Defining the Surface Oxygen Threshold That Switches the Interaction Mode of Graphene Oxide with Bacteria

As antimicrobials, graphene materials (GMs) may have advantages over traditional antibiotics due to their physical mechanisms of action which ensure less chance of development of microbial resistance. However, the fundamental question as to whether the antibacterial mechanism of GMs originates from parallel interaction or perpendicular interaction, or from a combination of these, remains poorly understood. Here, we show both experimentally and theoretically that GMs with high surface oxygen content (SOC) predominantly attach in parallel to the bacterial cell surface when in the suspension phase. The interaction mode shifts to perpendicular interaction when the SOC reaches a threshold of ∼0.3 (the atomic percent of O in the total atoms). Such distinct interaction modes are highly related to the rigidity of GMs. Graphene oxide (GO) with high SOC is very flexible and thus can wrap bacteria while reduced GO (rGO) with lower SOC has higher rigidity and tends to contact bacteria with their edges. Neither mode necessarily kills bacteria. Rather, bactericidal activity depends on the interaction of GMs with surrounding biomolecules. These findings suggest that variation of SOC of GMs is a key factor driving the interaction mode with bacteria, thus helping to understand the different possible physical mechanisms leading to their antibacterial activity.


S4
times to remove the suspended and unbound cells. Crystal violet solution (250 μL 0.1%) was added into each well and incubated at room temperature for 20 min, which allowed staining of the biofilm. The solutions were then discarded and the dye in the stained biofilm were released by adding 300 μL acetone/ethanol (1:4, v/v), and the absorbance at 540 nm (OD 540 ) was recorded under a microplate reader.
Readings from blanks were averaged out and subtracted from that with bacteria.
Biofilm formation was also determined using a LIVE/DEAD BacLight Bacterial Viability Kit.
Briefly, bacteria suspension (10 8 CFU/mL) were seeded in 35 mm confocal dishes containing GM suspensions (100 mg/L) in LB or M9 medium. After incubation for 48 h, the suspensions in the dishes were removed and the dishes were rinsed with PBS buffer (10 mM, pH 7.4) three times to remove the suspended and unbound cells. Then remaining biofilm were stained with the LIVE/DEAD BacLight Bacterial Viability Kit (Invitrogen Corp., Carlsbad, CA) according to the provided protocol. Images of the biofilm were taken on a confocal laser scanning microscopy (CLSM, Nikon A1, Tokyo, Japan).
For the plate-counting colony assay, E. coli bacterial cells (10 8 CFU/mL) were firstly incubated with GM suspension in M9 or LB medium for 3 h. At the end of the exposure period, the bacterial suspension was bath-sonicated for 10 min to break aggregates. Then, bacterial suspensions were immediately cultured on LB agar media and incubated overnight at 37 °C for colony forming unit (CFU) enumeration.
For G1 and G6 film exposure experiment, the E. coli bacterial suspension (10 8 CFU/mL) were exposed to films in M9 or LB medium for 3 h at room temperature. After discarding the excess bacterial suspension, the film was rinsed with 5 mL saline solution (PBS buffer, 10 mM, pH 7.4) to wash unattached cells from the film. The film was then transferred to 10 mL saline solution and sonicated for 10 min in an ultrasonic bath to detach bacteria from the film surface. The detached bacteria were subsequently cultured on solid LB agar and incubated overnight at 37 °C for CFU enumeration.

Examination of membrane integrity
Bacterial membrane integrity was measured by using a JC-1 assay kit (Beyotime Biotech, Nantong, China) as per the manufacturer's protocol. Briefly, bacteria suspensions (10 8 CFU/mL) were seeded in 35 mm confocal dishes containing 100 mg/L GM suspensions in LB or M9 medium, with each treatment in triplicates. After 48 h incubation, the bacterial cells were collected and washed with PBS thoroughly, and S5 then incubated with 10 μmol/L JC-1 dye for 15 min at 37 °C in the dark followed by washing with JC-1 staining buffer three times. The fluorescence was measured using a microplate reader (Infinite 200 Pro, Tecan, Switzerland) with the filters 490/20 (excitation) and 530/20 (emission) for green and 525/20 (excitation) and 590/35 (emission) for red. The ratio of red and green fluorescence was calculated. The biofilm was then imaged by confocal laser scanning microscopy (CLSM, Nikon A1, Tokyo, Japan) at 488 nm/530 nm excitation/emission wavelength.

Protein corona formation on GMs
100 µL of 1 mg/mL particles were incubated with 500 µL of LB medium for 3 h at 37 °C, shaking at 300 rpm. NMs and their associated corona were centrifuged for 10 min at 5000 g and the protein supernatant (soft corona) was removed. The particles were washed once with 250 µL of PBS buffer and then twice with 250 µL of ammonium bicarbonate buffer (ABC buffer) (100 mM, pH 8.0) (>99.5%, Thermo Scientific) by centrifuging for 10 min at 5000 g and removing the supernatant. NMs were transferred to a fresh vial after the second ABC wash.

Molecular dynamic simulations of interactions of GMs with lipid membrane
Molecular dynamics simulations of graphene and GM nanosheets interacting with a model E. coli outer membrane were performed in an aqueous environment. Due to the complexity of the membrane structure, we used palmitoyloleoylphosphatidylethanolamine (POPE) lipids, which is one of the most abundantly found in Gram-negative bacteria and has been used as a model membrane for studying nanomaterials interaction with bacteria 8 . Interactions of the membrane with G1, G4 and G6 as well as pristine graphene was simulated. For G1, G4 and G6, the carboxyl groups have been randomly attached to the carbon atoms of the edge of the nanosheet while the other two functional groups are attached at random to carbon atoms on the basal plane. The adopted construction strategy complies with the model of Lerf-Klinowski which describes outcomes from oxidation processes 9 .
The simulations have been carried out at atomistic resolution. A lipid membrane has been formed S7 by dissolving 340 POPE molecules in cubic box containing 31141 water molecules. The four nanosheets have the same dimensions of 4.92 nm × 1.99 nm. The Berger lipid force field has been adopted for the POPE molecules while water molecules have been represented by the SPC/E model. 10 The initial configuration of a single POPE molecule has been obtained from the ATB library. 11,12 The force field for pure graphene is given in by Gong et al., 13 while the OPLS-AA force field has been employed for the graphene oxide sheets. 14 All four nanosheets have been created using the GOPY tool. 15 Equilibration of the dissolved lipids in water has been performed for 30 ns in the isothermalisobaric (NPT) ensemble at 101.3 kPa and 300 K. The Nosé-Hoover thermostat and barostat was employed with coupling times of 0.1 and 1.0 ps. 16 A time step of 1 fs using the velocity-Verlet integration scheme has been used. A 1.1 nm atom-based cutoff for the summation of van der Waals interactions has been employed. The Coulomb interactions were computed using the Particle-Particle Particle-Mesh method. 17 Subsequently, the nanosheets were inserted either parallel or perpendicular to the surface of the lipid membrane at distances ranging from 1.5 nm to 0.1 nm with a step of 0.1 nm. Further equilibration for each system took place for 10 ns while keeping the distance between the nanosheets and the lipid membrane fixed. The simulations were performed with the LAMMPS code. 18 The potential of mean force (PMF) between the lipid membrane and a nanosheet with a given orientation has been determined by a series of constrained simulations. 19 The constraining force, , ( ) along the direction normal to the surface of the membrane at distance has been computed from the total forces exerted on the nanosheet, , and the membrane, : where and are the center of mass position vectors of the membrane and the nanosheet and 〈…〉 denotes time averaging. The PMF has been obtained by integrating the constraint force according to where is the reference PMF value and it has been set to zero at distances larger than 1.6 nm ( 0 ) between the lipid membrane and a nanosheet. In the production of the PMF profile, a series of constrained S8 simulations have been carried out with a decreasing step of 0.05 nm from 1.5 to 0.1 nm. Each simulation lasted for 4 ns. The first 3 ns constituted a pre-equilibrium stage, followed by a production run of 1 ns where the constrained force has been computed. The running average of the normalized force magnitude and potential energy exerted between the POPE membrane and the graphene nanosheet (size: 4.92 nm × 1.99 nm) at a distance of 0.3 nm between their axial center of mass coordinate were shown in the figure below. Both quantities are normalized by the average sampled between the 3.0 and 4.0 ns. It can be seen from these plots that both the force and the potential energy have reached their final value well before the 3 ns and that the system is well equilibrated.

Rigidity of GMs by computational modeling
Insights into the rigidity of the graphene and graphene oxide nanosheets have been gained by carrying out tensile loading simulations on the four systems of interest (G1, G4, G6 and graphene) in vacuum, using the loading scheme of Vijayaraghavan and Zhang. 20 Initially, the systems have been equilibrated at 300 K using the same settings as before for 0.1 ns. Subsequently, a cycle consisting of (i) displacement of the Subset of studies comparing the antibacterial effects of GO and rGO was shown in Table S1. Lack of SOC data renders the results from these studies difficult to compare.        (Table S2).         The ζ potential was higher for GMs with higher SOC in M9; the difference was diminished after 48 h.
The ζ potentials were similar for all GMs in LB.      OD 600 values in control group. OD g indicates OD 600 values after exposed to GMs suspension for 48 h.