Abstract

   In the competition between the pathogen infection and the host defense,
   infectious microorganisms may enter the host cells by evading host
   defense mechanisms and use the intracellular biomolecules as
   replication nutrient. Among them, intracellular Staphylococcus aureus
   relies on the host cells to protect itself from the attacks by
   antibiotics or immune system to achieve long-term colonization in the
   host, and the consequent clinical therapeutic failures and relapses
   after antibiotic treatment. Here, we demonstrate that intracellular S.
   aureus surviving well even in the presence of vancomycin can be
   effectively eliminated using an emerging cell-mimicking therapeutic
   strategy. These cell mimics with natural killer cell–like activity
   (NKMs) are composed of a redox-responsive degradable carrier, and
   perforin and granzyme B within the carrier. NKMs perform far more
   effectivly than clinical antibiotics in treating intracellular
   bacterial infections, providing a direct evidence of the NK
   cell–mimicking immune mechanism in the treatment of intracellular S.
   aureus.
     __________________________________________________________________

   An NK cell mimic that eliminates intracellular bacteria and activates
   antimicrobial immune responses was developed.

INTRODUCTION

   Staphylococcus aureus, the leading cause of bacterial infection in
   bodies worldwide, is a major threat to human lives ([36]1). In the past
   decades, the emergence and rapid spread of methicillin-resistant S.
   aureus (MRSA), a kind of bacteria resistant to common clinical β-lactam
   antibiotics, has made the S. aureus infections increasingly difficult
   to treat ([37]2). Disappointingly, the reduced susceptibility of MRSA
   to vancomycin has been reported in clinics ([38]3). MRSA is a pathogen
   capable of colonizing the skin and nasal mucosa of healthy individuals
   ([39]4). Once colonized by MRSA, patients can suffer diseases such as
   osteomyelitis, endocarditis, pneumonia, and sepsis ([40]5, [41]6).
   Numerous studies over the decades have illustrated that S. aureus is
   capable of invading cells and surviving thereinto, resulting in high
   resistance to the attacks by exogenous antibiotics or phagocytic cells,
   and the consequent ineffective bacterial clearance in vivo
   ([42]7–[43]12). At the same time, cellular nutrients will be used by S.
   aureus, which has been shown to reverse the innate immune response by
   inducing the proliferation of regulatory macrophages and even killing
   neutrophils ([44]13–[45]15). Together, these evidences suggest that the
   eradication of intracellular S. aureus to avoid recurrence is currently
   a major challenge in clinical infection therapy.

   As one of the major bacterial pathogens, S. aureus is responsible for
   approximately 80% of all cases of human osteomyelitis, which can invade
   and remain within osteoblasts, making osteomyelitis a chronic and
   unmanageable infectious disease of bone ([46]16). Resultantly,
   osteomyelitis shows especially low response to conventional antibiotic
   therapy, such as surgical debridement and antibiotic chemotherapy, due
   to the anatomical and physiological characteristics of bone,
   intracellular colonization of the bacteria, and biofilm formation on
   implants, leading to poor prognosis ([47]17, [48]18). Actually,
   approximately 40% of patients still suffer from recurrent and
   persistent infections after receiving conventional clinic treatments
   ([49]19). Other antibiotic alternative strategies, such as
   hyperthermia, photodynamic therapy, and nanomedicine ([50]20–[51]22),
   are also less satisfactory due to the limited penetration depth of
   light, or poor specificity to infected cells, or anti-immune responses
   of the bacteria, rendering them ineffective in eliminating deep bone
   infections.

   Nevertheless, in the self-protection mechanism of living organism,
   several immune pathways are considered to be effective in clearing
   bacteria from infected cells, such as cytotoxic T lymphocytes (CTL)
   ([52]23) and natural killer (NK) cells ([53]24). Most recently,
   artificial cells have attracted intensive attention as an alternative
   to natural cells, exhibiting great potentials in the development of the
   next generation of smart materials, autonomous micromachinery, and cell
   mimics, which may find applications in numbers of fields from medicine
   to the environment, and even in establishing novel theories about the
   origin of life ([54]25, [55]26). For example, integral biological cell
   imitators with cell-like structures have exhibited a number of main
   characteristics of living cells ([56]27), while comparatively
   engineered materials only mimic certain physical or structural features
   of cells (such as surface features, shape, morphology, or some specific
   functions) ([57]28). In addition, there have been few reports on
   establishing cell mimics with immune cell–related functions via
   nanotechnology and nanoparticle construction. The latest researches
   have shown that a number of therapeutic strategies with dual-targeting
   functions to infected cells and bacteria ([58]29–[59]31), as well as
   the design of immune mimics ([60]28, [61]32), are effective in
   destroying pathogens in infected cells. Therefore, a concept of cell
   mimics with NK-like activity is expected to produce notable efficacy
   for S. aureus treatments ([62]33).

   Here, we propose an emerging therapeutic strategy that could
   effectively kill intracellular S. aureus by a cell mimic with NK-like
   activity, named herein as NK cell mimics (NKMs). NKMs were constructed
   by loading perforin and granzyme B in modified inorganic mesoporous
   silica nanoparticles, which can be activated after recognizing the
   highly expressed redox potential microenvironment around infected
   cells. In brief, mesoporous silica carriers could be degraded to
   release perforin to perforate the membrane of infected cells and create
   channels for granzyme B to diffuse through and then kill infected
   pathogens within the cells, mimicking innate immunity of NK cells for
   intracellular pathogen clearance mechanisms. NKMs demonstrate
   remarkable treatment efficacy against vancomycin-insensitive
   intracellular S. aureus infection, based on the specific recognition
   mechanism of changes in redox potentials before and after osteoblast
   infections. Besides, NKMs evoke the immunostimulatory capacity of
   bacteria-infected osteoblasts and augment the host’s innate and
   adaptive immune responses to eliminate intracellular bacterial
   pathogens of bone tissue. NKMs treatment also shows a strong
   prophylactic activity against intracellular infection relapse by
   inducing robust bacteria-specific memory T cell responses. This kind of
   cell mimics simply composed of inorganic mesoporous silica as a carrier
   and engineered proteases as the cargoes features the advantages of easy
   mass production, high biosafety, and high and sustained performances
   against intracellular bacterial infections, which provides an
   innovative strategy for anti-infection and immune response activation
   by immune cell mimicking, exhibiting promising prospects in future
   clinical translation ([63]Fig. 1).

Fig. 1. NK cell mimics kill intracellular bacteria.

   [64]Fig. 1.
   [65]Open in a new tab

   (A) TEM image of NKMs (granules) added to S. aureus–infected
   osteoblast. Biodegradation of the NKMs to release perforin and granzyme
   B (orange border inset), triggering the disruption of the infected cell
   membrane (navy blue border inset). Red arrows point to damaged cell
   membranes with holes, and yellow arrows point to intact cell membranes.
   (B) Schematics showing that infected cells (the pale pink cells) with
   high redox expression induce the degradation of NKMs and the subsequent
   protease releases, while the oligomerization of perforin creates pores
   in the cell membrane, enabling the granzyme B diffusion through, which
   in turn targets the bacteria inside the infected cells.

RESULTS

Intracellular S. aureus is refractory to antibiotic treatments

   Intracellular survival of S. aureus has been recognized as a possible
   risk factor leading to antibiotic resistance and infection persistence
   in osteomyelitis ([66]16). To demonstrate the in vivo osteoblast
   infection by S. aureus, we analyzed the infected bone tissue of
   patients who were diagnosed of chronic peri-prosthetic S. aureus
   infections via Gram staining, Giemsa staining, bio–transmission
   electron microscopy (TEM), and confocal laser scanning microscopy
   (CLSM). The results show that in all four clinical samples collected
   from patients with confirmed recurrent S. aureus infections, bacteria
   were observed in osteoblasts located at the bone margin sites and were
   around 0.5 μm in size in the high-magnification TEM images, which
   provides direct evidence of chronic infection of osteoblasts by S.
   aureus (figs. S1 to S5). In addition, as bone cells provide a
   protective niche for S. aureus in the presence of antibiotic
   treatments, we evaluated the efficacy of vancomycin that is currently
   used as the standard of care for MRSA bone infections against bacteria
   sequestered inside the human osteoblast MG63 cell line ([67]Fig. 2F).
   CLSM and TEM results reveal that S. aureus can localize into the
   cytosol of MG63 cells, while the number and structure of S. aureus are
   hardly influenced by vancomycin therapy ([68]Fig. 2, A and B, and fig.
   S6), and the morphology of MG63 cells shows unnoticeable change after
   being infected by S. aureus ([69]Fig. 2B). Quantitative analysis of
   bacterial flow cytometry and colony-forming unit (CFU) count suggests
   that vancomycin has failed to kill MRSA that sequestered inside
   osteoblasts ([70]Fig. 2, C and D). These results demonstrate that the
   clinically available classical antibiotic for MRSA was ineffective in
   killing intracellular S. aureus due to the protection by the host
   cells, and alternative solutions are urgently needed ([71]Fig. 2E).

Fig. 2. Intra-osteoblastic S. aureus is resistant to vancomycin.

   [72]Fig. 2.
   [73]Open in a new tab

   MG63 cells infected with S. aureus were treated with or without
   vancomycin and imaged by (A) CLSM, (B) SEM, and bio-TEM. The red arrows
   indicate intra-osteoblastic S. aureus. (C and D) Representative flow
   cytometric analysis images and corresponding quantification of S.
   aureus–positive MG63 cells with or without vancomycin treatment (n =
   3). (E) CFU counting of intra-osteoblastic S. aureus with or without
   vancomycin treatment (n = 3). (F) Schematic showing that
   intra-osteoblastic S. aureus is resistant to vancomycin. Data in bar
   graphs are presented as mean ± SD (n = 3), and n represents independent
   experiments. Two-tailed Student’s t test was used for statistical
   analysis. Exact P values were provided.

Identification of enhanced reducibility as a cellular target for
intracellular infections

   How to discriminate between infected and uninfected cells presents a
   major obstacle in developing intracellular bacterial targeting
   strategies. It has been reported that pathogen invasion could change
   the cellular redox state of innate immune cells among plants and
   animals ([74]34–[75]36). Besides, changes in redox status have been
   observed in virus-infected antigen-presenting cells, which is proposed
   as a signal conceived by CD8^+ T cells to discriminate between
   uninfected and infected cells ([76]37). Inspired by this, we envisaged
   that the redox potential of bacteria-infected cells might serve as a
   promising target to treat intracellular infections ([77]Fig. 3D). We
   first investigate the redox-related gene expression in S.
   aureus–infected MG63 osteoblast using RNA-sequencing analysis.
   Hierarchical clustering reveals major divergences in nicotinamide
   adenine dinucleotide (NAD) pathway metabolisms between the infected and
   uninfected ([78]Fig. 3A). NAD synthesis–related genes such as NAPRT and
   IDO are significantly up-regulated, while NAD degradation–related genes
   such as NNMT and PAPR are down-regulated, indicating an overall
   increase in oxidation-reduction metabolic activities. To monitor the
   reductive and oxidative status of S. aureus–infected MG63 osteoblast,
   we measured the cell surface reduction activity of infected and
   uninfected MG63 cells using WST-1, a membrane-impermeable regent that
   forms a colored product following reduction ([79]38). S.
   aureus–infected MG63 osteoblasts exhibit higher cell surface reduction
   rates than uninfected MG63 osteoblasts ([80]Fig. 3B). Electrochemical
   measurements show that MG63 cells have been infected by S. aureus,
   resulting in enhanced reducibility of the cells and decreased open
   circuit potential (OCP) of the working electrode due to the electron
   supply by the reducing species. In addition, the cyclic voltammetry
   (CV) results after the infection shows the intensified reduction peak,
   i.e., increased current density at around 30, demonstrating the
   promoted electron-gaining reaction ([81]Fig. 3C). These results suggest
   that MG63 cells have most probably experienced certain changes in their
   redox species due to infection by S. aureus ([82]39, [83]40). The redox
   potential of a cell is mainly determined by reducible and oxidizable
   chemical pairs, including GSH [glutathione (reduced form)]/GSSG
   (oxidized glutathione), reduced thioredoxin/oxidized thioredoxin, NADH
   [reduced form of nicotinamide adenine dinucleotide (oxidized form)
   (NAD^+)]/NAD^+, and NADPH [reduced form of nicotinamide adenine
   dinucleotide phosphate (NADP^+)]/NADP^+ ([84]41). Notably, the
   proportion of NADH/NAD^+, NADPH/NADP^+, GSH/GSSG is significantly
   higher for S. aureus–infected MG63 cells than for uninfected MG63
   cells, which are responsible for the enhanced cell surface reduction
   rates of S. aureus–infected MG63 osteoblasts ([85]Fig. 3, E to G). We
   expect that the increased redox activity on the surface of S.
   aureus–infected MG63 osteoblasts could be used to develop an
   intracellular infection-specific antibacterial strategy by designing
   reduction-responsive nanoparticles.

Fig. 3. S. aureus invasion increases cell surface reduction of host
osteoblasts.

   [86]Fig. 3.
   [87]Open in a new tab

   (A) Heatmap showing hierarchical clustering of DEGs between uninfected
   and infected MG63 cells from three biological replicates. (B)
   Measurement of WST-1 cell surface reduction activity of S.
   aureus–infected MG63 cells in the presence of an intermediate electron
   acceptor. (C) OCP and CV curves of MG63 cells before and after being
   infected by S. aureus. (D) Proposed strategy for connecting increased
   surface redox activity of S. aureus–infected MG63 cells to enhanced
   protein release kinetics from a redox-responsive nanosystem. The
   proportion of (E) NADH/NAD^+, (F) NADPH/NADP^+, and (G) GSH/GSSG in
   uninfected and S. aureus–infected MG63 cells. Data in bar graphs are
   presented as mean ± SD (n ≥ 3), and n represents independent
   experiments. Two-tailed Student’s t test was used for statistical
   analysis. Exact P values were provided.

Designing the NK cell mimics

   These findings suggest that bacteria-infected cells exhibit a surface
   redox specificity, and a therapeutic modality aiming at identifying
   such a specificity for eliminating intracellular bacteria may be
   possible and promising in the prevention and treatment of intracellular
   pathogens. To verify this, we developed a cell-mimic therapeutic by
   creating an NK cell mimic that is able to be specifically activated
   upon the recognition of S. aureus–infected MG63 cells. Such NKMs were
   established by loading perforin and granzyme B in the mesopore channels
   of disulfide bond–doped dendritic mesoporous silica nanoparticles
   (DMONs) in the framework, followed by polyethylene glycol (PEG)
   modification on the DMON surface. These prepared NKMs were not directly
   cytotoxic to uninfected cells and will not selectively target these
   cells due to the unfavorable physicochemical properties based on the
   structural specificity of the mimics ([88]38, [89]42). Attractively,
   however, once in contact with bacterial host cells, NKMs will be
   cleaved by the high redox potential of the infected cell, promptly
   releasing perforin and granzyme B in active form. According to the
   properties of NK cells, we have found that more than 10 nanoparticles
   of NKMs can bind to a single infected cell and punch holes on the host
   cell membrane surface by perforin mediation, enabling sufficient
   amounts of granzyme B to penetrate into the phagosome to kill
   intracellular bacteria.

   The carrier and protease in NK cell mimics played their roles in the
   bacteria killing. Mesoporous silica nanoparticles, a kind of typical
   drug delivery carrier, demonstrated high-enough biocompatibility both
   in vitro and in vivo ([90]43–[91]45). The disulfide bond–incorporated
   DMON nanocarriers adopted in this study can specifically degrade via
   the -S-S- bond breakage in response to the reducing microenvironment
   within cells or tumor tissues containing high concentration of GSH
   ([92]46–[93]48), leading to subsequent structural collapse for
   responsive load release. Here, DMON was obtained by inserting organic
   groups containing disulfide bonds into the framework of mesoporous
   silica nanoparticles at the molecular level (fig. S7A) ([94]49–[95]52).
   Scanning electron microscopy (SEM) and TEM images reveal that the
   as-prepared DMONs are monodisperse and about 250 nm in diameter
   ([96]Fig. 4B and fig. S7, B and C) and present a dendritic mesoporous
   structure. N[2] adsorption-desorption isotherms further confirm the
   mesoporous feature of these DMON nanoparticles, which features an
   average pore size of 30 nm and a Brunauer-Emmett-Teller surface area of
   321.6 m^2 g^−1 ([97]Fig. 4E). The presence of disulfide bonds in DMON
   is confirmed by elemental mapping analysis ([98]Fig. 4C), which is
   consistent with the energy-dispersive x-ray photoelectron spectroscopy
   (XPS) results ([99]Fig. 4D and fig. S7H). To further confirm the
   incorporation of -S-S- bonds, DMON was characterized by ^13C
   cross-polarization magic-angle spinning (CPMAS) and ^29Si magic-angle
   spinning (MAS) solid-state nuclear magnetic resonance (NMR) spectra.
   From the ^13C NMR spectra, the peaks at 12.23, 23.88, and 42.32 ppm can
   be attributed to ^1C, ^2C, and ^3C, which correspond to the carbon
   atoms in the -Si-^1CH2-^2CH2-^3CH2-S-S-^3CH2-^2CH2-^1CH-Si- species,
   indicating the successful embedding of the organoalkoxysilane precursor
   bis(triethoxysilyl propyl)disulfide (BTES) within the framework of DMON
   (fig. S7F). In addition, the two peaks at −56.8 and −72.2 ppm in the
   ^29Si NMR spectrum are attributed to T[2] [C-Si(OSi)[2](OH)] and T[3]
   [C-Si(OSi)[3]] species derived from BTES, while the other peaks at
   −100.8 and −109.9 ppm are ascribed to the Q[3] [Si(OSi)[3](OH)] and
   Q[4][n(OSi)[4]] species from tetraethyl orthosilicate (TEOS) (fig.
   S7E). In addition, the Raman spectroscopy in combination with
   thermogravimetry–mass spectrometry was used to demonstrate the presence
   of disulfide bonds. The result shows the evident release of sulfide gas
   from the decomposition of the organic components within DMON during
   heating (fig. S7G). Meanwhile, Raman spectroscopy exhibits the specific
   stretching vibrations of -S-S- bonds at 437.4 and 488 cm^−1, and a
   typical stretching vibration of -S-C- bond at 636 cm^−1 ([100]Fig. 4F).
   These results suggest the successful framework hybridization of -S-C-
   and -S-S- bonds.

Fig. 4. Schematic and characterization of NK mimics.

   [101]Fig. 4.
   [102]Open in a new tab

   (A) Schematic showing in situ responsive NK mimics. (B) Representative
   TEM image of DMON. The inset is SEM image. The representative
   element-mapping images of DMON showing silicon (red), oxygen (green),
   and sulfur (yellow). Scale bar, 200 nm. (C) Linear scan element
   distribution of a single DMON. The inset shows the corresponding
   high-angle annular dark-field (HADDF) image. Scale bar, 200 nm. (D) XPS
   spectra of DMONs. (E) N[2] adsorption-desorption isotherms. The inset
   corresponds to pore-size distributions of DMON. (F) Raman spectrum of
   DMON. (G) Zeta potential profile of DMON and NK mimics (n = 5). (H)
   Schematic diagram of the release of NK mimics in the GSH environment
   and the change of the carrier disulfide skeleton structure. (I)
   Representative TEM images of DMON after 24 hours of incubation in GSH
   (0, 1, 5, and 10 mM). The inset shows the changes in the surface
   structure of a single nanoparticle. Scale bar, 100 nm. (J) Cumulative
   perforin and granzyme B protein release profile of NK mimics in pH 7.4
   and GSH (5 mM). Data in bar graphs are presented as mean ± SD (n ≥ 3),
   and n represents independent experiments.

   Next, the perforin and granzyme B proteins were loaded into DMON
   ([103]Fig. 4A and fig. S7D) with the loading rates of 24.42% and
   28.91%, respectively (fig. S7I). After loading the protease, the zeta
   potential of NKMs changed from −19.7 to −10.8 mV ([104]Fig. 4G).
   Subsequently, the responsive degradation of DMON was investigated in
   reducing condition (GSH, 0, 1, 5, or 10 mM), simulating normal and
   different post-infection intracellular environment. NKMs exhibit a
   rapid degradation in the reducing condition due to the reductive
   cleavage of disulfide bonds ([105]Fig. 4H). TEM images verify that DMON
   has completely collapsed into small fragments in the GSH solution in 24
   hours. In contrast, no obvious degradation can be found in the control
   group in 24 hours of incubation ([106]Fig. 4I). To verify whether the
   degradation of NKMs could induce the release of immobilized perforin
   and granzyme B, the NKM exposures in the reducing condition show a
   marked release of 75.94% of perforin and 82.77% of granzyme B proteins
   in 36 hours of incubation, which is in accordance with the DMON
   degradation process ([107]Fig. 4J).

NKMs eliminate intracellular S. aureus in vitro

   It is known that NK cells eliminate intracellular bacteria by releasing
   granzymes into infected cells through the pores punched by perforins
   ([108]53). We hypothesize that NKMs could kill intracellular S. aureus
   in a similar pattern. We first investigated the capability of NKMs in
   killing intracellular S. aureus in MG63 osteoblasts. Flow cytometry
   assay and bacterial CFU counting reveal that NKMs can significantly
   reduce the intracellular S. aureus loads in MG63 osteoblasts ([109]Fig.
   5, B and C). Bacterial immunofluorescence staining also shows that NKM
   treatment eliminates intracellular S. aureus to a much higher degree in
   comparison to either DP (DMON + perforin) or DG (DMON + granzyme B)
   treatments ([110]Fig. 5A and fig. S8A). In addition, CLSM images and
   counting results for other types of intracellular bacteria (e.g.,
   macrophages, neutrophils, and bone marrow stromal cells) after NKM
   treatment show that NKMs demonstrate much improved bacterial clearance
   efficacy compared to the high-dose rifampicin group ([111]29, [112]30)
   owing to the advantages of its spatiotemporal sequential treatment
   strategy (figs. S9 to S12). Meanwhile, the cytotoxicities of NKMs
   themselves as well as the supernatant after the treatment of
   intracellular bacteria by NKMs were evaluated on various types of
   cells, and the Cell-Counting-Kit-8 (CCK8) results demonstrate that at
   the doses higher than the therapeutic concentration, NKMs as well as
   the inflammatory factors in the posttreatment supernatant have
   negligible effects on the activity of normal cells (figs. S13 and S14).
   These outcomes suggest that NKMs could directly play an antibacterial
   role in treating intracellular infections.

Fig. 5. Immune-like eradication of intracellular S. aureus by NK mimics.

   [113]Fig. 5.
   [114]Open in a new tab

   (A) CLSM images of S. aureus–infected MG63 osteoblast with different
   treatments. (B and C) Intra-osteoblastic S. aureus bacterial load
   measured by flow cytometry (B) and CFU counting (C) (n ≥ 3). (D)
   Bio-TEM micrographs of S. aureus–infected MG63 osteoblast cells exposed
   to NKMs. Red dashed areas indicate intracellular bacteria. (E)
   Structure of perforin on the surface of a phospholipid bilayer (side
   view) and distance analysis to the nearest protein group in the
   phospholipid bilayer. (F) Interaction energies between perforin and
   phospholipid bilayers before and after pit formation. (G) AFM images
   showing the pore formation on the surface of MG63 cells exposed to
   NKMs. (H) Synchrotron SAXS of mammalian cell-mimicking liposomes after
   interacting with control (orange), NKMs (green), and NKMs* (red;
   containing the same components as NKMs, but not loaded) shows different
   structures. (I) CLSM images showing the intracellular uptake of
   granzyme B (Grz) by MRSA-infected MG63 cells. (J) GSEA pathway
   enrichment analysis of DEGs of infected MG63 osteoblast between the
   control and NKM group. (K) Secretion of TNF-α and IL-6 from MG63 cells
   after treatment with Ctrl, DP, DG, or NKMs (n = 3). (L) Schematic
   illustrating the intracellular bacterial killing process of NKMs and
   the elicitation of immunostimulating potential of osteoblast. Data in
   bar graphs are presented as mean ± SD (n ≥ 3), and n represents
   independent experiments. Two-tailed Student’s t test was used for
   statistical analysis. Exact P values were provided.

   Subsequently, we sought to identify specific antibacterial mechanism of
   NKMs. We first observed intracellular S. aureus morphology in MG63
   osteoblasts using bio-TEM and found that the border of these
   intracellular S. aureus becomes blurred upon NKM treatment ([115]Fig.
   5D). In contrast, the S. aureus bacteria structure inside the MG63
   osteoblasts remains clear and intact upon DP and DG treatments (fig.
   S8B). Besides, TEM images show that NKMs have degraded upon interaction
   with infected MG63 osteoblasts, indicating that NKMs biodegrade in an
   intracellular infection-responsive manner ([116]Fig. 5D). Alive
   bacteria actively acquire sulfur for bacterial protein synthesis and
   metabolism, while the uptake and utilization of S elements are of
   course no longer needed for dead bacteria ([117]54, [118]55). NKM
   treatment decreased the concentration of S element in the intracellular
   S. aureus, suggesting that NKMs have reduced intracellular bacterial
   viability effectively (figs. S16 and S17). In addition, pores of 10 to
   60 nm in diameter can be found on the surface of S. aureus–infected
   MG63 cells due to the presence of the NKMs at the cell membrane (fig.
   S8D), suggesting that the perforin released from NKMs have punched
   pores in S. aureus–infected osteoblast membranes. Atomic force
   microscopy (AFM) images further visualize the perforation effect of
   NKMs, in which pores ~50 to 100 nm in diameter can be observed on the
   surface of S. aureus–infected osteoblast membranes ([119]Fig. 5G and
   fig. S18, F and G). To understand the microscopic interactions between
   NKMs and osteoblast surfaces more clearly, we performed extensive
   molecular dynamic (MD) simulations for the perforation of perforin
   released from NKMs in the presence of lipid membranes. Notably,
   previous reports have demonstrated that perforins first form components
   of different sizes in the prepore intermediate state on the membranes
   of virus-infected cells and cancer cells ([120]56–[121]59). The prepore
   oligomers then become vertically collapsed in a cooperative manner,
   followed by insertion of subunits into the membrane to further form
   arc- and ring-shaped pores ([122]60). In accordance with the previous
   findings, a clear collapsing mechanism of the oligomers has been
   identified in our simulations of perforin-membrane interactions
   ([123]Fig. 5E). A molecular conformation with a cell membrane was
   constructed from 1152 1,2-dihexadecanoyl-rac-glycero-3-phosphocholine
   (DPPC) molecules (fig. S18, A to C), followed by further MD simulations
   to elucidate the effect of perforin’s perforation on the cell membrane.
   By placing the equilibrated perforin edge above the membrane surface,
   we observe that the perforin-attached cellular surface exhibits
   strengthened morphological fluctuations of the phospholipid bilayer in
   the region close to the perforin ([124]61, [125]62) (fig. S18D).
   Further analysis of the height of the phospholipid bilayer confirms the
   formation of the pore-like pits in the area next to the perforin and
   phospholipid bilayer, leading to the reduced thickness of the cell
   membrane by approximately 25.93% on 40 ns (fig. S18E). Next, the
   interaction energy levels between perforin and phospholipid bilayer
   before and after the formation of the hole were calculated. The van der
   Waals interaction energy between perforin and phospholipid bilayer is
   −737 kJ mol^−1, the electrostatic interaction energy is −53 kJ mol^−1,
   and the total interaction energy is −790 kJ mol^−1 before perforin
   approaches the surface of phospholipid bilayer. However, in 40 ns of
   perforin action on the surface of phospholipid bilayer, the van der
   Waals interaction energy between perforin and phospholipid bilayer
   reduces to −657 kJ mol^−1, the electrostatic interaction energy is −450
   kJ mol^−1, and the total interaction energy is −1107 kJ mol^−1
   ([126]Fig. 5F). The lower interaction energy of perforins with
   phospholipid bilayers suggests the more thermodynamically stable
   structure of depressed pores, manifesting the effectiveness of the
   spontaneous perforation through cell membranes ([127]63). Notably, due
   to the extremely high complexity of the present system, the current
   computing power can only satisfy the simulation in the time framework
   of tens of nanoseconds; resultantly, the simulation results of multiple
   perforin oligomerization and the complete process of intuitive pore
   penetration on the cell membrane have not been obtained.

   To gain more insight into the functioning mode of NKMs once in contact
   with mammalian cells, we used model liposomal unilamellar vesicles
   (LUVs) to mimic mammalian cells. Although the model liposomes are
   simple and partially lack the structural features of bacterial and
   mammalian cells, such as membrane asymmetry and the cytoskeletal
   network that plays an important role in endocytosis in mammalian cells
   ([128]64), experiments performed on the model LUV still faithfully
   captured the basic membrane remodeling response during coincubation
   with NKMs ([129]Fig. 5H). Small-angle x-ray scattering (SAXS) data show
   that DMON alone presents no signal, and dioleoylphosphatidylcholine
   (DOPC)/dioleoylphosphatidylglycerol (DOPG) membranes form vesicles only
   (control), while NKMs with membrane activity enable DOPC/DOPG membranes
   to form specific phases ([130]65, [131]66). As previously described,
   liposome-mimicking cell membranes are not redox responsive, so NKM*
   (with the same independent components as NKMs) is able to alter the
   membrane structure via perforin, inducing a membrane remodeling
   response and thus enhancing the characteristic peak intensity. The pore
   formation suggests that NKMs in contact with the cells may deliver
   granzyme to the infected MG63 cells. To validate this hypothesis, S.
   aureus–infected MG63 cells were investigated for possible staining by
   intracellular granzyme B. After the coculture with NKMs, MG63 cells
   were stained strongly by intracellular granzyme B. In contrast,
   granzyme B was barely detected in MG63 cells after coculture with DG
   ([132]Fig. 5I). These results indicate that NKM-induced killing of
   infected target cells is dependent on perforin-enabled pore formation
   and granzyme delivery at the synapse; this mechanism is highly similar
   with the model of NK cell–mediated intracellular bacterial killing.

   Although the primary roles of osteoblasts are to synthesize bone
   matrix, there is growing evidence that osteoblasts play an additional
   role in acting as nonprofessional phagocytic cells in perceiving
   intracellular bacterial invasion and initiating antibacterial immune
   responses by releasing an array of immunoregulatory molecules ([133]67,
   [134]68). To investigate whether NKMs could evoke the immune response
   of S. aureus–infected osteoblasts or not, we conducted RNA-sequencing
   and cytokine-releasing tests. NKM-treated S. aureus–infected MG63 cells
   exhibit a profoundly altered transcriptomic profile. Gene Set
   Enrichment Analysis (GSEA) analysis reveals that differentially
   expressed genes (DEGs) are mainly enriched in NK cell–mediated
   cytotoxicity and T cell receptor signaling pathway ([135]Fig. 5J).
   Besides, the genes that facilitate bacterial antigen presentation and
   osteoblast–T cell interactions in the above pathways, such as NFKBIA,
   TNF, CSF2, and FYN, were up-regulated (fig. S8E). The altered gene
   expression suggests that NKMs may evoke the T cell–and NK
   cell–stimulating potential of osteoblast, thus initiating the
   intracellular bacterial targeted host immune responses ([136]69).
   Cytokine-releasing test demonstrates that NKM treatment would elevate
   the levels of immunostimulating chemokines and cytokines such as
   granulocyte-macrophage colony-stimulating factor (GM-CSF),
   interleukin-6 (IL-6), and tumor necrosis factor–α (TNF-α), compared
   with DP or DG treatment ([137]Fig. 5K and fig. S8C). Given the great
   potential of these inflammatory molecules in recruiting macrophages and
   activating T lymphocytes to boost a T helper 1 (T[H]1)–type response,
   treatments with NKMs can greatly aid in activating host immune
   responses for inhibiting or even eliminating intracellular bacterial
   infections ([138]67, [139]68, [140]70) ([141]Fig. 5L).

   In brief, the NKM system not only can kill intracellular bacteria like
   the natural NK cells but also amplifies antibacterial immune response
   by inducing noncanonical osteoblast immunogenicity, which further
   evidences the NK cell–mimicking property of the NKM system.

NKMs eliminate intracellular S. aureus in osteomyelitis model

   Motivated by the intra-osteoblast bacteria-killing activity of NKMs
   observed in vitro, we examined the therapeutic potential of NKMs in
   murine osteomyelitis models. We first evaluated the in vivo biosafety
   of NKMs by analyzing routine blood and blood biochemistry, hematoxylin
   and eosin (H&E) and terminal deoxynucleotidyl transferase–mediated
   deoxyuridine triphosphate nick end labeling (TUNEL) histological
   staining, and Si elemental distributions of major organs at 1 week and
   1 month after administration of phosphate-buffered saline (PBS), NKMs
   (10 mg kg^−1), and NKMs (20 mg kg^−1), respectively (figs. S19, B and
   C, S21, and S22). No obvious variability was observed in hematological
   parameters and major organ sections in NKM-treated mice. In addition,
   in vivo fluorescence imaging showed a strong fluorescence signal of
   NKMs at the site of osteomyelitis infection lesions 24 hours after
   administration, which was consistent with CLSM observations in tissue
   sections (figs. S19A and S20). Subsequently, an osteomyelitis model of
   S. aureus infection in mice was established and then NKMs were
   intravenously injected via the tail vein ([142]Fig. 6, A and B). In
   vivo bioluminescent imaging reveals that the bacterial infection has
   been largely alleviated after NKM intervention, compared with the other
   treatments ([143]Fig. 6C).

Fig. 6. NKMs eliminate intracellular S. aureus and evoke antibacterial immune
responses in mice osteomyelitis.

   [144]Fig. 6.
   [145]Open in a new tab

   (A and B) Overall design of in vivo experiment. S. aureus–infected
   implants were inserted in the mice femur at the start of the
   experiment. At day 3, mice were allocated into four groups and DP, DG,
   NKMs, or PBS (10 mg kg^−1) was injected intravenously. Infection
   progression was monitors by in vivo imaging system (IVIS) imaging. (C)
   Bioluminescence imaging to monitor infection progression of
   osteomyelitis-bearing mice with different treatments on days 0, 5, 11,
   and 13 (n = 3 mice). (D) Confocal microscopy analysis of S. aureus and
   bone cells using anti–S. aureus antibodies (green) and phalloidin (red)
   in bone sections of osteomyelitis-bearing mice with different
   treatments. The border of bone is indicated by a white broken line, and
   the bacteria are indicated with white arrows. (E) TEM morphology of
   intracellular S. aureus in the bone sections of osteomyelitis-bearing
   mice with different treatments. The boundaries of osteoblasts are
   indicated by white dashed lines, and S. aureus is indicated by yellow
   circles and red arrows. (F) Intracellular morphology of S. aureus
   within OLCN locations in bone sections from osteomyelitis mice. S.
   aureus is indicated by yellow circles and red arrows. (G) Freshly
   dissected bone marrow from osteomyelitis-bearing mice with different
   treatments was analyzed by flow cytometry for immune composition (n = 3
   mice). Data in bar graphs are presented as mean ± SD (n = 3), and n
   represents independent experiments.

   In osteomyelitis, the S. aureus invasion and survival within
   osteoblasts, osteoclasts, and osteocytes constitutes a possible
   mechanism for infection persistence and recurrence ([146]71).
   Unfortunately, most previous studies regarding intracellular infection
   were performed in vitro, which are less relevant to the dynamic
   host-pathogen interactions in vivo. We monitored intracellular S.
   aureus infection in mouse femur bone marrow by fluorescence staining,
   Gram staining, and bio-TEM. As expected, NKM treatment substantially
   suppressed infection progression in comparison to the modest bacterial
   growth suppression by DP or DG treatment ([147]Fig. 6C and fig. S23, A
   and B). Tissue-wide visualization of bone marrow microenvironments
   reveals that the bacterial load is significantly reduced after NKM
   treatment. High-resolution CLSM images show the colocalization of
   bacteria (green), nucleus (blue), and cytoskeleton (red) in the
   osteoblast adjacent to bone margins, indicating the prevailing
   intracellular presence of S. aureus. After the NKM therapy, we observed
   a remarkable reduction in intracellular bacterial burden, while DP and
   DG control groups maintained persistent intracellular bacterial loads
   ([148]Fig. 6D). Bio-TEM also revealed typical morphology of osteoblasts
   infected with S. aureus adjacent to bone margins, which was described
   as an intracellular surviving form of S. aureus in osteomyelitis
   ([149]72). NKM treatment has led to a substantial decrease in the
   residual load of intracellular S. aureus, compared to the control
   groups ([150]Fig. 6E and fig. S23, C and D). Biofilm formation is also
   a major concern in osteomyelitis ([151]16); thus, effective
   antibacterial strategies should be able to destruct biofilms for
   osteomyelitis treatment. SEM observation reveals that bacteria can be
   hardly detected on the extracted implant of mice that received NKM
   treatments, in stark contrast with the dense bacterial biofilm formed
   on the implants of control groups (fig. S23E). Meanwhile, the results
   of both Masson and Gram staining on days 11 and 13 demonstrate the
   presence of staphylococcal abscesses in the osteomyelitis model, which
   were clustered in the medullary cavity of the long bones and associated
   soft tissues, and Gram-positive S. aureus cells were located at the
   center of the abscess ([152]16, [153]73). The S. aureus abscesses were
   effectively eliminated after the treatment with NKMs (fig. S25). The
   colonization of the osteocyte lacuno-canalicular network (OLCN) by S.
   aureus is one of the reasons for the long-term bacterial persistence of
   osteomyelitis and the failure of treatment, due to this unique
   environment that allows S. aureus to evade the attack by host immune
   cells ([154]74, [155]75). Bio-TEM imaging reveals the typical
   morphology of OLCN invaded by S. aureus in the infected bones
   ([156]Fig. 6F). After treatment with NKMs, no obvious bacterial
   invasion was observed within the OLCN, suggesting that the therapeutic
   strategy of NKMs is effective in removing bacteria in the OLCN and may
   be a promising protocol as an alternative to natural immune cells in
   the treatment of implant-induced chronic osteomyelitis.

   In addition, we evaluated the efficacy of NKMs in mouse osteomyelitis
   model against a cell membrane–penetrating clinical antibiotic to
   further highlight the clinical significance of NKMs. The results of in
   vivo fluorescence and bacterial enumeration show more pronounced
   clearance of bacteria around the implant in the high-dose rifampicin
   group and lower bacterial survival rate in the tissues around the site
   of infection than in the control group, suggesting that high-dose
   rifampicin has an antimicrobial effect on bacteria at the infected
   lesion and in the infected tissues. Comparatively, the proportion of
   bacteria in the infected implants and surrounding tissues is even lower
   in the group of mice treated with low concentrations of NKMs than with
   the high-dose rifampicin, suggesting that the strategy of
   cell-mimicking NKMs has a higher efficacy in controlling bone
   infections than the direct use of high-concentration, strongly cell
   membrane–penetrating clinically available antibiotics (fig. S26). In
   addition, we found that the therapeutic efficacy of NKMs is superior to
   the simple combination of DP and DG, confirming the necessity of a
   spatiotemporal “punch-bactericidal” sequential antimicrobial strategy
   with NKMs as cellular mimics.

   On the basis of the above results, we conclude that NKMs can both kill
   intracellular bacteria and damage biofilms effectively (fig. S15), and
   both perforin and granzyme are indispensable for killing intracellular
   bacteria, demonstrating the attractive advantages of this NK
   cell–mimicking system. We also analyzed the changes in the activation
   status of immune cells after NKM treatment. Compared with DP or DG, NKM
   treatment promoted NK cell activation and dendritic cell maturation of
   infection-draining lymph nodes while also inducing robust
   bacteria-specific CD4^+ and CD8^+ T cell responses ([157]Fig. 6G and
   fig. S24). We then tested the expressions of pro-inflammatory cytokines
   upon NKM treatment. The levels of interferon-γ (IFN-γ), GM-CSF, and
   TNF-α in the infection area were significantly elevated upon NKM
   treatment, suggesting the positive effects of NKMs on pro-inflammatory
   responses (fig. S23F). Among them, IFN-γ offers broad-spectrum pathogen
   resistance in most mammal cells through the induction of IFN-stimulated
   genes that encode antimicrobial factors ([158]76). Together, NKMs are a
   highly potent immune stimulator that activates the innate and adaptive
   antibacterial immunity in bacteria-bearing mice. The augmented NK cell
   activation and cytotoxic T cell responses benefit the eradication of
   intracellular bacterial infections by releasing perforin and granzymes,
   which results in largely enhanced therapeutic efficacy of NKMs
   ([159]53, [160]77). The efficacy achieved by this synergistic strategy
   through the spatiotemporally controlled release of NKMs is also
   stronger than that of multiple doses of antibiotics and the combined
   administration of antibiotics and DP (figs. S31 to S33). Additionally,
   we did not observe any noticeable signs of tissue damage in the heart,
   lung, liver, kidney, or spleen, as well as body weight after NKMs
   treatment (figs. S23G and S34), indicating high biosafety of NKMs
   therapy. To exclude potential damage to the bone elicited by the
   boosted immune responses, bone morphology by Masson, van Gieson,
   Safranin O/fast green, and TUNEL staining was characterized. We
   observed no notable pathological osteolysis, abnormal bone remodeling,
   or chondral and subchondral bone damage after NKM treatments compared
   to control groups (figs. S25 and S27). Overall, NKM therapy exhibits
   safe and potent antibacterial efficacy and induces robust antitumor
   immune responses toward intracellular bacteria in osteomyelitis model.

NKMs trigger durable immune memory responses to prevent infection recurrence

   Infection relapse is a major clinical sign of intracellular S. aureus
   infection of osteomyelitis ([161]78). To determine whether NKMs would
   trigger long-term protective immunity or not, we challenged surviving
   mice with a second injection of S. aureus 3 months after the initial
   bacterial challenge and NKM treatments ([162]Fig. 7A). The initial NKM
   treatments resulted in a significant reduction in the number of viable
   bacteria recovered from the implants and peri-implant tissues 14 days
   after bacterial rechallenge ([163]Fig. 7B). H&E and Giemsa staining of
   infected bone samples reveal substantial neutrophil and bacterial
   infiltration, signs of osteomyelitis, in control groups. Comparatively,
   treatment with NKMs resulted in improved pathological parameters
   ([164]Fig. 7, C and D). SEM observation shows that bacteria will adhere
   to bone marrow cells, indicating preferential invasion of the bacteria
   to find intracellular niche ([165]Fig. 7E). Flow cytometric analysis
   shows that NKM therapy vigorously expands CD44^highCD62L^low effector
   memory phenotypes and CD8^+ effector T cell subsets ([166]Fig. 7, F to
   H), which could clear up bacteria once they invade the cells ([167]79,
   [168]80). To avoid interference of immune effects from residual
   bacteria, immune effects were also evaluated in uninfected mice and in
   mice reinfected after clinical antibiotic treatment (figs. S28 and
   S29). The above results demonstrate the indispensable role of NKMs in
   priming progressive bacteria-specific vaccination to resist infection
   relapse in osteomyelitis.

Fig. 7. Enhanced bacteria-specific immune memory responses in the presence of
NKMs.

   [169]Fig. 7.
   [170]Open in a new tab

   (A) Workflow for bacterial rechallenge and immune memory responses
   induced by NKM treatment. (B) Bacterial CFU counting of infected
   tissues and implants 14 days after bacterial rechallenge (n = 3 mice).
   (C and D) Histology of the infected bone was evaluated by H&E (C) and
   Giemsa (D) staining 14 days after bacterial rechallenge. The cyan
   arrows indicate neutrophils, and the white arrows indicate bacteria.
   Scale bar, 10 μm. (E) Representative SEM images for infected implants
   from bacterial rechallenge mice. Scale bar, 1 μm. S. aureus is
   pseudocolored yellow. (F to H) The numbers of CTL (CD3^+CD8^+) and
   memory T cells (CD4^+CD44^highCD62L^low) were determined in the bone
   marrow on day 14 after bacterial rechallenge. Data in bar graphs are
   presented as mean ± SD (n = 3), and n represents independent
   experiments. Two-tailed Student’s t test was used for statistical
   analysis. Exact P values were provided.

DISCUSSION

   Biofilm formation and intracellular invasion are two typical features
   of S. aureus that make it a prevalent and hard-to-treat infection in
   human osteomyelitis ([171]16). Such an intracellular survival of S.
   aureus could be one of the major mechanisms of its resistances to the
   intravenous antibiotics as well as the host defenses such as engulfment
   by white cells ([172]81). NK cell is known to kill intracellular
   bacteria by selectively transferring antimicrobial granzymes into
   infected cells via the nanotubes formed by perforin ([173]53). However,
   the clinical translation of NK cell–based immunotherapy has remained
   challenging, at least in part due to the difficulty in NK cell
   isolation, high cost, and systemic toxicities ([174]82). To address
   this clinical challenge, we have developed herein an NK cell–mimicking
   platform that can be specifically activated upon interaction with S.
   aureus–infected osteoblast and kill the intracellular S. aureus
   ([175]Fig. 1).

   The present NK-mimicking platform features three biological findings.
   First, osteoblast will present increased cell surface reducibility
   following S. aureus intracellular invasion. The mechanisms underlying
   this redox regulation are most probably associated with the altered
   expressions of NAD synthesis and degradation-related enzymes ([176]Fig.
   3). Second, the perforin and granzyme releases from the NKM platform
   are essentially driven by the increased cell surface reducibility of
   the infected cells. Our in vitro studies have well revealed the NKM
   biodegradation and the perforation and granzyme releases from the
   system for intracellular bacterial elimination ([177]Fig. 4), which is
   in high similarity with the functions of NK cells against the
   intracellular bacteria ([178]53). This NK cell–mimicking process
   results in intra-osteoblastic S. aureus elimination both in vitro and
   in vivo, while in contrast, vancomycin, the current standard of care
   for MRSA infection, failed to kill these bacteria ([179]Fig. 2). Third,
   NKMs elevate the immunostimulating capacity of the infected osteoblast
   by activating NK cell–mediated cytotoxicity and T cell receptor
   signaling pathway, thus leading to the increased productions of
   immunoregulatory cytokines such as IL-6, TNF-α, and GM-CSF ([180]Fig. 5
   and fig. S8). The immunoactivated osteoblasts would recruit innate and
   adaptive immune cells and drive antigen-specific activation of T
   lymphocytes to eliminate intracellular bacterial pathogens of bone
   tissue ([181]67, [182]69, [183]70).

   Although intra-osteoblastic S. aureus has been extensively evidenced in
   vitro, the intracellular persistence of S. aureus in bone cells has
   rarely been validated in vivo, which challenges the development of
   intracellular bacteria-targeted therapeutics in chronic osteomyelitis
   ([184]75). We collected S. aureus–infected bone tissue samples from
   clinical patients who were diagnosed with chronic S. aureus
   peri-implant infections. Multi-parameter analysis on the infected bone
   demonstrated that osteoblasts can be chronically infected by S. aureus
   in vivo (figs. S1 to S5). These clinical data provide strong evidence
   for intracellular survival of S. aureus in osteomyelitis, which
   indicates the necessity and clinical value of developing intracellular
   bacteria-targeted therapeutic strategies. Besides, we have also
   evidenced the presence of S. aureus in bone cells of mice femur in vivo
   using histological analysis and bio-TEM ([185]Figs. 6 and [186]7), and
   found that the administration of NKMs can robustly suppress the
   infections in both primary mouse osteomyelitis and bacterial
   rechallenge settings. NKM treatment reduces both intracellular
   bacterial load and biofilm burden, thus restoring the homoeostasis and
   alleviating the tissue damage by the bacteria. Additionally, we found
   that NKM treatment can largely alleviate the formation of
   staphylococcal abscess communities and eliminate the bacteria invaded
   in the OLCN of bone, proving the high efficacy of NKM therapy in
   eliminating bacteria in the osteomyelitis model.

   The intracellular location is known to be a heaven for S. aureus to
   evade and compromise host immune system ([187]83). Therefore,
   alternative techniques of stimulating host innate immune responses
   would be useful in the treatment and prevention of intracellular S.
   aureus infection. We discovered that NKMs evoke antibacterial immunity
   by activating CD8^+ T cells and promoting the proliferations of NK
   cells ([188]Fig. 6). CD8^+ T cells and NK cells are both responsible
   for detecting infected cells and releasing cytotoxic granule proteins
   through the immune synapse between the immune cell and the infected
   cell ([189]53, [190]77). This result implies that NKMs not only can
   directly destroy intracellular bacteria but also evoke the
   antibacterial immunity of the host cells. We have also found that NKMs
   prevent the infection recurrence by evoking bacteria-specific T cell
   immunological memory responses ([191]Fig. 7), implying the
   translational potential of NKMs for preventing osteomyelitis relapse.

   We herein have tested the antibacterial efficacy of NKMs against
   intracellular S. aureus infection in the current study, while other
   kinds of prevalent human infections beyond S. aureus, which also rely
   on the similar intracellular protections like S. aureus in human
   bodies, remain to be explored in depth. Nevertheless, the present NKMs
   offer a paradigm of NK cell–mimicking platform to treat various
   intracellular infections.

MATERIALS AND METHODS

Study design

   The study was designed to validate an NK cell mimic to eliminate
   intracellular bacteria and activate the host’s immune response system
   for the treatment of osteomyelitis infections. In vitro
   characterization of formulated NKMs revealed the biological activity of
   the perforin and granzyme after encapsulation and released from the
   NKMs. Electrochemical experiments, enzyme-linked immunosorbent assay,
   bio-TEM, SAXS, and SEM micrographs were used to study the mechanism and
   behavior of NKMs during S. aureus–infected osteoblast-targeted
   degradation, perforation, and bacterial sterilization. In vivo
   antimicrobial efficacy of NKMs was evaluated using a previously
   established model of S. aureus–induced titanium implant osteomyelitis
   infection. To measure the efficacy of bacterial clearance after
   perforin and granzyme released from the NKM system, we used mCherry–S.
   aureus and in vivo imaging to monitor infection progression. The
   immunomodulatory efficacy of NKMs was assessed by flow cytometry
   analysis of the infected bone marrows. In addition, the antimicrobial
   efficacy of NKMs was determined by CFU counting, TEM, SEM, and
   histopathologic analysis of the infected tissues and implants. Finally,
   the translational potential of NKMs was assessed by analyzing the
   biosafety of the nanoparticles in the organs of interest. All clinical
   patient samples were obtained and used in accordance with the approval
   of the Ethics Committee of the Sixth People’s Hospital, Shanghai Jiao
   Tong University School of Medicine [approval no. 2023-KY-017(K)]. All
   animal studies conformed to the guidelines by the Animal Care Ethics
   Commission of Shanghai Sixth People’s Hospital, Shanghai Jiao Tong
   University School of Medicine (ethics approval number: DWLL2023-0469).

   The time point of infection inoculation, treatment group size, and
   endpoints were selected based on previous experience with the
   corresponding animal models. Animals were randomly assigned to
   treatment and time point groups and given numerical identifiers.
   Experimental group size (n) is indicated in all figure legends. All
   experiments involving animals and biological samples were approved by
   the Animal Care Center and Use Committee of the Shanghai Sixth People’s
   Hospital.

Statistical analysis

   All results are presented as mean ± SD unless otherwise indicated. The
   significance between two groups was evaluated by two-tailed Student’s t
   test. One-way analysis of variance (ANOVA) with Tukey’s post hoc test
   was used for multiple comparisons. Statistical analysis was evaluated
   using GraphPad Prism 6.0. P values of less than 0.05 were considered
   significant. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Acknowledgments