Redox-Triggered Release of Moxifloxacin from Mesoporous Silica Nanoparticles Functionalized with Disulfide Snap-Tops Enhances Efficacy Against Pneumonic Tularemia in Mice

Effective and rapid treatment of tularemia is needed to reduce morbidity and mortality of this potentially fatal infectious disease. The etiologic agent, Francisella tularensis, is a facultative intracellular bacterial pathogen which infects and multiplies to high numbers in macrophages. Nanotherapeutics are particularly promising for treatment of infectious diseases caused by intracellular pathogens, whose primary host cells are macrophages, because nanoparticles preferentially target and are avidly internalized by macrophages. A mesoporous silica nanoparticle (MSN) has been developed functionalized with disulfide snap-tops that has high drug loading and selectively releases drug intracellularly in response to the redox potential. These nanoparticles, when loaded with Hoechst fluorescent dye, release their cargo exclusively intracellularly and stain the nuclei of macrophages. The MSNs loaded with moxifloxacin kill F. tularensis in macrophages in a dose-dependent fashion. In a mouse model of lethal pneumonic tularemia, MSNs loaded with moxifloxacin prevent weight loss, illness, and death, markedly reduce the burden of F. tularensis in the lung, liver, and spleen, and are significantly more efficacious than an equivalent amount of free drug. An important proof-of-principle for the potential therapeutic use of a novel nanoparticle drug delivery platform for the treatment of infectious diseases is provided.


Introduction
Francisella tularensis is a highly infectious bacterium that causes a life threatening disease, tularemia. Inhalation of as few as 25 bacteria is sufficient to cause severe illness. [1] Its extremely high infectivity, ease of dissemination by the air borne route, and capacity to cause severe disease motivated its development as a biological weapon by Japan during the second World War [2] and by both the U.S. and the former Soviet Union during the cold war. [3] Although effective antibiotics for treatment of tularemia are available, intensive care is frequently required, relapse and complications are frequent, and the infection can be fatal even with appropriate treatment. Concern over its potential for use as a biological weapon has led to its federal classification as a Tier 1 Select Agent. It has been estimated that deliberate dispersal of F. tularensis over a large city would overwhelm health care facilities and result in thousands of deaths. [4] Development of more effective treatment for tularemia has the potential to reduce the number of patients requiring intensive care and to reduce the duration that such care is required.
Because F. tularensis causes disease primarily by replicating intracellularly within host macrophages, [5] a delivery strategy that targets macrophages and delivers high concentrations of antibiotic to the macrophages has the potential to provide more effective treatment.
After systemic administration, nanoparticles are avidly taken up by macrophages of the mononuclear phagocyte system in the lung, liver, and spleen. [6][7][8] Because these are the cells infected by F. tularensis, a nanoparticle delivery system has the potential to deliver high concentrations of antibiotic to the site of infection while minimizing systemic exposure.
Nanoparticles also have several other advantages over free drug, including shielding the drug from metabolism and excretion and providing more favorable pharmacokinetics. While several different nanoparticle delivery platforms have been studied for antibiotic delivery, including 5 liposomes, solid lipid particles, poly-L-lactide (PLGA), and biological materials such as gelatin, chitosan, and alginates, [9,10] mesoporous silica nanoparticles (MSNs) offer several important advantages, including structural and chemical stability, uniformity, inherent lack of toxicity, capacity to encapsulate exceptionally high concentrations of different types of cargo, and versatility in incorporating rational design features, including stimulus responsive drug release systems. [11][12][13][14][15][16][17] In this work, we have developed a stimulus-responsive MSN platform for treatment of tularemia that delivers the antibiotic moxifloxacin (MXF) intracellularly in response to the intracellular redox potential.
Living cells have more reducing power than extracellular medium or plasma because of numerous redox couples that are kept primarily in the reduced state by metabolic processes such as glycolysis, mitochondrial electron transport, and the pentose phosphate pathway. These redox couples include NADH/NAD; NADPH/NADP; thioredoxin/oxidized-thioredoxin, cysteine/cystine, and glutathione (GSH)/GSSG, with the latter redox couple being quantitatively the most abundant inside cells, with cytosolic GSH concentrations in the 1 -10 mM range. [18] Extracellularly, in culture medium and in plasma, the cysteine/cystine redox couple is quantitatively the most important. Disulfide snap-top MSNs release cargo selectively intracellularly because the redox potential is much lower in the intracellular than in the extracellular environment. [19,20] On the basis of the intracellular glutathione/glutathione disulfide ratio, the redox potential is estimated to range from -250 mV in rapidly dividing cells to -200 mV in differentiating cells to -160 mV in cells undergoing apoptosis. [21] Different compartments within the cell also maintain different ambient potentials; for example, based on the thioredoxin redox poise, the cytoplasm, nucleus, and mitochondria exhibit redox potentials of -280, -300, and -340 mV, respectively. [20] On the other hand, the GSH/GSSG redox couple in plasma has a redox 6 potential of -140 mV [22] and the much more abundant cysteine-cystine is even more oxidized, with a redox potential of -80 mV. [23] A similar situation is replicated in cell culture model systems, as human cell lines regulate the redox state of the cysteine-cystine couple in their culture medium to approximately -80 mV. [24] Prior to addition to cultured cells, cysteine-free RPMI-1640 has a relatively high redox potential of -37 mV and RPMI supplemented with 0.45 mM cysteine has a redox potential of -182 mV.
Disulfide snap-top MSNs utilizing cyclodextrin as the cap have been reported previously. [19,25,[27][28][29] The original version involved a rotaxane where α-cyclodextrin (α-CD) threaded on a stalk was held in place by a bulky stopper on the end of the disulfide-containing stalk. [19] Subsequent versions used direct covalent bonding of β-cyclodextrin (β-CD) to a stalk [27] and noncovalent bonding of β-CD to adamantine at the end of stalks of different lengths. [25,28,29] In this paper the functionality of the snap-top with the shortest (propyl group) stalk is discussed.
Although streptomycin and aminoglycosides are historically considered the treatment of choice for tularemia, they cross membranes poorly, have relatively high minimum inhibitory levels against F. tularensis, have side effects of ototoxicity and nephrotoxicity, and are difficult to administer. Doxycycline or ciprofloxacin are recommended for post-exposure treatment in a mass casualty setting. [3] In contrast to aminoglycosides, fluoroquinolones cross membranes readily and have much lower minimal inhibitory concentrations against F. tularensis.
Ciprofloxacin has been used successfully both in animal models of tularemia [30] and in the treatment of clinical tularemia infections. [31] In a mouse model of pneumonic tularemia comparing ciprofloxacin, gatifloxacin, and MXF, while all three fluorquinolones showed efficacy during the treatment phase, both MXF and gatifloxacin were superior to ciprofloxacin in preventing relapse, indicating greater efficacy in eradicating the F. tularensis. [32] Because of its 7 potent antimicrobial activity against F. tularensis as well as potent activity against many other important intracellular human pathogens, including Mycobacterium tuberculosis, [33] Listeria monocytogenes, [34] Mycoplasma, Chlamydia, Shigella, and Salmonella, we developed our redoxresponsive disulfide snap-top MSNs (MSN-SS-MXF) for delivery of MXF.
In this study, we demonstrate that our MSN-SS-MXF delivery platform releases its antibiotic cargo intracellularly in macrophages, is effective in killing F. tularensis in infected macrophages in a cell culture model, and is a much more effective treatment than an equivalent amount of free drug in a mouse model of pneumonic tularemia. 8

Synthesis of Disulfide Snap-top MSNs
To utilize MSNs to deliver MXF into macrophages and release the drug intracellularly in a controlled fashion, we developed a disulfide snap-top attached to the surface of the MSN so as to trap drug inside mesopores. The synthesis procedure is illustrated in Figure 1. A silane stalk (3mercaptopropyl) trimethoxysilane was attached on the surface of MSN first and then 1adamantanethiol reacted with the silane linker in the presence of the oxidant thiocyanogen to form a disulfide bond ( Figure S1, Supporting Information). The modified MSNs preserve mesoporous structures after all surface modification and surfactant template extraction procedures ( Figure S2), and the particles exhibit a mean diameter of 90 nm by DLS measurement in H2O. Disulfide modified MSN was then mixed with MXF PBS solution for 24 h, followed by adding β-CD as the capping molecule which formed a stable complex with the adamantyl group.
In reducing environments (e.g. after addition of glutathione or after uptake by macrophages), the disulfide bond is cleaved and cargo is released. It's possible that in the intracellular environment competitive binding for the β-CD may also occur and contribute to cargo release. The strong binding affinity between the adamantyl group and β-CD ensure that cargo is trapped inside the pores and prevents premature leakage before reaching target cells.
An alternative method reported by others for constructing redox-sensitive valves is to attach an admantyl group covalently to the particle surface . [27] In that case, one more chemical reaction is required, after drug loading, to form the amide bond. This extra synthetic step requires that the drug-loaded particles (without capping) be suspended in solution in order to attach the caps. During this step, drugs can diffuse out of the pores and catalyst molecules can diffuse into the pores and contaminate the cargo. 9 The snap-top used in this paper contains a short (propyl group) linker in order to hold the β-CD caps close to the pore openings and inhibit leakage. The chemicals used in the synthesis procedure, illustrated in Figure 1, are both commercially available. MXF is a fourth generation fluoroquinolone active against both Gram-positive and Gramnegative bacteria. It has a UV-Vis maximum absorption peak at 288 nm in PBS allowing spectroscopic measurement of its concentration. We measured the absorbance of MXF in solution before and after loading the nanoparticles and used the difference in concentration to calculate the amount of MXF taken up by the particles (including inside pore channels and on external surfaces). The mass of MXF taken up by particles divided by the mass of MSNs is defined as "uptake capacity" (expressed in wt%). After washing mechanized MSN with PBS sufficiently to remove MXF from the outer surface, the nanoparticles were dispersed in deionized water or PBS and then an excess amount of 2-mercaptoethanol or glutathione was added to cleave the disulfide bond and release the drug ( Figure 2). The mass of released MXF divided by the mass of the particle is defined as "release capacity" (expressed in wt%).
Moreover, the flat baseline before adding the reducing agent indicates that no premature release occurs and that there is strong binding between the adamantyl group and β-CD.

Optimization of Uptake and Release Capacity
Release capacity of a nanoparticle delivery system is an important factor that impacts in vivo efficacy, as a higher release capacity allows a greater amount of drug to be delivered to target cells with the same number of MSNs. We exploited charge interactions between the cargo molecules and the MSN inner pores to achieve a high uptake and release capacity. MXF has two ionizable groups with pKa's of 6.3 and 9.3, and the extent to which the drug is positively charged, neutral, or negatively charged is pH-dependent. Hence, the pH of the loading solution markedly impacts uptake capacity. In PBS buffer with pH 7.4, 87.8% of MXF molecules are zwitterionic species, 7.3% molecules are positively charged, and 4.8% are negatively charged.
We modified the MSNs with either amine groups or phosphonate groups to make both the inner pore and outer particle surface positively or negatively charged, respectively. Positively charged cargo interacts electrostatically with negatively charged pores, thereby increasing the uptake capacity; however, strong electrostatic interaction between cargo molecules and pore channels may also slow the rate of cargo release. [35] On the other hand, positively charged pores electrostatically repulse the positively charged cargo molecules, thereby decreasing the uptake capacity but facilitating and increasing the rate of cargo release.
Before attaching snap-top caps, we measured the uptake capacity of MSNs with different pore charges and found that with positively charged mesopores the uptake capacity was near zero, indicating that it is too difficult for MXF molecules with a positive net charge to diffuse into positively charged MSN channels. Use of negatively charged pores dramatically increased the uptake capacity to 30 wt% and the release capacity to 3 wt% (10 mM MXF in a volume of 1 mL PBS) (Table S1A, Supporting Information). Other experiments showed that a further increase in negative charge on pores does not improve uptake and release capacity. Pore modification was achieved by co-condensation of two silanes, in which diethylphosphatoethyltriethoxysilane (DEPETS) was mixed with tetraethyl orthosilicate (TEOS) and then added to heated base solution in a dropwise fashion. Different amounts of DEPETS (10 μL, 25 μL and 35 μL) were mixed with TEOS (60 μL) to make more negatively charged pores, and these nanoparticles showed similar release capacity of ~2-3 wt% under the same loading conditions (Table S1B, Supporting Information). This result suggested that the amount of phosphonate groups inside the pores is saturated and hence the attraction of positively charged MXF molecules is maximized.
We also tested loading MSN-SS in solutions of different pH because in acidic solutions, most of MXF molecules are positively charged and interact with negatively charged inner pores, resulting in a higher uptake capacity. However, lowering pH may also render phosphonate groups on inner pores partially protonated and thus less negatively charged, resulting in a lower uptake capacity. Experiments showed that loading with pH 3 MXF solution (1 mL 10 mM) 13 resulted in 9.6 wt% uptake capacity, which is much lower than the 22.2 wt% uptake capacity obtained when loading with pH 7.4 MXF solution (1 mL 10 mM). The enhanced uptake capacity at pH 7.4 is due to more negatively charged mesopores at this pH (Table S1C, Supporting   Information).
Moreover, we compared the uptake capacity of MSNs (10 mg) with 10 μmol disulfide stalk surface coverage with that of MSNs with 20 μmol surface coverage. We hypothesized that the higher surface coverage would cap more MXF molecules inside the pores. However, we obtained uptake capacities of 22.2 wt% and 19.7 wt% with surface coverage of 10 μmol and 20 μmol, respectively, which indicated that higher surface coverage with the silane stalks may increase the surface hydrophobicity of MSNs and lower the uptake of the hydrophilic drug MXF (Table S1D, Supporting Information). Therefore, 10 μmol disulfide stalk surface coverage provided a satisfactory balance between hydrophobicity and capping MXF within pores so as to achieve high uptake.
To obtain a higher uptake and release capacity, we loaded the same amount of MSN-SS with a more concentrated MXF PBS solution (40 mM MXF in a volume of 1mL PBS vs. 10 mM MXF in a volume of 1mL PBS). This yielded an uptake and release capacity of 135 wt% and 51 wt%, respectively, the highest release capacity yet obtained (Table S1E,

Measurement of MSN-SS-MXF Release Capacity
We used a F. novicida bioassay ( Figure S3, Supporting Information) to determine the maximum amount of MXF released from particles. With this assay we measured the amount of drug released from MSN-SS-MXF in PBS or DMSO with and without 2-mercaptoethanol by determining the inhibition of F. novicida growth in broth cultures. We measured a release capacity for MSN-SS-MXF of 12 wt% in PBS and a total of 18 wt% after addition of reductant.
When MSN-SS-MXF was dispersed in DMSO with 2-mercaptoethanol, the bioassay measurement showed a release capacity of 51 wt%. The higher release capacity in DMSO with 2-mercaptoethanol indicates that not all MXF molecules were released from mesopores in PBS.
In comparison, the highest release capacity obtained from using the pH-sensitive nanovalve was around 8 wt%. [11]

Disulfide Snap-Top MSNs Release Cargo at Physiological GSH Concentrations
Quantitatively, GSH is the major reducing agent in cells, with intracellular concentrations of approximately 10 mM in healthy cells. [36,37] To determine whether the disulfide snap-tops   Based on the drug release capacity of 27.4 wt%, the impact of MXF delivered by various doses of the MSN-SS-MXF in killing of F. tularensis LVS was compared with that of free drug using a median-effect plot. [38] As shown in Figure 5D,

MSN-SS-MXF is Much More Efficacious than an Equivalent Amount of Free MXF in a Mouse Model of Pneumonic Tularemia
We assessed the efficacy of the MSN-SS-MXF in a mouse model of pneumonic tularemia established previously for evaluation of vaccine candidates. [39][40][41] In the first of two experiments (Experiment 1), mice were infected by the intranasal route (i.n.) with 4000 CFU of    Figure 8B). In both cases, the silica of the MSNs is found predominantly in the lung, liver, and spleen, the same three organs that are preferentially targeted by F. tularensis.
Organs from infected mice that received repeated i.v. injections of PBS were also subjected to ICP-OES analysis. The amount of silica found in these control organs was negligible, indicating a very low background silica level in the organs ( Figure 8C).

Discussion
Numerous serious human infections, including those caused by Mycobacterium tuberculosis, Salmonella, Brucella, Legionella pneumophila, and F. tularensis, are caused by microbes that replicate intracellularly in macrophages of the mononuclear phagocyte system. These pathogens exploit the intra-macrophage niche as a source of nutrients and a shelter against host defenses.
The macrophage can also pose an obstacle to conventionally administered antibiotics that must cross its plasma membrane and often additional intracellular membranes enclosing the pathogen.
Because nanoparticles are preferentially internalized by macrophages of the mononuclear phagocyte system, they are attractive as a drug delivery platform for infections cause by these pathogens. A nanoparticle delivery platform that releases drug exclusively intracellularly has the potential to release high concentrations of drug into infected cells, thus providing for a greater killing efficacy relative to free drug and at the same time limiting systemic exposure to the drug and off-target toxicities. The nanoparticle delivery platform also has the potential to improve the pharmacokinetic profile of the drug by shielding it from excretion and metabolism before it reaches its target cells. Key to the success of such a nanoparticle delivery system is a disulfide snap-top mechanism that releases the drug cargo only after uptake of the nanoparticle into the host cell. Several different mechanisms have been developed to provide for autonomously controlled release of drug cargo from mechanized nanoparticles in response to the intracellular environment, including pH, competitive binding, enzymatic activation, and redox potential. [17,19,35,[42][43][44][45][46][47][48] Each system has unique chemistry and must be optimized for its drug cargo to achieve maximum loading and controlled release. In the case of the important antibiotic MXF, we have demonstrated that we can achieve very high loading and controlled intracellular release at physiological GSH levels using MSNs functionalized with disulfide snap-tops. Indeed, loading 28 and release of MXF into MSN-SS-MXF were far superior (release capacity 5-fold greater) to that achieved in a previous study in which we utilized pH-gated MSNs. [11] Consistent with this, the MSN-SS-MXF were more efficacious than the MXF-loaded pH-gated MSNs in treating pneumonic tularemia in mice, as evidenced by higher efficacy ratios vs. free drug in the lung, spleen, and liver. [11] MSNs taken up by macrophages will enter the endosomal-lysosomal pathway, which may have a lower concentration of GSH than the cytosol. In addition, γinterferon (often elevated in infections) has been shown to lower GSH levels in macrophages. [49] However, lysosomes have a powerful γ-interferon-inducible lysosomal thiol reductase (GILT) [50] capable of cleaving disulfide linkages, including those present in β-CD-based polyrotaxanes therapeutics for lysosomal storage disease. [51] Modification of the mesopores with phosphonate groups has allowed us to increase the loading and release capacity of our MSNs, and functionalization of the MSNs with a disulfidecleavable capping system provides for very tight closure of the mesopores under extracellular conditions, preventing premature release of drug cargo, yet allows for ready opening of mesopores and release of mesopore-bound drug cargo in response to the intracellular environment. While redox-responsive disulfide gate mechanisms have been described, [19,25] they have not previously been tested in vitro or in vivo for safety or efficacy in the delivery of an antibiotic for treatment of an intracellular pathogen. Ma et al. used a similar cap and thread system for delivery of doxorubicin by disulfide snap-top MSNs in a cell culture system and in zebra fish, [29] although we have used a different synthetic route for attaching the adamantane.
Most of the previously reported MSN disulfide-snap-tops have used a different chemistry for their redox sensitive gates. [52][53][54][55][56][57][58][59][60][61][62] 29 We have shown that our disulfide snap-top MSN loaded with MXF is safe and well tolerated in vitro and in vivo. Importantly, we demonstrated the successful treatment of a serious infectious disease, pneumonic tularemia, using the MSN-SS-MXF. In our cell culture model, the MSN-delivered MXF showed efficacy equivalent to that of free MXF. In contrast, in our in vivo mouse model of pneumonic tularemia, the MSN-delivered MXF was three to five times more efficacious than free drug. The difference in efficacy ratios for our in vitro vs. in vivo models likely reflects the fact that with the in vitro model, the macrophages in cell culture wells are exposed to a constant concentration of drug over the course of the experiment whether it is released from the MSNs or administered as free drug. In contrast, in the mouse model of pneumonic tularemia, the efficacy of the MXF administered as free drug is reduced because it is subject to metabolism and excretion and there is no preferential targeting of free drug to tissues or cells that are infected by F. tularensis. Hence, the MSN-delivered MXF can achieve higher levels in the infected tissues and host cells than free MXF. Indeed our ICP-OES analysis demonstrated preferential uptake of the MSN by lung, liver, and spleen, which are the main tissues infected by F. tularensis. In addition, because MSN-encapsulated drug is shielded from metabolism and excretion, it is likely to have a more favorable Area Under the Curve/Minimal Inhibitory Concentration (AUC/MIC) ratio compared with free drug. The 3-to 5-fold enhanced efficacy of MSN-SS-MXF compared with free drug serves as proof-of-principle that this platform has potential to provide more effective treatment for tularemia as well as other important infections caused by bacteria that multiply intracellularly in macrophages. The average efficacy ratio of MSN-SS-MXF in the lung vs. free moxifloxacin of ~4:1 was superior to that of our previously described pH-gated nanoparticle (MSN-MBI-MXF) which had an average efficacy ratio in the lung of ~3:1 vs. free moxifloxacin. With our current design, the MSNs 30 passively target infected macrophages, but it is likely that even greater enhancement of therapeutic efficacy can be achieved by surface modifications (e.g. targeting to specific cellular receptors) that further enhance targeting to infected tissues and uptake by macrophages or by use of an aerosol delivery device that delivers the MSNs directly to the lung, as has recently been demonstrated for liposomally encapsulated ciprofloxacin in treatment of tularemia. [63] 31

Conclusions
We have developed a redox-responsive disulfide snap-top MSN-drug delivery platform that achieves high uptake and release capacity for the broad spectrum antibiotic MXF by optimizing the MSN inner pore charges, the loading pH, stalk surface coverage density, and the loading concentration of MXF. These MXF-loaded disulfide snap-top MSNs are taken up avidly by F.

tularensis-infected human macrophages and kill the bacterial pathogen in macrophages in vitro.
Most importantly, these nanoparticles are much more effective than an equivalent amount of free drug in treating lethal pneumonic tularemia in a mouse model. Our study demonstrates the utility of nanotherapeutics in treating serious and potentially fatal infectious diseases caused by intracellular pathogens.  Fraction of inhibition = 1 -(log CFU from sample treated with a known concentration of MXF or releasable MXF from MSN-SS-MXF / log CFU from untreated sample). A median-effect plot [38] for MXF or MSN-SS-MXF was generated using MXF or MXF equivalent (MSN) dose in base-10 logarithm as the X-axis and the fraction of surviving bacteria divided by the fraction of killed bacteria in base-10 logarithm as the Y-axis.

Materials and Reagents
Statistics: Statistical analyses were performed using GraphPad Prism software (version 5.01) and R 3.2. [64] Means were compared across groups using one way analysis of variance ( Study approval: All experiments with mice were conducted within the guidelines and according to the protocol approved by the UCLA Institutional Animal Care and Use Committee.  Figure S1. Adamantyl group attachment was confirmed by 13 C-CPMS NMR spectroscopy