Nonleaching Antibacterial Concept Demonstrated by In Situ Construction of 2D Nanoflakes on Magnesium

Abstract In bone implants, antibacterial biomaterials with nonleaching surfaces are superior to ones based on abrupt release because systemic side effects arising from the latter can be avoided. In this work, a nonleaching antibacterial concept is demonstrated by fabricating 2D nanoflakes in situ on magnesium (Mg). Different from the conventional antibacterial mechanisms that depend on Mg2+ release and pH increase, the nanoflakes exert mechanical tension onto the bacteria membranes to destroy microorganisms on contact and produce intracellular stress via physical interactions, which is also revealed by computational simulations. Moreover, the nanoflake layer decelerates the corrosion process resulting in mitigated Mg2+ release, weaker alkalinity in the vicinity, and less hydrogen evolution, in turn inducing less inflammatory reactions and ensuring the biocompatibility as confirmed by the in vivo study. In this way, bacteria are killed by a mechanical process causing very little side effects. This work provides information and insights pertaining to the design of multifunctional biomaterials.

. Morphology of the samples determined by SEM (Scale bar = 1 μm).

Experimental Section
Hydrothermal treatment of Mg. The pure Mg samples (99.8%) were cut into plates with dimensions of 10 mm × 10 mm × 5 mm, ground with SiC abrasive paper (from 400 to 2000 grits) successively, and ultrasonically cleaned in alcohol. After drying in nitrogen, they were put into a 25 mL Teflon-lined autoclave containing 10 mL of NaOH (pH = 12). The autoclave was sealed in a metal shell and treated hydrothermally in an oven at 120 ˚C at a heating rate of 10 ˚C min -1 for 4 h, 8 h, and 12 h (samples designated as HT4, HT8, and HT12, respectively). After the reaction, the autoclave was cooled in air and the samples were ultrasonically cleaned with deionized (DI) water, dried in nitrogen, and stored in vacuum until further use. Electrochemical tests. The corrosion tests were carried out in the LB medium at 37 ˚C on an electrochemical workstation (Zennium, Zahner, Germany) with a 3-electrode configuration.
The sample with an exposed area of 1 cm 2 , platinum rod, and saturated calomel electrode (SCE) were the working electrode, counter electrode, and reference electrode, respectively.
The open circuit potential (OCP) was first monitored as reference prior to the electrochemical test for about 300 s. EIS was conducted at the OCP by applying an alternating potential of 10 mV between 100 kHz and 1 Hz. Afterwards, the samples were dried in nitrogen at room temperature. The Tafel curves were acquired at a scanning rate of 10 mV s -1 from -2 V to -1 V with respect to SCE and the degradation rate was determined from the corrosion current (I corr ) obtained from the polarization curves by extrapolating the cathodic branch of the polarization curve to the corrosion potential. The data were analyzed and simulated with ZSimpWin and the measurements were carried out at least three times to improve the statistics.
Immersion experiments. The corrosion behavior was evaluated by examining the surfaces and physicochemical properties following immersion. The samples were sealed with silicone rubber with one side (10 mm × 10 mm) exposed and immersed in 1 mL or 0.1 mL of LB on a 24-well plate. The samples were also immersed in the LB medium with a pH of 6 to assess the stability. After 6 h, the samples were taken out, washed with alcohol, dried in nitrogen, and examined by SEM. In the meantime, the leachate solution was collected for the subsequent tests of pH and ORP using a pH meter (STMICRO5, Ohaus, USA) and ORP meter (STORP1, Ohaus, USA). The concentration of Mg 2+ in the immersion liquid was determined by inductively-coupled plasma mass spectrometry (ICP-MS, Agilent 7500, Agilent, USA) and the sample weight was recorded before corrosion. After immersion for 6 and 24 h, the corrosion products formed on the surface were treated with chromic acid (CrO 3 , 200 g L -1 ; AgNO 3 , 10 g L -1 ) for 5 min, rinsed with distilled water and alcohol, and dried in nitrogen. The weight was measured again to calculate the weight loss and degradation rate.
The hydrogen peroxide assay kit (ab102500, abcam, UK) was also used to evaluate the behavior of the samples in H 2 O 2 .

All-atom simulation of Mg(OH) 2 (001) with a bilayer.
In addition, all-atom molecular dynamics (MD) simulation was performed to investigate the interactions between the hydrothermal coating and solution as described. The trigonal lattice of Mg(OH) 2 was adopted as brucite (centrosymmetric space group ) under ambient conditions. The lattice constants of Mg(OH) 2 were , , , and and the supercell had 14 × 14 × 3 crystallographic unit cells corresponding to 588 Mg atoms and 1176 OH groups [1] . The initial coordinates were generated from the in-house MATLAB codes. The fully hydroxylated surface did not have charge redistribution about the OH group on the surface and was aligned parallel to the x-y plane. Above this solid substrate surface was a bilayer 3 higher along the z-axis and the bilayer was about 50 thick. The simulation box dimensions were 4.25 nm × 4.00 nm × 8.00 nm and the bilayer and solid substrate were placed in an aqueous environment with sodium and chloride ions. The details are listed in Table 1. We chose the modified ClayFF force field [2] to parameterize Mg(OH) 2 [3] which had been successfully used for many systems of oxides, hydroxides, and clay minerals [4] . The ClayFF force field considered the entire interactions within and describes the structure completely by the nonbonded Lennard-Jones and Coulomb potentials except for the hydroxyl groups in the clay layer [5] .
Regarding the bilayer membrane, the CHARMM36 force field [6] was chosen to model POPC and sodium and chloride ions when interacting with the Mg(OH) 2 layers and this force field was also compatible with ClayFF [7] . The simple point charge (SPC) potential was used for water molecules and consistent with both the CHARMM36 and ClayFF force field. The Lorentz-Berthelot mixing rules were applied to both the CHARMM36 and ClayFF force fields for van der Waals interactions in this study.
Molecular dynamics (MD) simulation was performed using GROMACS version 2016.4 [8] with the electrostatic and van der Waals cutoff distance set to 1.2 nm. The periodic boundary conditions were applied to all three dimensions and the similar MD simulation pipeline was adopted through energy minimization using a steepest descents algorithm with convergence obtained once the maximum force on any one atom is less than 100 kJ mol -1 nm -1 .
Isothermal-isochoric (NVT) and isothermal-isobaric (NPT) equilibration at 310 .  were cultured in the LB medium for 12 h in an incubator (shaking rate = 220 rpm and temperature = 37 ˚C) to the exponential growth state. The bacterial solution was doubly diluted with the fresh medium and cultivated for another 3 h for reactivation to achieve OD 600 = 0.3 and 0.6, respectively. Afterwards, the bacteria solution was diluted to a concentration of 2-3×10 5 CFU mL -1 for the following tests and 100 μL of each solution was added to the surface of the samples. The antibacterial test was carried out on 24-well plates with the gaps between the wells filled with autoclaved water to avoid evaporation of the medium. At time points of 1, 3, 6, and 18 h, the adhered bacteria were detached from the surface with 900 μL of the PBS, diluted to the proper concentration, spread on a solid agar plate, and cultivated for another 16 h to count the CFU. The antibacterial rate was determined by the following formula: Antibacterial rate = ( ) ×100%. The antibacterial effects of the samples in the 1 mL and 2 mL system were determined with the same number of bacteria and more details can be found from our previous paper [9] .
In (v/v) alcohol sequentially. The specimens were dried in vacuum at 37 ˚C, mounted on the specimen stage, and sputter-deposited with a 10 nm thick Au layer before SEM examination.
The bacteria images were taken and the membrane area was calculated.

Coarse Grain (CG) Model for E. coli outer membrane. Calculation was performed to
determine the ratio of the change of the membrane area for different surface tensions, which mimicked the external surface tension when the membrane was in contact with the 2D nanoflakes. In E. coli which is gram-negative, the cell envelop contains a lipopolysaccharide (LPS) rich outer membrane [10] and LPS acting as a selective barrier is crucial to innate immunity in diverse eukaryotic species. Lipid A is a disaccharide-bound lipophilic domain and considered the primary immunostimulatory center of the LPS [11] . In this work, we utilized a coarse grain MARTINI model of E. coli outer membrane composed of lipid A in the outer leaflet and 1,2-dipalmitoyl-3-phosphatidyl-ethanolamine (DPPE) in the inner leaflet to investigate the impact of different surface tensions on the E. coli outer membrane [12] .
The building blocks of E. coli lipid A were coded in the freely distributed script and the outer leaflet of the membrane is pure lipid A and the inner leaflet was DPPE. The 10.3 nm × 10.3 nm membrane was solvated with standard water (W) as specified in Table 2 for faster and less expensive simulation. Ten percent of the CG water beads were of the anti-freeze type and the simulation box dimensions were 10.3 nm × 10.3 nm × 9.69 nm. The system was charge neutral by adding Na + counterions and periodic boundary conditions were applied to all three dimensions.
Energy minimization was performed using the steepest-decent algorithm [8a] with a 20 fs time until the maximum force on any bead was below the tolerance parameter of 100 kJ mol - After the 10 µs run without surface tension, the system was subjected to 14 different surface tensions by using surface-tension pressure coupling and taking reference pressure as 5 bar, 100 bar, 200 bar, 300 bar, 400 bar, 500 bar, 600 bar, 800 bar, 1,000 bar, 1,200 bar, 1,400 bar, 1,600 bar, 1,800 bar, and 2,000 bar (0. 2, 4.2, 8.3, 12.2, 16.2, 20.0, 23.6, 29.5, 34.2, 37.8, 40.5, 42.5, 43.9 and 44.9 in dynes·cm -1 , respectively) for coupling in the membrane. These surface tension simulations were performed for 1 µs with a time step of 10 fs. Inner structure of the bacteria examined by TEM. After incubation for 3 h, the specimens were treated ultrasonically for 5 min in PBS to dislodge bacteria from the sample surface.
The solution was centrifuged for 5 min (4,000 × g) to collect the bacteria from the bottom.
The bacteria were fixed successively with 2.5% glutaraldehyde and 1% OsO 4 at room temperature overnight. After washing with PBS and dehydration with alcohol and acetone with gradient concentrations, the samples were embedded in Spurr's resin (Spurr Embedding Kit, Spurr, USA) before slicing into sections (<100 nm thick) with a glass knife and staining with uranylacetate. The stained samples were placed on a copper wire mesh and examined by TEM (TecnaiG 2 12 BioTWIN, FEI company, USA) at 120 kV.

Membrane integrity evaluation.
The integrity of the bacteria membrane was evaluated by detecting the membrane potential or comparing the leakage of intracellular compounds. The membrane potential of the bacteria was measured by a membrane potential kit (B34950, Invitrogen, USA) and the bacteria treated with CCCP served as the positive control. The bacteria depolarization level was calculated as the red/green fluorescence ratio and more information about the procedures can be found elsewhere [13] . To determine the concentration of the leaked compounds, the samples with bacteria were ultrasonically treated in PBS and centrifuged to acquire the supernatant, and the concentration of the extracellular protein in the suspension was determined by the BCA protein assay kit (Sigma, USA). The concentration of released DNA/RNA was quantitatively measured by detecting the absorbance of the bacteria solution at 260 nm on a NanoDrop spectrophotometer (ND-1000, Thermo Fisher Scientific, USA).
Antibiofilm evaluation. The antibiofilm effects were qualitatively evaluated by CLSM (Leica SPE, Germany) as well as SEM and quantitatively examined by crystal violet staining [14] . The bacteria with an initial concentration of 2-3 × 10 6 CFU mL -1 were cultured on the samples for 48 h. In the quantitative analysis, the specimens were gently rinsed in PBS, stained with 0.1% crystal violet for 20 min, and rinsed in a deionized water bath. The bound crystal violet was eluted with 1 mL of 100% alcohol and then the optical density was determined on a multimode reader (EON, BioTek, USA) at 590 nm [15] . After culturing for 4 and 24 h, the cells were rinsed twice with PBS and 500 μL of the live/dead staining solution were added to each well. After incubation at 37 ˚C with protection from light for 30 min, the samples were observed under a fluorescence microscope (Leica TCS SP5 Matrix, Leica Microsystems, Germany) and more details about the procedures can be found elsewhere [16] . Biofouling resistance assessment. The whole blood obtained from the rats were centrifugated at 1500 × g for 15 min to collect the platelet-rich plasma (PRP). 150 μL of PRP were added to the surface of each sample and the samples were kept in a standard cell culture incubator for 2 h before washing with PBS three times. Afterwards, the samples were fixed with paraformaldehyde (4%) for 10 min, rinsed with PBS three times, and stained with phalloidin-fluorescein isothiocyanate (Sigma, USA) for 30 min before observation by fluorescent microscopy. The images were captured and fouling resistance effect was quantitatively evaluated by counting the adhered platelets in the specific region.
In vivo assessment of the antibacterial performance, anti-inflammatory effects, and biocompatibility. Sprague Dawley rats (200 g, Female) were employed and the in vivo tests were approved by the Ethics Committee for Animal Research, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. The rats were housed for one week to acclimate to the new environment. They were divided into 4 groups assigned as Normal, Control, Mg, and HT12, respectively. They were anesthetized with pentobarbital sodium (45 mg kg -1 ) via intraperitoneal injection before the hair was shaved from a 3 cm × 5 cm area and sterilized with povidone iodine. After incising the skin layer-by-layer parallel to the spine, the samples (5 mm × 5 mm × 2 mm) were implanted in the subcutaneous soft issue. The skin incisions were sutured before S. aureus and E. coli in 100 μL of PBS (10 7 CFU mL -1 ) were injected around the implant. The inflammatory response was examined on a daily basis.
After 3, 7, 10, and 14 days, the rats were euthanized. The implants were collected, put in PBS, and shaken for 2 min on a vortex shaker to count the implant-related CFU. Meanwhile, the surrounding soft tissues were immersed in PBS and homogenized (Scientz-IID, Ningbo, Zhejiang, China) for CFU counting. In the histological observation, other parts of the soft issues were fixed with 10% buffered formalin, washed with PBS, dehydrated in gradient alcohol, embedded in the paraffin, and sectioned. The sections were deparaffinized and stained with H&E before observation by optical microscopy. In the toxicity evaluation, the major organs such as heart, liver, kidney, lung, and spleen were harvested at the time of explantation and prepared similar to the historical observation.
Statistical analysis. The data were analyzed by the Student t-test and presented as mean ± standard deviation (SD) with a difference of P < 0.05 being significant and P < 0.01 being highly significant.