Small-molecule-hosting nanocomposite films with multiple bacteria-triggered responses

We report pH/bacteria-responsive nanocomposite coatings with multiple mechanisms of antibacterial protection that include the permanent retention of antimicrobials, bacteria-triggered release of antibiotics and bacteria-induced film swelling. A novel small-molecule-hosting film was constructed using layer-by-layer deposition of montmorillonite (MMT) clay nanoplatelets and polyacrylic acid (PAA) components, both of which carry a negative charge at neutral pH. The films were highly swollen in water, and they exhibited major changes in swelling as a function of pH. Under physiologic conditions (pH 7.5, 0.2 M NaCl), hydrogel-like MMT/PAA films took up and sequestered ∼45% of the dry film matrix mass of the antibiotic gentamicin, causing dramatic film deswelling. Gentamicin remained sequestrated within the films for months under physiologic conditions and therefore did not contribute to the development of antibiotic resistance. When challenged with bacteria (Staphylococcus aureus, Staphylococcus epidermidis or Escherichia coli), the coatings released PAA-bound gentamicin because of bacteria-induced acidification of the immediate environment, whereas gentamicin adsorbed to MMT nanoplatelets remained bound within the coating, affording sustained antibacterial protection. Moreover, an increase in film swelling after gentamicin release further hindered bacterial adhesion. These multiple bacteria-triggered responses, together with nontoxicity to tissue cells, make these coatings promising candidates for protecting biomaterial implants and devices against bacterial colonization. A thin film coating for medical implants and other biomedical devices can be triggered to fight bacterial infection when required. Svetlana Sukhishvili of Stevens Institute of Technology in the United States and collaborators in Kazakhstan and the Netherlands developed the film, which is composed of montmorillonite clay nanoplatelets and polyacrylic acid. The film can take up and retain the antibiotic gentamycin at up to 45% of its dry mass. When bacteria adhere to the film, the antibiotic is released and the film swells to prevent bacterial penetration. Some antibiotic remains bound within the film, offering lasting antibacterial protection. The system is less likely to promote antibiotic resistance than routine antibiotic use as its effects are triggered only if a significant bacterial challenge occurs. It could also be adapted to carry a variety of other antibiotics. Antimicrobial-containing nanocomposite coatings can respond to bacterial challenge through multiple mechanisms and show unprecedented antibacterial efficacy. Coatings composed of similarly charged montmorillonite (MMT) clay nanoplatelets and polyacrylic acid (PAA) can take up and sequester large amounts of antibiotics under normal physiologic conditions, preventing the development of antibiotic resistance. When challenged with media-acidifying bacteria (Staphylococcus aureus, Staphylococcus epidermidis or Escherichia coli), they release PAA-bound gentamicin and increase their water uptake, while retaining MMT-bound antibiotics. These multiple bacteria-triggered responses, together with biocompatibility to tissue cells, make these coatings promising candidates for protecting biomaterial implants and devices against bacterial colonization.


INTRODUCTION
Polymer nanocomposites uniquely combine the properties of inorganic and organic components that is central to the development of novel advanced materials. The naturally occurring smectite clays, which are chemically and thermally stable and biocompatible, are attractive nanocomposites. Exfoliated clay nanoplatelets strongly impact the mechanical, transport, optical, fire retardant and gas barrier properties of nanocomposite materials. 1,2 For many biomedical applications, however, hydrophilic rather than hydrophobic nanocomposites are of great interest, and responsive properties are essential. 3 Reports on responsive hydrophilic nanocomposites are rare, and those few known to us advantageously use the complementary properties of temperature/pH-responsive polymers and clay nanosheets to yield environmentally controlled free-standing materials. 4 A promising way to leverage the favorable properties of hydrophilic-responsive nanocomposites on surfaces is to use the layer-by-layer (LbL) technique to construct 'smart' surface coatings. 5 To construct clay-containing LbL films, electrostatic interactions of positively charged polymers with negatively charged basal planes of silicate nanosheets, 6,7 neutral polymer/nanoplatelet hydrogen bonding 8 or a combination of the two have been explored. 9 Nanoplatelets of smectite family clays such as montmorillonite (MMT), which have a high aspect ratio and a high in-plane elastic modulus, can be included within LbL films at a much higher percentage than their clay-polymer bulk counterparts, resulting in ultrathin LbL assemblies with exceptional mechanical 8 and fire retardant/oxygen barrier properties. 10,11 However, the response properties of nanocomposite LbL constructs, which are central to the biomedical applications of these materials, have remained largely unexplored. We have recently reported on highly swollen, hydrogellike nanocomposite LbL films composed of neutral, temperatureresponsive polymer and MMT nanoplatelets. 12 Although these surface-bound hydrogels demonstrated large-amplitude reversible swelling/deswelling transitions as a function of temperature, these films were not suitable matrices to host small molecules for controlled release. Organoclays trapped within a polyelectrolyte capsule wall have been used to endow responsiveness to clay-containing LbL assemblies. 13 Here, we sought to explore assemblies of unmodified MMT platelets and a weak polycarboxylic acid that could be directly assembled within robust LbL films. These assemblies served as 'smart' matrices that could host and release small bioactive molecules in response to bacterial stimuli.
Our clay-polyacid surface coatings present a unique combination of properties, including (1) like charge of film components and extremely high degrees of film swelling, (2) pH control of film swelling and (3) the availability of polyacid and clay platelet-binding sites for pH-dependent and pH-independent antibiotic binding, respectively. We demonstrate that these nanocomposite coatings are ideal matrices for hosting functional molecules, including antibiotics with multiple functionalities to protect against bacterial colonization.
Bacterial colonization of surfaces, especially when such colonization progresses to biofilms, reduces the effectiveness of antibiotic therapies 14 and presents a serious problem in biomaterial implants and device surgery. To alleviate this problem, a number of approaches 15 have been proposed to modify surfaces with polymer coatings that are anti-adhesive, 16,17 kill bacteria on contact, 18 elute antimicrobial compounds with time 19 or provide antibacterial protection through enzymatic degradation of a biofilm matrix. 20 The assembly of surface coatings via LbL deposition enables the inclusion of a broad range of components within conformal coatings and has been explored to construct contact-killing 21 or release-killing films. 22 Owing to the complexity of bacterial and biofilm metabolism and the limitations of each individual antibacterial defense path, reliable antibacterial protection of surfaces remains elusive. Therefore, new multifunctional coatings are being explored that combine releasekilling and contact-killing capabilities. 23,24 Moreover, antibacterial surfaces have been designed that can respond to light or pH. 25,26 The present study goes beyond this prior work by describing a new type of coating that was engineered to combine permanent, contactkilling protection with bacteria-triggered, on-demand release of antimicrobials. By exploring clay-containing nanocomposites rather than all-polymer films as matrices to host an antibiotic, we were able to prevent low-level tail release of antibiotics that consequently minimizes the development of antibiotic resistance. Simultaneously, we combined antibiotic retention with a bacteria-triggered release mechanism, a concept recently introduced by our group, 26,27 such that antimicrobials remain sequestered within coatings until bacteria approach the surface. In our model, the secretion of lactic acid by staphylococci 28 or acetic acid by Escherichia coli 29 locally acidifies the environment to induce a pH-triggered release of antibiotics. Similarly, antibiotic-free 'self-defensive' coatings have been developed for use with a different trigger, that is, a bacteria-secreted enzyme. 30 Although we have previously demonstrated the concept of 'self-defense' antibacterial protection using a strongly charged, toxic peptide, a clinically relevant cationic antibiotic for local release, such as gentamicin, could not be sufficiently retained within all-polymer gel films and microgels in 0.2 M salt solutions at pH 7.5. 26 Here, by using a clay-polymer nanocomposite matrix, we were able to achieve a combination of strong antibiotic retention with bacteria-triggered release under physiologic conditions.

LbL film deposition
All chemicals were used as purchased without any further purification. Milli-Q system (Millipore, Billerica, MA, USA) filtered water with a resistivity of 18.2 MO was used in all experiments. Silicon (110) wafers were prime grade, p-type with boron dopant, 525 ± 25 mm thick, with a native oxide layer B2 nm thick and were purchased from Cemat Silicon SA (Warszawa, Poland). The wafers were cut with a Fletcher steel wheel glass cutter (Fletcher, East Berlin, CT, USA). The silicon wafers were precleaned under a quartz UV lamp for at least 2 h, soaked in concentrated sulfuric acid for 1 h and then carefully rinsed with Milli-Q water. The silicon wafers were then dried under a flow of nitrogen. To enhance the attachment of multilayers to the silica surface, a layer of branched poly(ethyleneimine) (M w 65 kDa, Sigma Aldrich, St Louis, MO, USA) was deposited as a precursor. Poly(ethyleneimine) was allowed to adsorb at the surface from 0.2 mg ml À1 polymer solutions in 0.01 M phosphate buffer (monobasic and dibasic sodium phosphate, Sigma Aldrich ) at pH 5. The poly(ethyleneimine) deposition step was followed by rinsing the surface with a buffer solution of the same pH value or by centrifuging and redispersing silica particles in a fresh 0.01 M phosphate buffer solution (PBS) at pH 5. For LbL deposition of particles, monodisperse silica particles with a diameter of 4.0 ± 0.2 mm were purchased from Polysciences Inc. (Warrington, PA, USA) as 10% dispersions in water. These dispersions were diluted with water to 0.1% before depositing the nanocomposite multilayers. Deposition of MMT/polyacrylic acid (PAA) multilayers on both flat and particulate substrates was performed using 0.2 mg ml À1 PAA (M w 450 kDa, Scientific Polymer Products, Inc. Ontario, NY, USA) and B0.4-0.5 mg ml À1 MMT (Cloisite Na þ from Southern Clay Products Inc., Gonzales, TX, USA) aqueous solutions at pH 2.2. A stock solution of MMT (5 g l À1 ) was prepared at least 2 weeks in advance. The MMT deposition solutions were prepared by diluting a stock MMT solution 10-fold with water, followed by sonication overnight. During multilayer deposition, PAA and MMT were allowed to adsorb for 6 min. For silica dispersions, the adsorption and rinsing steps were performed under continuous shaking. After each deposition step, three rinsing/centrifugation cycles were performed using water at pH 2.2. All films contained MMT as an outermost layer.

Incorporation of gentamicin within MMT/PAA films
To load an antibacterial agent, MMT/PAA films were exposed to gentamicin solutions (0.2 mg ml À1 in 0.2 M NaCl, 0.01 M phosphate buffer at pH 7.5) for at least 30 min to achieve complete absorption of gentamicin within the films.
The pH-triggered release of gentamicin from MMT/PAA films To study the pH-triggered release of gentamicin from MMT/PAA hydrogel films, the loaded nanocomposite matrices were exposed to 0.01 M phosphate buffer solutions containing 0.2 M NaCl with the pH set at various values between 3.5 and 7.5.

Colorimetric assay for gentamicin quantification
To generate a calibration curve, 0.5-ml aliquots of gentamicin solutions at concentrations ranging from 0.01 to 0.5 mg ml À1 were mixed with 0.15 ml 1.25% (w/v) aqueous ninhydrin solution. The reaction was allowed to proceed for 15 min in boiling water before absorption measurements at 400 nm.

Bacterial strains and growth conditions
Staphylococcus aureus ATCC 12600, Staphylococcus epidermidis ATCC 14990 and E. coli ATCC 25922 were purchased from the American Type Culture Collection (ATCC, Wesel, Germany). Each strain was taken from a frozen stock (stored at À80 1C), streaked on an agar plate and incubated overnight at 37 1C. E. coli was grown on an agar plate with brain-heart infusion broth (Oxoid, Basingstoke, UK), and S. aureus ATCC 12600 and S. epidermidis ATCC 14990 were grown on a tryptic soy broth (Oxoid) agar plates. After incubation, the plates were stored at 4 1C for up to 1 month. Precultures of bacteria were made by inoculating a single bacterial colony in 10 ml of brain-heart infusion broth for E. coli and tryptic soy broth for S. aureus and S. epidermidis and incubated at 37 1C in ambient air for 24 h. These precultures were sonicated at 30 W (Vibra Cell model 375, Sonics and Materials Inc., Newtown, CT, USA) on ice (3 times for 10 s each with 30 s between intervals), counted using a Bürker Türk counting chamber (VWR International BV, Amsterdam, The Netherlands) and then diluted with the appropriate growth medium to the desired concentration.

Flow chamber experiments
Bacterial precultures were used to inoculate 200 ml of the appropriate growth medium and grown for 16 h. The bacteria were harvested by centrifugation at 5000 Â g for 5 min at 10 1C and washed twice in sterile PBS. The bacteria were suspended in PBS at concentrations of 3 Â 10 8 and 3 Â 10 6 bacteria per ml. Before use, silicon substrates with deposited coatings were rinsed with 70% ethanol for 20 s and sterile PBS for 30 s and then dried in air under sterile conditions.
Silicon substrates affixed to the bottom plate of the parallel plate flow chamber 31 were exposed to continuous flow of bacterial suspension for 2 h at a flow rate of 1 ml min À1 . Subsequently, unattached bacteria were flushed out by rinsing with PBS buffer for 15 min, and adhering bacteria were stained with live/dead staining (BacLight, Molecular Probes Europe BV, Leiden, The Netherlands). The numbers of live and dead bacteria were determined by fluorescence microscopy (Leica Microsystems BV, Rijswijk, The Netherlands).
Osteoblast cell seeding on LbL films. Silicon wafers coated with (MMT/PAA) 9.5 films were disinfected with 70% ethanol for 1 h followed by three washes in sterile PBS. These samples were loaded with gentamicin for 2 h. As controls, wafers coated with gentamicin-free (MMT/PAA) 9.5 films and bare Si wafers were used. All wafers were placed in a 12-well plate, and a volume of 40 ml of HFOB cell suspension containing 2 Â 10 4 cells was pipetted onto each wafer. The cells were allowed to attach for 2 h, and then 1 ml of medium was added to each well. The cultures were maintained for 7 days with medium replacement every other day.
Quantification of osteoblast cell numbers on LbL films. Cell numbers were determined using an MTS assay kit (Promega, Madison, WI, USA) after culturing for 1 and 7 days. Briefly, samples were transferred to a fresh 12-well plate. A volume of 200 ml of the combined tetrazolium salt (MTS)/phenazine methosulfate solution (20:1) was pipetted into each well containing 1 ml of culture medium. The ratio 20:1 is the ratio of the volumes of MTS to phenazine methosulfate. The plate was then incubated for 2 h at 37 1C in a humidified atmosphere containing 5% CO 2 . Cell proliferation was expressed as the absorbance at 490 nm measured using a microplate reader (BioTek, Winooski, VT, USA). Each group was measured in triplicate. The cell number was determined using a standard calibration curve.
Osteoblast cell immunofluorescence staining. Fluorescence staining was performed to observe the formation of actin filaments after 1 and 7 days of cell culture. In brief, cells cultured on LbL nanocomposite coatings were fixed with 4% paraformaldehyde (Sigma-Aldrich) in PBS at 4 1C for 30 min. The cells were permeabilized for 5 min in 1% Triton X-100 (Sigma-Aldrich) in PBS, followed by blocking with 1% bovine serum albumin in PBS for 30 min. The cells were then treated for 30 min with fluorescein isothiocyanate-labeled phalloidin (1:100; Sigma-Aldrich) to stain actin fibers. Images were captured using a Zeiss LSM confocal microscope (North Chesterfield, VA, USA).
Live/dead osteoblast cell viability. Cell viability was quantified using a live/dead cell viability assay kit (Invitrogen) after 1 and 7 days. The cells were stained based on membrane integrity and intracellular esterase activity as per the manufacturer's protocols. Briefly, cell-seeded scaffolds were incubated at room temperature for 30 min in PBS containing 2 mM calcein AM (green) and 4 mM ethidium homodimer-1 (red). Cells stained green (live) and red (dead) were imaged using a Zeiss LSM confocal microscope.

Characterization methods
Zeta potential measurements were carried out for 0.05% silica particle suspensions using a Zetasizer Nano ZS (Malvern Instruments, Malvern, UK) at 25 1C.
Confocal laser scanning microscopy imaging of coatings and osteoblasts was performed using a Zeiss LSM 5 PASCAL confocal scanning system (Carl Zeiss MicroImaging Inc., Thornwood, NY, USA) equipped with a C-Apochromat 63 Â 1.2W corr water immersion objective. For film visualization, multilayer coatings were immersed into a very dilute Alexa 488 hydrazide sodium salt (Alexa 488, Molecular Probes, Inc., Grand Island, NY, USA) solution. For confocal laser scanning microscopy studies, the hydrogel films deposited at the surface of the Si wafers were placed vertically in a Lab-Tek chamber slide (Sigma-Aldrich). The chambers of the Lab-Tek slide were then sequentially filled with buffer solutions at a certain pH.
Scanning electron microscopy images were obtained using a Zeiss Auriga Dual-Beam FIB-SEM (North Chesterfield, VA, USA). Cross-sectional specimens were prepared by fracturing. Silicon wafers with sample films were fixed to scanning electron microscopy specimen holders with conductive tape. Gold-platinum alloy was sputtered onto sample surfaces with an RF-plasma chamber for 10 s. The applied voltage varied from 1 to 3 kV.
Fourier transform infrared spectroscopy (FTIR) analysis was performed on Bruker Tensor-27 (Billerica, MA, USA) spectrometer equipped with an MCT detector and OPUS 6.5 software (Bruker Optics Inc., Billerica, MA, USA). The total number of scans was 256 at 4 cm À1 resolution. Origin Pro 8.0 software (OriginLab Corporation, Northampton, MA, USA) with a curve-fitting procedure was utilized to quantify the amount of released gentamicin.
Atomic force microscopy (AFM) measurements were performed in air at room temperature using a NSCRIPTORdip pen nanolithography system (Nanoink, Chicago, IL, USA) operating in contact mode. AFM measurements of film thicknesses were obtained by scanning the AFM tip over a scratch in the film made with a razor.
Ellipsometry measurements of the thickness of dry polymer-antibiotic films were performed using a homebuilt instrument, assuming a constant film refractive index of 1.5.
Statistical analysis. All quantitative data are reported as mean±s.d. At least three samples per time point were evaluated for statistical analysis. Significant differences among the groups of scaffolds were determined by performing Student's t-test. A confidence interval of Po0.05 was considered statistically significant.

Responsive nanocomposite assemblies
We first assembled a hydrogel-like LbL matrix for hosting therapeutic molecules. Figure 1 illustrates the direct assembly of MMT platelets with a weak polyanion, PAA, within the LbL films. The success of this assembly is somewhat counterintuitive, as both MMT and PAA carry a negative charge at neutral pH. To circumvent electrostatic repulsion between film components, assembly was performed at pH 2.2, at which PAA carries no negative charge (pK a of PAA is B6.8). AFM measurements of film thickness by scanning the AFM tip over a razor scratch in the film confirmed that, similar to other clay-polyelectrolyte multilayers, 15,19 the dry thickness increased linearly with the number of deposited layers (Figure 1a), suggesting a low diffusivity of film components during deposition. The bilayer thickness for the MMT/PAA system of B15 nm is not unusual for polyelectrolyte/MMT, for which larger bilayer thicknesses have been reported, 32 with differential thicknesses of 11 ± 1 nm for clay and 4 ± 1 nm for PAA. These data suggest the assembly of partially exfoliated MMT platelets 7 and reflect a loopy conformation of uncharged PAA chains under deposition conditions at pH 2.2. The neutrality of PAA molecules during the assembly conditions follows from z-potential measurements, when MMT/PAA shells were constructed around 4-mm-diameter silica particles at pH 2.2 ( Figure 1b). The z-potentials oscillated between B À35 ± 1 mV and B0 ± 1 mV for the deposition of the MMT and PAA layers, respectively (see Figure 1b). The linear growth of nanocomposite films was also confirmed by FTIR (Supplementary Figure S1) that showed a linear increase in the intensities of a 1717 cm À1 4C ¼ O stretching vibrational band of PAA and several bands assigned to MMT nanosheets (high-intensity 1047 cm À1 band and lower-intensity 1120, 922, 890, 850 and 800 cm À1 bands) with the number of MMT/PAA bilayers.
MMT/PAA constructs have unprecedented degrees of swelling that has not been previously observed with uncrosslinked LbL films. Figure 1c shows the equilibrated swelling degrees of (MMT/PAA) 150.5 films stained with Alexa 488 for visualization by confocal laser scanning microscopy. The swelling kinetics of the MMT/PAA films in solutions at various acidities and the corresponding contact angle changes are shown in Supplementary Figure S2. Even at pH 2.5, where PAA does not carry a charge, the film takes up a large amount of water (swelling degree B4). At pH 42.5, the water uptake dramatically increases as the PAA becomes more ionized, reaching a swelling degree as high as 23 at pH 7.5.
The like charges of both film components and the stability of the highly hydrated MMT/PAA assemblies are unique features of these films. Although interactions between PAA and MMT solutions have been previously observed in bulk aqueous solutions at low pH, 33 the assembly of PAA and MMT within LbL constructs usually required the use of a 'binder' polycation layer that had to be alternated between PAA and MMT during the construction of polycation/PAA/ polycation/MMT hybrid films. 34 To the best of our knowledge, this is the first report of LbL films built solely of like-charged components. Similar to bulk MMT/polymer hydrogels, clay nanoplatelets can be considered as large 'crosslinkers' , and a smaller number of crosslinks may be required to stabilize nanocomposite hydrogel structures compared with all-polymer hydrogels. 4 At the same time, unlike bulk hydrogels, MMT/PAA LbL films are prepared via sequential adsorption from MMT and PAA solutions on a solid; therefore, their structure and composition are governed by different (that is, adsorption) laws than those applicable to their bulk counterparts.
We hypothesize that nanocomposite film components are held together via hydrogen bonding of unionized PAA groups within the basal plane and via electrostatic interactions between charged PAA units and the positively charged edges of MMT nanoplatelets. Whereas MMT nanoplatelets present negative charge at the basal plane, their edges are positively charged in acidic media because of the protonation of -AlOH groups (pK a of AlOH 2 þ ¼ B5). In agreement with our hypothesis on the significant role of clay-edge-positive charges, a MMT/PAA assembly could only be constructed at a pH p4. In Multiresponsive bacteria-triggered surface coatings S Pavlukhina et al addition to ion pairing between PAA and AlOH 2 þ edge groups, assembly may also be driven by dipole-cation interactions between carbonyl groups of PAA and positive edge charges. Remarkably, when assembled under acidic conditions and exposed to high-pH buffer solutions between 2.5 and 8, nanocomposite films were stable for at least 45 days. At pH48, however, the films deconstructed as the edge AlOH and SiOH groups of MMT nanoplatelets acquired a negative charge (pK a of AlOH and SiOH ¼ 7.9-8.5). Thus, although the pH window for deposition of clay/PAA films was controlled by the pK a of AlOH 2 þ at the edges, the stability of the formed multilayers correlated with the pK a of the AlOH and SiOH groups.
The stability and highly open, swollen structure of MMT/PAA assemblies at pH 7.5 are beneficial for the use of these films as matrices to host and release bioactive molecules, such as positively charged antibiotics. We used gentamicin, a broad-spectrum antibiotic clinically used against Gram-positive and Gram-negative bacteria, as a guest molecule for loading within MMT/PAA nanocomposite films. An increase in the dry film thickness (Figures 2a and d) indicated an inclusion of a large amount of antibiotic, that is, B45±5% of the dry matrix thickness of gentamicin within the film corresponded to high loading densities of B450 mg mm À3 . The loading was accompanied by an increase in the water contact angle of the 9.5-bilayer nanocomposite matrix from 55±41 for as-deposited to 78±41 for gentamicin-loaded MMT/PAA films. The cross-sectional SEM images of (MMT/PAA) 90.5 coatings are consistent with earlier reports of the preferred orientation of clay nanosheets parallel to film surfaces. 8,32 In addition, the presence of gentamicin within the nanocomposite films was monitored by FTIR using a characteristic vibrational band of gentamicin at B1557 cm À1 (Figure 2b).
Supplementary Figure S3A shows AFM images of dry (MMT/ PAA) 90.5 films, demonstrating a root-mean-square roughness of B30±5 nm for these B1.4-mm-thick films containing MMT platelets within the outermost film layer. The film roughness was almost independent of the number of deposited layers (data not shown). After gentamicin loading, the film roughness remained unchanged (Supplementary Figure S3B), indicating a laterally homogeneous inclusion of gentamicin. Absorption of gentamicin by the highly swollen nanocomposite matrix was accompanied by drastic, 15-fold film deswelling (Figure 2c), that occurred because of a neutralization of the negative charges within the MMT/PAA matrix by ionic pairing with gentamicin molecules. After the uptake of gentamicin, the swollen film thickness comprised 50% MMT/PAA, B23% gentamicin and B27% water.
Obvious candidates for gentamicin sequestration are the carboxylic groups of the assembled PAA that can form ionic pairs with the protonated amino groups of gentamicin. Basal planes of MMT are also known to bind to positively charged molecules, including cationic antibiotics, via an exchange with native cations. 35,36 To confirm the binding of gentamicin with MMT nanoplatelets, we directly assembled gentamicin and MMT from pH 7.5 solutions and used these films as a control. Supplementary Figure S4 shows that unlike (MMT/PAA) n films, the MMT/gentamicin films were very rough (Supplementary Figure S4), suggesting uneven deposition of components. However, these films could be consistently constructed at pH 7.5 and were invariable in their morphology in various buffer solutions. Therefore, we conclude that gentamicin binding occurs with both components of the film (PAA and MMT). However, despite the binding of gentamicin with PAA, films composed solely of PAA and gentamicin could not be constructed because of solubilization of the adsorbed gentamicin by the excess amounts of PAA added at each sequential deposition step (Supplementary Figure S5).
It was important to probe the long-term behavior of gentamicincontaining films under bacteria-free physiologic conditions, that is, in PBS at pH 7.5. Most of the currently used antibiotic-containing coatings elute their content as a function of time with a low tail release, contributing to the development of antibiotic-resistant bacterial strains. In contrast, gentamicin-loaded MMT/PAA films showed exceptional stability in PBS at pH 7.5, and they did not elute measurable amounts of gentamicin for up to 45 days, even when exposed to large volumes of buffer that was refreshed daily (Figure 3). The 45-day stability was also confirmed by the invariability of the AFM-measured dry thickness of the film.
However, when challenged by acidic buffer solutions, gentamicinloaded films released some of the sequestered antibiotic. The amount of gentamicin released was strictly controlled by the environmental pH and leveled off after B5 min with each drop in pH. This suggests that release is determined by the charge balance of gentamicin ionic pairing with PAA units within the film and occurs as ionic pairing is disrupted because of the protonation of PAA in low-pH solutions. In contrast to our previously reported single-component films of all-polymer chemically crosslinked poly(methacrylic acid) LbL hydrogels, 26 clay/PAA layers exhibit significantly enhanced retention of antibiotics within the host matrix in salt solutions, most likely because of favorable dipole-dipole interactions of antibiotics with MMT nanoplatelets. Increasingly, large percentages of gentamicin were released from MMT/PAA films in solutions with lower pH, as confirmed by FTIR and AFM (Figures 4a and b). The pH-triggered gentamicin release was also confirmed in experiments with gentamicin-loaded (MMT/PAA) 3.5 films deposited on silica particles, for which the amount of released gentamicin was determined by monitoring the supernatant absorbance of a ninhydrin-gentamicin product (Supplementary Figure S6).
All techniques indicated that even at a pH as low as 3.5, approximately half of the included gentamicin remained bound within the nanocomposite matrix. Because the control MMT/ gentamicin films did not release gentamicin when the pH was varied between 3 and 8 (Figure 4a), the released gentamicin accounts for that what was bound with PAA, and the fraction of unreleased gentamicin at pH 3.5 represents that what was sequestered by the MMT nanoplatelets. The presence of two types of antibiotic-binding sites in MMT/PAA nanofilms is advantageous for the antibacterial activity of these coatings, as it supports a dual mechanism of antibacterial defense via a combination of pH-triggered release and contact killing.
After completing gentamicin release at pH 3 and re-exposure to PBS solutions at pH 7.5, the films drastically enhanced their water uptake capability and increased their swelling ratio to B9 (Figure 4c). Such an increase is a result of pH-induced desorption of gentamicin from the PAA units and its replacement with inorganic counterions. Dissociation of these inorganic counterions from the binding sites led to swelling of the nanocomposite. Postrelease swelling was less pronounced than the initial swelling of the MMT/PAA matrix, and this is probably because of the residual crossbinding of charged gentamicin molecules with PAA and clay nanoplatelets. The z-potentials of the MMT/PAA-coated silica particles before and after gentamicin release were also consistent with two types of gentamicin binding within the matrix. An initial, strongly negative z-potential of the MMT/PAA coating of the silica particles shifted to almost zero after gentamicin uptake (Figure 4d) and then increased its negativity after the particles were incubated at lower pH values, separated by centrifugation and redispersed in buffer solutions at pH 7.5 ( Figure 4d). More negative z-potentials indicate a loss of positively charged gentamicin by the coating after pH-triggered antibiotic release. Concurrently, the z-potential does not reach the higher negative values measured with the original unloaded MMT/PAA matrix, indicating partial retention of gentamicin within the coating as a result of strong, pH-independent binding of gentamicin by the MMT platelets.

Inhibition of bacterial growth
The antibacterial activity of the gentamicin-loaded nanocomposite coatings was then studied by exposing the coatings to S. aureus ATCC 12600, S. epidermidis ATCC 14990 and E. coli ATCC 25922. We explored the antibacterial properties of the antibiotic-loaded MMT/PAA films via two types of experiments that employed either Petrifilm agar plating or flow chamber protocols to probe the antibacterial activity. Figure 5 illustrates the bacterial survival during the Petrifilm assay that employed gentamicin-loaded MMT/PAA films with different numbers of bilayers in the nanocomposite matrices. The results are expressed as 'bacterial survival' or 'complete killing' that constitutes a highly stringent, 'binary' evaluation of the coating ( Figure 5). Figure 5 shows that the antibacterial activity of the gentamicinloaded films was remarkably high: even when challenged with such a high bacterial number-8 Â 10 7 bacteria per cm 2 (close to complete bacterial monolayer coverage)-the gentamicin-loaded MMT/PAA 10 À5 % coverage of the surface by bacteria. The antibacterial activity at such a low surface coverage clearly shows that local acidification of the film by a single adhering bacterium induces sufficient gentamicin release for antibacterial activity. To further confirm the film's antibacterial activity and to identify any live bacteria that remained undetected in the Petrifilm assays, samples of (MMT/PAA) n films after the assays with the highest bacterial challenge concentration were immersed in liquid growth media for possible bacterial regrowth at 37 1C. No growth was detected on the 9.5-bilayer gentamicin-loaded MMT/PAA films, confirming complete killing of staphylococci by the antibiotic-loaded matrix. In contrast, gentamicin-loaded films with a smaller number of MMT/PAA bilayers (n ¼ 0.5, n ¼ 1.5, n ¼ 3.5 and n ¼ 6.5) and gentamicin-free MMT/PAA films exhibited staphylococcal growth. Figure 5 shows that complete killing was dependent on both the staphylococcal concentration and on the number of MMT/PAA bilayers in the nanocomposite matrix. The data clearly show that bacterial survival decreases with an increasing number of MMT/PAA bilayers within the gentamicin-loaded nanocomposite films. This observation suggests that gentamicin loaded within the entire film thickness, rather than only deposited within the film top layers, became bioavailable as a result of the release triggered by the lowered pH associated with bacterial metabolism. (MMT/Gent) n films (n ¼ 4, 7 and 10) were much less active and only showed complete killing at smaller bacterial challenges of 80, 160 and 1600 bacteria per cm 2 . Pre-exposure of gentamicin-containing films to PBS  solutions for 9 days did not negatively affect the antibacterial activity of (MMT/PAA) 9.5 coatings (data not shown). The Petrifilm assays were also performed with gentamicinloaded coatings using two other bacterial strains: Gram-positive S. epidermidis ATCC 14990 and Gram-negative E. coli ATCC 25922. (MMT/PAA) 9.5 coatings completely killed these pathogens at challenges of 1.6 Â 10 5 and 1.6 Â 10 7 bacteria per cm 2 (Supplementary Figure S7).
The Petrifilm assays mimic interfacial gaps that are present, for example, between a bone implant and a bone. However, because of the deliberately small fluid volume (25 ml) in the assay, these experiments could not elucidate whether bacterially induced, local pH changes are still effective in a flowing bulk medium. To address this question, we performed bacterial adhesion experiments in a parallel plate flow chamber, assuming that the pH of flowing bulk medium is not affected by staphylococci on the coating. Figure 6 shows that gentamicin-free MMT/PAA matrices and gentamicinloaded 9.5-bilayer MMT/PAA assemblies induced an approximately fourfold reduction in staphylococcal adhesion compared with control, even under a constant 2 h flow of concentrated suspensions (3 Â 10 8 bacteria per ml, yielding a surface coverage by staphylococci of B10%) of S. aureus at a flow rate of 1 ml min À1 . However, a similar observation was made at a 100-fold lower bacterial concentration (3 Â 10 6 bacteria per ml, corresponding to a surface coverage after 2 h of flow of B0.1%). The antibacterial efficacy at such a low concentration under flow conditions confirms the conclusion drawn from the Petrifilm assays: acidification by a single adhering staphylococcus induces sufficient local gentamicin release for antibacterial activity. Importantly, the antibacterial activity of the coating in the flow experiment can be enhanced such that an B15-fold reduction in staphylococcal adhesion is observed by increasing the number of bilayers within the MMT/PAA matrix to 15.5 (Figure 6a). These results confirm that bacteria can trigger gentamicin release from the nanocomposite coating via highly localized pH changes in the absence of a pH change in the bulk fluid surrounding the coating and that these pH changes induce antibiotic release from the entire film thickness rather than just the film surface.
Cytocompatibility of the coatings A combination of antibacterial activity with good tissue-cell integration is a central requirement for the development of the next generation of totally internal implants and devices made from biomaterials. 37 HFOB cells were used for in vitro cytocompatibility experiments. Figure 7 shows the number of osteoblasts on the left and the cytoskeletal changes in HFOB 1.19 cells on the right. The HFOB 1.19 cells were deposited on gentamicin-free and gentamicin-loaded (MMT/PAA) 9.5 nanocomposite coatings or on the surface of bare silicon wafers that served as the control.
Compared with Si wafers, the HFOB 1.19 cells for both MMT/PAA groups exhibited a slightly lower increase in cell number (Po0.05) at day 7 that may be attributed to differences in surface chemistry and roughness. The effects of surface chemistry, charge, roughness, stiffness and wettability on cell adhesion and proliferation are well known. [38][39][40] For all three substrate types after day 1, HFOB 1.19 cells exhibited a star shape with focal contacts or a spindle shape indicating healthy cells. After 7 days, an increased F-actin density indicated good

CONCLUSIONS
Here, we explored hydrogel-like nanocomposite LbL films as matrices for the controlled sequestration and delivery of antibacterial compounds. The coatings were composed of like-charge polymers and inorganic nanosheets bound together via noncovalent interactions, and they exhibited extremely high degrees of swelling in water. Furthermore, the LbL constructs were mechanically robust because of the presence of clay nanosheets. The nanocomposite films were able to host large amounts of gentamicin, an antibiotic with a broad spectrum of antibacterial activity that is clinically used in local drug delivery systems, that remained safely sequestered and did not elute until the coating was challenged by pH or bacterial stimuli. This ondemand self-defense strategy minimizes side effects and prevents the continuous elution and low-level tail release of preloaded antibiotics that reduces the risk of developing antibiotic-resistant bacterial strains. Gentamicin-loaded films demonstrated a high efficacy against several types of bacteria known to cause infections associated with medical implants, such as S. aureus, S. epidermidis and E. coli. We suggest that the high efficacy of the coatings is likely a result of the multiresponsive and self-defensive nature of the coatings that provide a multifaceted antibacterial response through a combination of bacteria-triggered release, contact killing and film swelling associated with antibiotic release (provided as an animation in Supplementary Figure S9). These multiple responses were enabled through a synergistic combination of inorganic (clay nanosheets) and organic (weak polyacid) components within the LbL assemblies. Additional advantageous features of the coating include facile control of the total amount of sequestrated and released antibiotic via the number of clay-polymer bilayers assembled within a nanocomposite matrix and the conformal nature of the coatings to substrates of complex shape. Finally, we demonstrated that gentamicin-loaded nanocomposite coatings inhibited bacterial colonization while simultaneously promoting a healthy tissue response. Taken together, these properties make these coatings promising candidates for fighting bacterial infections associated with biomedical and implantable devices.