Comparing Strategies for Magnetic Functionalization of Microbubbles

: The advancement of ultrasound-mediated therapy has stimulated the development of drug-loaded microbubble agents that can be targeted to a region of interest through an applied magnetic ﬁ eld prior to ultrasound activation. However, the need to incorporate therapeutic molecules while optimizing the responsiveness to both magnetic and acoustic ﬁ elds and maintaining adequate stability poses a considerable challenge for microbubble synthesis. The aim of this study was to evaluate three di ﬀ erent methods for incorporating iron oxide nanoparticles (IONPs) into phospholipid-coated microbubbles using (1) hydrophobic IONPs within an oil layer below the microbubble shell, (2) phospholipid-stabilized IONPs within the shell, or (3) hydrophilic IONPs noncovalently bound to the surface of the microbubble. All microbubbles exhibited similar acoustic response at both 1 and 7 MHz. The half-life of the microbubbles was more than doubled by the addition of IONPs by using both surface and phospholipid-mediated loading methods, provided the lipid used to coat the IONPs was the same as that constituting the microbubble shell. The highest loading of IONPs per microbubble was also achieved with the surface loading method, and these microbubbles were the most responsive to an applied magnetic ﬁ eld, showing a 3-fold increase in the number of retained microbubbles compared to other groups. For the purpose of drug delivery, surface loading of IONPs could restrict the attachment of hydrophilic drugs to the microbubble shell, but hydrophobic drugs could still be incorporated. In contrast, although the incorporation of phospholipid IONPs produced more weakly magnetic microbubbles, it would not interfere with hydrophilic drug loading on the surface of the microbubble. evaluate them for the purpose of diagnostic ultrasound imaging. The iron content of MBs was measured using inductively coupled plasma optical emission spec-troscopy (ICP − determine the quantity of IONPs still into the MBs after cleaning of the suspension. The of ﬂ ow under the in ﬂ uence of a ﬁ eld was also measured. the toxicity of examined by measuring the levels of hemolysis recorded in horse blood following incubation with MBs at 37 ° C. supernatants was measured in triplicate (120 μ L per well, in a Greiner UV-Star 96-well plate) at 541 nm. This procedure was performed twice. Statistical Analysis. Statistical analyses were performed using Microsoft Excel and GraphPad Prism with signi ﬁ cance de ﬁ ned as p < 0.05. One-way analysis of variance was used for experiments investigating the di ﬀ erence between more than two groups with post-hoc Tukey honestly signi ﬁ cant di ﬀ erence test. Data is presented as mean ± standard deviation.


■ INTRODUCTION
Gas microbubbles (MBs) stabilized by a surfactant or a polymer shell were initially developed as contrast agents to enhance ultrasound imaging of the vasculature. 1 The highly nonlinear response of the MBs to moderate ultrasound pressures produces acoustic radiation over a range of frequencies that can be readily distinguished from those due to the surrounding tissue. 2 At higher pressures, the large amplitude oscillations exhibited by the MBs can produce substantial biological effects, 3−7 and these are being actively investigated for their potential application in ultrasoundmediated therapy.
A further complementary application of MBs is as therapeutic carriers of highly potent drugs. By encapsulating drugs within a MB and releasing them "on demand" using ultrasound, the risk of toxic side effects can potentially be reduced. Numerous formulations have been proposed to facilitate the loading of drugs within the MB coating and/or on its surface in the form of liposomes, nanoparticles, or single molecules. 8 Extending the functionality of MBs, however, generates a series of potentially conflicting requirements for their design. In particular, MBs must have adequate stability both in storage and in circulation. They must be sufficiently responsive to ultrasound to promote therapy and ideal to be imaged during treatment. They must be able to carry sufficient quantities of drug to be effective; and it is also desirable to be able to target them to specific sites, for example, tumor endothelium.
With respect to the last of these requirements, there are a number of different approaches that can be applied either alone or in combination. The attachment of targeting species, such as antibodies, to the MB surface enables specific binding to target regions. 9 The short half-life of MBs in circulation, however, can hinder the efficiency of this approach. To address this, acoustic radiation force has been used to concentrate and slow down the MBs at a site of interest 10 to increase binding. 9 A further possibility is to incorporate magnetic material into the MB, enabling localization at a target site using an externally applied magnetic field. The aim of this study was to investigate different methods previously reported for incorporating iron oxide nanoparticles (IONPs) and compare them to new formulations. The different types of magnetic MBs were assessed in terms of their stability, acoustic response, and targeting efficiency.
Multiple formulations have been presented in the literature to produce magnetic MBs stabilized by polymers, surfactants, and lipids. 11,12 In the case of polymer-coated MBs, IONPs are typically embedded within the MB shell during manufacture. 13−18 However, several studies have also investigated the performance of MBs with IONPs conjugated to their outer surface. This external loading approach was found to be advantageous in terms of improving the magnetic properties while preserving the acoustic response and also improving the cellular uptake. 19−22 Whereas polymer MBs provide good physical stability, 23 cross-linked polymers are more resistant to compression and expansion, thus producing less echogenic MBs. 24 Most commercial contrast agents use phospholipids and lipid-mix systems, which offer excellent echogenicity, biocompatibility, and circulatory stability. 24,25 The spontaneous self-assembly of lipids into monolayers at interfaces, with hydrophobic tails facing the gas phase, facilitates MB formation and manufacturing and promotes physical stability. 24,26 Previous studies on magnetic lipid-coated MBs have incorporated hydrophobic IONPs on the internal surface of the shell enabling the targeted delivery of genes or thrombolytic drugs, 27 interactions. MBs coated with ionic surfactants and charged IONPs have also been reported, 37−39 but the toxicity concerns related to these systems limit their clinical feasibility. In this study, the characteristics of phospholipid MBs loaded with IONPs using different methods are compared, and the implications for drug-delivery applications are discussed.

■ EXPERIMENTAL SECTION
Materials and Methods. MB Formulations. Five different MB formulations were tested. The first formulation (OilMB) consisted of phospholipid MBs loaded with hydrophobic IONPs suspended in an isoparaffin oil phase that partitions into the bubbles upon sonication. The OilMB method, previously presented, 27 is similar to the acoustically activated lipospheres reported by others, where a layer of soybean oil is embedded in the bubbles to act as a reservoir for a hydrophobic drug. 40,41 The IONPs used for this formulation have a diameter of 10 nm and are suspended in isoparaffin. The ferrite particles are reported to take approximately 10% by volume of the solution, therefore providing a magnetic flux density of 400 gauss, according to the manufacturer, corresponding to a saturation magnetization of 6.3 emu/g. Two further hydrophobic IONP formulations (LipMB-DBPC and LipMB-LαPC, collectively referred to as LipMB hereafter) were prepared using phospholipid-coated IONPs directly incorporated into the bubble coating. LipMB-LαPC has been previously described, 42 and LipMB-DBPC was chosen to test the impact of the IONP lipid coating on magnetic MBs. The method relies on the phospholipid coating to transfer the IONPs to the MB shell upon sonication and allows loading of hydrophobic IONPs into MBs without introducing an oil phase. The lipid-coated IONPs had a hydrodynamic diameter of 50 nm with a crystal core of 30 nm made of magnetite.
In the fourth formulation (SurfMB), hydrophilic dextran IONPs were biotinylated by reacting the carboxylic acids on the Concentration measured immediately after production, prior to cleaning. b Percentage difference in the B-mode image intensity of a channel with flowing MBs ± the application of a magnetic field. c Obtained by fitting an exponential curve to the measured change in the MB concentration over time at 37°C.

ACS Applied Materials & Interfaces
Research Article carboxymethyl dextran coating with biotin−PEG−NH 2 and attached to the surface of biotinylated MBs via an avidin bridge. The motivation behind this method was to produce IONPs with a surface chemistry similar to that of the clinical product Feraheme, 43 which is made of IONPs coated with a carboxymethyl dextran sorbitol ether. The carboxylic acids on Feraheme could be used as handles for biotinylation to enable loading onto the external surface of biotinylated MBs. The dextran IONPs used for this formulation had a hydrodynamic diameter of 50 nm and crystal core of 30 nm made of magnetite. The fifth formulation represented the control population and consisted of phospholipid MBs without IONPs (unloaded).
The properties of the different MB formulations are summarized in Table 1.
The different IONPs were chosen to have a hydrodynamic diameter of 50 nm for all formulations except for OilMB, which used IONPs of 10 nm in diameter in isoparaffin to allow a comparison with the method used earlier. 27 Surfactants were used to promote the formation of phospholipid MBs. To ensure the biocompatibility of the final product, they were chosen from the ingredients of clinically approved MBs. 44,45 The different magnetic MBs shown in Figure 1 were compared in terms of their physical stability and acoustic and magnetic responses. The physical stability of MBs was quantified in terms of their concentration decay over time at 37°C and under flow. It should be noted, however, that this method does not reflect all considerations for in vivo stability as it does not incorporate clearance mechanisms found in living organisms. Langmuir trough measurements were used to measure the mechanical properties of the lipid monolayer systems at a gas/water interface corresponding to the MB shell. 46 In particular, these measurements were conducted to capture the changes in the surface tension of lipid-mix monolayers upon compression and expansion, and the findings were compared with the MB physical stability results. The acoustic properties of the different MB formulations were assessed at both 1 MHz, typical of therapeutic ultrasound applications, and 7 MHz to evaluate them for the purpose of diagnostic ultrasound imaging. The iron content of MBs was measured using inductively coupled plasma optical emission spectroscopy (ICP−OES) to determine the quantity of IONPs still incorporated into the MBs after cleaning of the suspension. The quantity of MBs retained under flow under the influence of a magnetic field was also measured. Finally, the toxicity of the formulations was examined by measuring the levels of hemolysis recorded in horse blood following incubation with MBs at 37°C.
To assess the hemolytic potential of the different formulations, horse blood in acid-citrate-phosphate was used. A 100 mL sample was purchased (ref: HB046, Lot 34701900) from TCS Biosciences Ltd. (Botolph Claydon, Buckingham, United Kingdom). Positive controls were achieved by adding a solution of Triton X-100 purchased from Sigma-Aldrich Ltd. (Gillingham, Dorset, UK), and they were diluted in deionized water.
MB Manufacture. General Protocol. A mixture of lipids and surfactants dissolved in chloroform was added to a glass vial to produce a 5 mL batch of MBs at a total concentration 4 mg/mL. The sample was covered with a pierced parafilm and set on a hot plate at 50°C for 12 h to evaporate the chloroform. Once all of the solvent evaporated, the dried lipid film was suspended in 5 mL of Dulbecco's PBS for 1 h at 80°C under constant magnetic stirring. The magnetic stir bar was then removed.
The lipid mix solution was sonicated at a low intensity (QSonica Q125, 20 kHz, 3 mm probe tip, amplitude: 20%, 1 min) with the sonicator probe tip immersed in the solution. The sonicator probe tip was moved to touch the air and water interface, and a light flow of SF 6 gas was added to fill the headspace of the sample vial. The sonicator was then turned on at high intensity (amplitude: 80%, 20 s). The samples were capped and cooled on ice for 10 min after which a layer of foam was visible at the top of the sample as well as a thick layer of densely packed MBs underneath the foam.
Cleaning of MBs was done through centrifugation using 10 mL syringes modified to fit into 50 mL centrifuge tubes. The sample was aspired into the syringe using a blunt 18-G needle. The syringe was capped, placed into a 50 mL centrifuge tube, and centrifuged at 100g for 5 min at 4°C. The subnatant was removed, the bubbles were resuspended in the original volume of PBS at 4°C, and the washing procedure was repeated three times.
Preparation of Unloaded MBs. Unloaded MBs were prepared following the general protocol using a 9:1 molar ratio mixture of DBPC/PEG40S to prepare the lipid mix.
Preparation of LipMB (Phospholipid-Mediated Loading). To make magnetic MBs through phospholipid-mediated loading, the lipid mix was prepared according to the protocol for unloaded MBs. A volume of LαPC or DBPC IONPs corresponding to 3.75 mg of iron oxide was added to the lipid mix between the low-intensity and highintensity sonication. Cleaning was carried out according to the general protocol.
Preparation of OilMB (Oil Loading). To make magnetic MBs through oil loading, the lipid mix was prepared according to the protocol for unloaded MBs. A volume of isoparaffin IONPs corresponding to 3.75 mg of iron oxide was added after the two sonication steps. The vial was capped and cooled on ice for 5 min before repeating the two sonication steps to form magnetic MBs. Cleaning was carried out according to the general protocol.
Preparation of SurfMB (Surface Loading). To make magnetic MBs through surface loading, the lipid mix was prepared with a 82:9:9 mixture of DBPC/PEG40S/DSPE-PEG2000-biotin. The general protocol was followed until the cleaning step. The suspension was first cleaned to remove excess lipids in solution via centrifugation. A solution of avidin (2 mL, 2.5 mg/mL) in PBS was added to the resuspended biotinylated MBs and left to mix on ice on a rotary shaker for 5 min. The MBs were then centrifuged again before adding 300 μL of biotin−dextran IONPs (12.5 mg/mL) to the resuspended avidin-loaded MBs. The sample was left to mix on ice on a rotary shaker for 5 min again. The magnetic MBs were centrifuged once more and finally resuspended with PBS at 4°C.
MB Characterization. MB Size and Concentration. MB size and concentration were determined through the analysis of optical images 47 as previous studies have confirmed the reliability of this method compared to other options available. 48,49 For this, the MB suspension was diluted 1:20 in PBS, and 10 μL of this was loaded onto a hemocytometer with a cover slide. Thirty microscope images were acquired through an optical microscope (Leica DM500 optical microscope, Larch House, Milton Keynes, MK14 6FG, UK) with a 40× objective lens at room temperature. The images were then analyzed using purposely written MATLAB code (R2016b, The MathWorks, Natick, MA, USA) to determine the MB mean size and concentration. 47 Stability Studies. The stability of the different MB formulations was investigated by measuring the changes in the MB concentration

ACS Applied Materials & Interfaces
Research Article over time under flow at 37°C in PBS. Flow was induced by a peristaltic pump used with a 1.6 mm inner diameter latex tube producing a flow rate of 10 mL/min. Measurements were made at t = 15, 30, and 45 min and 1, 1.5, 2, 2.5, and 3 h. Each formulation was tested four times, and the collected data was fitted to a one-phase exponential decay nonlinear curve model using GraphPad Prism (Version 6, GraphPad Software Inc., La Jolla, CA, USA), which showed randomness of residuals.
Langmuir Trough. Langmuir trough experiments were conducted at room temperature and at 37 ± 1°C on a MicroTrough XL model from Kibron (Kenilworth, Warwickshire, UK) connected to a recirculating chiller-heater for temperature control. A lipid monolayer was formed at the gas/water interface of 75 mL of deionized water by adding 1 μL of lipids dissolved in chloroform (25 mg/mL) and letting it equilibrate for 5 min, after which it was compressed (56.1 mm 2 /s) from 11 000 to 2000 mm 2 . The average area per lipid molecule was calculated using the average lipid molecular weight. The surface pressure (Π) of the monolayer was then measured over the molecular area (Å 2 /molecule).
The effect of phospholipid-mediated loading of IONPs on a lipid monolayer was studied by adding IONPs to the deionized water subphase at a concentration of 20 μg/mL before forming the lipid monolayer. Isoparaffin IONPs were added in corresponding amounts on top of the lipid monolayer before compression.
Transmission Electron Microscopy. Transmission electron microscopy (TEM) was used to visualize the incorporation of IONPs onto the MBs. To obtain dried MB samples for TEM imaging, a previously described method was used. 50 Briefly, a carbon-coated copper grid (C267/050 thin clean films of carbon 300 mesh 3 mm copper grid, TAAB Laboratories Equipment Ltd, Berks, RG7 8NA, UK) was exposed to a glow-discharge cycle, and the carbon side of the copper grid was placed onto a 10 μL drop of the diluted MB sample on a piece of parafilm for 2 min. The TEM grid was then lifted and placed onto a 20 μL drop of 2% w/v uranyl acetate for 10 s. The grid was again lifted, carefully dried with a piece of filter paper, and left to dry for 10 min under an aluminum cover for reduced light exposure. The samples were imaged through a FEI Tecnai 12 transmission electron microscope operated at 120 kV, and the images were acquired using a Gatan OneView CMOS camera.
MB Acoustic Properties. At 1 MHz. The acoustic emissions from MBs exposed to 1 MHz ultrasound were recorded using a passive cavitation detector (PCD, 7.5 MHz centre frequency, 12.7 mm diameter, 75 mm focal distance, Olympus NDT, Essex, UK). A 2 MHz high-pass filter was used to remove the drive frequency from the recorded PCD traces before preamplifying (SR445A, SRS, Sunnyvale, CA, USA), digitising (Handyscope HS3, TiePie Engineering, Sneek, Netherlands), and saving on a computer drive.
Acquired PCD traces were analyzed by fast Fourier transform using MATLAB, and harmonics (multiples of the drive frequency ±100 kHz, >2 MHz), ultraharmonics (half-integer harmonics of the drive frequency ±50 kHz, >2 MHz), and broadband (remaining signal >2 MHz) components were extracted for each acquisition. The power in these frequency subsets was calculated for each acquisition over the exposure time.
At 7 MHz. The magnetic MBs were then tested at 7 MHz to give an indication of their behavior while being exposed to diagnostic ultrasound. Deionized water was used as a diluent and flown through a 1.6 mm diameter channel in a 2.5% agar phantom at a rate of 0.2 mL/min. A bolus of 200 μL of cleaned (9 ± 1) × 10 6 MB/mL was injected through a three-way valve and imaged using a commercially available ultrasound system (iU22, Philips, Bothell, WA, USA) and a linear array (L12-5, 7 MHz). A mechanical index (MI) of 0.04 was found to minimize MB destruction for the least stable formulation and used for all experimental runs. Each MB formulation was injected into the phantom and imaged three times. Commercially available SonoVue MBs (Bracco Imaging, Milan, Italy) were also imaged to provide a comparison. Images were analyzed with MATLAB (R2016b, the MathWorks, Natick, USA) by computing grayscale intensities in predefined regions of interest (ROIs) for all acquisitions to determine the point at which the injected MBs reached a constant concentration. Sets of 100 frames at constant MB concentration were examined to measure the average mean intensity within the ROIs.
Magnetic Properties. ICP−OES. The iron content of the different MB formulations was detected using ICP−OES (PerkinElmer, Optima 8000). These measurements were used to determine the amount of IONPs loaded onto MBs to compare with MB magnetic retention results. Iron measurements were performed at a wavelength of 238.204 nm, and a 7-point calibration curve was completed for iron in 2% nitric acid solution (0−1 mg/L) with a R 2 value of 0.9999 prior to sample measurements.
Magnetic Retention under Flow. The same setup as for the acoustic characterization of MBs at 7 MHz was used to study the magnetic retention of MBs under flow when exposed to a magnetic field. The magnetic field was applied using a five-element Halbach array magnet placed 1 cm below the channel. The magnet was previously described by Stride et al. and consisted of five rectangular N52 grade NdFeB permanent magnets (25 × 10 × 10 mm, supplied by NeoTexx, Berlin, Germany) with transversal magnetizations of 1.50 T, oriented at an angle of 90°from one another, and held by an aluminum frame. 27 A bolus of 100 μL of 5 × 10 7 MB/mL was injected through a three-way valve, and MBs were left to flow for 1 min above the magnet before acquiring ultrasound images with a linear array (L12-5, 7 MHz) angled perpendicular to the magnet. An ultrasound MI of 0.04 was used to minimize MB destruction. Tests were completed for cleaned magnetic MBs, and flashes of increased acoustic drive level were used to destroy and confirm the retention of MB material. As grayscale intensities of magnetic MBs flowing through the system without magnet have been acquired for their acoustic characterization, quantification of magnetic retention was performed through analysis of image intensity within the ROIs with and without magnet over 10 frames and compared to background measurements. ROIs were automatically segmented from the bottom of the channel with the first one-third of the channel being without magnet, and the remaining two-third being directly 1 cm above the Halbach array magnet (Figure 2).
Hemolytic Potential. Horse blood was used to assess the hemolytic potential of the different magnetic MB formulations. The blood was stored at 5°C and used the day following delivery at room temperature (21°C). The blood (200 μL) was diluted with 163 μL of MBs at 1.8 × 10 7 MB/mL or with PBS for positive and negative controls. The samples were then incubated at 37°C for 5 min. An aliquot of 80 μL of each blood sample was added to 1.12 mL of PBS. The positive control was further diluted with 120 μL of 1% Triton X-100 in water as a 100% hemolysis reference, whereas the other samples were further diluted with 120 μL of PBS. All samples were then centrifuged at 500g and 4°C for 5 min. The absorbance of the

ACS Applied Materials & Interfaces
Research Article supernatants was measured in triplicate (120 μL per well, in a Greiner UV-Star 96-well plate) at 541 nm. This procedure was performed twice.
Statistical Analysis. Statistical analyses were performed using Microsoft Excel and GraphPad Prism with significance defined as p < 0.05. One-way analysis of variance was used for experiments investigating the difference between more than two groups with post-hoc Tukey honestly significant difference test. Data is presented as mean ± standard deviation.

■ RESULTS AND DISCUSSION
MB Size Distribution and Concentration. The concentration and mean diameter of MBs (n = 3) for the different formulations are shown in Figure 3. OilMB exhibited a significantly lower yield and larger mean diameter compared to all other groups, with approximately 10 times fewer bubbles produced of double the size. This effect could be due to the excess surfactant a present in the isoparaffin IONP suspension disrupting the phospholipid coating or due to the oil phase itself reducing the coating integrity. The remaining MB formulations tested presented comparable concentrations and sizes.
Centrifugal cleaning of MBs enabled removal of excess starting materials in solution, although the final yield was lower after cleaning ( Figure S2), indicating that some MBs were also lost during processing. An average final concentration of (4 ± 1) × 10 8 MB/mL was obtained for unloaded MBs, LipMB, and SurfMB, whereas OilMB presented a lower final yield of (4 ± 0.5) × 10 7 MB/mL.
MB Stability under Flow. The results of the stability studies on MBs exposed to flow at 37°C are shown in Figure  4. One-phase exponential decay nonlinear regression models were fitted to the datasets, which indicate that LipMB-DBPC and SurfMB both showed significantly improved stability by more than doubling their half-lives to 2 and 0.9 h, respectively, compared to 0.4 h for unloaded bubbles. In contrast, the OilMB exhibited a significant decrease in stability (half-life of 0.1 h).
MBs were observed for over 3 h circulating in a 2 mL flow loop at 37°C, which is a longer timeframe than that reported in the in vivo models. This difference in timescale is explained by the absence of clearance mechanisms that exist in vivo and by the fact that the MBs were undiluted in the flow loop. 51 The incorporation of nanoparticles into MBs has been shown previously to prevent gas diffusion and MB coalescence 52 through a form of Pickering stabilization, which might explain the stability of LipMB-DBPC and SurfMB. The solid nanoparticles reduce the surface area available for gas diffusion, and physical jamming of the particles can inhibit MB shrinkage. The addition of LαPC IONPs had a destabilizing effect, which could be due to the presence of acyl chains of different lengths in LαPC (approximately 33% 16:0, 13% 18:0, 31% 18:1, and 15% 18:2 PC), reducing the molecular packing on the MB surface and hence the resistance to gas diffusion. Isoparaffin IONPs also had a destabilizing effect on MBs, which was probably the reason for the low production yield of this formulation ( Figure  3A). This may similarly have been due to the effect of isoparaffin on the packing of the phospholipid coating. The Langmuir trough measurements discussed in the next section support this.
The relative instability of this formulation was further indicated by the results shown in Figure 5. Additional measurements were made at 37°C and under flow of the number of magnetic droplets settled at the bottom of the hemocytometer and the number of buoyant magnetic MBs at the top of the sample underneath the coverslip. Figure 5 indicates a decrease in the buoyant OilMB over time, while the number of magnetic droplets sunken at the bottom of the haemocytometer increased. The size of the droplets suggests that this might have been caused by diffusion of the encapsulated gas out of OilMB, forming magnetic droplets that then sank because of the absence of the gas core.
Langmuir Trough. Surface pressure−area (Π−a) isotherms of DBPC/PEG40S monolayers without IONPs were characterized and imaged at 18°C for different pressure ranges

ACS Applied Materials & Interfaces
Research Article ( Figure S3). The isotherms show an initial gradual rise in the surface pressure followed by an intermediate plateau around 35 mN/m, which has been reported as the collapse pressure of PEG40S within the mixture. 46 The presence of this plateau was associated with the "squeeze-out" of the emulsifier from the monolayer. 46,53 Following PEG40S elimination, the slope of the isotherm steepens because of the dominant elastic behavior of the remaining DBPC monolayer, which finally collapses at ∼60 mN/m at 18°C. A broadening in the hysteresis was observed beyond 35 mN/m, indicating a loss of material into the water phase, which can be due to the formation of the bilayer, lipid budding, and vesicles from the remaining DBPC ( Figure S3C1,C2).
The effect of adding isoparaffin ( Figure 6A) and DBPC IONPs ( Figure 6B) to the monolayer was then tested at 37°C, and isotherms similar to those for the control were obtained. The addition of isoparaffin IONPs did, however, lower the final collapse pressure of the system, and the DBPC IONPs suppressed the "squeeze-out" plateau to some extent. The addition of LαPC IONPs ( Figure 6C) also reduced the surfactant "squeeze-out" plateau, indicating retention of PEG40S in the monolayer beyond 35 mN/m, and led to a lower monolayer collapse pressure. These results suggest that the addition of LαPC IONPs to DBPC/PEG40S MBs may reduce the loss of PEG40S during compression. Upon formation of MBs, the lipid-mix coating is compressed because of the Laplace pressure driving the system toward dissolution. Thus, the lipid packing increases, squeezing the emulsifier out of the coating until the remaining condensed-phase lipids reach a metastable state with a very low surface tension enabling bubble stabilization. 46 Any remaining PEG40S within the coating would hinder the system reaching this metastable state.
MB Acoustic Properties. The acoustic emissions from MBs exposed to 1 MHz ultrasound were investigated, as this frequency is relevant for therapeutic ultrasound drug delivery, and the energy and duration of emissions have been shown to be correlated with the therapeutic response. 54−56 The acoustic emissions from each magnetic MB formulation were measured for suspensions [(1 ± 0.5) × 10 7 MB/mL] exposed to 1 MHz ultrasound, 0.85 MPa pk−pk , and 30% duty cycle for 30 s ( Figure  7A). The signals were analyzed for harmonics, ultraharmonics, and broadband emissions separately over the exposure time, and the results in Figure 7B−D indicate that there was no significant difference between the formulations.
The MBs were also examined at 7 MHz through analysis of B-mode images of the MBs [200 μL, (9 ± 1) × 10 6 MB/mL] under 0.2 mL/min flow at room temperature. The results shown in Figure 8A indicate that all magnetic formulations behaved similarly at an MI of 0.04. Although all MBs were clearly visible, the image intensity was lower than that for the clinical agent SonoVue for the same concentration of MB injected and comparable size distribution ( Figure 8B). This was attributed to the stiffer shell of DBPC/PEG40S MBs 57 preventing bubble expansion compared to softer systems made of 1,2-distearoyl-sn-glycero-3-phosphocholine phospholipid as for SonoVue. Contrastingly, no acoustic difference was observed between the magnetic MB formulations and the unloaded systems, indicating that the lipid composition governed the acoustic oscillations of the system more than the presence of the nanoparticles.
MB Magnetic Properties. Individual MBs produced via different loading methods were examined using electron microscopy ( Figure 9). The images suggest that hydrophobic isoparaffin and LαPC IONPs remain in clusters on MBs (A1− 3, B1−3), unlike DBPC and hydrophilic dextran IONPs (C1− 3, D1−3). Additionally, there appeared to be larger quantities of dextran IONPs based upon visual inspection of images D1− 3 compared to other formulations.
Quantification of iron loading in the different formulations is presented in Figure 10A. These data indicate a significantly higher iron loading in OilMB (0.61 pg/MB) and SurfMB (0.71 pg/MB) compared to that in LipMB (∼0.025 pg/MB) and the background measurement in the unloaded control (0.006 pg/MB). The lower loading of LαPC and DBPC IONPs in MBs could be due to the low incorporation of IONPs during MB production or the removal of IONPs from the MB coating during the washing procedure. Nevertheless,

ACS Applied Materials & Interfaces
Research Article the LipMB-DBPC still exhibited a detectable response to magnetic fields, as shown in Figure 10B.
All magnetic MB formulations were tested for their ability to be retained under flow at a rate of 0.2 mL/min in a 1.6 mm channel in a tissue-mimicking phantom and imaged using a commercially available diagnostic ultrasound device. The channel size was chosen to fit the human physiological range for small arteries and veins that surround the tumor lesions with a flow rate of 0.2 mL/min. 58 The concentration of MBs injected into the system was kept fixed with 100 μL bolus injections of clean MB stock at 5 × 10 7 MB/mL tested for each formulation. MBs were left to circulate at 0.2 mL/min for 1 min with the Halbach array magnet situated 1 cm underneath the channel. The diagnostic ultrasound probe was kept off during this 1 min period to avoid MB destruction. After 1 min of circulation, a series of Bmode images were captured during which a "flash" of increased ultrasound intensity was used to destroy the retained material

ACS Applied Materials & Interfaces
Research Article and obtain background images. As the magnetic MBs produced similar grayscale intensities while exposed to 7 MHz diagnostic ultrasound and an MI of 0.04, their magnetic retention was compared by analyzing the intensity of a sequence of 10 frames for each formulation. Average image intensities of selected ROIs were compared to the control regions within the same frame. The regions chosen for analysis were located at the bottom of the channel to avoid noise from the remaining MBs flowing through the system (Figure 2).
The results in Figure 11 confirm the results from the ICP− OES measurements, with SurfMB showing significantly higher retention (61% increase in the B-mode image intensity) than the other MBs tested (10−19% increase under the same conditions). OilMB did not exhibit a significant difference in retention compared to LipMB, which might be due to their lower stability compared to the other formulations. In fact, OilMB produced a retained bolus that was comparable to

Research Article
SurfMB only on the first imaging frame, after which the bolus was destroyed due to the imaging ultrasound. No bolus of retained MBs was observed when using LipMB but only a downward movement of the MBs toward the magnet.
Hemolytic Potential. The toxicity of the different formulations was assessed in horse blood, and the results ( Figure 12) showed no hemolysis following the addition of different magnetic MBs. The samples were incubated at 37°C for 5 min, reflecting the short half-life of MBs in circulation. 59,60 Additionally, several in vivo studies using OilMB 27,28 and LipMB-LαPC 42,61,62 have been performed with no reported toxicity.
Evaluation. A summary of the findings is presented in Table 2. Different loading methods were compared for producing magnetic MBs, with the aim of prolonging their stability and enabling strong magnetic and acoustic responses. Although, OilMB enabled high loading of iron, this method produced unstable MBs with a substantially reduced half-life at 37°C and under flow compared to unloaded MBs. In contrast, SurfMB was found to be more stable than unloaded systems and magnetically more responsive than all other formulations. Whereas LipMB-DBPC presented a lower retention efficiency during magnetic targeting, this system still demonstrated a significant improvement in stability compared to unloaded MBs. It should also be noted that magnetic retention could be affected by the different magnetic properties of the IONPs used in different formulations, and further optimization should be carried out to investigate the effect of IONP size, charge, and magnetic properties.
For drug delivery purposes, the most versatile way of loading MBs with hydrophilic drugs or liposomes relies on the postfunctionalization of MB shell phospholipids. SurfMB could present a challenge for this type of additional surface functionalization, although it would not interfere with the loading of hydrophobic drugs within the shell. Contrastingly, LipMB-DBPC remains a versatile method to produce stable magnetic MBs, which allows the attachment of hydrophilic drugs to shell phospholipids but challenges the loading of hydrophobic drugs. b Although the incorporation of IONPs can potentiate oxidative stress because of the release of free iron from the particles followed by iron-catalyzed Fenton reaction, 63−65 the iron content after washing was found to be low, with a maximum of 0.7 mg/mL for a MB concentration of 10 9 MB/ mL. Additionally, considering the recommended dose of SonoVue MBs of 2−2.4 mL in adults, 66 which equates to approximately 10 9 MB administered, the corresponding dose of iron would be 0.7 mg. Compared to clinical intravenous injections of Feraheme (ferumoxytol) recommending two doses of 510 mg of iron over 8 days in adults with a chronic kidney disorder, 43 the administered dose of iron within magnetic MBs is orders of magnitude lower than the currently approved dose for anemia. Nevertheless, the composition of material used is important to ensure minimal toxicity even at low doses. The phospholipids used to make the MB shell and to coat phospholipid-stabilized IONPs were chosen to be saturated to minimize the chance of lipid peroxidation during ultrasound-induced MB collapse in the presence of IONPs. 67−69 The externally conjugated dextran IONPs were chosen to mimic Feraheme, but additional improvements in biocompatibility could be implemented to inhibit macrophage clearance and improve the efficiency of this method in vivo. 70

■ CONCLUSIONS
Magnetic MBs produced using different IONP loading methods were investigated to compare their stability, acoustic emissions, and magnetic retention. External binding of IONPs to the MB coating was found to produce a significant improvement in MB stability compared to unloaded systems while preserving their acoustic behavior. This method also produced the most magnetically responsive MBs. The disadvantage of this method is that although it allows for hydrophobic drug loading within the MB coating, it renders further attachment of hydrophilic drugs to MB shell phospholipids challenging. The phospholipid-mediated loading Figure 12. No hemolysis was observed when the magnetic MB formulations were added to the horse blood. Results were compared to 100% control (addition of 1% Triton X-100). The samples were incubated at 37°C for 5 min, and hemolysis was characterized with absorbance measurements at 541 nm. The entire procedure was done twice, and the samples were transferred in triplicate onto a 96-well plate for absorbance measurement.
The isoparaffin IONPs used in this work are stabilized with a surfactant. We unfortunately could not obtain information on the nature of the surfactant from the manufacturer. Preliminary experiments showed that a lower loading of paclitaxel was obtained in LipMB compared to unloaded MBs (data not shown).