Kinetic, Metabolic and Macromolecular Response of Bacteria to Antimicrobial Agents

Bacteria have the extraordinary capability to modify their phenotype in response to stress agents. Response mechanisms to stressor (such as nanoparticles or light treatments) are the result of an evolutionary process in which bacteria can become resistant, or adapt to the stressor. Therefore, it is essential to elucidate the bacterial adaptation mechanisms to metal nanoparticles or pulse ultraviolet light since these processes not only can compromise the efficacy of these treatment methods but also has implications to public health issues. Silver nanoparticles (AgNPs) are one of the most commonly used nanomaterials in consumer products and medical applications due to their antimicrobial properties. Also, pulsed lights (PL) applications for water treatment purposes are gaining increasing attention. PL has showed higher inactivation of microorganisms and degradation of PAHs, because of its rich and broad-spectrum UV content, high energy peaks and predictable treatment outcomes. However, the fate of microorganisms after exposure to AgNPs and pulsed lights as well as their negative impacts possible on the environment and public health are growing concerns. There are knowledge gaps related to: the studying molecular level change of microorganism after exposure to AgNPs by Fourier-transform infrared spectroscopy (FTIR); Bacterial adaptation to chronic exposure to nanoparticles in continuous culture: kinetic and macromolecular response; and the effect of pulsed ultraviolet light system for removal of polycyclic aromatic hydrocarbon and disinfection of pathogens in drinking water. In this study, Escherichia coli K-12 MG1655 (E. coli) responses to AgNPs was assessed under batch and continuous conditions. Further, we investigated the response of the same bacteria to PL. In detail, we evaluated antimicrobial agent’s impacts on metabolic functions and cell structure such as, colony formation units, membrane permeation, respiration, growth, gene regulation and changes in cellular composition. The results of batch culture for AgNPs toxicity test showed that bacteria developed resistance toward AgNPs and resulted in changes in the genotype and expression in the phenotype. Moreover, in continuous culture, results showed that culture growth conditions significantly affect bacterial response to nanoparticle exposure. Finally, from PL exposure to bacteria we obtained that the antimicrobial efficiency of PL depends on the PL lamp cut-off. This study provides data that have a predominant role to determine the performance of toxicological tests. Hence, the knowledge of nanoparticles fate in different growth condition minimizes the upcoming environmental and public health issues due to releasing nanoparticles release on the ecosystems. Also, understanding the bacterial responses to antimicrobial agents help us to select the more sensible agents for antimicrobial purposes.

. Comparison of altrations in biomolecular group between PL exposed and non-exposed bacteria using FTIR. .  16 The absorbance of bacterial regrowth at OD 600 nm. a) Raw data of bacterial regrowth b) linear trend line of slope and regression for regrowth bacteria at different conditions. Marks are: • controls (non-exposed bacteria), Δ regrowth of PL1 exposed bacteria, ♦ regrowth of PL2 exposed bacteria, x blank which is LB media. Impacts of PL1 and PL2 on bacterial remaining respiration percent (RRP). Bacteria exposed to PL2 for 76 joule/cm 2 and bacteria exposed to PL1 for 95 joule/cm 2 . Control is non-exposed bacteria. Error 18 Impacts of PL1 and PL2 on bacterial membrane permeability that shows undisturbed membrane cell percent. Bacteria exposed to PL2 for 76 joule/cm 2 and bacteria exposed to PL1 for 95 joule/cm 2 . Control is non-exposed bacteria. Error 19 Hieratical analysis of fatty acids, protein, carbohydrates, and nuclei acids between regrowth bacteria and exposed bacteria to PL1 for 76 joule/cm 2 . Control is non-exposed bacteria. Regrowth control is regrowth of non-exposed bacteria. ........ 143 Fig. 20 Hieratical analysis of fatty acids, protein, carbohydrates, and nuclei acids between regrowth bacteria and exposed bacteria to PL2 for 95 joule/cm 2 . Control is non-exposed bacteria. Regrowth control is regrowth of non-exposed bacteria. ........ 144

Abstract
Fourier transform infrared (FTIR) is a spectroscopy method that can identify variations in the total composition of microorganisms through the determination of changes of functional groups in biomolecules. FTIR measures the vibration and rotation of molecules influenced by infrared radiation at a specific wavelength. This technique allows the identification of structural changes of the molecular binding between microorganisms and metal atoms, which can provide information about the nature of their interactions. In this review article, we will describe the state of the art in current uses of FTIR for the elucidation of bacteria-nanoparticle interactions. We will describe advantages for the application of FTIR in the field of nanotoxicology, including higher signal-to-noise ratio, high energy throughput, as well as high accuracy and stability which are applicable to solid phase samples but not recommended for assays in the liquid phase. Limitations such as multiple background scans and post processing analysis are not deniable.
Comparison of FTIR with other commonly used tools such as raman spectroscopy, mass spectrometry, nuclear magnetic resonance spectroscopy, and X-ray photoelectron spectroscopy are also discussed. Finally, we present an application of FTIR for the assessment of bacterial changes in response to the exposure to silver nanoparticles (AgNPs). The results showed that the AgNPs-induced structural changes in the peptide and amino acids region may lead to alterations of conformation and/or composition of Amid B and Amid III. These results showed that bacteria developed resistance toward AgNPs and resulted in changes in the genotype and expression in the phenotype. Here, ATR-FTIR provided the evidence of the AgNPs cytotoxicity induced intracellular level alterations in bacteria.

1-Introduction
Rapid and specific analytical tools to characterize the interactions between bacteria and nanoparticles are essential for the development of safe and effective nanomaterials. These tools can also help to prevent the unintentional negative effects  Recently, FTIR has been used to characterize bacteria exposed to nanoparticles From a biochemical perspective, it is important to determine interactions at the molecular level as they could provide fundamental information on the effects observed at a systemic level. Due to the technological advances, simplicity of the sample preparation, and high speed of analysis, FTIR spectroscopy could fill the growing demand of fast and reliable toxicity screening in the field of nanotoxicology.

-Overview of FTIR fundamentals
The emitted radiation from an IR source passes through an interferometer composed of a beam-splitter, a fixed mirror, and a moving mirror (Fig. 1). The interferometer measures the wavelength of emitted light via interference patterns that help to increase accuracy. IR spectra are obtained by applying IR radiation to a sample and measuring 5 the intensity of the passing radiation at a specific wavenumber (Fig. 1). The number of   scans can be adjusted based on the quality requirement for the sample analysis; currently, the most common number of scans used is 2 8 . IR radiation of certain molecular groups can be detected at a specifics wavenumbers. The x-axis of the spectrum represents the wavenumber while the y-axis represents absorbance or The most common FTIR based methods for bacteria characterization are transmittance FTIR, attenuated total reflectance (ATR-FTIR), and micro-spectroscopy FTIR.

1.1.1-Transmittance FTIR
FTIR analysis can be carried out in solid, liquid or gas samples, however here only solid and liquid phase sample applications will be discussed. Solid samples should be ground with potassium bromide (about 5% of the weight of the sample) and pressed to 6 form a hard pellet. Then the sample is placed between two infrared-transparent plates.
There are various types of transparent material used to analyze different types of samples. For liquid samples, selected compatible IR transparent windows could be zinc selenide or diamond glasses are more appropriated. In both cases, the IR beam passes through the sample, as previously described. The low noise-to-signal ratio in samples makes the transmittance method advantageous because the pressed sample has a low number of random fluctuations of the baseline, resulting in higher sensitivity (Coates 2006  In general, the observed intensity of the absorption bands along the ATR-FTIR spectrum is less than the transmitted FTIR spectrum. The main difference between transmittance FTIR and ATR-FTIR is their depth of penetration. Transmittance FTIR measures a spectrum that is an average of the bulk properties of the sample; however, ATR-FTIR can only probe through samples up to 300 nm in thickness (Lasch and Naumann 2000; Winder and Goodacre 2004; Burgula et al. 2006). These two methods differ as well in sample preparation requirements. In the case of ATR-FTIR, the sample is directly placed on to the crystal surface, but for transmittance FTIR, the sample must be placed between two transparent glass (Lasch and Naumann 2000; Winder and Goodacre 2004; Burgula et al. 2006).

1.2-Advantages, disadvantages, and limitations
All FTIR techniques present advantages and disadvantages for the analysis of bacteria exposed to stress conditions, such as exposure to nanoparticles. FTIR is a time-  (Lin et al. 2004). FTIR is also less expensive for bacterial identification compared to other commonly applied methods (Davis and Mauer 2010). Moreover, FTIR has three remarkable advantages: (1) higher signal-to-noise ratio, (2) high energy throughput, and (3) high accuracy and stability. A higher signal-to-noise ratio is possible because the wavelengths are measured simultaneously, which is called "Fellgett advantage".
Prevention of the light dispersion in FTIR causes high energy throughput which is called the "Jacquinot advantage". Regarding accuracy and stability, a remarkable advantage of FTIR is the use of a HeNe (helium neon) laser which acts as an internal reference for each scan and provides accurate and stable wavenumber scales of an interferometer which is referred to as the "Connes advantage". These three properties of FTIR are effective for nanotoxicoly assays performed (Naumann 2006). All of these three properties of FTIR are effective in the solid phase rather the liquid phase of nanotoxicological studies.
However, there are also disadvantages. Multiple background scans and sample scans are necessary to avoid artifacts and variations in the spectra due the surrounding environmental conditions in sample heterogeneity. For instance, measuring the sample in culture media at different temperatures can influence the FTIR spectra of the sample (Cadet and de la Guardia 2000). Pretreatment of the samples may be required to purify the sample and to prevent peaks from overlapping on the spectra. For example, water from bacteria samples in liquids can overlap the band of amide compounds and produce a loss of information due to the large absorption of water molecules in the 1637 cm -1 wavenumber (Lasch and Naumann 2000). In some cases, this can be avoided by preparing a dried solid sample. To identify strains of bacteria, a library for characterization and identification is required, however this can be purchased from various scientific companies. Finally, the raw data can require extensive postprocessing analysis.

1.3-Additional tools to complement FTIR analysis in observing changes in intracellular composition
FTIR limitations can be compensated for by combining the analysis with other techniques. The most commonly used techniques to determine the composition of bacteria include raman spectroscopy, mass spectrometry, nuclear magnetic resonance spectroscopy (NMR), and X-ray photoelectron spectroscopy. Therefore, in order to access Raman spectroscopy, the outputs of Raman spectroscopy require filtration by notch or band pass filters before further analysis (Colthup 2012).

1.3.1-Raman Spectroscopy
The main benefit of Raman spectroscopy is that it requires less sample preparation compared to some FTIR modest (Colthup 2012). Although Raman spectroscopy is highly sensitive to some molecular vibrations (Larkin 2011), ambient noise, such as fluorescence, can interfere with the ability to obtain an accurate Raman spectra (Parker 1983).
Raman spectroscopy has been used to investigate the toxic effect of nanoparticle exposure to microorganisms and tracking the location of the nanoparticles with cells Raman spectroscopy has it combined with FTIR can provide valuable data about the microorganism's interactions with nanoparticles as well as nanoparticle characterization. This is possible because Raman and FTIR spectroscopies are sensitive to homo-nuclear (C-C, C=C and C≡C bonds) and hetero-nuclear (OH stretching in water) functional groups, respectively (Larkin 2011).

1.3.2-Mass Spectrometry
Mass spectrometer characterizes the mass of a molecule by determining the masscharge ratio (m/z) of its ion. Ions are produced by enforcing either the loss or gain of a charge from a sample, called the ionizing process. The ionizing process can be performed through electron bombardment, which produces the charged molecules (ions) of chemical compounds in a sample. Ions are electrostatically guided into a mass analyzer in which they are separated according to their mass-charge ratio and characterized. The results of molecular ionization, ion separation, and ion detection can be observed by a spectrum that provides molecular mass and structural information (Lay 2001). 14 A specific advantage of mass spectrometry over FTIR is the capacity to quantify the structural information, although the data collected by mass spectrometry is difficult to interpret (Gopal et al. 2013). On the other hand, the analysis of compounds with multiple functional groups is very difficult when using mass spectrometry (Fang et al.

1.3.3-Nuclear magnetic resonance spectroscopy (NMR)
NMR characterizes the physical and chemical properties of certain atomic nuclei. As all electrically charged nuclei have an associated spin, the applied external magnetic fields transfer energy at a specific wavelength from the base energy to a higher energy level. The wavelength corresponds to radio frequencies; hence, when the spin returns to its base level, energy is emitted at the same frequency, and the released signal results in a NMR spectrum for the nucleus (Shah et al. 2006). NMR spectroscopy can determine the molecular structure, content, and purity of a sample. NMR can detect an unknown compound's molecular structure and match it against spectral libraries or directly infer them from basic structures of molecules (Serber and Dötsch 2001). In the case of known compounds, NMR can characterize the molecular structure as well as physical properties at the molecular level. These properties include structural changes, phase changes, solubility, and diffusibility (Serber and Dötsch 2001).
In order to analyze more complex materials such as bacteria (Serber and Dötsch 2001), NMR is more appropriate than the previously mentioned techniques. Some advantages of this tool are versatile identification in terms of material composition and high chemical sensitivity (Shah et al. 2006). NMR can measure the long-range heteronuclear distances, which can be used to measure a distance between an antibiotic and a specific site in the membrane (Cegelski 2015). In addition, NMR can determine protein structures, and the location of every atom in the space (Schanda et al. 2014).
One of the disadvantages is that NMR has the potential to change the genetic code of bacteria and consequently the behavior of the protein, due to the use of synthetized amino acids for labeling (Reckel et al. 2005). The combined NMR and FTIR approach was applied to investigate the interaction of different structures of peptides (α-helices and β-sheets) with the bacterial membrane and its diastereomer (Oren et al. 2002). NMR revealed the interactions between peptides within membranes, while FTIR characterized the composition of α-helical structures within the peptide. NMR also detected the ratio of α-helical and β-sheets within membranes. The combined approach showed which peptide organized more selectively with the membrane (Oren et al. 2002).
To our knowledge, no study has used NMR and FTIR combined for nanotoxicology. A spectrometric analysis of bacteria exposed to nanoparticles by NMR can provide comprehensive information about the molecular composition of cells and complement the FTIR results.

1.3.4-X-ray photoelectron spectroscopy (XPS)
In XPS, photoelectrons are emitted from the sample surface due to excitation with mono-energetic Al kα x-rays. The energy of emitted photoelectrons is measured by an electron energy analyzer. The photoelectron peak determines the binding energy and intensity of the elemental identity, chemical state, and quantity of an element. XPS has different detection limits and analysis depths to detect the chemical composition of In terms of studying toxicity effects, XPS has been applied to characterize the nanoparticle surface properties, such as characterizing the surface coating of Ag-TiO 2 nanoparticles and then complementing this analysis by studying the bactericidal effect of these nanoparticles via FTIR (M. Lopez Goerne, 2011).

2.1-Fatty acid region
The chains of several bacterial membrane amphiphiles (e.g., phospholipids) and sidechain vibrations can be characterized by the peaks in the region around 3100-2800 cm -1 , which are observed due to the -CH stretching vibrations of CH 3

2.2-Protein Region
The peaks at 3200 cm -1 and 3060 cm - Fang et al. reported protein conformational changes in treated gram-negative bacteria exposed to quantum dot nanoparticles 21 . Here, clear decreases of peak intensity were observed after exposure in the spectral peaks at 1645 cm -1 , 1540 cm -1 , and 3288 cm -1 .
In another study, where gram-positive bacteria were exposed to oxide nanoparticles, the bands at 1600 cm -1 and 1408 cm -1 showed increases in peak intensity compared to the control 15 . A possible explanation for the mechanism in the gram positive bacteria exposed to oxide nanoparticles is that the carboxyl group concentrations increased due to the contribution of the carboxylic groups forming an inner sphere complex with the oxide metal center. Another explanation is that the ester bond of D-alanine branch is connected to the exposed gram-positive bacteria, then detached to form D-alanine acid The results of the discussed report clarified that the peptide and amino acids, regardless of their position (intracellular or extracellular), were modified in the nanoparticle-treated bacterial FTIR signature.

2.3-Carbohydrates Region
The spectral changes after bacteria exposure to nanoparticles may take place in the

2.4-Fingerprint region
The fingerprint falls between wavenumbers 900 cm -1 to 600 cm -1 and indicates unique weak bands that correspond to nucleic acids, i.e. phenylalanine, tyrosine, tryptophan, ). Additionally, due to a pH change in bacteria exposed to nanoparticles, the functional groups of the nucleic acid also undergo changes, which can be revealed in the fingerprint region of the FTIR spectra (Buszewski 2015).

Example of ATR-FTIR use for the study of E. coli exposure to
AgNPs using batch reactors E. coli K-12 (ATCC 23716), a non-pathogenic strain, was selected for this study. E.
coli is a gram-negative bacterium that has been found to be metabolically active and Microplates with six wells were used to grow bacteria until log phase, measuring the optical density at 600 nm (OD600) every 15 minutes. After 7.5 hours of bacterial growth, suspensions of AgNPs in deionized water were injected to achieve a concentration of 15 mg/L inside of the exposed condition (media plus bacteria+ AgNPs). Controls wells were included to detect contamination (media with no bacteria), and comparison between the non-exposed condition (media plus bacteria) and exposed condition. After that, plates were run for 7.5 additional hours to assess AgNPs toxicity at 2.5 hours, 5 hours, and 7.5 hours.
Liquid samples were prepared for ATR-FTIR (Nicolet iS50 FTIR, Thermo Scientific) analysis by fixing the optical density (OD600) to 0.8 using a UV−vis spectrophotometer (Genesis, 10UV, Thermo Scientific). Following this, bacteria were centrifuged at 13,000 rpm for 10 minutes, and the supernatant was removed. The pellets were suspended in 10 μL of PBS 10% (Zhang 2013), and the suspension of bacteria with AgNPs was directly placed onto the crystal surface (Gurbanov et al.

2015).
Spectra were the result of 256 scans with a resolution of 4 cm −1 in the 4000-350 cm −1 spectral range. The data was provided by Omnic software (Thermo Scientific) and processed using Matlab (Mathworks Software).
A unique FTIR spectrum is detected for the AgNP treated bacteria after each contact time (Fig. 2). Increased contact time resulted in decreased peaks in the spectra, where the untreated bacteria had the most intense peaks throughout the spectrum compared to the bacteria exposed for 7.5 hours, which had the weakest intensity peaks. To detect the specific responses within the treated bacteria, the data were analyzed region by region on the spectra. Fig. 3 shows the spectra for the fatty acid region, including the E. coli profile and E.
coli exposed to AgNPs. The obvious changes were related to shifting due to the deformation of >CH 2 and ⱱ s (C=O) in lipids after 7.5 hours of treatment ( Table 1). The change in the asymmetry vibration of phosphate groups also disappeared after 7.5 hours of exposure. The observed changes in the fatty acid region can be due to the alterations in the fluidity of the cytoplasm membrane or cell wall. Fig. 4 shows the ATR-FTIR spectra in the 1800-1200 cm −1 range of the E. coli as a control and E. coli exposed to AgNPs. The ATR-FTIR protein region peaks of E. coli before and after treatment are shown in (Table 2).
The band at 1284.8 cm −1 showed increased peak intensity in order to express the presence of amide III components of proteins in untreated E. coli. This band was shifted to 1287.3 cm −1 after 2.5 hours and 5 hours of exposure, but disappeared after 7.5 hours of exposure. Furthermore, amide A and amide B bands were observed at 3098.6 cm −1 and 3277.9 cm −1 in the untreated bacterial profile, respectively. However, the greatest shifting happened for the amide B band after 2.5 hours of treatment and remained consistent until 7.5 hours of treatment ( Fig. 3 and Table 2). exposure. In addition, a band at 968.6 cm −1 appeared in the spectra of untreated E. coli and exposed E. coli after 5 hours, while the exposed E. coli after 2.5 hours did not exhibit this band. All peaks of this region disappeared after 7.5 hours of nanoparticle exposure. The deformation of bacterial cell walls can be the reason of the shifting in the carbohydrate region (Fig. 5a), which shows P=O symmetric stretching in DNA, RNA, and phospholipids bands that shifted from 1064 cm −1 in untreated E. coli to 1078 cm -1 after 2.5 hours and 5 hours of exposure.
In the region between 900 cm -1 and 600 cm -1 , weak bands appeared (Fig. 5b). The    The interactions of AgNPs that caused the changes in the fatty acids, specifically -CH deformation, can be the reason for alteration in membrane permeability, which suggests the formation of pits in the bacterial cell wall. Another reason for differences in membrane permeability is the modification at the cellular ATP level, which was observed in exposed bacterial spectra due to the dehydration of phospholipids. In another study where bacteria were exposed to ZnO nanoparticles, the phosphodiester while another reason could be that a decrease in ATP levels inside the bacteria occurs due to the lack of nutrients and oxygen after 7.5 hours in a batch growth system.

Conclusion
FTIR is an extremely rapid technique compared to conventional techniques. FTIR has uniform applicability to various bacteria and a high specificity for differentiating toxic effects at intracellular levels. Thus, it can provide clear discrimination between chemically exposed bacteria in comparison to controls.

Abstract
Nanoparticles with antimicrobial properties are used in thousands of nano-enabled consumer products. Therefore, it is important to understand the response mechanisms of bacteria that are exposed to these nanoparticles at different conditions. Moreover, it is necessary to evaluate possible microbial adaptation mechanisms. In our study, Escherichia coli K-12 MG1655 (E. coli) were grown continuously in bioreactors at two specific growth rates (0.1 h -1 and 0.2 h -1 ) and then exposed to chronic concentrations of casein-coated silver nanoparticles (AgNPs) [1 mg/L] for about 180 generations. After initiating the injection of AgNPs, the results showed a change in growth kinetic parameters between non-exposed and exposed systems. Maximum yield (Y max ) decreased by 33%, while the maintenance coefficient (m s ) increased by 52%. This was evidence indicating the versatility of the culture to growth in the exposed conditions and even the ability to achieve a new stationary state. However, the adaptation was achieved at a metabolic cost. Comparing the concentration and composition of extra-cellular substances (ES) that were produced showed differences between the non-exposed and exposed conditions, and also between the exposed systems in the two growth conditions. In the AgNPs-exposed bioreactor (EB) growing at 0.1 h -1 , AgNPs-ES complexes showed that the ratio of the area representing β-sheets to the area representing α-helix proteins was higher than 2.4, which implies the formation of a protein corona, while at an exposed growth rate of 0.2 h -1 this ratio was showed that culture growth conditions significantly affects bacterial response to nanoparticle exposure. Therefore, these growth parameters should be determined and reported when performing toxicological tests.

Introduction
Antimicrobial nanoparticles are used to inhibit and deactivate unwanted microorganisms, 1-3 however, bacteria stress response mechanisms can hinder the efficacy of these nanoparticles. Therefore, it can be inferred that the physicochemical properties of the nanoparticles will also likely change throughout the experimental period. 9 Continuous reactors can be used to overcome these limitations. These reactors can achieve steady-state conditions where the concentration of the growthcontrolling substrate and density of the culture do not change significantly with time. 10 Continuous cultures have been used extensively to study bacteria stress response and to investigate the development of antibiotic resistance. [11][12][13] In a previous study, we compared bacteria exposed to pulses of AgNPs using continuous and batch reactors, and our results showed that in terms of membrane permeability, there were marked differences between both conditions. 1 Numerous studies have reported the effects of AgNPs in terms of multiple parameters related to cell viability. 14,15 However, to the best of our knowledge, changes in the physicochemical properties of the nanoparticles at different continuous culture conditions have not been reported.
Additionally, most studies have focused on the antimicrobial effects of nanoparticles, but have not considered that in nature, microorganisms grow in the presence of inhibitors. Therefore, in addition to nutrient uptake, growth may be controlled by the presence of anthropogenic inhibitors, such as nanoparticles.
Therefore assessing the impact of nanoparticles on kinetic parameters that commonly occur in the ecological context is extremely important. [16][17][18] The main objective of this study was to elucidate the impact of bacterial growth conditions and their response to chronic exposure to silver nanoparticles. We used continuous bioreactors to determine the inhibitory effect of nanoparticles in terms of kinetic parameters, nanoparticle-ES interactions, and transcriptomic analysis. Additionally, culture samples from the effluent of the exposed bioreactor (EB) and control (CB) bioreactors were re-cultured in batch mode, and then acutely exposed to AgNPs in order to compare membrane permeation, respiration activity, and reactive oxygen species (ROS) generation responses.

Bacteria culture.
In this study, we used a slightly modified version of the multiplex bioreactor system from our previous study 1 ( Figure S1 in the ESI). The vessels were fed with sterile M9 minimal medium adjusted to pH 7. After reaching steady state in terms of bacteria optical density (OD), the four bioreactors were inoculated with E. coli, and two bioreactors were injected continuously with AgNPs using syringe pumps (model 100 from Kd-Scientific)  Figure S2 in the ESI). The operational conditions in terms of specific growth rate selected for this study were to 0.1 h -1 and 0.2 h -1 .
Biomass concentration was followed by measuring OD600 according to previously published protocols. 47 Additionally, dry biomass weight was obtained for all bioreactors and normalized based on the internal volume of the continuous bioreactor vessels.
Glucose concentration (S) in the bioreactors was determined thru a glucose (HK) assay reagent 25

Nanoparticle Characterization.
AgNPs were continuously injected into each bioreactor to achieve a concentration of 1 mg/L inside the exposed bioreactors.
( Figure S4 and details provided in ESI).
The nanoparticle hydrodynamic diameter size distribution and zeta potential (ζ) of the nano-suspensions were determined by using dynamic light scattering (DLS) using a Zetasizer Nano (Malvern, ZEN 3600). Inductively-coupled plasma spectroscopy (ICP-OES optima 3100, Perkin Elmer) and inductively coupled plasma mass spectrometry (iCAP Q ICP-MS) were used to measure the concentrations of AgNPs and silver ions (Ag + ). The ionic release from AgNPs at each condition was quantified as Anaya et al., 1 Ag ions released was quantified for the conditions: AgNPs with ES (from continuous bioreactors), AgNPs with M9 minimal medium, and AgNPs with distilled water (control). Digestion in 2% nitric acid was required for all samples before analysis.

EPS quantification and characterization. Initially, ES was separated from
the bacterial culture suspension of continuous bioreactor using a previously described methodology by Seo and Bailey. 46 Dried ES were characterized using ATR-FTIR (Nicolet iS50 FTIR, Thermo Scientific). Spectra of the samples were the result of 256 scans with a resolution of 4 cm -1 in the 1800-900 cm -1 spectral range (Omnic software, Thermo Scientific) and processed using MATLAB (MathWorks Software). Hierarchical cluster analysis (HCA) was applied to discriminate the compositional differences between the ATR-FTIR spectra of ES from exposed and control bioreactors at the different conditions tested. 29 For HCA, a data set was collected from the pair-wise similarity coefficients of all spectra as a matrix of correlation coefficients, which contains total number of spectra (N entries). 30 Between two spectra, each correlation coefficient can range from 0.0 for completely different spectra to 1.0 for identical spectra. The similar spectra were obtained by the recalculation of correlation matrix. Then, the identical spectra were merged into a new object and the merging process repeated until all spectra are combined into a small number of clusters.
Thermogravimetric analysis (TGA) of dried ES was performed in a thermogravimetric apparatus with N 2 atmosphere by using a TA instrument Q500 TGA following previously reported methodology. 31 Data were also obtained from the first derivative of the TGA line. Raw data of TGA were smoothed by a moving average and the Gaussian fit to first derivative of smooth data was found by MATLAB software.
Formation of protein corona on the AgNPs surface was determined as well by using the ATR-FTIR. 32 It was assumed that the observed peak is the summation of two Gaussian functions representing β sheet and α helix structure, 3,32 coefficients for these functions have been determined by a parametric fitting of the data by MATLAB on the observed peak in range of 1800-1550 cm -1 . After subtraction of the protein from the casein layer of AgNPs, the ratio between areas under β sheet to α helix structure of proteins was altered after corona formation. 35 The ES without AgNPs was exposed to different concentration of AgNPs to validate the protein corona formation onto AgNPs surface.

Reverse-Transcription Quantitative Polymerase Chain Reaction (RT-qPCR)
Reverse transcript (RT)-qPCR is a valuable tool widely used for analysis of gene expression. [33][34][35][36][37] Previously, dominant genes have been identified to response AgNPs inhibitory effects including outer membrane porin (ompF), copper efflux 70 oxidase (cueO), copper/silver efflux system (cusA), and copper transporter (copA) genes. 4,19 These genes are responsible for the following metabolic pathways; outer membrane porin F, oxidizes model substrate dimethoxyphenol, silver and copper efflux of membrane transporter, and silver-translocating P-type ATPase efflux pump. 19 Also, a previous report showed that at the transcriptional level, fatty acid synthesis inhibited by the AgNPs inhibitory effects 35

Comparison of cultures to acute exposure to AgNPs
Samples were collected from the effluent of the bioreactors EB and CB at 92 hours and 60 hours for runs with specific growth rates of 0.1 h -1 and 0.2 h -1 , respectively.
Bacteria from the samples were harvested and exposed to two concentrations of AgNPs (1mg/L and 10mg/L) to compare the response of the resultant EB and CB cultures for each condition in terms of respirometric activity, membrane permeation and ROS production.

Respirometric analysis. Cell respiration was quantified in non-growing
conditions (media consisting on PBS solution and glucose, but not other nutrients), to compare electron transports activity due to the aerobic metabolization of the carbon 72 source. Respiration activity was determined through the reduction of tetrazolium dye 41 measuring absorbance at 590 nm every 0.25 hour for 5 hours.

Membrane permeation analysis.
The membrane permeation of E. coli was determined using the Baclight kit (propidium iodide and SYTO 9) with a microplate reader. Propidium iodide becomes intercalates to the DNA only when the membrane is disrupted, while SYTO 9 indicates intact membranes. 41 The green/red fluorescence ratio between the EB and CB at the given AgNPs concentration (1 mg/L and 10 mg/L) was calculated as previously reported. 60 The membrane permeation of E. coli was determined using the Baclight kit (propidium iodide and SYTO 9) with a microplate reader. Propidium iodide becomes intercalates to the DNA only when the membrane is disrupted, while SYTO 9 indicates intact membranes. 41 The green/red fluorescence ratio between the EB and CB at the given AgNPs concentration (1 mg/L and 10 mg/L) was calculated as previously reported. Non-fluorescent 2',7' -dichlorofluorescin diacetate (DCFDA) was formed by deacetylation of cellular esterases. 33 Abcam protocol 42 was employed after slight modifications (details in the ESI). Intensity was measured at an excitation wavelength of 488 nm and at an emission wavelength of 535 nm for 12 hours using a fluorescence microplate reader.

Statistical analysis
The data were analyzed by student's t-test using MATLAB. Differences between means were considered statistically significant at p < 0.05. Data are presented as mean ± standard error of the mean (SEM) of at least three independent experiments, unless otherwise stated. T-test analysis was performed to detect the differences between kinetic parameters of control bacteria and AgNPs-exposed bacteria as well as the differences between toxicity effects of nanoparticles on the resultant culture of bioreactors (control bacteria and AgNPs-exposed bacteria). Also, the changes in negative charge of nanoparticles surface in different solutions were assessed by t-test analysis.  Figure S5 in the ESI). However, in all exposed bioreactors the concentration of AgNPs was lower than the desired value due to possible trapping in the pellets of bacteria during the centrifugation step of the sampling process. Table 3 shows the concentration of bacteria inside the biore`actors (X), concentration of glucose inside the bioreactors (S), saturation constant (K s ), yield coefficient (Y x/s ), maximum yield coefficient (Y max ), and maintenance coefficient (m s ).

Effect of growth conditions on particle size and zeta potential.
The hydrodynamic diameters of the AgNPs in pristine M9 minimal medium and in the ES of the exposed bacteria were measured at 32 hours for both growth rates conditions (Table S5 in the ESI). Also, the stability of the AgNPs in distilled water (DI) was determined as a control ( Figure S7  Since the biomass evolution profiles of the EB-0.1 cultures were less inhibited by the AgNPs, Cryo-TEM images were only collected for this condition (Fig. 7).

80
For reference, images of AgNPs in DI water were also collected (Fig. 7a). ES from EB-0.1 images showed that the nanoparticles were trapped in the ES after 8 hours in the bioreactor (Fig. 7b). The images also showed that agglomeration increased after 32 hours at EB-0.1 (Fig. 7c). These results suggested that nanoparticle-ES interactions can lead to biological corona formation on the surfaces of the nanoparticles due to the high affinity of the nanoparticle surface with sulfur-or nitrogen-containing compounds, especially amino acids. 19 Since protein corona is an unstable form of assembly on the surfaces of the nanoparticles, 47,48 different AgNPs sizes can be observed in Fig. 7d. Therefore, the surface of the AgNPs had been modified and these nanoparticles could not pass through cell membranes (Fig. 7e).

Formation of protein corona on the surfaces of the nanoparticles.
Protein conformation of α-helices and β-sheets were examined in samples collected HCA of FTIR data from ES composition showed that ES extracted from EB-0.1 segregated distinctly from the ES from control groups of 0.2 h -1 ( Figure S10 in the SI).
Therefore, the formation of protein corona on the surface of AgNPs depended on the composition of ES released by bacteria at different specific growth rates.
TGA was conducted (weight loss vs. temperature) to determine thermal stability of the ES that were interacting with the AgNPs. The analysis was performed using derivative thermogravimetric analysis (DTG) from the TGA results. The thermal stability of the ES at CB-0.1 showed a double peak at DTG, which exhibited 30% weight loss at 100 °C followed by a minor peak at 300 °C with less than 5% weight loss (Fig. 9a) and ( Figure S11a in the ESI). In addition, DTG of AgNPs-ES at EB-0.1 (Fig. 9a) showed a 82 major peak at 100 °C to 200 °C associated with 20% weight loss, and this was followed by a minor peak at 400 °C with less than 10% weight loss. On the other hand, ES at CB-0.2 had a double peak with 20% weight loss at 100 °C and 220 °C (Fig. 9b). However, AgNPs-ES at EB-0.2 (Fig. 9b) had two more minor peaks with 2% weight loss that were observed at 450 °C to 550 °C in addition to the peaks at 100 °C to 180 °C and 220 °C with 12% weight loss. Thermograms at 700 °C indicated that the AgNPs had the lowest weight loss at EB-0.1, while the ES had their maximum weight loss due to the degradation of protein at the higher temperatures.

Gene expression level at two specific growth rates
The results of the analyses indicated that there were several mechanisms by which bacteria responded to AgNPs exposure in both specific growth rates. The results, including various quantities of ES and their compositional characteristics as well as the AgNPs-ES interaction at different specific growth rates, indicated the necessity of investigating both specific growth rates on the gene expression level of eight target genes (ompF, cueO, cusA, soxS, cpsB, zwf, copA, and fabR) normalized using an internal reference gene, rrsB (Fig. 10). In this study, the membrane in EB was protected by the response of upregulated genes, such as copA and cusA to AgNPs at both specific growth rates. These genes are responsible for silver and copper efflux of membrane transporter, lipid biosynthesis, and silver-translocating P-type ATPase efflux pump. Nanoparticles mediate the generation of ROS and also modulate the antioxidant activities of ROSmetabolizing enzymes, such as slow electron transport, the NADPH-dependent flavor enzyme, catalase, glutathione peroxidase, and superoxide dismutase. 33 In general, the soxS and cueO 19,33 genes, which are responsible for ROS expression, were not affected by the AgNPs at EB-0.1, however, only cueO genes were up-regulated at EB-0.2.
Also, zwf gene showed higher expression at EB-0.2 in compare with EB-0.1 which may be associated with changes in the metabolic pathway of glucose-6-phosphate 86 dehydrogenase in the respiration process 33 and/or the relatively impenetrable membrane of EB-0.1. 50 At EB-0.1, the nanoparticles could not cause the generation of ROS, so the cells were not involved in compensating for the disruptive impacts of the nanoparticles due to the formation of ROS. However, the activation of the copA, cpsB, fabR, and cusA regulons prevented irreversible damage to EB-0.1.

Comparison of cultures to acute exposure to AgNPs
Since the continuous culture results showed different responses at the two specific growth rates, we compared the response of the resultant culture at these two conditions to acute exposure to nanoparticles in terms of: respiration activity, membrane permeation, and intracellular ROS production. These tests were performed in batch conditions using the resultant cultures from the control bioreactors (CB) and exposed bioreactors (EB). In addition, no nanoparticles for the negative control and 10 mg/L for the positive control were applied for the inhibitory effect of AgNPs. Fig. 11 shows no differences between the percent of remaining respiration (PRR) of EB and CB groups at both specific growth rates (p > 0.05) when they were exposed to different nanoparticle concentrations.

Inhibitory effect on respiration.
Silver ions released from the nanoparticles are the main mechanism for metabolic disruption of bacteria. These results showed that at batch growing conditions, the resultant cultures respond similarly to nanoparticle exposure.

Evaluation of the generation of ROS.
When CB-0.1 was exposed to 1 mg/L of AgNPs, the fluorescence increased, although the exposure of EB-0.1 to AgNPs (1 mg/L) did not show any significant generation of ROS (p = 0.89) (Fig. 13a). ROS generation and UMC were observed between CB and EB, which contrast with PRR between CB and EB in the same condition. This suggested that ROS generated during electron transport is slowed by high mitochondrial membrane potential. 51 Hence, oxygen radicals react with oxygen dissolved in the membrane and cause the membrane disruption.

Discussion
In terms of AgNPs impacts on kinetic parameters, our results showed that the microbial population continues to grow in the presence of the AgNPs, however, metabolic activity is impaired. This is demonstrated by changes in the kinetic parameters (decrease of Y x/s and Y max as well as increase of K s , and m s ). The metabolic changes led to a new steady state for a continuous culture in which 92 the resultant concentration of biomass is lower than that obtained at the steady state in non-exposed conditions. Thus, the resultant biomass concentration was inversely dependent on the specific growth rate (high specific growth rate, low end biomass concentration).
Kinetic parameter analysis helps us to understand the fate of the microbial population as a whole, but it does not provide insight into the mechanism(s) governing the observed effect. As such, bio-macromolecule analysis was Notably, the ES composition of the AgNPs-exposed bioreactors were different than the controls at the same condition as evidenced by the hierarchical cluster analysis of FTIR data (HCA, Figure S10  In the context of transcriptomic analysis, the main genes that respond to chronic levels of continuous stress due to the nanoparticles are zwf, CusA, and copA. Specifically, the up-regulation of the zwf and copA genes was in agreement with the changes in growth parameters such as K s , Y max and m s in the exposed bioreactors . The zwf gene could lead to a K s increase, while decreasing Y max, in two ways, inactivation of phosphomannose by the AgNPs, and a drop in the efficiency of sugar metabolism. 43 The copA gene in the exposed cultures translocated silver thru the ATPase efflux pump by increasing m s . Furthermore, the upregulation of the cpsB gene, production of capsular polysaccharide and canonical regulatory transcription, 39 was faster at EB-0.1 than EB-02, which showed a faster metabolic pathway response to the stressors for cultures growing at lower specific growth rates. 53 These results are evidence that cell requirements increase due to chronic exposure to AgNPs. Finally, the acute test results showed that even if bacteria are able to adapt to nanoparticles, they cannot transfer this adaptability to the following generations over 114 generations at 0.2 h -1 (number of E. coli generation during 32 hours of nanoparticles exposure). However, at 0.1 h -1 , the number of generations was 206, which increases the probability of bacterial adaptation to nanoparticles, which was in agreement with the acute tests. In this content, UCM and ROS generation between CB and EB showed that at a EB condition, nanoparticles did not significantly disrupt the cell membrane, and did not produce oxidative damage by ROS. Hence, EB are not compensating for a lower disruptive effect from nanoparticles than CB when they are exposed to nanoparticles of 1 mg/L. However, the low PRR for the 0.1 h -1 resultant culture when exposed to AgNPs indicates the impact of culturing conditions on the bacterial response to AgNPs.  54 This study is in agreement with the results obtained from the higher growth rate condition. However, another study claimed that genomic analysis of AgNPs exposed E.coli showed resistance by generation 200, where three mutations smoothly occurred in AgNPs resistance bacteria 2 .
These studies indicate that despite previous claims to the contrary, bacteria can easily evolve resistance to AgNPs, and this occurs by relatively simple genomic changes in a few generations. In conventional batch tests, since the contact time was 5 hours, the number of generations that were exposed to nanoparticles was 17, which means that at a batch culture condition the probability of adaptation is even lower than that for a continuous culture, unless the specific growth rates of a culture are high enough to decrease the number of generations. Hence, at lower specific growth conditions, the probability of producing nanoparticle-resistant bacteria will increase. Therefore, the bacterial culture condition influences the inhibitory effect of nanoparticles by changing their physiochemical properties, and also caused permanent bacterial resistance.

Conclusions
The fate of nanoparticles and their inhibitory effects in a continuous culture depends on the bacterial specific growth rate associated with different concentrations and the composition of ES produced at each growth condition.
ES at a lower growth rate are more effective in reducing the inhibitory effect of the nanoparticles. This occurs thru consumption of ROS, immobilization of the nanoparticles, and the formation of protein corona on the surfaces of the nanoparticles. Cultures exposed to nanoparticles are able to growth and achieve new stable conditions (steady state) at higher energy consumption than unexposed cultures. This is due to the activation of several regulons, such as zwf, CusA, and copA, which occurs to prevent irreversible damage from the stress condition.

Conflicts of interest
The authors declare no conflicts of interest.

SUPLEMENTARY INFORMATION
Additional details for method Figure S1. Condition and tests performed in this study

Maximum specific growth rate
First we performed the batch culture to find the optimum temperature and optimum substrate concentration for E.coli Mg1655 growth. The µ max depends on the culture temperature, 1 hence, temperatures in the range of 37 °C -40 °C were examined in a batch bioreactor to find the optimum temperature for E. coli growth. Also, at high substrate concentration, growth will occur at µ max . Therefore, to determine the optimum substrate concentration, two concentrations of glucose (8 g/L and 20 g/L) as the only carbon source. 1,2 Then, the µ max of E. coli was determined by batch culture in optimum growth condition (temperature = 37 and glucose concentration = 8 g/L). For this purpose, direct measurements of µ max carried out in a batch bioreactor (Error! eference source not found.b).
Also, in order to obtain the wash out dilution factor the experiment was performed with six bioreactors: four bioreactors inoculated with 300 µl of E. coli (OD at 600nm was 1.8) and 2 bioreactors as controls contained only M9 minimal media. A concentration of 8 g/L of glucose as the only carbon source was added into minimal M9 media container. Initial µ was 0.1 h -1 which increased to 0.2 h -1 after 20 hours and finally increased to 0.3 h -1 after 32 hours. As shown in Error! Reference source not ound.a, after increasing specific growth rate to 0.3 h -1 , the cells start to wash out.
As the optimum specific growth rates for the investigating the response of cells at different specific growth rates to AgNPs, 0.1 h -1 and 0.2 h -1 were selected.

Cell dry weight
Then, we found the cell dry weight for each condition by multiplying cell dry weight with the absorbance at OD600. In detail, 1 ml of cell suspension was centrifuged in pre-dried and pre-weighed 1 ml test tubes at 13000 × g for 5 min. After removal of the supernatant, the samples were measured for cell wet weight and then dried at vacuum for at least 24 hours. 47 The dry biomass weight obtained for all four bioreactors. Then, they normalized based on the inside volume of continuous bioreactor's vessels.

Substrate concentration
We determined the outlet substrate concentration (S out ) by using Glucose (HK) Assay

Saturation constant
We normalized the outlet substrate concentration (S out ) by bacterial absorbance (OD 600). Then we converted the outlet substrate concentration (S out ) unit from mg/mL to mmol/L. We prepared a chart calculating K s , with 1/µ as y axis and 1/ S out as x axis. K s was determined by multiplying slope with 1/intersect. Also, we checked K s values with K s =S*((µ max /µ)-1) equation (Table S 1 and Table S 2).

Substrate mass balance
Also, mass balance for substrate in bioreactors at steady state checked by the following equation:

Maximum Yield and maintenance coefficient
The maximum yield and maintenance coefficient calculated by the linear regression of 1/µ as x axis and 1/yield as y axis. Slope is the maintenance coefficient and 1/intersect is the maximum yield. Figure S3. Relation of yield coefficient (Y x/s ) of E. coli to specific growth rate (µ) using double reciprocal linear transformation. Δ, Grey line and grey box represents control bacteria (CB). ▲, Black line and black box represent AgNPs-exposed bacteria (EB).

Continuous injection of nanoparticles to bioreactors
AgNPs are continuously injected in to the system to achieve 1 mg/L concentration in bioreactors. Precisely, three single syringe infusion pumps were utilized for continuous injection of AgNPs into bioreactors ( Figure S4 in the ESI). Based on conservation of mass for AgNPs in distilled water and continuity of the flow we can write: 1 + 2 = 3 1 1 + 2 2 = 3 3

= 2 2 / 3
Where Q 1 is M9 minimal culture media inflow rate, C 2 and Q 2 are AgNPs concentration and flow rate, from syringe infusion pump, respectively, as well as Q 3 is the outflow and C 3 is concentration of AgNPs in outflow. Based on this formulation the concentration of AgNPs in outflow is determined and assumed to be equal to AgNPs concentration inside the reactor (Table S 3).  and AgNPs solution were used for measuring the particle hydrodynamic diameter size distribution and zeta potential (ζ) of the suspensions as well as measuring AgNPs and silver ions concentrations.

Testing designed primers for transcriptomic analysis
To ensure that the primer dimer would not be formed, multiple primer pairs were analyzed for the possibility of primer dimer formation. We ordered the primer pairs which showed the lowest possibility to form self and/or hetero dimers. Next, a series of dilutions for the cDNA template was tested for qPCR efficiency and the melting 112 curves were carefully analyzed. Since, there was only a single melting point for each primer pair it was established that the amplification was specific and only one target amplicon for each primer pair was being generated. The melting temperatures of the qPCR product ranged from 81.81 °C to 86.84 °C. The primer dimer formation was ruled out by the absence of melting point at temperature between 65 °C and 70 °C.
Furthermore, we used ΔΔC T method for gene regulation calculation. First, for both control bioreactors and AgNPs-exposed bioreactors, we determined the ΔC T by subtracting the mean C T value of each gene from mean C T value of housekeeping gene (rrsB). Then, in order to calculate the ΔΔC T we subtracted the ΔC T of genes at control bioreactors from the ΔC T of genes at AgNPs-exposed bioreactors. Finally, the regulation of each gene (R) is obtained by 2 -ΔΔCT  and incubated for 1 hour at 37 °C. Finally, cells were washed with buffer 1X and exposed with given concentration of AgNPs. Procedures were performed in the dark.
Each 96-well microplate consists of blanks (AgNPs 1 mg/L and 10 mg/L), AgNPs exposed bacteria, and control bacteria. The blanks absorbance were subtracted from AgNPs exposed bacteria, allowing us to make consistent comparison with control bacteria.

Dissolution rate of nanoparticles in continuous bioreactors
Dissolution experiments consisted of measuring the concentration of Ag ion released in ES from both growth rates. Ag ions release was quantified for ES with AgNPs     ES can affect the inhibitory effectiveness of AgNPs through two mechanisms;

ES concentration and characteristic at different specific growth rates
physiochemical alteration of nanoparticles' surface modifications and nanoparticles' immobilization in ES matrix. 7 In here, it was considered the heterogeneity between the compositions of ES at different specific growth rates ( Figure S10). Therefore, after characterization of ES compositions by ATR-FTIR, 8  the control group ES concentration and composition was function of growth rate. In addition, ES from AgNPs exposed bacteria changed in composition at the higher growth rate condition and obtained similar compositions to ES from exposed bacteria at low growth rate condition. Compared with CUV, PL has showed higher inactivation of microorganisms and degradation of PAHs, because of it rich and broad-spectrum UV content, high energy peaks and predictable treatment outcomes 4 . Typical PL lamps produce high intensity light pulses in a wavelength spectrum between 100 to 1100 nm. For water treatment purposes, short pulses (100-400 μs) in the wavelength range below 400 nm are commonly applied 1,5 .
In terms of microbial disinfection, several studies have shown that PL produce DNA damage [3][4][5][6][7][8][9] . High intensity light pulses of PL cause thymine dimerization in the DNA chain and prevent replication 4 . Kramer and Muranyi 5 reported that with fluence up to 0.27 joule/cm 2 , only about 80% of DNA was formed relative to the non-exposed cells.
They also reported that observed colony count reduction after exposure to PL in L.
innocua is directly corresponded to the occurrence of DNA low quantity 5 .
DNA damage have been reported as the main mechanism, however other studies have showed the high complexity of the microbial response to PL to be a multi-target inactivation process 8,9 . Transcriptomic analysis have shown that several stress related proteins upregulated in PL exposed cells, implying an increase in transcriptional and translational processes in response to PL 6 . Other reported effect are increase of bacterial mutation frequency as well as changes in the abundance of 19 proteins as revealed by a global proteome analysis 6 and whole genome DNA microarray analysis 7 .

127
The effect of PL on cellular components and metabolic cell activity [6][7][8] have been described while other studies documented structural damages to the cell membrane or cell wall after PL exposure [11][12][13][14] . For instance, structural cell damage has been observed by propidium iodide uptake 11,12 , protein or other cellular component leakage 12,13 , as well membrane shrinkage 13 . Effects of metabolic activities and cellular dysfunctions, low glucose uptake 6,7 activity, and reduction of esterase activity 12,15 were reported 6,7 as well as less cultivability and more cellular damage on respiration, enzyme activity, and membranes 7 . Kramer and Muranyi also assessed less cultivability and more cellular damage on respiration, enzyme activity, and membrane 5 . On the other hand, Garvey et al. 14 suggested that PL exposed cells are viable but nonculturable and that colony count reduction cannot be directly linked to membrane disruptions. There are incomplete understandings among previous reports that need to comprehensively investigate including the PL impacts on bacterial inactivation and bacterial cellular components. In addition, any direct effects of PL on cellular biomolecules, such carbohydrates, nucleic acids, proteins and fatty acids, were not investigated. To the best of our knowledge, detecting the heterogeneity between the cellular components (nucleic acids, carbohydrates, proteins, and fatty acids) of PL exposed bacteria and regrowth of PL exposed bacteria have not been reported.
It is well known that UV light photolyases PAHs 3,4,16,17 . Rates of photolysis generally decrease with decreasing PAH molecular weights 17 . At wavelengths present in sunlight, higher molecular weight PAHs, such as pyrene, possess higher reactivity because of their higher extinction coefficients 16 . There is evidence for phenanthrene 128 degradation when irradiated with PL 18 , but the effect of PL on other PAHs remains unclear 18 .
The impetus for this study was two-fold: (1) to explore the impacts of pulsed light (PL) on Escherichia coli MG1655 in terms of membrane disruption, impact on bacterial respiration, membrane permeation, regrowth lag phase duration and the biomolecular alterations of fatty acids, protein, carbohydrates, and nuclei acids, (2) degradation kinetics of polycyclic aromatic hydrocarbons by PL. We consider the results of this study of interest to water treatment professionals seeking an alternative to conventional water disinfection and treatment technologies.

Materials
A non-pathogenic strain of Escherichia coli K-12 strain MG1655 (E. coli) was selected for this study, because is a Gram-negative bacterium extensively used in disinfection studies 19,20 .  Table 4. Both lamps are lowpressurized and mercury free.

130
The lamps were installed in a laboratory scale SteriPulse®-RS 4000 system (XENON Corporation, Wilmington, MA) (Fig. 14  and nucleic acids between exposed, non-exposed, regrowth of exposed, and regrowth of non-exposed bacteria. The comparison was mainly based on exposed and regrowth 132 of exposed to non-exposed, but regrowth of non-exposed was also added to verify its similarity to non-exposed and to consider possible effects of compositional differences due to bacterial regrowth.

Culturability
In order to compare between the culturability of PL1 exposed bacteria and PL2 exposed bacteria, two tests such as colony forming and bath growth were performed.
The log removal of bacteria were obtained for PL1 and PL2 through counting the number of colony forming unit (CFU) based on previous study 24 .

Batch growth
10 µL of PL1 and PL2 exposed bacteria were regrown in 5 mL of LB media at 37 °C in a six well microplate. Absorbance at 600 nm was recorded every 0.25 hour for 18 hours. The slope of the log phase showed the specific growth rates of the bacteria. The control and blank were non-exposed bacteria and LB media, respectively. Experiment was run in two 6 well microplate and each microplate contains duplicates of each condition. Afterwards, samples from each well were collected for FTIR analysis.

Respirometric test
Cell respiration was quantified in non-growing conditions of PL1 and PL2 exposed bacteria and non-exposed bacteria to compare their metabolic activity of electron 133 transports by tetrazolium dye. Non-growing condition was selected for metabolic activity detection in order to determine only the respiration of bacteria and avoid growth, since, bacteria are not able to grow in PBS. The remaining respiration rate were determined using previously reported protocols 25 . In brief, 96-well microplate was prepared in final volume of 100 μL in each well. Then the following list added respectively; glucose [80 mg/L], tetrazolium dye, PL exposed bacteria solution and each well was mixed thoroughly by pipetting at least 10 times. Controls and blanks were respectively non-exposed bacteria and only PBS instead of bacterial solution.
The experiment was run in two microplates and each microplate contains triplicate of each condition. A Microplate reader recorded the absorbance every 0.25 hour for 12 hours. The percentages of remaining respiration (PRR) of the bacteria exposed to PL1 and PL2 were compared by dividing the slope value rates of PL2 exposed bacteria and PL1 exposed bacteria with the slope values of the non-exposed bacteria only first 4 hours was considered (Equation 3).

= ( / ) × 100 Equation 3
Where, P t is the slope from the absorbance-time graph for PL exposed bacteria and P c is the slope from the absorbance-time graph for non-exposed bacteria (control).

Epifluorescence staining membrane permeability analysis
A membrane permeability test was also performed for the PL1 and PL2 exposed bacteria to compare the membrane permeability with non-exposed bacteria. The cell membrane permeation of PL1 and PL2 exposed E. coli was quantified by propidium iodide and SYTO 9 dyes in a microplate reader based on previous report with slight modifications 22 . A calibration curve of live and dead bacteria was prepared to quantify and compare the membrane disruption on bacteria before and after PL exposure. Each plate contained triplicate wells for each condition to quantify the disturbed cell membrane. The green/red fluorescence ratio between the PL exposed bacteria and non-exposed bacteria were calculated. Undisturbed cell membrane percent (UCM) obtained by Equation 4.
Where, P t is green/red fluorescence ratio for PL exposed bacteria and P c is green/red fluorescence ratio for non-exposed bacteria (control)

Fourier transform infrared spectroscopy (FTIR) analysis
Bacterial compositions from PL1 and PL2 exposed, non-exposed, regrowth of PL1 and PL2 exposed, and regrowth of non-exposed conditions were characterized by directly placing specimens on the crystal surface for attenuated total reflection (ATR) mode of FTIR (Nicolet iS50 FTIR, Thermo Scientific) in the 3300-600 cm -1 spectral range same as previously reported protocol 26 . The software for the data providing was Omnic software (Thermo Scientific) and for data processing was MATLAB (MathWorks Software).
The heterogeneity of bacterial various biomolecular groups such as fatty acids, proteins, and nucleic acids between exposed, non-exposed, regrowth of exposed, and regrowth of non-exposed bacteria were determined by HCA 27 . For HCA, a data set was collected from the pair-wise similarity coefficients of all spectra as a matrix of correlation coefficients, which contains the total number of spectra (N entries). 28 Between two spectra, each correlation coefficient can range from 0.0 for totally 135 different spectra to 1.0 for identical spectra. The similar spectra were obtained by recalculation of the correlation matrix. Then, the identical spectra were merged into a new object, and the merging process repeated until all spectra were combined into a small number of clusters.

Degradation of aqueous phase PAH
Stock solutions of PAHs were produced by mixing reagent-grade, powder forms of the model PAHs with methanol (1 mg/L) and stirred for 24 hours in the dark. PAHs sample aliquots were prepared by transferring 10 mL of the stock solution to 1 L of distilled water and stirred for another 24 hours in the dark. Then, 20 ml of PAHs sample aliquots were added to Teflon plates and were exposed to PL1 and PL2 for a range between 0 to 190 fluences (joule/cm 2 ). Controls were run for each condition and consisted of non-exposed PAHs which placed on the chamber uncovered and without turning on the PL. Dichloromethane (DCM) was considered as blank Blanks for analyzed for quality control purposes.
After exposure, a 30 µL of deuterated PAH mixture with a individual PAH concentration of 50 mg/L was added to the samples as an internal standard. Then, 2 mL of dichloromethane (DCM) was added to the samples and mixed completely through agitation for 1 min. After 2 min of settling time, the DCM phase containing the PAH (1 mL) was extracted and analyzed by the GC-MS (QP2010S, Shimadzu).

PAHs Degradation Kinetics
The degradation of PAHs following PL1 and PL2 exposure were modeled as first and second order decays on Equation 5 and Equation 6.

Equation 6
where α is the decay rate coefficient, C 0 is the initial concentration of PAHs, C t is the concentration of PAHs at time t.

Toxicity of degraded PAH solution
For respiration test of the exposed PAHs on E. coli, the remaining respiration was quantified in the same conditions that mentioned above. The only difference was measuring the respiration of E. coli in presence of the PL1 and PL2 exposed PAHs.
Previously exposed PAHs from three different fluences (0, 19, 190 joule/cm 2 ) were selected to follow the effects of the PAHs and byproducts of exposed PAHs on bacterial respiration. Table S1 in the SI contains the used PAHs concentration in this test.

Disinfection mechanisms of PL1 and PL2 on E. coli K-12
Initially we performed the kinetic study of bacterial exposure to several fluences including 0, 19, 38, 57, 76, 95, 190 joule/cm 2 using both lamps. Then, we selected 76 joule/cm 2 for PL1 and 95 joule/cm 2 for PL2, based on the kinetic study of colony forming test and membrane permeation results ( Figure S1 in the SI).

Culturability of PL1 and PL2 exposed E. coli K-12
The influence of PL1 and PL2 on culturability of exposed and regrowth bacteria were determined by different analysis. These analyses include membrane filtration 137 (capability of to form colonies), as well as growth rate determination and lag phase duration from regrowth of exposed bacteria.

Colony forming units
PL1 treatment with fluence of 76 joule/cm 2 and PL2 with 95 joule/cm 2 were shown to be enough bacterial removal as shown the log removal for each case ( Figure S1). Fig.   15 shows the reduction in the bacterial colonies in PL1 and PL2 by demonstrating reduction of colony forming unit in terms of log removal. The log removal obtained by the log 10 of divided non-exposed bacteria to PL exposed bacteria. The reduction of colony forming unit of cultivation and enumeration for PL1 and PL2 exposed bacteria was in agreement with those obtained by other studies. PL has been reported to result in a 0.5 to 8 log 10 2-5 reduction of colony forming unit. The differences between the PL1 and PL2 lamps in reduction of colony forming unit related to the ozone diffusion in the PL1 by oxidation of the lipid bi-layer of bacterial membrane 15 . was the non-exposed bacteria, hence no colony unite reduction occurred. Error bars indicate the standard deviation of triplicate samples.

Batch regrowth
Bacterial regrowth showed a longer lag phase and slower growth rate after exposure to PL1 compared with PL2 (Fig. 16). With growth rate for PL1 and PL2 at 0.1 h -1 and 0.13 h -1 , respectively, versus the no-exposed case at 0.21. However, both a short lag phase and fast growth rate were observed for non-exposed bacteria regrowth. With lag phase for PL1 and PL2 at ~10 hours and ~8 hours compared to non-exposed at ~2 hours. Previous studies describe photoreactivation as an enzymatic DNA-repair mechanism in different microbial species and its occurrence after PL exposure 7,11 .
Farrell et al. assessed that 4 h of sunlight illumination of Candida glabrata exposed with various fluencies caused a higher recovery up to 1 log 10 compared to dark stored samples 11 . Similarly, Kramer et al. 7 reported recovery was increased by up to 2 log 10 exactly after E. coli or L. innocua were exposed to artificial daylight directly. Also, a possible explanation for the differences between culturability of PL1 exposed and PL2 139 exposed bacteria is ozonation 29 . Ozonzation is the process that ozone combined with UV in the chamber and increases the PL1 disinfection performance. conditions. Marks are: • controls (non-exposed bacteria), Δ regrowth of PL1 exposed bacteria, ■ regrowth of PL2 exposed bacteria, x blank which is LB media. Error bars are the standard deviation of triplicate samples. Fig. 17 shows a statistically significant PRR difference between the bacteria exposed to both PLs and non-exposed bacteria (p < 0.05). Based on previous report metabolic activity is vulnerable to PL than non-exposed cells due to damage to the DNA or other cellular structures like lipids and proteins 7 . The differences in metabolic PRR of the PL1 and PL2 is due to ozone presence in PL1 which first damages the cell membrane, leading to a low metabolic activity and then inhibition of bacterial growth 30 .

Metabolic activity
140 Fig. 17 Impacts of PL1 and PL2 on bacterial remaining respiration percent (RRP). Bacteria exposed to PL2 for 76 joule/cm 2 and bacteria exposed to PL1 for 95 joule/cm 2 . Control is nonexposed bacteria. Error bars are the standard deviation of triplicate samples.

Membrane permeability
The undisturbed cell membrane (UCM) results (Fig. 18 ) indicated that the effect of the PL2 on the membrane permeation of bacteria was lower than PL1. The statistical analysis confirmed the disparity of impacts of PL1 and PL2 on bacterial membrane disturbance (p = 6.81E-07). Our finding was in agreement with Krishnamurthy et al. 13 report. They demonstrated S. aureus cell membrane damage, cytoplasmic membrane shrinkage, and cellular content leakage after PL exposure 13 . Otherwise, Ferrario et al. 12 found that PL provoked rupture of the S. cerevisiae membrane allowing propidium iodide to penetrate cells as well as progressive loss of esterase activity.
However, the significant differences between the PL1 and PL2 effect on the UCM of 141 exposed bacteria explain by ozone ability to increase the permeability of the cell membrane and causing the efflux of intracellular substances 30 . Fig. 18 Impacts of PL1 and PL2 on bacterial membrane permeability that shows undisturbed membrane cell percent. Bacteria exposed to PL2 for 76 joule/cm 2 and bacteria exposed to PL1 for 95 joule/cm 2 . Control is non-exposed bacteria. Error bars are the standard deviation of triplicate samples.

Cell composition
ATR-FTIR spectra were used to compare the regrowth of PL1 exposed bacteria, PL2 exposed bacteria, and non-exposed bacteria demonstrated changes in the functional groups of bacteria. Comparison of the FTIR results for exposed bacteria and regrowth of exposed bacteria showed significant changes at wavenumbers: biomolecular groups: C=O Stretching, >CH deformation, C=O vibration, and C-O-C vibration for PL1 and PL2. We further investigated each of these wavenumbers in Table 5. Table 5. Comparison of altrations in biomolecular group between PL exposed and nonexposed bacteria using FTIR. The vibration of C-O-C in PL exposed bacteria is corresponding carbohydrate backbones and the C-O-C group of sugar derivatives. Hence, the possible explanation for shifting at this region is the possibility of trivial reducing sugar adsorbing 31 . These results assessed that unlike PL2, the negative impacts of PL1 on protein-like amphoteric polymers (N-H and C-N in amide II) and carbonyl groups (C= O) were not compensated by both PL exposed bacteria after regrowth. Also, HCA results in Fig. 19 shows that the carbohydrate region between regrowth of exposed bacteria to PL1 and exposed bacteria to PL1 segregated distinctly. However, HCA and FTIR composition results demonstrated the distinct segregation on the carbohydrates, protein and fatty 143 acids region between regrowth of PL2 exposed bacteria and PL2 exposed bacteria ( Fig. 20). Overall, a higher heterogeneity values in the dendogram were obtained between regrowth of exposed bacteria and exposed bacteria in nucleic acids, carbohydrates and fatty acids regions compared to proteins region. These results indicate that there is more similarity in the protein regions of the regrowth bacteria and exposed bacteria than other cellular components.

Fig. 19
Hieratical analysis of fatty acids, protein, carbohydrates, and nuclei acids between regrowth bacteria and exposed bacteria to PL1 for 76 joule/cm 2 . Control is non-exposed bacteria. Regrowth control is regrowth of non-exposed bacteria.
` Fig. 20 Hieratical analysis of fatty acids, protein, carbohydrates, and nuclei acids between regrowth bacteria and exposed bacteria to PL2 for 95 joule/cm 2 . Control is non-exposed bacteria. Regrowth control is regrowth of non-exposed bacteria.
To summarize the microbial disinfection results, the regrowth of PL1 exposed bacteria are still growing with low specific growth rate and high lag time. On the other hand, our observations corresponding to significant damage on cell membrane permeability of PL1 exposed bacteria (Fig. 18) and lowest heterogeneity of proteins exposed and control groups of bacteria in all conditions ( Fig. 19 and Fig. 20) may explain by the Garvey et al., finding which suggested no damage on cell membrane regarding the protein leakage in PL exposed cells 32 . The distinct segregation of carbohydrate region in PL1 exposed bacteria in compare with regrowth of PL1 exposed bacteria indicates that the bacterial respiration process (Fig. 17) had the most disturbance compared to other cellular mechanisms. In this content, the shifting of carbohydrate backbone and significant reduction of PRR might be related to reducing sugar adsorbing 31 .
Furthermore, the important similarity between PL1 and PL2 exposed bacteria was the heterogeneity of fatty acids of PL exposed bacteria and regrowth of exposed bacteria which was the same for both lamps. Ferrario et al., assessed that PL agitated rupture of the cytoplasm membrane and progressive loss of esterase activity 12 . These results are in agreement with previous study that cell structural properties such as cellular content leakage, cell membrane damage and cytoplasmic membrane shrinkage occurs after PL exposure 13 . In addition, the comparison between PL1 and PL2 exposed bacteria showed that PL1 has statically higher removal percent than PL2 (p < 0.05).
The effectiveness of PL1 in bacterial removal may be explained by ozone diffusivity.
In this content, Sharrer et al. 29 reported that bacteria tend to embed within particulate matter or that form bacterial aggregates that provides shielding from oxidation by UV.
They also, pointed out the high CFU reduction when UV irradiation was followed by ozonation (2.5-4.3 log 10 reduction) 29 . Thus, combining ozone with a UV irradiation which provided in PL1 consistently reduced bacteria counts to near zero. Also, the comparison of the impacts of these two lamps on cellular components showed that the FTIR composition analysis and HCA results from regrowth of PL1 exposed bacteria lost their similarity to non-exposed bacteria. The possible explanation for this phenomena might be reacting to residual activity and molecular responses 8,33 .
However, PL2 exposed bacteria showed ability to regrowth after exposure. Moreover, high heterogeneity between nucleic acids of PL2 exposed bacteria and its regrowth as well as similarity between regrowth of PL2 exposed bacteria and non-exposed bacteria point to the ability of PL2 exposed bacteria to recover itself.

PL2
PAHs rate of destruction tests have been conducted using the ozone generating UV lamp, using the same exposure fluences as for the bacteria (0, 19,38,57,76,95,190 joule/cm 2 ) ( Figure S2 in SI). The results of these experiments are displayed in Fig. 21.
Results were in agreement with previous studies in terms that high molecular weight PAHs (pyrene) are degradable faster than low molecular weight PAHs (fluorene, naphthalene and anthracene) 17 . Overall, all of the PAHs are removed completely after exposure to 57 joule/cm 2 using the PL1. The pathway of PAHs oxidation in the both PL system can develop according to the structure of individual PAHs.

Removal kinetic order of PAHs concentration after exposure to PL1 and PL2
During PAHs exposure to PL1 and PL2 for 0, 19, 38, 57, 76, 95, 190 joule/cm 2 , the decay kinetic orders were investigated by finding the closest linear regression for their degradation constant (Table 6 and Figure S2 in SI). Fluorene and naphthalene was positively assigned to first-order decay rate constants (k) for both PL1 and PL2, but, they were negatively correlated with the normalized first-order decay rate constants of pyrene and anthracene. However, regardless of exposure to PL, the removal rates of pyrene and anthracene were better explained with estimation of second order kinetics. Also, no significant differences observed between the degradation rate of the PL1 and PL2. These results oppose Ledakowicz et al. 3 study which compared efficiency of three systems (UV, UV with O 3 ) for PAHs 148 degradation. The authors recommend UV system in combination with O 3 as the most effective in PAHs degradation processes 3 .

Toxicity of PAHs byproducts on E. coli in term of respiration
The respiration test of the exposed PAHs on E. coli showed that the byproduct compounds which were released by the PAHs as a result of exposure to PL1 and PL2 didn't show any toxicity effect on bacteria. Also, the highest bacterial RRP was observed when the high concentration of PL1 exposed PAHs was exposed to bacteria.
These results are not conclusive in terms of the toxicity effects of PAHs byproducts after exposure, since the scope of this study didn't include these analyses. (Table S2 in the SI).

Conclusion
In summary, PL1 and PL2 treatments have shown its effectiveness in inactivating E.
coli and degrading PAHs in this study. The results show the potential of PL1 for microbial disinfection and both PLs for the fast degradation of PAHs. However, further testing for daughter PAH products that may be released as byproducts from exposure of the parent PAHs to PL1 and PL2 is needed for optimizing this system.
With regard to PAHs, near total degradation was achieved for all four PAHs when working with the SteriPulse®-RS 4000 system and PL1 and PL2 lamps. Moreover, degradation kinetics reflected the molecular weight of the PAHs, with pyrene, a high molecular weight PAH, degrading faster than the other three low molecular weight PAHs. Dash lines and ■ marks represent fluorene exposed to PL1. Long dash dot lines and Δ marks represent naphthalene exposed to PL1. Solis line and ◊ marks represent fluorene exposed to PL1. b) Second order kinetic orders of the pyrene and anthracene during exposure to PL1 and PL2. Round dot lines and • marks represent anthracene exposed to PL2. Dash lines and ■ marks represent pyrene exposed to PL1. Long dash dot lines and Δ marks represent anthracene exposed to PL1. Solis line and ◊ marks represent pyrene exposed to PL1. 157

IV -CONCLUSIONS
We investigated the antimicrobial agent's inhibitory effects in terms of kinetic, metabolic, and cellular component response of bacteria in three studies. In first part of this study, based on the literature review of comparing FTIR with four other tools (Raman, NMR,XPS, mass spectrometry), we selected the more appropriate characterization tool to screen the inhibitory effect of nanoparticles on bacteria. We found FTIR as an appropriate tool to study toxic effects on cellular components of bacteria. FTIR was able to differentiate between exposed and non-exposed bacterial components through change of spectra due to changes of functional groups in biomolecules. In the second part of this study, we applied this tool along with kinetic and metabolic tests for several experimental conditions and found that the inhibitory effects of nanoparticles in a continuous culture depend on the bacterial specific growth rate. At different growth conditions bacteria released different a concentration and composition of extracellular substance (ES). Bacteria at lower growth rate produce ES that are more effective reducing the inhibitory effect of the nanoparticles by changing the physicochemical properties of their surfaces. In the third part of this study, we demonstrated that the recovery of exposed bacteria depends on the range of ultraviolet wavelength applied by comparing the results of kinetic, metabolic and cell compositions of exposed bacteria when lamps with different cut off wavelength were used. At lower wavelength, we observed permanent damage to cellular composition which might be explained by generation of ozone at low ultraviolet wavelength and due to the high adsorption of this wavelength on DNA molecules.
Although experimental conditions in these three studies may diverge from each other (continuous culture vs. batch culture), our results from these three experiments allows us to assess the antimicrobial agent's effects on kinetic, metabolic and cellular composition of bacteria.
The main impacts of this study, is not only on the current protocols used to assess the toxicological response of nanoparticle on microorganism but also determine unintended impacts such as antibiotic resistance, which can have a high impact to public health. Since resistant bacteria not only can compromise the efficacy of antimicrobial agents but also could have implications on public health in terms of metal-antibiotic co-resistance. Hence, if resistant bacteria are released to the environment, it will become a threat to public health.
Overall in this study, we observed that the antimicrobial agents can cause permanent compositional changes of bacteria components. However, the inhibitory effects of antimicrobial agents on bacteria depend on the number of generations that the bacteria culture is exposed. For this purpose, future research should include detection of permanent gene mutations in exposed bacteria by RNA sequencing. The extracted RNA will be sequenced by next generation sequencing (RNAseq). Once the sequencing data has been obtained the sequences will be analyzed using currently by Breseq 0.24rc6 pipeline. Breseq pipeline uses three types of evidence to predict mutations, read alignments (RA), missing coverage (MC), and new junctions (JC).
Comparing the sequence of RNA between exposed and non-exposed bacteria at different bacterial generations will assist the scientific community to better understand the number of generation of bacteria that corresponds to gene mutations. Additionally, 161 investigating bacterial resistance to antimicrobial agents for a larger number of generations of the bacteria culture could provide complementary results to those obtained on this dissertation. New findings in that subject can further support decision making and develop better policies regarding use of antimicrobial agents and their potential negative effects in the environment and public health.