety of physical, chemical and biological factors. Chemical factors relate mainly to fluid composition, and include ionic strength (Kim et al., 2009; Wang et al., 2011), type and content of natural organic matter (Yang et al., 2012), presence ofclay particles (Vasiliadou & Chrysikopolous, 2011) and nature of dissolved metal ions (Kim & Walker, 2009; Kapetas et al., 2012). Properties of the porous medium are dominated by grain shape, size, composition and surface charge (Dong et al., 2002; Syngouma & Chrysikopolous, 2011; Kapetas et al., 2012). Biological properties relate to those of the bacteria and include cell types (Chen & Walker, 2012),motility (Camesano & Logan, 1998; De Kerchove & Elimelech, 2008),growth phase (Walker et al., 2005), surface macromolecules (Liu et al., 2007; Tong et al., 2010) and cell viability (Kuzmar & Elimelech, 2005; Asadishad et al., 2013a).The overarching conclusion that arises from studying all these factors is that for a given set of physical parameters of the porous medium (grain shape, size and packing), transport of bacteria is controlled by the magnitude of surface charge of the bacterial cell relative to that of the porous medium grains.
Meanwhile, lack of access to clean, potable drinking water is a worldwide issue, with the World Health Organisation estimating at least 1.8million deaths a year due to drinking water contaminated with bacteria (WHO, 2004). These problems are particularly acute in rural areas of developing countries lacking large-scale water treatment infrastructure (Grabow, 1996), as well as during disasters when potable water supplies are in short supply due to contamination and damage to infrastructure (Faruque et al., 2005; Roig et al., 2011). In both cases, small scale, potable water treatment technologies under the banner of point of use technologies (Sobsey et al., 2008) have become the default choice for providing access to clean drinking water (Loo et al., 2012).
Despite their difficulties with deployment, biosand filters, constructed by filling a container with sand and/or gravel (Mahmood et al., 2011) constitute one of the cheapest such technologies (Loo et al., 2012). It has been shown that their microbiological removal performance can be improved by coating the sand with iron oxides (Murphy et al., 2010; Ahammed & Davra, 2011), which increases attachment efficiencies of cells (Ahammed & Davra, 2011). Similar improvements in performance were also shown by coating the sand with silver nanoparticles (Mahmood et al., 1993), while alternative filtration devices incorporating silver nanoparticles ((Jain and Pradeep, 2005; Oyanedel-Craver and Smith, 2008; Lv et al, 2009; Dankovich and Gray, 2011; Lin et al, 2013) have also been shown to provide improved bactericidal properties by supplying ionic silver and through production of reactive oxygen species (Jain and Pradeep, 2005; Savage and Diallo, 2005; Kim et al, 2007; Lv et al, 2009; Marambio-Jones and Hoek, 2010; Lin et al, 2013; Mpenyana-Monyatsi et al, 2012).
In this study, we tested whether coating iron oxide coated sand with silver nanoparticles could improve its bactericidal properties, based on the hypothesis that the improved cell attachment due to iron oxide coating (Johnson et al., 1996; Ryan et al., 1999; Li et al., 2004; Abudalo et al., 2005; Ahamed & Davra, 2011; Metge et al., 2011; Kapetas et al., 2012) increases the effective exposure time of cells to nanoparticles. We used breakthrough analysis of Escherichia coliJM109 cell transport in columns filled with sand containing different coatings, with clean sand as control, coupled with viability assessment of effluent cells.
2. Methodology
2.1 Sand preparation and characterisation
General-purpose silica sand from Fisher Scientific was sieved to collect the 120-350µM fraction and first heated in an oven at 450ºC for 4 hours to remove organic matter, followed by soaking in 20% v/v nitric acid to desorb trace metals (Mpenyana-Monyatsi et al., 2012). The sand was then rinsed in deionized water repeatedly to remove any fine sediment and raise the pH back to neutral. The sand was dried overnight at 60ºC and a portion kept for use in control experiments (CS). The rest was treated by coating with iron oxide and silver nanoparticles as described below.
A portion of the cleaned sand was coated with silver nanoparticles (CS-NP) by first soaking in 1M ammonia solution to raise the pH to above 9 and hence deprotonate silanol functional groups on the surface ((Kim et al, 2007). The deprotonated sand was soaked in 5mM silver nitrate solution at a mass to volume ratio of 1 to 8 overnight, which led to silver ions adsorbing to the deprotonated sand (Dankovich and Gray, 2011; Mpenyana-Monyatsi et al, 2012). The silver adsorbed to the sand was then reduced to nanoparticulate silver by exposure to UV light overnight (Huang et al., 1996; Spadaro et al., 2010) while still soaked in silver nitrate solution. Subsequently, the solution was decanted and unadsorbed nanoparticles were removed by repeated washing in deionized water until the rinses were clear before drying the coated sand overnight at 60ºC.
Iron oxide coated sand (IOCS) was made following Kapetas et al (2012), by mixing 30g of Fe(NO3)3.9H2O in 300ml of deionized water and titratingdrop wisewith NaOH to pH of 6 (Yee & Fein, 2002). The suspension was left to mix overnight, after which the supernatant was drained and the sand washed with deionized water until the supernatant was clear, before drying the coated sand at 60ºC overnight.
Two further portions of sand were prepared. In one (IOCS-NP), silver nanoparticles were precipitated on a portion of the iron oxide coated sand by the same treatment as that for nanoparticle coated clean sand. In another portion nanoparticle-coated clean sand was further coated in iron oxides to produce NP-IOCS. The rationale behind this was to test whether the order in which nanoparticles and iron oxides were coated affected both cell transport and viability.
Coatings were characterized using a Field Emission Gun-scanning electron microscope (FEG-SEM) on carbon coated samples mounted on conducting tape. The analysis was carried out on a CarlZeiss Sigma HDVP microscope at an accelerating voltage of 5 kV.Silver nanoparticle suspensions decanted from the UV-treated sands were subjected to zeta potential measurements following dilution in 10mM NaClat pH 7 using the Malvern Zetasizer Nano ZS. The different types of sand were also analysed for their zeta potential on suspensions of ~300mg/L in 10mM NaCl adjusted to pH 7. The instrument was set up to measure two readings, each of which consisted of 20 separate scans.
2.2. Column flow experiments and sampling
All packing and experimental manipulations were conducted in a laminar flow cabinet in order to maintain sterile conditions. Column preparation and packing followed the method of Kurlanda-Witek et al (2014). Glass columns (12cm in length and 1cm diameter) with matching top and bottom end-caps and fittings (Diba Omnifit) were used. All tubing and column parts used in the experiment were autoclaved and dried under UV light prior to setting up the experiment. Fluid flowed in (bottom) and out of the column through 1/16 inch outer diameter (OD) and 1/8 inch internal diameter (ID) PTFE tubing with ¼-28 tpi UNF fittings (PP) in both end caps. The 1/16 inch tubing was further connected to L/S 13 platinum-cured silicone pump tubing, 5mm OD and 0.8mm ID (Masterflex). Columns were packed with porous media in 10mM NaCl electrolyte (adjusted to pH 7 using 0.1M NaOH/HCl) using the wet packing method(Deshpande and Shonnard, 1999). Glass beads of 0.5mm diameter were placed at the top and bottom to prevent the fine sand from clogging the inflow and outflow of the column. Electrolyte and bacteria suspensions made in 10mM NaCl and adjusted to optical density (OD) of 0.2, chosen to prevent excessive pore clogging in the columns,were injected at a constant flow rate of 0.4ml/min. The column was primed with the background electrolyte for an hour, equivalent to ~7 pore volumes (PV) before initiating the injection of a bacteria suspension.E. coliwas chosen as it is often used in studies as an indicator of fecal contamination in water (Dankovich and Gray, 2011). Cells were grown in 1L of nutrient broth at 30°C on a shaking table and washed after 24 hours (early stationary phase). The input suspensionwas continuously mixed throughout each experiment using a magnetic stirrer. Bacteria injection was restricted to less than 12 hours to avoid having to constantly change the input suspension as cell death in the influent suspension started around this time and replenishing influent solutions would have changed the reference point for viability assessments.
Effluents were collected every 4 minutes. The absorbance was measured on 1 ml of the effluent using a Camspec M501 single beam scanning UV/visible spectrophotometer at a wavelength of 600nm to determine optical density of the bacteria suspension as a basis for determining bacterial cell breakthrough.Due to the large number of effluent samples generated in each experiment (100+), we used the bacteria breakthrough curve to select samples at three critical time points for analysis: at time zero when flow switched to cell suspension(to), when OD was ~50% of the influent (t0.5) and at end of the experiment when OD exceeded 70% of the influent end (tend). The last criterion was used because for the iron oxide coated sands, we were not able to attain full breakthrough of bacteria within the 12-hour time window constrained by influent cell viability. Influent suspensions were also collected and analysed at the same time points to monitor temporal changes in input cell viability. Column parameters were calculated from the breakthrough curve of a conservative tracer injected at the same flow rate as for bacteria suspensions. Bromothymol blue dye was used as a tracer with its breakthrough measured using the same UV-Visible spectrometer at550 nm(Kurlanda-Witek et al., 2014).
Since the focus of our study was to characterise the viability of cells in the context of water disinfection through filters containing sand doped with iron oxides and silver nanoparticles, all our experiments were run with only the rising limb (step input) of the breakthrough curve (see also Mthombeni et al., 2012). However, we run an additional experiment with one of the iron oxide coated sands containing silver nanoparticles (IOCS-NP) that included the falling limb in order to investigate whether cell attachment in these columnswas also irreversible,as observed for iron oxide coated sand alone (e.g. Scholl & Harvey, 1992; Abudalo et al., 2005; Metge et al., 2011). For this experiment, influent injection was switched back to the pure electrolyte immediately after the rising limb breakthrough curve attained C/Co = 0.5. Subsequently, sampling continued as for the other tests but cell viability tests were not carried out on effluent suspensions.
2.3. Cell viability in influent and effluents
Cell viability in influent and effluent fluids was determined by plating on agar to measure colony forming units (CFU/ml) and by fluorescence microscopy following LIVE/DEAD staining. To determine CFU, 0.1ml of each sample was diluted serially to 10-4 and 20 l of the diluted suspension was plated in duplicate on nutrient agar. The plates were incubated overnight at 37°C. Fluorescence microscopy was carried out using a Zeiss AxioImager Z1 microscope following staining of samples with Cyto9 and propidium iodide (Invitrogen), using FITC (490/530 Ex/Em) and TRITC (547/572 Ex/Em) filters. Samples were prepared by mixing 1 ml of the suspension with 3μl of the combined (premixed) stain in sterile Eppendorf tubes and incubating for 15 minutes (Zhu and Xu, 2013). Stained suspensions were centrifuged for 10 minutes and rinsed threetimes in 10mM NaCl in order to remove the stain and hence avoid artefacts due to cell death post-staining. Centrifuging also generated a dense pellet for spreading on a glass slide coated with 1% agarose gel to immobilize cells during imaging.
2.4.Celltransport modelling
We used HYDRUS-1D (Šimůnek et al., 2009) to model the transport of E. coli cells in the different sand columns using the OD data. The data was fitted to an advection-dispersion equation (ADE) including sorption terms and attempts were made to account for different processes that may contribute to the shape of the breakthrough curve (Bradford et al., 2007; Foppen et al., 2007; Tufenkji, 2007). One or two site models were tested including different bacterial removal processes such as straining, depth dependent sorption, ripening or Langmuir-dynamics adsorption.Based on the lowest objective function and minimal amount of optimized parameters (most parsimonious model), we found that we needed two different models to fit the different types of sands. Transport in clean sands (CS and CS-NP) was modelled witha reversible (attachment-detachment)interaction with one site, hence (Tufenkji, 2007):
(1)
(2)
Data for iron oxide-coated sands (IOCS, IOCS-NP and NP-IOCS) required a two site model involving irreversible attachment to one site (site 2, S2) and a Langmuir adsorption process (Tobler et al., 2014), thus:
(3)
(4)
In these formulations, C is the concentration (normalised OD) in the fluid, x is distance (outlet) along the column, v is the average linear flow velocity, ka is the attachment rate coefficient, kd is the detachment rate coefficient, S is the concentration of cells on the porous medium (with those on site 1 designated as S1 where Smax1 is the maximum permissible adsorbed cells to S1), k1is the Langmuir coefficient,k2is the attachment rate coefficient for S2, ρbis the dry bulk density of the porous medium, εis the porosityand D is the hydrodynamic dispersion coefficient, defined thus:
(3)
where is the dispersivity andD0is the molecular diffusion or bacterial motility coefficient (Tobler et al., 2014). Based on the ADE model fit to the tracer data, the value of the hydrodynamic dispersion coefficient was calculated and was kept constant across models. A porosity value of 40% was calculated experimentally and confirmed during the simulation of the dye tracer breakthrough curve. During modelling, the boundary conditions were set to constant flux.
3. Results
3.1 Physicochemical properties of sands
Examination of the sand showed that coatings were present on grain surfaces as manifest by differences in colour (Figure SI.1), although it is also apparent that some of the iron oxides and silver particles are not attached to grains. The SEM results show that unlike in clean sand controls (Figure 1a and 1ai, note that for all images in Figure 1, larger versions are shown in Figure SI.2 of the supporting information to highlight features described herein) lacking bright spots, silver particles of variablesizes ranging from nanoparticulate to microparticulate aggregates were present on sand grains. However, the distribution was patchy and most particles were associated with pitted grains and rough surfaces (Figure 1b and bi), suggesting that the surface particles represent trapped particles which likely precipitated homogeneously rather than through the nucleophilic substitution mechanism proposed by Kim et al (2009).Iron oxide coatings were equally patchy (Figure 1c and ci, d and di) but generally resulted in a rough surface that trapped more silver particles (Figure 1e and 1i). The presence of silver and iron coatings was confirmed by qualitative EDX analysis (Figure SI.3).
Measured zeta potentials were negative for all sands produced (Figure SI.4), although values were statistically different amongst clean sand (CS = -30.2±1.2mV), nanoparticle coated clean sand (CS-NP = -14±2.5mV) and iron oxide coated sand (IOCS = -5.98±0.7 mV). The transition to more positive values when sand is coated in iron oxides is consistent with other studies (e.g. Abudalo et al., 2005). As expected (e.g. Terada et al., 2012), suspensions of E. coli have negative zeta potentials (-33.4±1.13 mV), as are silver nanoparticle suspensions (-17.3±6.3 mV). The latter imparts a more negative zeta potential to iron oxide coated sands whereas clean sand is rendered slightly more positive by silver nanoparticle coatings. These changes are likely to have significant impact on cell transport.
3.2. Cell breakthrough curves
Optical densities of the individual effluent samples (symbols) normalized to the mean of the influent samples at the three different time points are plotted against pore volume (PV) in Figure 2.Lines are fitted curves based on transport modelling (section 3.4), except for the column that included the falling limb (IOCS-NP-R) which was not modelled and where the line connects the data points. E. coli transport in clean sand is only marginally slower than the conservative tracer with breakthrough at 1.6PV, breakthrough being defined as pore volume at 50% of signal and equals 1 for the conservative tracer (Fetter, 1998).