Nanoparticles Toxicity on Escherichia coli: Batch and Kinetic Approach

Silver and dysprosium oxide are two examples of materials used for the manufacturing of nanoparticles with current and future commercial relevance, respectively. Silver nanoparticles (nAg or AgNPs) are one of the most commonly used nanomaterials in consumer products and medical applications due to their antimicrobial properties. Dysprosium oxide nanoparticles (nDy2O3) are gaining interest for biomedical applications because of their fluorescence and paramagnetic properties, which can be used as contrast agents in magnetic resonance analysis. However, the fate of nAg and nDy2O3 and their possible negative impacts on the environment and public health are growing concerns. Nanoparticles entering and accumulating in different environmental compartments will very likely interact with native bacteria in soil and aquatic environments. There are knowledge gaps related to: the exposure of novel nanomaterials on microorganisms in different water chemistry conditions; the effect of reactor configuration to assess nanotoxicology; and the effect of the specific growth rate on the response of microorganisms exposed to nanoparticles. In this study, nanoparticles toxicity on Escherichia coli (E. coli) was assessed under batch and continuous conditions, and evaluated their impacts on metabolic functions and cell structure such as, viability, membrane permeation, respiration, growth and changes in intracellular composition. The results showed that several methodologies are needed to obtain a comprehensive understanding of the toxicological of the exposure of nanoparticles on microorganisms. At growing conditions, chemostat systems can provide a better assessment of the nanoparticle inhibitory effects on microorganisms in comparison to batch systems. However, there is not control of the contact time and the specific growth rate and contact time effects are combined. Longer term exposure and chronic studies are suggested to separate the growth rate effect from the contact time. The data produced during this study is relevant to determine the real world implications on ecosystems and public health when the nanoparticles are released into the environment. With an understanding of the fate of nanoparticles in aqueous media, a more careful selection of toxicological methodologies and testing conditions can be made. This will allow for more accurate studies that measure the responses of microorganisms to the exposure of nanoparticles.

: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.  Table 6: Comparison of proteins region of untreated E. coli and exposed E. coli with AgNPs using ATR-FTIR for DF of 0.  Table 7: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.  Table 9: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.  Table 11: Comparison of proteins region of untreated E. coli and exposed E. coli with AgNPs using ATR-FTIR for DF of 0. mg/L and c) 50 mg/L in the chemostat reactors. Δ represents bacteria control without AgNPs. ▲represents bacteria control exposed to AgNPs. • represents a LB media control to detect contamination. □ represents LB media exposed to AgNPs to study nanoparticles stability and control to quantify the OD 670 from the AgNPs. The arrow x and the red line show the time when AgNPs were injected into the system. OD 670 was read every each 2.5 hours (150 min) after samples were re-suspended in PBS 10%. Bars   The inhibitory effect of nanoparticles on bacteria depends of physicochemical properties of the nanoparticles (e.g., charge, aggregation, cell-nanoparticle ratio and dissolution) which can differ among each other due to changes in size, charge, coating agent, manufacturer, reagents used during the synthesis and at the same time the water chemistry conditions influence the nanoparticle performance and bacteriananoparticle interaction. Release of ions from nanoparticles to bacteria promotes lysis, which makes nanoparticles, including silver a widely used antimicrobial agent.
AgNPs may serve as a medium to deliver Ag + more effectively (binding and reduced bioavailability by common natural ligands to the bacteria membrane and cytoplasm is less susceptible), whose proton motive force would reduce the local pH (close to pH 3.0) and improve Ag + release 9  3) Quantify the inhibitory effect of nAg on components of bacteria at different contact times and growth rates. The variations of specific functional groups in biomolecules allow to identify changes in the total composition of bacteria exposed to nAg through chemostats coupled with Fourier transform infrared (FTIR) and study the effect of the physiological stage of the bacterial population.
These studies produced relevant data about the implications on the environment and public health when the nanoparticles are released into the environment. In addition, a better understanding for the selection of toxicological methodologies with the purpose to increase the accuracy of future nanotoxicological studies will be assessed.

Abstract
There is increasing interest in the study of dysprosium oxide nanoparticles (nDy 2 O 3 ) for biomedical applications due to their fluorescent and paramagnetic properties. However, the fate of nDy 2 O 3 , and their effects on natural biological systems, are a growing concern.
This study assessed the toxicity of nDy 2 O 3 on Escherichia coli for concentrations between 0.02 and 2 mg/L, exposed to three concentrations of NaCl (8,500; 850 and 85 mg/L) and three glucose concentrations (35, 70, 140 mg/L). The ranges of these variables were selected to cover manufacturer recommendations of analytical methodologies for toxicity assessment, environmental and industrial nDy 2 O 3 effluent concentrations, and metabolic activity. Two array-based toxicity techniques were used to evaluate the 27 combinations of conditions. Fluorescent dyes (Live/Dead) and respirometric assays were used to measure the undisturbed cell membrane (UCM) and remaining respiration percentage (RRP), respectively.
Respirometric tests showed a higher toxic effect than Live/Dead test assays, indicating that metabolic processes are more affected than the physical structure of the cell by exposure to nDy 2 O 3 . After exposing the bacteria to concentrations of 2.0 mg/L uncoated nDy 2 O 3 for 2 hrs at 85 mg/L of NaCl and 140 mg/L of glucose, the RRP and UCM decreased to 43% and 88%, respectively. Dysprosium ions (Dy +3 ) toxicity measurement suggested that Dy +3 was the main contributor to the overall toxicity.

Introduction
Gadolinium, holmium and dysprosium belong to the lanthanide oxide-based nanoparticles (LnONps), which have acquired more relevance in recent years in regard to the locating, diagnosing and treating of diseases [1][2][3] . LnONps have unique paramagnetic properties that allow greater spatial and temporal resolution through a higher signal-to-noise ratio. These properties play a fundamental role in acquiring and enhancing the contrast in T 1 or T 2 magnetic resonance images (MRI) 4,5 . Due to the higher sensitivity provided by the LnONps, the MRI contrast is improved and the T 1 or T 2 relaxation times are discriminatorily shortened in the region of interest 6 .
Dysprosium oxide nanoparticles (nDy 2 O 3 ) have recently received increasing attention due to their potential applications in the biomedical field 4,5 including cancer research, new drug screening, and the delivery of drug applications 2,7,8 . However, the fate of nDy 2 O 3 and their effects on natural biological systems are growing concerns 9 . nDy 2 O 3 will enter into aquatic and land environments through wastewater treatment facility effluent and wastewater sludge due to an inability to retain and or remove these nanoparticles completely 10 . Moreover, the release of nDy 2 O 3 into land environments from agricultural applications could transport nanoparticles to surface waters via stormwater runoff and to groundwater via infiltration through the soil 10,11 .
Previous studies have provided limited insight into the toxic effects of nDy 2 O 3 and Dy ions on natural systems. Kattel et al. 12 investigated the in vitro toxicity effect of ultra-small spherical dysprosium oxide and dysprosium hydroxide nanorods 5 . Both nanoparticles were coated with D-glucuronic acid and exposed to DU 145 and NTC 1469 cell lines. These studies showed that the nanoparticles were not toxic to the human cells for concentration values ranging from 0 to 37.3 mg/L.
Harper et al. 13 tested 11 types of metal oxide nanoparticles, including nDy 2 O 3, and found that high mortality of embryonic zebrafish was observed when they were exposed to 250 mg/L of nDy 2 O 3 for 5 days of continuous waterborne conditions. In addition, concentrations of 250 mg/L for nDy 2 O 3 produced morphological malformations of the zebrafish's jaw and eyes.
Toxicological assessment of nanoparticles can be studied in terms of their impact on metabolic functions and cell structure such as cell viability, membrane permeation, growth and respiration. Live/Dead assay (BacLight viability kit) is a commonly-used method to measure cell viability 14 and membrane permeation on bacteria through the integrity of cell membranes. The manufacturer of the reagents used for the Live/Dead test recommends that experiments and samples have to be prepared in specific water chemistry conditions (8,500 mg/L of NaCl) to avoid a decrease in staining efficiency 14 . Previously, the metabolic activity of bacteria has been measured using a traditional respirometric bottle test (RT). Water chemistry conditions with monovalent and divalent cations have been successfully used in the range of 10 to 1,000 mg/L; however, high concentrations of glucose (in the order of 300 mg/L) and bacteria (in the order of 10 9 CFU/mL) were required to quantify a toxic response 15 . This type of test can be used to measure the interaction and effect of nanoparticles on microorganisms. Nevertheless, each methodology required its own range of optimal conditions, which makes it a complex process to assess toxic effects when identical water chemistry conditions are used. This is highly relevant to the evaluation of the toxic effect of nanoparticles, since the physicochemical properties of the nanoparticles (e.g., charge, aggregation, cell-nanoparticle ratio and dissolution) can differ among each other and also their properties are influenced by the physicochemical characteristics of the aqueous solution.
In this study, we propose to evaluate the use of array-based dyes methods in identical water chemistry conditions and observe the effect of nDy 2 O 3 and exposure on E. coli metabolic activity and structural integrity of E. coli under variable water chemistry conditions.

Materials
A non-pathogenic wild strain of E. coli (IDEXX laboratory) was selected for this study. E. coli is a Gram-negative bacterium that has been found to be metabolically active in saline solution without growth 16 and has been extensively studied in nanotoxicological research 15,17,18 . Reagents used to prepare the growth media for the bacteriasodium chloride (NaCl), yeast extract, and tryptonewere purchased from Sigma Aldrich. Glucose was purchased from Sigma Aldrich and used as received. Tetrazolium dye (Redox Dye Mix A) was purchased from Biolog and used to measure the respiratory responses of E. coli. Cell membrane permeation was measured using SYTO 9 and propidium iodide; both reagents were purchased from Invitrogen.

Methods
Hydrodynamic diameter and zeta potential were measured by Malvern Zetasizer Nano ZS, ZEN 3600, dynamic light scattering (DLS). Data was collected at 0.25 hrs and at 2 hrs after nanoparticles exposure to bacteria to differentiate the effect of aggregation of nanoparticles. Shape characterization of the nanoparticle was obtained by using JEOL JEM-2100 LaB6 transmission electron microscope (TEM) imaging.
Ionic release from nDy 2 O 3 for each condition was quantified as per Liu and Hurt 19 using centrifugal ultrafilter devices (ultra-4,3K) purchased from Amicon. Inductivelycoupled plasma spectroscopy (ICP-OES optima 3100, Perkin Elmer) was used to measure the concentration of nDy 2 O 3 and Dy ions before and after the contact times established for each of the water chemistry conditions tested. Samples were digested in nitric acid (2% v/v, HNO 3 ) before analysis.
Growth media consisted of 10 g/L NaCl, 5 g/L yeast extract, and 10 g/L tryptone.
After the solution was prepared, it was autoclaved and then inoculated with E. coli. E.
coli was grown for 12 hrs in a culture media at 37°C. Bacteria were harvested during the logarithmic growth phase and centrifuged at 2000 rpm (751 g) for 0.25 hrs. The supernatant was discarded and the pellet re-suspended in the respective NaCl  Blanks without nDy 2 O 3 , glucose and NaCl were analyzed for each scenario. correction, the value obtained in the respective well, was subtracted from the experiment values and also served as a secondary control to confirm that experimental conditions had no reducing effect on the tetrazolium dye without the presence of bacteria.
Each plate has been set up in quadruplicate wells for each condition, and the plates were run in duplicate to quantify the percent of remaining respiration (PRR).
The PRR (Eq. 1) is the ratio of slopes between bacteria exposed to nDy 2 O 3 and the blank bacteria (samples containing bacteria that were not exposed to nDy 2  Where, P t = Green/red fluorescence ratio for bacteria exposed to nDy 2 O 3 P c = Green/red fluorescence ratio for bacteria control without nDy 2 O 3

Toxicity tests for nDy 2 O 3 ion release.
Additional experiments using the respective ions concentration, released at the highest concentration of nDy 2 O 3 , were performed. These experiments allow us to determine the contribution of Dy ions to the overall toxicity on E. coli. Similar method was used to prepare the plate, but dysprosium ions were used instead of nDy 2 O 3 .

Statistical analysis
The results from each data set were analyzed with SAS statistical software, version 9.1.2. A generalized linear mixer model (GLIMMIX) was used to identify statistical differences among glucose, NaCl, and nDy 2 O 3 concentrations because the response was not necessarily normally distributed. A p value of less than 0.05 was considered to indicate significant difference.

Size and zeta potential without bacteria. Uncoated nDy 2 O 3 characterization
consisted of size and zeta potential measurements in two of the three water chemistry conditions. Tests with 8500 mg/L NaCl were discontinued due to interference between tetrazolium dye and NaCl. Details will be explained in Section 3.2. Figure

Shape and pH.
TEM imaging confirmed the shape of the nDy 2 O 3 to be spherical ( Figure 2). Nanoparticles were found to have an average size of 74.8 + 5 nm.
Changes in H + ions before and after nanoparticle exposure were recorded periodically with a pH meter. The pH measurements at t = 0 hrs and t = 2 hrs ranged from 5.5 to 6.2, indicating that pH did not function as an additional stress on bacteria performance. Moreover, there was no statistically significant change.

Ion release.
Ion release experiments were conducted for the highest concentration of nDy 2 O 3 , (2.0 mg/L), which was the most toxic condition for E. coli.
Dy ions were measured in all water chemistry and glucose conditions to determine the amount of dissolution over time. Table 2 shows a sudden increase in Dy ions measured at 0 hrs and 0.25 hrs, with a plateau occurring at 2.0 hrs.       The results showed that cell viability was predominantly lost due to interactions of Dy +3 ions with E. coli rather than nDy 2 O 3 ( Figure 5). This suggests that nDy 2 O 3 could have caused damage to the cell membrane, and Dy +3 could have entered into the cell and disturbed intracellular activities, as previously presented.

Comparison between toxicology methodologies
This study performed a comparison between two toxicological tests using nDy 2

Conclusion
The results showed that respirometric and permeation membrane tests can be used to provide a comprehensive assessment of nanoparticle toxicity on microorganisms.
This study evaluated the performance of two toxicity methodologies: the Live/Dead assay to evaluate the membrane permeation, and the respirometric assay to evaluate the metabolic activity of bacteria. The respirometric microarray test proved to be more sensitive than the Live/Dead test in measuring nanoparticle toxicity.
With an understanding of the fate of nanoparticles in aqueous media, a careful selection of appropriate toxicological methodologies can be made to improve the accuracy of future nanotoxicological studies.   In chemostats, the constant inflow of fresh media and aeration resulted in less AgNPs aggregation, thus increased the AgNPs-bacteria interactions, in comparison to batch conditions. AgNPs at 1mg/L, 15mg/L, and 50mg/L inhibited the growth (OD 670 reduction) by 0%, 11% and 16.3%, respectively. Membrane extracts exposed to 1mg/L, 15mg/L, and 50mg/L of AgNPs required greater changes in area by -0.5cm 2 , 2.7cm 2 and 3.6cm 2 , respectively, indicating that the bacterial membranes were disrupted and bacteria responded by synthesizing lipids that stabilize or strengthen membranes.

Acknowledgements
This study showed that the chemostat is more appropriate for the testing of nanotoxicological effects when testing bacteria at growing conditions.

Introduction
Silver nanoparticles (AgNPs) are one of the most commonly used nanomaterials in consumer products due to their antimicrobial properties . Cell membrane acts as a permeability barrier to the cytoplasm and is able to regulate the transport of macro-and micro-nutrients from the media to the cytoplasm as well as the osmotic pressure through the plasma membrane. The osmotic pressure influences the integrity and hydration of cells and their intracellular compartments.
Inflowing of water and swelling is governed by a decrease in external osmotic pressure (Wood, 2015), whereas an increase in osmotic pressure results in outflowing of water and dehydration. However, water fluxes coming simultaneously from opposite directions can disturb several cellular properties, including cell volume, turgor pressure, strain and cytoplasmatic membrane tension. Attenuation of water fluxes, by the accumulation or release of solutes, is one mechanism by which cells will respond to changes in external osmotic pressure (Wood, 2015).
AgNPs can damage bacterial membrane via three mechanisms. First, the electrostatic interaction between cell membranes and nanoparticles can interrupt transmembrane electron transfer, and produce break formation (pit formation).
Through this mechanism, AgNPs can also penetrate into the cell membrane producing an increase in the permeability, resulting in an uncontrolled plasma-membrane transport and even leading to cell death . Secondly, AgNPs can release silver ions (Ag + ) through cooperative oxidation with both protons and dissolved O 2 (Liu and Hurt, 2010). Ag + can be transported in the bacterial membrane by the potential disruption of nanoparticles to the cell wall and membrane. Bacterial membrane permeability can be affected by the Ag + mechanism described above and cause the release of lipopolysaccharides (LPS) and membrane proteins . Moreover, Ag + can cause the release of phosphate, mannitol, succinate, proline and glutamine from the cytoplasm and disrupt the respiration cycle by inhibiting the uptake of phosphorous, thus impairing the formation of energy-regulating compounds such as nicotinamide adenine dinucleotide (NADH) or damaging molecules, such as DNA .
Finally, the interaction between Ag + and thiol groups in proteins, in addition to inactivating the respiratory enzymes, can lead to the production of undesirable compounds, such as reactive oxygen species (ROS) . Intracellular oxidative stress can then occur as a result of high amounts of ROS, which can cause changes in the permeability of the cell membrane, protein structure, mitochondrial activity, and DNA replication .
Bioreactors are used to grow bacteria in continuous or batch mode. Previous studies have provided insight into the potential use of continuous reactors (chemostats) to assess stress conditions on bacteria. The effect of pH, osmotic stress, antibiotic resistance, and temperatures on bacteria have been studied extensively using chemostats   (Leenheer and Cogan, 2008). In chemostats, bacteria response in terms of cell growth and adaptation can be studied under single and multiple conditions, such as competition with nutrient recycling and antibiotic treatment . Comparatively, batch reactors have been used broadly to quantify the antimicrobial properties of nanoparticles in terms of their impacts on metabolic functions and cell structure such as, viability, membrane permeation, growth and respiration (Anaya et al., 2015;Mirzajani et al., 2011;Roe et al., 2008;Choi et al., 2008;Oyanedel-Craver, 2013, 2012).
The objective of this work is to compare batch and chemostat systems to assess the toxicity of casein-coated AgNPs on Escherichia coli (E. coli) based on membrane permeability, which is an indicator of membrane integrity and cells ability to adapt its membrane lipid composition. The effects of AgNPs on E coli have been widely studied, and thus the results produced in this research can be compared to those obtained previously. Spherical casein-coated AgNPs have been characterized and used in our research group and others Kvitek et al., 2009) to examine nanoparticle interactions with synthetic lipid monolayers, but they have not been used to examine lipid monolayers from membrane extracts to assess AgNPs exposure (Guzmán et al., 2013;Peetla and Labhasetwar, 2008;Torrano et al., 2013).
Monolayer film balance analysis yields surface pressure-area isotherms that can be used to assess the biophysical properties of lipids as well as lipid composition (Kurniawan et al., 2013;Bothun et al., 2016;Venkataramanan et al., 2014). The influence of the bacteria growth conditions coupled with membrane permeability assays and membrane extract analysis provides new methodologies and testing conditions that may be used to more accurately examine the response of microorganisms to nanoparticle exposure.

Materials
A non-pathogenic strain of E. coli K-12 (ATCC 23716) was selected for this study. E. coli is a Gram-negative bacterium that has been extensively used in nanotoxicological studies Choi et al., 2008;. Reagents used to prepare the growth media for the bacteriasodium chloride propidium iodide used for cell membrane permeation were purchased from Invitrogen.
Standard casein-coated AgNPs were obtained from Argenol Company, Spain.

Nanoparticle characterization
Hydrodynamic diameter and zeta potential were measured using dynamic light scattering (DLS) (Malvern Zetasizer Nano ZS, ZEN 3600). Inductively-coupled plasma spectroscopy (ICP-OES optima 3100, Perkin Elmer) was used to measure the concentrations of AgNPs and Ag + ions. Digestion in 2% nitric acid was required for each sample before analysis. The ionic release from AgNPs at each condition was quantified as per Liu and Hurt (2010) using centrifugal ultrafilter devices (ultra-4,3K) purchased from Amicon. Three concentrations of AgNPs -1 mg/L, 15 mg/L and 50 mg/Lwere used to assess the changes in permeation and surface pressure in chemostat and batch reactors.

Growth media and Bacteria culturing
Lysogeny Broth Miller (LB) growth media consisted of 10 g/L NaCl, 5 g/L yeast extract, and 10 g/L tryptone . Culture media was immediately autoclaved after preparation. For each experiment, a fresh bacteria culture was grown for 12 hours in the LB media at 37 o C. After that, optical density at a wavelength of 670 nm (OD 670 ) was measured separating the bacteria from the culture media by centrifuging it at 2500 rpm (1174 g

Bioreactors
A multiplexed chemostat arrays and a Synergy TM MX microplate reader (BIOTEK, VT) were used to perform the continuous and batch tests, respectively. All experiments were run in duplicate, including controls to detect contamination (media with no bacteria), non exposed condition (media plus bacteria) and exposed condition (media plus bacteria plus AgNPs). The ratio between number of AgNPs (estimated from DLS size distribution measurements) and bacteria was kept constant for both chemostat and batch reactors to compare the bacteria response. However, it will change either through the influent of fresh media in the chemostat or growth of bacteria in the case of batch test Samples were taken from each reactor every 2.5 hours using the sample needle ( Figure 6). pH was measured immediately after sample collection. The OD 670 was measured using a spectrometer once the culture media had been removed from the sample through centrifugation and resuspending the bacteria pellet in 10 mL of PBS (10%).

Batch tests
Batch tests were run using a microplate at similar conditions than those used during the chemostat tests. For both tests the ratio of bacteria/nanoparticle was kept constant. A 10 µL of bacteria stock solution was inoculated into each of the six wells containing 5 mL of LB media. After approximately 7.5 hours, when the culture reached log phase with an OD 670 was around 1 then AgNPs were injected to achieve the desire concentrations of 1 mg/L, 15 mg/L and 50 mg/L. After the injection of the nanoparticles, the plates were incubated for 5 hours to assess the AgNPs effect on bacteria concentration. OD 670 was determined using the same procedure described in the chemostat section.

Langmuir film balance
The membranes of the bacteria were extracted before injection of the AgNPs and after 5 hours of exposure to the respective nanoparticle concentration in the chemostat. In this case, the batch tests samples collected at 5 hours were used for the extraction of the membrane. After bacteria membranes were extracted according to the protocol of Bligh and Dyer (Bligh and Dyer, 1959) lipid monolayers were analyzed by Langmuir film balance based on the surface pressure-area isotherms.
Lipid packing, based on the total area occupied by the lipid extract at the air/water interface at constant temperature, was altered with moveable barriers in a Langmuir film balance. All results are presented for compression isotherms. Surface pressure was calculated as  =  o  where  o is the air/water interfacial tension and  is the air/water/membrane extract (lipid) interfacial tension. The surface tensions were measured using a Wilhelmy plate (Figure 7).

Epifluorescence staining membrane integrity test
A membrane integrity test was also performed for the batch reactors to compare the permeability membrane and the surface pressure changes. The cell membrane permeation of E. coli was determined using the Backlight kit (propidium iodide and SYTO 9) with a microplate reader. Propidium iodide becomes intercalated to the DNA within cells, and indicates whether bacteria that have a damaged membrane while SYTO 9 indicates intact cell membranes . 1) is the green/red fluorescence ratio between bacteria exposed to AgNPs and the blank bacteria (bacteria not exposed to AgNPs) at given AgNPs concentration. Data was analyzed after 5 hours of AgNPs exposure to quantify inhibitory effect of AgNPs on the bacteria. Where, P t = Green/red fluorescence ratio for bacteria exposed to AgNPs P c = Green/red fluorescence ratio for bacteria control without AgNPs

Physicochemical characterization
Characterizations of AgNPs consisted of size measurements in fresh media and bacteria-free media collected after 12 hours of bacterial growth (used growth media).
The used growth media reduced the AgNPs stability producing aggregation. A concentration of 15 mg/L AgNPs was the more suitable concentration to measure size, using the DLS, because 1 mg/L was too close to the lower detection limit and 50 mg/L was above the ideal range of the instrument. Figure   Hours respectively). Two bioreactors were used as controls, which contained bacteria without nanoparticles as controls to compare the growth rates between bacteria with and without exposure to nanoparticles, and two with only growth medium to detect possible contamination, and to study the interaction of AgNPs in the LB medium. Figure 9 shows the time when AgNPs were injected with a red arrow. Afterwards, the systems were operated for at least 5 additional hours (300 min). The OD 670 was measured to estimate bacteria concentration every 2.5 hours (150 min).
Only when the concentration of nanoparticles inside the chemostat was 50 mg/L a reduction on bacteria concentration was detected. In this case, a sustained  mg/L in the chemostat reactors. Δ represents bacteria control without AgNPs. ▲represents bacteria control exposed to AgNPs. • represents a LB media control to detect contamination. □ represents LB media exposed to AgNPs to study nanoparticles stability and control to quantify the OD 670 from the AgNPs. The arrow and the red line show the time when AgNPs were injected into the system. OD 670 was read every each 2.5 hours (150 min) after samples were re-suspended in PBS 10%. Bars represent the error between duplicates AgNPs concentration was quantified after the injection in each reactor using the ICP-OES. Figure 10 shows that the AgNPs were inside the bioreactors between 5   Bacteria grew for 7.5 hours (450 min) until log phase was achieved. After that, AgNPs were injected to reach the required concentrations (1mg/L, 15mg/L and 50 mg/L) inside the microplate wells. Then, the test run for additional 5 hours (300 min), which corresponds to hydraulic retention time (HRT) in the chemostat reactor ( Figure   11).

Batch
Nevertheless, AgNPs concentration of 50 mg/L was not suitable for the batch tests due to the fast aggregation and sedimentation leading to false readings in the results. Details are presented in detail in the SI.
The AgNPs inhibition on bacteria measured through the ratio of slopes based on OD 670 growth curves between bacteria exposed to AgNPs and the control bacteria (bacteria not exposed to AgNPs) were 20% and 0%, at 1 mg/L, and 15 mg/L of AgNPs, respectively. These values showed that 1 mg/L was the condition that produced a slight reduction in terms of bacteria concentration, compared to the other conditions.
The EPS most likely accumulated inside the batch reactors (microplate wells) promoting destabilization and sedimentation of the nanoparticles on the bottom of the well. Furthermore, less interaction between AgNPs and bacteria at 15mg/L due to high level of aggregation and sedimentation of AgNPs, can be the reason for the null or slight inhibitory effect detected. Additionally, AgNPs and the silver ions can be either chelated or coated by the EPS preventing the Ag + release. Figure 12 showed that EPS decrease the release of ions between 8 and 12 times for 50 and 15 mg/L (1 mg/L was not suitable for analysis because the concentration of ions was close to the ICP-OES detection limit) respectively, in comparison to Ag ions released into the culture media.

Chemostat
Surface pressure was calculated as  =  o  where  o was the air/water interfacial tension and  is the air/water/membrane extract (lipid) interfacial tension.  shifted to lower areas relative to the control. In these cases, the lipid monolayers required more compression to achieve comparable surface pressures to the control.

Increases in
This result infers that the membrane lipids where cells were exposed to AgNPs were occupied less area at the air/water interface. The most plausible explanation for this observation is that the lipids from extracts where cells were exposed to AgNPs contained more saturated than unsaturated acyl tails. Lipids with saturated acyl tails do not exhibit "tail kinking" and occupy less interfacial area than lipids with unsaturated acyl tails do. At 50 mg/L there was no statistical difference in the A isotherms with and without AgNPs exposure. These results suggest that in chemostat cultures membrane disruption caused by AgNPs was a destabilizing or fluidizing (disordering) effect where the cells responded by shifting their lipid composition to more rigid, saturated lipids to counteract membrane fluidization. Figure 13b shows the results for A isotherms conducted on membrane lipids obtained from batch cultures. At 15 mg/L and 50 mg/L, the A isotherms where cells were exposed to AgNPs were shifted to higher areas relative to the control. In contrast to the results for membrane lipids from chemostat cultures, the lipid monolayers required less compression to achieve comparable surface pressures to the control,

Epifluorescence staining membrane integrity test
Undisturbed cell membrane (UCM) results are displayed in Figure 14. Using t test, it was found a statistical difference between 1mg/L and 50 mg/L (p= 0.026) and between 1mg/L and 15 mg/L, (p=0.027), however no statistical difference was found between 15 mg/l and 50 mg/L (p=0.152). The statistical analysis confirmed that low concentrations are a higher inhibitory effect in comparison with the high concentrations.

Discussion
This study compared the use of chemostat and batch reactors to assess the casein coated AgNPs exposure to E. coli in terms of bacteria growth and surface tension changes. Conditions from chemostat reactors were replicated in the batch reactor to compare AgNPs inhibition effects in both systems. In our study, it was found that there is only a small or null inhibitory effect at all the AgNPs concentration used in batch conditions. Therefore, these results differ from previous studies simulating natural water conditions (Zhang and Oyanedel-Craver, 2013;Mirzajani et al., 2011). This may be due that bacteria are growing condition (rather than non-growing in tests using natural water conditions) and a lower In the case of chemostat reactors, the constant feeding and aeration seemed to increase the stability of nanosuspension compared to batch conditions. Constant inflow of fresh media could reduce the accumulation of EPS inside the reactor, while aeration could reduce the aggregation of nanoparticles and therefore increasing the contact between AgNPs and bacteria.
Greater inhibitory effect of AgNPs observed in chemostat cultures was consistent with the membrane lipid monolayer results that showed that AgNPs had a destabilizing effect which forced the bacteria to counteract this effect by synthesizing lipids that are known to strengthen cellular membranes. Langmuir Blodgett and epifluorescence staining membrane integrity test for batch conditions showed a similar trend. Here, 1mg/L of AgNPs had a higher inhibitory effect on bacteria compared to 15 mg/L and 50 mg/L of AgNPs.

Conclusions
This study shows that chemostat systems can be used to provide a better and more comprehensive assessment of the nanoparticle inhibitory effects on microorganisms compared to batch systems at growing conditions. Langmuir Blodgett was used to evaluate the surface tension changes in bacterial membranes, this test provided additional information to the staining membrane integrity test. The formation of EPS during the growth of E. coli influenced the level of the AgNPs effect on the cell membranes, and therefore the bacterial growth. With an understanding of the fate of nanoparticles in aqueous media, a more careful selection of appropriate toxicological methodologies and testing conditions can be made. This will allow for more accurate studies that measure the responses of microorganisms to the exposure of nanoparticles.

ACKNOWLEDGMENTS
This research work was supported by the National Science Foundation under the grant numbers CBET-1350789 and CBET-1055652.

Comparative study between chemostat and batch reactors to quantify membrane
permeability changes on bacteria exposed to silver nanoparticles

Batch tests at 50 mg/L
The batch test at 50 mg/L was not suitable due to the fast aggregation and sedimentation of AgNPs in contact with the EPS. It seems initially from the figure SI 2, that the bacteria in contact with 50 mg/L of AgNPs are growing continually even faster than the bacteria without AgNPs. However, there is a mix of factors that can lead to this misconception. First, the epifluorescence test indicates that the all of the bacteria are alive at this condition ( Figure 9). Second, the increase in terms of OD 670 corresponds to the AgNPs sedimentation, which can be corroborated by the well image that appears in the figure SI 2 but it does not correspond to bacterial growth.
Additionally, the control (composed by EPS and AgNPs with 50 mg/L) showed an atypical behavior because the OD 670 did not keep the direct correlation in time, instead of that dramatically decreased.

Introduction
Silver nanoparticles (AgNPs) are one of the most commonly used nanomaterials in consumer products and medical applications due to their antimicrobial properties . Nevertheless, AgNPs antimicrobial activity can be compromised, due to bacteria capability to modifying their phenotype in response to stress agents by genetic mutation or mobile genetic material acquisition from another bacterium (Fraser and Kaern, 2009;He and Chen, 2010;Leenheer and Cogan, 2008a;De Gelder et al., 2008;Palmer and Kishony, 2013).
AgNPs, and silver ions (Ag + ) are stress agents that can cause changes in the permeability of the cell membrane, protein structure, respiration cycle, and DNA replication .
Bacterial adaptation to Ag + at the molecular level has been previously reported as three main mechanisms: Ag + accumulation, efflux Ag + pump and halide ions effect.
TEM and EDX analysis have revealed that a bacteria Pseudomonas stutzeri strain (AG259) can be resistant to Ag + stress through the formation and accumulation of dense metal deposits (Slawson et al., 1992;Silver et al., 1999). A second mechanism consist on the pump of Ag + from the cytoplasm to the bulk and the hydrogen ions pumping in reverse direction occur due to the activation of sil genes, which includes nine sil genes (silA,silB,silC,silE,silF,silR,silS,silP ORF 105 and silAB ORF96) for both efflux system, SilCBA and SilP. Two periplasmic protein (SilE and SilF) act as a molecular chaperon and transport Ag + to SilCBA to continue injection (Silver, 2003). Finally, the role of halide ions in the silver resistance mechanism is to decrease the availability of Ag + through binding with halide ions, in culture media (Gupta et al., 2001).
Continuous systems (chemostat) have been previously used to study the bacteria response to stress conditions in terms of cell growth and adaptation. Leenheer and Cogan, 2008b;. Chemostats allow bacteria to growth in a defined, ideally constant and controllable set of physico-chemical conditions. Moreover, continuous cultures have the advantage that time-independent concentration (steady state) can be achieved and the contact time between stressor and bacteria can be manipulated changing the rate of supply of the limiting substrate reflected (flow rate).
Previous studies have reported that organisms growing at slow specific growth rates could adapt easier to stress conditions than those growing at a faster rate. (Van Hoek et al., 1998).
To our knowledge stress-response to AgNPs has not been systematically evaluated using continuous cultures as function of concentration and contact times.
The objective of this work is to elucidate the response of Escherichia coli (E. coli) to the exposure casein-coated AgNPs at different contact times using multiplex chemostat reactors. At steady conditions the specific growth rate is equal to the dilution factor (DF), which is the ratio between the flow rate of medium injected and the culture volume, and where lower DF corresponds to greater contact times in comparison with higher DF.
To gain additional mechanistic insight about AgNPs -cell interactions, we have included Fourier transform infrared (FTIR) analysis. The changes of specific functional groups in biomolecules can be related to different toxicological mechanism affecting bacteria exposed to AgNPs. The influence of different bacterial growth conditions coupled with FTIR technique can be used as a new approach to examine the response of microorganisms to nanoparticles exposure.

Materials
During this study a non-pathogenic strain of E. coli K-12 (ATCC 23716) was selected. E. coli is a Gram-negative bacterium that has been extensively used in nanotoxicological studies Choi et al., 2008;. Reagents used to prepare the growth media and phosphate buffer solution (PBS) for the bacteria were purchased from Sigma Aldrich: sodium chloride, yeast extract, tryptone; monobasic potassium phosphate, dibasic potassium phosphate, ethylenediaminetetraacetic acid (EDTA), respectively. The chemical oxygen demand (COD) was used to measure biomass using the TNT 822 kit from Hach Company.
SYTO 9 and propidium iodide used for cell membrane permeation were purchased from Invitrogen. Standard casein-coated AgNPs were obtained from Argenol Company, Spain.

Nanoparticle characterization
Hydrodynamic diameter and zeta potential were measured using dynamic light scattering (DLS) (Malvern Zetasizer Nano ZS, ZEN 3600). Inductively-coupled plasma spectroscopy (ICP-OES optima 3100, Perkin Elmer) was used to measure the concentrations of AgNPs and Ag + ions. Before the analysis of each sample a digestion in 2% nitric acid was required. The ionic release at each condition from the AgNPs was quantified as per Liu and Hurt (2010) using centrifugal ultrafilter devices (ultra-4,3K) purchased from Amicon. The two concentrations of AgNPs inside the vessel were used: 15 mg/L and 50 mg/L. AgNPs exposure was measured at two contact times (hydraulic retention time) using multiplex chemostat reactors.

Growth media and Bacteria culturing
Lysogeny Broth Miller (LB) growth media consisted of 10 g/L NaCl, 5 g/L yeast extract, and 10 g/L tryptone . After the culture media was prepared it was immediately autoclaved. For each experiment, a fresh bacteria culture was grown for 12 hours in the LB media at 37 o C. Then, bacteria was separated from the culture media by centrifuging it at 2500 rpm (1174 g) for 15 min and additional pellet resuspension in PBS (10%) solution . The PBS (10%) solution consisted of 1.12 g/L K 2 HPO 4 , 0.48 g/L monobasic potassium phosphate KH 2 PO 4 and 0.002 g/L EDTA. The biomass concentration was measured indirectly through chemical oxygen demand (COD) using Hach TNT 822 with a Hach DR 2800 spectrometer after separating bacteria from culture media by centrifugation and followed resuspension in PBS (10%). The COD reduction is measured after the AgNPs injection. Since COD cannot differentiate between alive and dead biomass, it is very likely that a delaying in the detection of COD changes will be expected. An optical density at a wavelength of 670 nm (OD 670 ) was measured to follow the bacteria concentration evolution in each reactor. An optical density reading at this wavelength indirectly reflects the number of bacteria.

Bioreactors
Multiplexed chemostat reactors were used to perform the continuous tests. All experiments were run in duplicate, including controls, to detect possible contamination (media with no bacteria), non exposed condition (media plus bacteria) and exposed condition (media plus bacteria plus AgNPs).
The bioreactors and sample vessels were sterilized twice before use in these experiments. The array consisted of six small vessels fed with LB medium and air.
Two needles were attached to each vessel for air and culture media injection ( Figure   15). A volumetric flow of air of 0.7 L/min was passed through a trap to remove humidity. This was then followed by a 0.2 µm filter before being pumped into the bioreactors, which corresponds to centrifuge tubes of 50 mL. The airflow maintained

FTIR tests
Liquid samples were prepared for ATR-FTIR (Nicolet iS50 FTIR, Thermo Scientific) analysis by fixing the OD 670 to 0.75 using an UV−vis spectrophotometer (Genesis, 10UV, Thermo Scientific). Following this, bacteria were centrifuged at 13,000 rpm (18,894 g) for 10 minutes, and the supernatant was removed. The pellets were suspended in 10 μL of PBS 10%, and the suspension of bacteria with AgNPs was directly transferred onto the crystal surface (Gurbanov et al., 2015).
Spectra were the result of 256 scans with a resolution of 4 cm −1 in the 4000 cm −1 -350 cm −1 spectral range. The data was analyzed by Omnic software (Thermo Scientific) and processed using Matlab (Mathworks Software).
A unique FTIR spectrum is detected for bacteria and each of their components and vibration modes. The full spectra range is divided in three specific regions; nucleic acid (900 cm -1 to 600 cm -1 ), carbohydrates plus proteins (1800 cm −1 to 900 cm −1 ) and the fatty acid region (3300 cm −1 to 2800 cm −1 ); and each region is divided in specific vibration modes. AgNPs effect on bacteria is detected through the peak shifting and peak intensity analysis of specific functional groups in biomolecules which can be related to different toxicological mechanism (Al-Holy et al., 2006;AlRabiah et al., 2013;Arakawa et al., 2001).

Epifluorescence staining membrane integrity test
The cell membrane permeation of the E. coli was determined using a BacLight kit (propidium iodide and SYTO 9) with a TM MX microplate reader (BIOTEK, VT).
Propidium iodide becomes intercalated to the DNA within cells, and indicates whether bacteria have a compromised membrane while SYTO 9 indicates intact cell membranes . SYTO 9 stains all cells green and propidium iodide can interact with DNA in cells with compromised membrane. A stain solution composed of SYTO 9 and propidium iodide fluorescent nucleic acid stains was mixed at a 1:1 (v/v) ratio with a subsequent dilution in DI water (12 μL of stain mixed solution in 2 mL of DI water). 100 μL of bacteria samples from each of the six wells with a fixed OD 670 of 0.06 were added to the wells of a 96 well flat-bottom black microplate. Thereafter, 100 μL of mixed stain solution was added and mixed into each well by thoroughly pipetting at least 10 times. Additional incubation was required in a dark at room temperature for 15 min before reading with the microplate. A calibration curve of live and dead bacteria before and after nanoparticle exposure was necessary to quantify and compare the membrane disruption on bacteria. Each plate was run in duplicate and contained triplicate wells for each condition to quantify the Undisturbed Cell Membrane (UCM). The UCM (Eq. 1) is the green/red fluorescence ratio between bacteria exposed to AgNPs and the blank bacteria (bacteria not exposed to AgNPs) at given AgNPs concentration. After each sample was collected, the data was analyzed (Eq. 1) to quantify inhibitory effect of AgNPs on the bacteria. Where, P t = Green/red fluorescence ratio for bacteria exposed to AgNPs P c = Green/red fluorescence ratio for bacteria control without AgNPs

Statistical analysis
The results from each data set were analyzed with SAS statistical software, version 9.1.2. A generalized linear mixer model (GLIMMIX) was selected due to the response was not necessarily normally distributed. A p value of less than 0.05 was considered to indicate significant difference among different DF, and AgNPs concentrations.

Nanoparticle characterization
Characterizations of AgNPs consisted of size measurements in fresh media and bacteria free media collected after 15 hours of bacterial growth (used growth media).
The used growth media reduced the AgNPs stability, producing aggregation. A concentration of 15 mg/L AgNPs was the more suitable concentration to measure size, using the DLS, because 50 mg/L was above the ideal range of the instrument (1mg/L to 20 mg/L). Figure 16 shows the average hydrodynamic diameter of the AgNPs suspended in different media between 0 hours and 28 hours for 15 mg/L of AgNPs using the DF of 0.1 h -1 . It is expected that at 50 mg/L greater sizes of aggregate will be formed due the greater number of nanoparticles in the solution, thus increasing the nanoparticle-nanoparticle interaction.
AgNPs suspended in DI water were stable at a size of 55.3+1 nm and zeta potential measurements ranging between -29.6 mV and -32.3 mV. Once AgNPs were contacted with LB media, their size increased to 84+5 nm. Likewise when AgNPs were contacted with used growth media, aggregation of nanoparticles was detected.
The size of AgNPs increased from 52 nm at time zero to 217 nm after 5 hours showing that exopolymeric substances (EPS) released during the bacterial growth (composed of lipids, proteins and nucleic acids) could affect the AgNPs stability and could influence the inhibitory effect.  showed that EPS can trap the AgNPs outside the membrane as protective mechanism and decrease the inhibitory effect on bacteria due to the less interaction between bacteria and nanoparticles.
The dissolution of ions over time in the used media at room temperature was 0.068% and 0.055% of the total silver concentration from reactors with 50 mg/L and DF of 0.1 h -1 and 0.6 h -1 , respectively. The ion release reduction showed that AgNPs and the silver ions can be either chelated or coated by the EPS, preventing the Ag + release (Anaya et al., 2016).

Time (hours)
AgNPs were injected only into the test bioreactors after steady conditions were reached to achieve the desired concentrations (15 mg/L or 50 mg/L). Afterwards, the systems were operated for 35 hours and 10 hours for the DF of 0.1 h -1 and 0.6 h -1 , respectively. The OD 670 was measured to track bacteria concentration and COD to quantify biomass reduction after bacteria separation from culture media by centrifugation.
For a DF of 0.1 h -1 and 15 mg/L of AgNPs a reduction of 2% COD was detected at 25 hours after the AgNPs injection. After that, COD reduction was not observed.
In the case of 0.1 h -1 and 50 mg/L there was a reduction of COD from 5 hours to 25 hours of 13%, after 25 hours the COD reduction was not more than 5.6%.
For the DF of 0.6 h -1 ,a small reduction (1.8%) in biomass concentration was detected with 15 mg/L of AgNPs at 2.5 hours and 5hours, while at 50 mg/L a decrease in COD (biomass) was observed at 5 hours and 10 hours up to 14.5% (Figure 17). .

AgNPs concentration
AgNPs concentration was quantified after the injection in each reactor using ICP-OES. Figure 18 shows that AgNPs interact with bacteria between 25 hours and 35

FTIR results
The finger print (nucleic acid region) bands of E. coli are found in the region between 900 cm -1 and 600 cm -1 , (SI Figures SI1a, SI2a values. The changes of specific functional groups in biomolecules are expressed in terms of shifting and peak intensity, which can be related to different toxicological mechanism affecting bacteria exposed to AgNPs. Bacteria exposed to 50 mg/L showed greater changes in functional groups in comparison with 15 mg/L in both DF. Results summarized in table 11 showed that the main changes were found in the fatty acid and protein region. For the condition in which 0.1 h -1 dilution factor and 50 mg/L of AgNPs was applied, the untreated bacteria had the most intense peaks throughout the spectrum compared to the bacteria exposed for 35 hours The changes for the fatty acid region were shifting due to the deformation of asymmetric vibration of (C=O) and the asymmetric vibration of (P=O). On the other hand, in the protein region were shifted the peaks for amide III.
For the 0.6 h -1 dilution factor for 50 mg/L of AgNPs, the same trend as 0.1 h -1 dilution factor and 50mg/L of AgNPs was found, where the untreated bacteria had the most intense peaks throughout the spectrum compared to the bacteria exposed for 10 hours. Changes were detected only in the fatty acid region, a shifting in the symmetric vibration of CH 3 was found. Table 3: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.1 h -1 and 50 mg/L. *Peak was shifted to a different wavenumber, **Peak was not observed.  Table 5: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.6 h -1 and 50 mg/L. *Peak was shifted to a different wavenumber, **Peak was not observed. Table 6: Comparison of proteins region of untreated E. coli and exposed E. coli with AgNPs using ATR-FTIR for DF of 0.6 h -1 and 50 mg/L Table 7: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.1 h -1 and 15 mg/L. *Peak was shifted to a different wavenumber, **Peak was not observed. Table 8: Comparison of proteins region of untreated E. coli and exposed E. coli with AgNPs using ATR-FTIR for DF of 0.1 h -1 and 15 mg/L Table 9: Comparison of fatty acids region of untreated and exposed E. coli with AgNPs using ATR-FTIR for a DF of 0.6 h -1 and 15 mg/L. *Peak was shifted to a different wavenumber, **Peak was not observed. Table 10: Comparison of proteins region of untreated E. coli and exposed E. coli with AgNPs using ATR-FTIR for DF of 0.6 h -1 and 15 mg/L *Peak was shifted to a different wavenumber, **Peak was not observed  White bars represent samples from the reactor injected with 15 mg/L and black bars with 50 mg/L. Zero time (0 hours) corresponds right before AgNPs injection. Each value represents an average of 4 wells from two duplicate plates.

Discussion
Formation of EPS during the growth of E. coli influenced the level of the AgNPs effect on the cell membranes. EPS accumulation inside the bioreactors greatly affected the stability of AgNPs and inhibited the Ag + release, thus reducing the cellnanoparticle interactions. However, biomass concentration and staining fluorescence tests confirmed that AgNPs produced inhibition in terms of COD mass reduction and disturbance in cell membrane permeation even though the EPS formation. A higher inhibitory effect were found at the high AgNPs concentration (50 mg/L) and the low DF (0.1 h -1 ), these conditions implies that the contact time between bacteria and nanoparticles is higher in comparison to the high DF (0.6 h -1 ). For the DF of 0.1h -1 , 13.6% of COD reduction and 22% of membrane permeation were detected at 50 mg/L of AgNPs in comparison to 1.9% reduction of COD and 7.3 % at 15 mg/L of AgNPs and the same DF. The same trend was observed for the DF of 0.6 h -1 , but with lower inhibitory effect, 1.9% of COD reduction and 7.7% of membrane permeation.
The FTIR results also showed that the spectral regions changed based on the dilution factor and the AgNPs concentration. FTIR results showed that AgNPs also could cause changes in the fatty acids, specifically -CH deformation, which can be correlated to the alteration in membrane permeability. FTIR results also suggested that membrane permeability changes can be due to the dehydration of phospholipids.
The most remarkable functional groups changes of the exposed E. coli with DF 0.1 h −1 and 0.6 h −1 were in the fatty acid and protein region. This could suggest that part of the common changes between these two DF were due to the damage of conformational/compositional alterations in some of the components of the protein structures that could be intracellular proteins or cell wall peptides (Jiang et al., 2010;Nadtochenko et al., 2005).
DF of 0.6 h −1 showed alterations in the fluidity of the lipids produced by the symmetric vibration of CH 3 , which have been analyzed previously (Hu et al., 2009;Nadtochenko et al., 2005). These impacts on methyl groups' stretching may be caused by the ROS, which identify malfunctions in the respiratory chain enzymes and other membrane proteins and lipophospho-polysaccharids (Jiang et al., 2010).
FTIR results shows that AgNPs can impact the structure and function of proteins Rizzello and Pompa, 2014;Wigginton et al., 2010) and can damage nucleic acid molecules with a higher inhibition at high AgNPs concentration (Choi et al., 2008b;Graves et al., 2015). The effect of membrane alterations in the treated E coli was high compared to the observed changes in proteins. These changes were enough to damage the outer cell membrane by AgNPs, which causes more entry of AgNPs into the cells could be one of the reasons that AgNPs have cytotoxic effect on E. coli (Arora et al., 2008;Wigginton et al., 2010;Zhang and Oyanedel-Craver, 2013b) Statistical analysis confirmed that the effect of different AgNPs concentrations are significant in terms of the COD reduction (p<0.001) and permeation (p=0.0324) in comparison with DF, which was not significant with COD (p=0.8161) but significant for permeation (p=0.0157). For all of the conditions, AgNPs concentration was the most influential variable and determined the magnitude of the exposure response. This is consistent with the results obtained previously, which showed higher inhibitory effect in conditions with high AgNPs concentrations. Also a strong correlation was observed when the combined effect between DF and AgNPs concentration was analyzed, which is very significant to assess permeation response (p=0.0013). A similar trend was observed in the high toxicity effect of AgNPs on E. coli when bacteria were in longer contact with high AgNPs concentration.

Conclusion
This study showed that chemostat systems can be used to evaluate the inhibitory effect of nanoparticles in continuous culture at different growth rate of bacteria. The results did not agree with previous studies with regard to the specific growth rate due to the different contact times that were achieved in the chemostats at different specific growth rates. With the current conditions, there is not control of the contact time and the specific growth rate and contact time effects are combined. Longer term exposure and chronic studies are suggested to separate the growth rate effect from the contact time.
AgNPs were more statistically significant in comparison with the DF, although nanoparticles instability produced by the interaction with the EPS substance.
FTIR provided additional information to differentiating toxic effects at intracellular levels, being the protein and the fatty acid groups the most affected at 50 mg/L of AgNPs.
Continuous bioreactors coupled with FTIR represent a toxicological technique that can be used to assess systematically the response of microorganisms at different bacterial growth conditions to stress conditions such as nanoparticles, providing a clear understanding about the inhibitory effect produced on each bacteria component.