Carbon dot incorporated multi-walled carbon nanotube coated filters for bacterial removal and inactivation

Multi-walled carbon nanotube (MWCNT) filters incorporated with carbon quantum dots (CDots) or single-walled carbon nanotubes (SWCNTs) were produced for bacteria removal from aqueous solutions and also for inactivating the captured bacteria. TMTP Millipore membranes were used as the base of these filters. The results showed that filters with higher MWCNT loading had higher bacterial removal efficiencies. Filters with a MWCNT loading of 4.5 mg were highly effective at removing bacteria from aqueous solution, resulting in a log reduction of 6.41, 6.41, and 5.41 of E. coli cell numbers in filtrates compared to MWCNT filters without coating, MWCNTs filters with 0.15 mg CDot coating, and MWCNTs filters with 0.15 mg SWCNT coating, respectively. Ionic strength played an important role in bacteria removal. A higher NaCl concentration resulted in higher bacteria removal efficiencies of the filters. Both CDot coatings and SWCNT coatings did not significantly affect the MWCNT filter effects (P > 0.05). The coatings, especially CDot coatings, significantly inhibited the activities of bacteria retained on the filter surfaces (P < 0.05). The inhibitory rates were 94.21% or 73.17% on the MWCNT filter surfaces coated with 0.2 mg CDots or SWCNTs, respectively. These results demonstrated that MWCNT filters with CDot coatings were highly effective to remove bacteria from water and to inhibit the activities of the captured bacteria on filter surfaces.


Introduction
Waterborne diseases caused by pathogenic microorganisms in contaminated water systems lead to 3.4 million human deaths each year. 1 The World Health Organization (WHO) stated that safe water supplies, hygienic sanitation, and good water management are fundamental to global health. To protect public health, the U. S. Environmental Protection Agency (EPA) has National Primary Drinking Water Regulations which set standards on the maximum contaminant level (MCL) of these pathogens. For example, as Escherichia coli O157:H7 is the most problematic pathogen, 2,3 the EPA regulations require routine sampling of drinking water for testing total coliforms and E. coli, and the MCL level is zero for total coliforms. 1 Various methods and technologies have been investigated to eliminate, inactivate, or directly remove pathogens from water. One of the most commonly used methods is to add antimicrobial chemicals into water. These chemicals include gaseous chlorine (Cl 2 ), liquid sodium hypochlorite, chloramine, chlorine dioxide, chloramines, hydrogen peroxide, bromine, etc.
Although these chemicals have shown high efficacies to inactivate microorganisms, they pose risks in environmental safety and public health, 4 causing cancers and adverse reproductive disorders by some of these chemicals and their by-products. In other cases, direct contacts with high concentrations of antimicrobial chemical reagents cause extensive damage to human tissues. 5 Because of the risks posed by these chemicals, the EPA initiated a rule in 1992 to evaluate the need for additional controls for microbial pathogens and disinfectants, with the goal to develop an approach that would reduce the level of exposure to disinfectants, but still keep the high efficacy for controlling microbial pathogens. 6,7 However, in many cases, low concentrations of the chemical reagents for lower health risks are insufficient for the required antimicrobial effect. 8 The use of lters without the need to add any chemicals to water is a considerably safer way to remove microorganisms in contaminated water. The traditional membrane microltration is effective in the removal of suspended solids, but its usefulness to bacteria removal is limited, showing only about 61% in the bacteria removal rate. 9 Recent studies have demonstrated that nanomaterials-modied lters could signicantly increase the efficiency in microorganism removal. 4 Specically, carbon nanotubes (CNTs) have been explored for water purication and desalination purposes to take advantage of the features such as fast water transport, large surface area, ease of functionalization, 10 and high adsorption capacity. Our previous study and those by others have demonstrated that lters modied by multi-walled CNTs (MWCNTs) could effectively remove bacteria from water samples. 4,11 On the other hand, single-walled CNTs (SWCNTs) are known to exhibit antimicrobial activities against bacteria and viruses, and are much more efficient than MWCNTs. 12,13 In order to not only capture or remove bacteria from water, but also inactivate the captured bacteria, the development of effective lters with antimicrobial function is highly valuable. In this regard, the incorporation of antimicrobial agents with CNTs in the modication of lters should be a feasible strategy. Filters coated with MWCNTs and SWCNTs or incorporation of other agents were reported to show some antimicrobial activities. 4,14 Carbon dots (CDots), another class of carbon-based nanomaterials, have recently been discovered as potent antimicrobial agents with visible-light activation. CDots are small carbon nanoparticles with surface passivation, each having a core-shell structure of a carbon nanoparticle core and a thin shell of so materials (organic or biological species), with a size prole of less than 10 nm in diameter. 15,16 There is now substantial experimental evidence suggesting that CDots are similar to conventional nanoscale semiconductors in photoexcited state properties and associated redox processes, which afford CDots to exhibit photocatalytic activities, with diverse catalytic reactions induced by photogenerated electrons (and corresponding holes) on the dot surface. 17 Our group has reported that the photoinduced redox processes in CDots are responsible for their strong photo-activated antibacterial activities, 18 and that the antibacterial effectiveness is correlated with uorescence quantum yields of CDots. 19 It has also been demonstrated that the optical and redox properties of CDots can be altered or improved substantially by various surface passivation and/or functionalization schemes. [20][21][22][23] The goal of this study was to combine the excellent adsorption characteristics of MWCNTs with the antimicrobial activities of CDots for lters with the dual function of capturing or removing bacteria from water and inactivating the captured bacteria on the lters. Commercially acquired Millipore membranes were used as the base lters. MWCNTs were used to coat the membranes, and the coated lters were graed with antimicrobial CDots. For comparison, lters co-coated with both MWCNTs and SWCNTs (for additional antimicrobial activity) were also evaluated. The loading amount of MWCNTs, SWCNTs, and CDots, and the effects of lter's operation conditions were tested in terms of the bacteria removal and inactivation efficiencies.

Materials and methods
Preparation of MWCNT lters, MWCNT-SWCNT lters, and MWCNT-CDots lters MWCNTs were purchased from NanoIntegris Inc. (Skokie, IL, USA) and used as received without further purication (purity > 95 wt%). According to the description from the manufacturer, the tube outer diameter, inner diameter, and length were 10-20 nm, 3-5 nm, and 10-30 mm, respectively. The specic surface area was 233 m 2 g À1 .
The preparation procedures of MWCNTs coating on lters were similar to those described in our previous publication. 4 Briey, isopore polycarbonate hydrophilic membrane lters (TMTP membranes) were purchased from EMD Millipore (Billerica, MA) with a diameter of 25 mm and a pore size of 5 mm. MWCNTs solutions at the concentration of 3 mg mL À1 were prepared by suspending MWCNTs in 50% dimethylsulfoxide (DMSO), followed by sonication for 10 min. The MWCNTs lters were produced by depositing MWCNTs solutions onto TMTP membranes with desired loadings. The lters were air dried for 4 h. DMSO residues on the lters were removed by ltering 2.5 mL of 100% ethanol, and then the ethanol residues were removed by ltering 5 mL deionized water (DI-H 2 O). The ow rates on both ltering steps were controlled at 0.5 mL min À1 using a syringe pump.
To prepare SWCNTs-coated MWCNTs lters, SWCNTs solutions (1 mg mL À1 ) with functional group -OH on SWCNT surfaces were purchased from Nanolab. Inc. (Newton, MA). The SWCNT lengths were 1-5 mm. The manufacturer synthesized SWCNTs using a chemical vapor deposition process with high yield and purity, containing little or no amorphous carbon. These SWCNTs contained 95.93% weight percentage of carbon and 4.07% of other elements (Na, Al, Si, S, and Fe). SWCNT-MWCNT lters were produced by depositing the SWCNTs solution with desired volumes onto the pre-made MWCNTs lters. The MWCNT-SWCNT lters were air dried for 1 h and then rinsed with 2.5 mL DI-H 2 O by syringe ltering to remove unattached SWCNTs.
To prepare CDots-coated MWCNTs lters, CDots with 2,2 0 -(ethylenedioxy)bis(ethylamine) (EDA) as the surface functional molecule were synthesized using the same procedure described in our previous publication. 18 Briey, carbon nanopowder sample (1 g) was reuxed in an aqueous nitric acid solution (5 M, 90 mL) for 48 h. The reaction mixture was cooled to room temperature, followed by dialyzing against water for 3 days. The desired carbon nanoparticles were obtained by removing the water from the post-dialysis mixture through centrifugation at 1000g. To synthesize the EDA-CDots, the carbon nanoparticles were reuxed in neat thionyl chloride for 12 h, and then the excess thionyl chloride was removed via purging with nitrogen. The post-treatment carbon nanoparticles (50 mg) were mixed carefully with dried EDA (500 mg) in a ask, heated to 120 C, and stirred vigorously under nitrogen for 3 days. The reaction mixture was cooled to room temperature, dispersed in water, and then centrifuged at 20 000 Â g to collect the supernatant as the EDA-CDots in an aqueous solution. Free EDA and other impurities were removed from the solution by dialysis in membrane tubing (cutoff molecular weight $ 500) against fresh water. Detailed procedures and characterization of EDA-CDots were previously reported, size-wisely, EDA-CDots were 4-5 nm in average diameter. 24,25 Similar to the SWCNT-MWCNT lter preparation, the MWCNT-CDots lters were produced by depositing with desired volumes of EDA-CDots solutions at desired concentration onto the pre-made MWCNTs lters. The MWCNT-CDots lters were air dried for 1 h and then rinsed with 2.5 mL DI-H 2 O by syringe ltering to remove unattached CDots.
Bacteria cultures and ltration E. coli cells or B. subtilis cells were freshly grown in nutrient broth at 37 C overnight. The cells were harvested by centrifugation, washed twice with 0.85% NaCl solution, and then resuspended in 0.85% NaCl, except those stated specically in the results section, for further experimental uses.
To evaluate the bacteria capture efficiencies of the lters, 2 mL of cell suspension were ltered through the freshly prepared lters at the velocity of 0.5 mL min À1 using a syringe pump. The ltrates were collected and the bacterium numbers in ltrates were determined using the traditional surface plating method on Luria-Bertani (LB) agar plates. The bacterium numbers captured by the lters were calculated by using the cell numbers before ltration subtracting the cell numbers in ltrates aer ltration.
To determine the inactivation effect of the coated lters on the captured bacteria on the lter surfaces, each lter membrane was sit at room temperature for 30 min and then immersed in 2 mL PBS buffer in a centrifuge tube, followed by 5 s sonication at an ultrasonic water bath and vigorously vortexing until MWCNTs were detached from the membrane. The obtained suspensions were used to determine the viable cell numbers captured on the lter surfaces by the surface plating method. The inactivation efficiency to the captured cells were calculated as the following equation:

Imaging of captured cells on lter surfaces
Scanning electron microscopy (SEM) imaging was used to examining the morphologies of E. coli cells retained on MWCNTs lters, MWCNTs-CDots lters, and MWCNTs-SWCNTs lters. All the tested lters had 3 mg MWCNTs loadings with or without 0.15 mg CDots or SWCNTs coating on the surface. The captured cells on each lter were rst sat at room temperature for 1 h, then xed overnight by immersing into 1 mL of 4% formaldehyde and 2% glutaraldehyde solution in a 1.5 mL centrifuge tube at 4 C. The xative was removed and the lters were gently rinsed with 1 mL DI-H 2 O. The lters were then air dried and coated with gold using Denton Vacuum Desk IV (Czech Republic). The FEI XL30 microscope (Netherlands) was used to take SEM images at the Shared Materials and Instrumentation Facility (SMIF) in Duke University.

Statistical analyses
Statistical analyses were performed to compare the effects of the lters by the use of the general linear model (GLM) procedure of the SAS System 9.2 (SAS Institute Inc., Cary, NC, USA), with P < 0.05 being considered as signicant different.

Bacteria removal efficacies of MWCNTs and MWCNTs-CDots coated lters in different buffers
MWCNTs coated and MWCNTs-CDots coated lters were used to test the bacterial removal efficiency to E. coli cells in different buffers. Both types of the lters had 3 mg MWCNTs loadings, while the later had 0.15 mg CDots coating on the surfaces. The uncoated TMTP membranes without MWCNTs and CDots were used as the controls. E. coli cells were suspended in PBS, 0.85% NaCl, or LB broth. Aer the ltering process, the viable cell numbers in the ltrates were determined by the surface plating method. Fig. 1 shows the logarithmic value of E. coli cell numbers in the ltrates aer the ltration using respective lters. The controls were the uncoated TMTP membranes with 5 mm pore size which was much larger than E. coli cell size, so all the cells should pass through the TMTP membranes. As such, the controls showed no cell number decrease from the original cell suspensions ($2.5 Â 10 6 CFU mL À1 ) before ltration. The use of TMTP membranes as the base lters for MWCNTs coating allowed high ow uxes at low operating pressures while providing a sturdy support for the MWCNTs coating layer.
As shown in Fig. 1, in PBS, the ltrations using MWCNTs and MWCNTs-CDots coated lters both resulted in 1.24 log reduction in the cell numbers compared to the controls. In LB broth, the same log reductions (1.23 log) were achieved by both types of lters for ltering E. coli cells, whereas in 0.85% NaCl, the ltrations by both types of lters reached 2.24 log, approximately 1 log more than those achieved in the other two buffers.
It is noted that in each buffer, there was no signicant difference in the cell number reduction between using the MWCNTs coated lters and the MWCNTs-CDots coated lters (P > 0.05).
These results indicated that the buffer played a role in bacteria attachment on MWCNTs' and MWCNTs-CDots' surfaces on the coated lters. It is known that the nature of bacterial attachment to a surface might be determined by the properties of the bacterial cell and the surface properties of the material, as well as the surrounding liquid phase and its inuence on the substratum. 26 Material surface properties that affect bacteria attachment include hydrophobicity/ hydrophilicity, roughness, charge, functional group, etc. 27,28 Surfaces conditioned by the migration and adsorption of organic and inorganic molecules, also known as conditioning lm, could change their physicochemical properties such as hydrophobicity and surface charge. These property changes could be inuenced by the bulk liquid phase surrounding the material and affecting bacterial adhesion. 27,29 CNTs are highly effective for bacteria adsorption. Solution composition could affect the aggregation state of CNTs as well as bacterial retention in porous media. 30,31 It was reported that diffusion kinetics of bacterial cells in CNT solution was dependent on the concentration and average diameter of the CNT aggregates and also on the type of CNTs. 32 In the case of MWCNTs coating on the lters, it was possible different conditioning lms with different components were formed and changed the surface properties of the MWCNTs when ltering cells in different buffers, leading to the different extend of bacterial attachment on the MWCNTs or MWCNT-CDots coatings, thus resulting different bacterial removal efficacies when ltering cells in different buffers.

Inactivation efficiency ¼ total number of captured cells À the number of viable captured cells total number of captured cells Â 100%
Another possible factor that could contribute to the buffer effect on ltration efficiency was the ionic strength in different buffers. To test the ionic strength effects on bacterial removal efficiency by MWCNTs coated lters, E. coli cell resuspension in tap water with the addition of 0, 0.2, 0.4, or 0.6% NaCl were tested. The cells in these solutions were ltered through MWCNTs coated lters (3 mg MWCNTs loading). The ltration of the cell suspensions without NaCl resulted in the cell number reduction from 6.11 log to 4.79 log, indicating that MWCNTs coated lters could remove 1.32 log of E. coli cells, approximately removing 95.25% of E. coli cells (Fig. 2). Cell numbers in all the ltrates were also compared among all the ltrations of cell suspension in water with different NaCl concentrations, the log reductions of E. coli cell numbers in ltrates were signicantly increased (P < 0.05) in the cell suspensions with increasing concentrations of NaCl, as the log reductions of cell number were 1.47, 1.48, and 2.17 in cell suspensions with addition of 0.2, 0.4, and 0.6% NaCl, respectively. It suggested that the increased ionic strength in cell suspensions due to the increasing NaCl concentration improved the ltration efficiency.
This observation demonstrated that MWCNTs coated lters captured more bacteria cells at higher iconic strength, partially explaining the results above on the ltration efficiency in different buffers. Based on literature, bacterial adhesion to surfaces is oen explained by the principles of the DLVO theory in dilute NaCl solutions (0.0 to 1.17%), by which the interaction between the cells and the surface is described in terms of attractive van der Waals forces and electrostatic interactions. 33 Since CNTs are neutral or slightly negatively charged and E. coli cells are slightly negatively charged, 34,35 it was possible that higher ionic strength improved electrostatic attraction force between MWCNTs surfaces and the bacterial surfaces, leading to more cell captures on MWCNTs. Besides, the ionic strength of the surrounding medium might have a direct effect on the cells, thus modifying their rigidity, which in turn inuenced their damping effects, and changing the cell surface molecules that mediated the attachment. Similar observations were reported by other studies regarding bacteria-surface interactions, which showed that stronger ionic strength in the solution resulted in greater bacteria adhesion on metal surfaces, 28 quartz crystal surfaces, 33 and CNTs surfaces. 36 Yang et al. 36 studied the inuence of CNTs on the transport and retention behavior of E. coli cells in packed porous media at both low and high ionic strength in NaCl and CaCl 2 solutions. Their results demonstrated that CNTs increased cell retentions at high ionic strength (25 mM NaCl and 1.2 mM CaCl 2 ), whereas CNTs at low ionic strength (5 mM NaCl and 0.3 mM CaCl 2 ) did not affect the  Paper retention and transport of the cells. 36 Brady-Estevez et al. reported that MWCNTs lters produced by depositing MWCNTs on 5 mm pore size PTFE membranes could remove more MS2 bacteriophages by increasing ionic strength, from 5.06 log removal at 1 mM NaCl to greater than 6.56 log removal at 100 mM NaCl. 37 Our results in this study are consistent with these results reported in literature.

Effects of MWCNTs loading, CDots and SWCNTs loading on bacterial removal efficiency of the coated lters
To test the effect of MWCNTs loadings on bacteria removal efficacies of different lters, MWCNTs coated lters, MWCNTs-CDots coated lters, and MWCNTs-SWCNTs coated lters, each with three different MWCNTs loadings of 1.5 mg, 3.0 mg, and 4.5 mg were tested. The coating amount of CDots or SWCNTs on MWCNTs-CDots lters or MWCNT-SWCNTs lters was 0.15 mg. Fig. 3A shows the log reductions in cell numbers aer the ltrations using the three types of coated lters with different MWCNTs loadings. As shown in Fig. 3, at each level of MWCNTs loading, all three types of coated lters showed signicant bacterial removal compared to the controls, with the lters of higher MWCNTs loadings showing higher bacterial removal efficiency, despite there being no signicant difference between the lters with MWCNTs loading at 1.5 mg and 3.0 mg (P > 0.05). The lters with the highest (4.5 mg) MWCNTs loadings were the most effective and signicantly higher than those with lower MWCNTs loadings (P < 0.05) for E. coli cell removal, even reaching a complete removal by MWCNTs coated lters and MWCNTs-CDots coated lters. The average log reductions in cell numbers achieved by MWCNTs lters, MWCNTs-CDots lters, and MWCNTs-SWCNTs lters with 4.5 mg MWCNTs loading were 6.41, 6.41, and 5.41, respectively. These lters showed high bacterial removal capacities due to the high MWCNTs loading. The results are consistent with previous studies by achieving the similar level of bacterial removal. For example, MWCNTs/Trix buckypapers prepared on 5 mm pore sized PTFE membranes could remove >99% of E. coli cells in 0.9% NaCl solution. 38 Vecitis et al. 39 demonstrated that an anodic MWCNTs microlter was effective for a complete removal of bacteria. Wang et al. 40 produced lters with the coating of co-poly(propionylethyleneimine)-co-ethyleneimine (PPEI-EI) functionalized MWCNTs, and were able to capture higher than 4 log (up to 6 log) of bacterial cells. 40 Sharing the same characteristics of the outer walls with MWCNTs, SWCNTs lters were also observed to be able to retain E. coli cells completely on the lter surfaces. 14 However, at each level of MWCNTs loading, there was no difference in bacterial removal efficiency when the MWCNTs lters coated with or without CDots or SWCNTs, indicating that additional coating of CDots or SWCNTs did not contribute to the bacterial removal efficiency.
To further conrm the effects of CDots coating or SWCNTs coating on bacteria removal efficacies of the coated lters, lters with three different loading levels of CDots or SWCNTs (at 0.1, 0.15, or 0.2 mg) depositing on MWCNTs lter with 3 mg MWCNTs loadings were tested. Fig. 3B shows the log reductions in cell number aer the ltrations using the MWCNTs-CDots lters and MWCNTs-SWCNTs coated lters with the three levels of CDots and SWCNTs loadings. The results indicated that all the lters signicantly removed E. coli cells (P < 0.05) in comparison to the control samples, with the log reductions in cell number of $2.3 to $2.5 log. However, there was no statistically signicant differences (P > 0.05) among the MWCNTs-CDots lters with different CDots loading, and among the MWCNTs-SWCNTs lters with different SWCNTs loadings, conrming the observation from the test above and proving that CDots or SWCNTs coatings did not affect MWCNTs lters' efficacies on bacterial removal.
Although MWCNTs and SWCNTs have different diameters, they have the same seamless, cylinder structure of graphene layer on the surface. SWCNTs and MWCNTs had identical zeta potential values around À4 mV when dispersing in PBS buffer solution, and around À18 mV when dispersing in cell culture medium. 41 It suggested that the diameter of CNTs did not affect their charge distribution and the dispersion stability in PBS and cell culture medium. Besides these similarities between MWCNTs and SWCNTs, the amount of SWCNTs coating (0.1-0.2 mg) was not high enough to signicantly increase the adsorption capacity and change the capability of bacterial capture by MWCNTs lters in this study, therefore, no signicant difference on bacteria captures (P > 0.05) by MWCNTs lters and MWCNTs-SWCNTs lters was observed. Previous publications about bacterial adsorptions on SWCNTs and MWCNTs were controversial. For instance, Sweetman et al. 38 reported that MWCNTs buckypapers captured more E. coli cells than SWCNTs buckypapers, while Choudhury et al. demonstrated that SWCNTs were better candidates for adsorption on microorganisms than MWCNTs. 42 Similar controversies were also observed on biomolecules binding with CNTs at different diameters. For instance, Morikawa et al. 43 observed that MWCNTs with smaller diameters had more adsorptions of osteoblast-like cells than the ones with larger diameters; on the contrary, Mu et al. 44 demonstrated that MWCNTs with larger diameters ($40 nm) generally exhibited stronger protein binding compared to those with smaller diameters ($10 nm). These differences might be caused by different bacteria adsorption conditions, such as working on CNTs lters vs. in CNTs solutions, ltering cells on the CNTs lters produced with different procedures, using CNTs with different functional groups, performing in different medium solutions, etc.
As for CDots coatings, the average size of EDA-CDots was less than 5 nm in diameter with a carbon core of 3-4 nm in diameter in this study. CDots passivated by hydrocarbon chains on their surfaces exhibited high affinity to bacterial cells aer short incubation. 45 The bacteria-CDots binding proles were inuenced by bacterial strains, among which were involved with different lipid compositions, molecular organization, and macroscopic structure of bacterial surfaces. 45 However, when small amount (0.1-0.2 mg) of CDots were coated on MWCNTs lters, the coating did not affect the lters' capabilities in bacteria capture.

Effect of CDots and SWCNTs loading on bacterial inactivation
The CDots or SWCNTs coating on the MWCNTs layer was intended to afford the coated lters with antimicrobial activity which can be used to inactivate/inhibit the bacterial cells captured on the lters, as CDots and SWCNTs have both demonstrated strong antimicrobial activities against bacterial cells, and they have compatibility with MWCNTs. Fig. 4 shows the inactivation efficiencies of MWCNTs lters with three different levels of CDots and SWCNTs costing. Aer the cells were captured and retained on the lters for 30 min, the MWCNTs coated lters showed about 62.3% of bacteria inactivation, while the MWCNTs-CDots (0.2 mg) lters and MWCNTs-SWCNTs (0.2 mg) lters exhibited 94.2% and 73.2% inactivation efficiency, respectively, demonstrating that CDots or SWCNTs coating on the MWCNTs did afford the coated lters with antimicrobial activity. Between CDots and SWCNT coating, CDots was more effective on bacteria inactivation than SWCNTs coating, especially when both were at the same loading of 0.15 mg or 0.2 mg. Between the different levels of CDots coatings, higher CDots loading showed higher bacteria inactivation efficiency, as shown in Fig. 4, when CDots loading was increased from 0.1 mg to 0.15 mg, the bacteria inactivation percentage were signicantly increased (P < 0.05) from 68.6% to 94.0%. Further increase of CDots coating to 0.2 mg was slightly more effective than 0.15 mg coating, but not statistically difference (P > 0.05). The effect of SWCNTs coating showed a slight increase trend, with the inactivation percentage only increased from 70.0% to 73.2% as SWCNTs deposit was increased from 0.1 mg to 0.2 mg. When the MWCNTs loading was 4.5 mg on all these lters, no obvious difference was observed on the inactivation efficiency in comparison to its counterpart lters with 3 mg MWCNTs loading.
The results have clearly demonstrated that the coatings of CDots or SWCNTs on the MWCNTs lters can enhance the inactivation function of the resulting lters, although studies have reported that MWCNTs lters itself without other coatings could inactivate the bacteria retained on their surfaces, despite the inactivation efficiencies being not remarkable. For example, Park reported that 83.7% of the E. coli cells were inactivated on a glass ber air lter with MWCNTs deposition; 46 Dong and Yang 4 indicated that MWCNTs lters caused 18.9% inactivation of B. anthracis cells on the lter surfaces; Kang et al. 47 demonstrated that MWCNTs lters led to 70% loss of metabolic activity of E. coli cells. The inactivation came mostly from the direct physical contact between CNTs and bacterial cells, as CNTs could damage bacterial membranes, disrupt their activities, and eventually destroy their viabilities. 47 MWCNTs with smaller diameters generally display stronger toxicity to bacteria, while their toxicity was inuenced by many factors including bacteria types and the electrostatic repulsion between MWCNTs and bacteria. 48 SWCNTs appear to have greater antimicrobial effects than MWCNTs. The inactivation of E. coli cells attached to SWCNT aggregates in solution (80 AE 10%) was much higher than for the cells attached to MWCNTs (24 AE 4%). 47 Due to the higher antimicrobial effects of SWCNTs, the additional coating of SWCNTs onto MWCNTs lters afforded higher inactivation efficiency to the cells retained on the mixed coating. In the case of CDots coating, the inactivation efficiency was enhanced by CDots as CDots have been demonstrated for very strong photo-activated antimicrobial activity. 18,49,50 We rst reported CDots' photo-activated antimicrobial activity in 2016, in which EDA-CDots with visible light illumination was observed to inactivate $4 log of E. coli cells while the same concentration of EDA-CDots without light illumination only inactivated less than 1 log of cells. 18 A further study demonstrated the EDA-CDots' photo-activated antimicrobial activity is correlated to its optical property-uorescence quantum yield (F F ), with CDots with higher F F having higher photoactivated antimicrobial activity. 19 EDA-CDots' photo-activated antimicrobial activity can also be enhanced by combination with other chemicals 49 CDots with other surface passivation, such as polyethyleneimine (PEI), also exhibited photo-activated antimicrobial activity to bacterial cells. 51 As such, EDA-CDots was selected for coating on MWCNTs in this study. Compared to MWCNTs alone, the minimal inhibitory concentration (MIC) of CDots was 64 mg mL À1 on both E. coli and B. subtilis cells, 49 while the MIC value of MWCNTs was 500 mg mL À1 and 2000 mg mL À1 when MWCNTs were dispersed in unsaturated phospholipid and in polysorbates, respectively. 52 CDots' higher antimicrobial effects than MWCNTs, especially when the treatment was performed under light, afforded the higher bacterial inactivation efficiency on MWCNTs-CDots lters.
Morphologies of E. coli cells retained on different coated lters E. coli cells showed different morphological changes on different coated lter surfaces (Fig. 5). Aer 1 h retention on MWCNTs lters, the majority of the E. coli cells were intact and full in shape (Fig. 5A). This observation was similar to those in some of the previous studies, which demonstrated that there were no remarkable morphological changes in the cells retained on the MWCNTs lters. 4,47 However, some other studies reported that obvious cell membrane damages were observed on bacterial cells trapped on SWCNTs or MWCNTs lters. These different observations might be due to different contact time or different lter operation procedure. Fig. 5B presents the cells on the MWCNTs-CDots lters, showing that most cells were wrapped by MWCNTs, partially attened, and not as full as the ones on the MWCNTs lters. Dong et al. 49 observed that 10 mg mL À1 CDots treatment on E. coli cells in PBS caused cells aggregation, but did not change cells' morphology. The morphological changes of cells on MWCNTs-SWCNTs lters were in between the ones on MWCNTs and MWCNTs-CDots lters (Fig. 5C). These morphological changes correlated to the cells' inactivation efficiencies by the lters, among which the MWCNTs-CDots lters showed the highest inactivation efficiency.

Bacterial removal effect of the coated lters on Gram positive bacteria
The MWCNTs, MWCNTs-CDots, and MWCNTs-SWCNTs lters were also used to test their effects on Gram positive Bacillus subtilis bacteria removal. All the lters had 3 mg MWCNTs loading, while MWCNTs-CDots lters and MWCNTs-SWCNTs lters had 0.15 mg of CDots and SWCNTs coatings, respectively. Fig. 6 shows the log reduction of B. subtilis cells in the ltrate aer the ltration with the three types of coated lters. The results indicated that all three types of lters signicantly decreased B. subtilis cell numbers in ltrates (P < 0.05). B. subtilis cells were decreased from 6.19 log to 3.68 log by ltering through MWCNTs lters, with a reduction of 2.51 log in viable cell number. This observation was close to that in our previous publication, which indicated that MWCNTs lters with 1.5 mg of MWCNTs captured 2.44 log of B. anthracis cells. 4 CDots or SWCNTs coating did not evidently affect B. subtilis bacteria removal efficiency, similar to the observations on E. coli cells. Compared to ltering E. coli cells, the lters showed similar efficacies in removal of B. subtilis cells. This is consistent with other observations that bacterial adsorption to MWCNTs occurred spontaneously in solution, and MWCNTs' adsorption capacities were nearly the same regardless of the types of strains. 53

Conclusions
This study demonstrated that incorporation of antimicrobial materials in MWCNTs-coated lters afforded these lters with dual functions: removal of bacterial cells from aqueous solution and inactivation of cells retained on the coated lters. Bacterial removal efficiency was largely related to the loading level of MWCNTs on the base membranes, but was not affected by the loading level of CDots or SWCNTs; whereas the inactivation function of the coated lters was related to the loading level of CDots or SWCNTs as they dominated antimicrobial activity on the lters. MWCNTs lters with 4.5 mg MWCNTs loading achieved bacterial removal from water at 5.41-6.41 log reduction on E. coli cells or B. subtilis cells, which is considered highly effective. Additional coating of 0.2 mg CDots or SWCNTs on MWCNTs lters could inactivate more than 90% and more than 70% of the cells retained on the lter, respectively, affording the lters with considerable bacterial inactivation function. The lters developed in this study have the application potential to remove both Gram positive and Gram negative bacteria cells in aqueous solution. Such multifunctional features of the coated lters are benecial to many biological applications, including bacteria removal, isolation, concentration, water purication, detection, and decontamination of pathogens.

Conflicts of interest
There are no conicts of interest to declare.