High-Speed Optical Characterization of Protein-and-Nanoparticle − Stabilized Microbubbles for Ultrasound-Triggered Drug Release

Objective: Ultrasound-triggered bubble-mediated local drug delivery has shown potential to increase therapeutic ef ﬁ cacy and reduce systemic side effects, by loading drugs into the microbubble shell and triggering delivery of the payload on demand using ultrasound. Understanding the behavior of the microbubbles in response to ultrasound is crucial for ef ﬁ cient and controlled release. Methods: In this work, the response of microbubbles with a coating consisting of poly(2-ethyl-butyl cyanoacrylate) (PEBCA) nanoparticles and denatured casein was characterized. High-speed recordings were taken of single microbubbles, in both bright ﬁ eld and ﬂ uorescence. Results: The nanoparticle-loaded microbubbles show resonance behavior, but with a large variation in response, revealing a substantial interbubble variation in mechanical shell properties. The probability of shell rupture and the probability of nanoparticle release were found to strongly depend on microbubble size, and the most effective size was inversely proportional to the driving frequency. The probabilities of both rupture and release increased with increasing driving pressure amplitude. Rupture of the microbubble shell occurred after fewer cycles of ultrasound as the driving pressure amplitude or driving frequency was increased. Conclusion: The results highlight the importance of careful selection of the driving frequency, driving pressure amplitude and duration of ultrasound to achieve the most ef ﬁ cient ultrasound-triggered shell rupture and nano-particle release of protein-and-nanoparticle − stabilized microbubbles.


Introduction
Microbubbles are of great interest for ultrasound-mediated local drug and gene delivery [1−4].Locally delivering drugs using microbubbles may result in increased efficacy and reduced systemic side effects compared with traditional drug administration, as the microbubbles are driven only in the local field of the ultrasound burst [1,5].Two approaches to ultrasound-mediated local drug delivery have been investigated: co-administration of the therapeutic load with the bubble suspension, and binding of the therapeutic load onto or into the bubble shell [2,5,6].
In the first approach, the microbubbles are driven locally using ultrasound, so that the bubble oscillations may induce a mechanical effect on their surroundings, for example, on the vessel endothelium and interstitial tissues [4,7].As such, the permeability of the affected endothelial cell membranes is temporarily increased (termed sonoporation or sonopermeation), which then allows the co-administered drugs to enter, or to pass, the endothelial cells [8−10].
The second approach, in which drugs are attached to or incorporated into the bubble shell, has been found to result in increased drug uptake [7,11].Nano-sized particles, such as liposomes, can be used to encapsulate hydrophilic drugs with a high loading capacity [5].This results in a low degree of degradation and clearance of the drug, which, in combination with the local release, reduces toxicity [12,13].For this approach, the loaded microbubbles are driven into oscillation such that the payload is released from the microbubble.The payload is thereafter either deposited locally onto cells in patches (sonoprinting) and subsequently taken up [7,14], or is allowed to extravasate by sonopermeation after release [10,15].
In both approaches, the oscillatory response of gas microbubbles to an incident pressure field is strongly dependent on bubble size, driving frequency, and driving pressure amplitude through the resonance behavior of the bubbles [16,17].Microbubbles are coated to prevent dissolution and coalescence, and the coating material consists of phospholipids, polymers or proteins [18,19].The viscous and elastic properties of the lipid shell change the resonance behavior [20], thereby directly affecting the efficacy of payload release.In contrast, protein and polymer shells typically are rigid [21,22].Rigid-shelled microbubbles have a much higher resonance frequency at low acoustic driving pressures [20], whereas shell rupture occurs at higher pressure amplitudes [23].After shell rupture, the core gas may be released and a free gas bubble remains in the surroundings, which has been reported for various microbubbles [23−25].The typical heterogeneity in the viscoelastic shell properties of rigid-shelled microbubbles (see, e.g., Supplementary Fig. S8B from Snipstad et al. [26]) poses a significant challenge for characterization methods performed on a population, for example, acoustic absorption measurements [27,28].
In this work, we focus on the optical characterization of microbubbles that are developed for ultrasound-mediated drug delivery.Their coating is composed of a denatured protein and polymeric nanoparticles, which stabilize by self-assembly during production to form the shell [29].The nanoparticles can be used as drug carriers, and the nanoparticle-stabilized microbubbles in combination with focused ultrasound have been shown to improve delivery of the encapsulated drug and to improve the therapeutic response of breast and prostate tumors in mice [26,30] compared with drug-loaded nanoparticles alone.In vitro studies have demonstrated improved uptake into cells compared to incubating cells with microbubbles and nanoparticles separately during ultrasound exposure, suggesting the occurrence of sonoprinting [11,14].Yemane et al. [31] investigated the kinetics and distance of penetration of the nanoparticle into the extracellular matrix using real-time intravital microscopy during ultrasound exposure, which indicated that extravasation of the nanoparticles occurs mainly at vessel branching points.Furthermore, nanoparticle-stabilized microbubbles in the presence of focused ultrasound were reported to increase the permeability of the blood−brain barrier [32].
How these protein-and-nanoparticle−stabilized microbubbles respond to ultrasound is largely unknown, and with that the acoustic parameters for optimal nanoparticle release from the microbubble.The aim of this work is therefore to understand in detail the acoustic response of these novel drug delivery agents, so as to find the optimal ultrasound parameters for an efficient and well-controlled drug delivery system.This work also aims to evaluate the efficiency of nanoparticle release from the protein-and-nanoparticle−stabilized microbubbles by recording their response to ultrasound at varying ultrasound frequencies and amplitudes.High-speed imaging is used to record the oscillation dynamics of the microbubbles, while high-speed fluorescence imaging is employed to study the release of the nanoparticle payload, to relate the oscillation dynamics to payload release.

Theory
The dynamics of a coated gas microbubble are described by the Rayleigh−Plesset equation [17,33,34].The Rayleigh−Plesset equation can be linearized to provide an analytical solution by assuming small-amplitude oscillations of the microbubble, while being driven by sinusoidal driving P(t) = P a exp(iωt) at an angular frequency of ω and amplitude P a .The Rayleigh−Plesset equation is then reduced to a damped harmonic oscillator equation, which provides an expression for the relative radial excursion R e (t) [17] where R(t) is the microbubble radius, R 0 the microbubble radius at rest, ρ L the density of the medium and δ tot the total damping coefficient of the microbubble due to reradiation of the ultrasound, viscous dissipation and thermal damping [35].The angular eigenfrequency ω 0 of the microbubble is given by [20] ω with κ the polytropic exponent of the gas, P 0 the ambient pressure, σ(R 0 ) the surface tension at rest, and χ eff the effective shell elasticity.For phospholipid-coated bubbles, the effective shell elasticity is maximum at low oscillation amplitudes.At high oscillation amplitudes, the microbubble shell ruptures upon expansion and buckles upon compression, during which the shell elasticity vanishes, resulting in a reduced effective shell elasticity [36,37].The eigenfrequency of the microbubble is related to the angular resonance frequency ω res , at which the radial oscillation of the microbubble is the largest, through the damping of the shell: In contrast to phospholipid-coated microbubbles, which are typically flexible (low stiffness), polymer-and protein-coated microbubbles typically are rigid-shelled with a high stiffness [38].The nanoparticle-andprotein−stabilized microbubbles were previously observed to have a higher stiffness than SonoVue (see Supplementary Fig. S8A in Snipstad et al. [26]), which are lipid-coated microbubbles.The surface tension σ(R) for a rigid-shelled microbubble can be modeled using a model for solid-shell contrast agent bubbles [22].According to this model, the rigid-shelled microbubble buckles when the bubble is compressed to the point at which the radius is lower than the critical buckling radius, while the shell irreversibly ruptures when the bubble is expanded until its radius surpasses the critical breakup radius.
The validity of these theoretical models can then be tested, the bubbles characterized and the physical parameters derived, by comparison with experimental data probing the bubble resonance behavior obtained from high-speed optical imaging, acoustic scattering, or acoustic attenuation measurements.

Protein-and-nanoparticle−stabilized microbubbles production
The drug delivery agents under investigation are protein-and-nanoparticle−stabilized microbubbles, developed at SINTEF (Trondheim, Norway) [11,29,31].The microbubbles are filled with perfluoropropane gas and have a shell consisting of the denatured protein casein and poly (2-ethyl-butyl cyanoacrylate) (PEBCA) nanoparticles that have been coated with polyethylene glycol (PEG).The PEBCA nanoparticles were produced using mini-emulsion polymerization as described previously [29].The nanoparticles are fluorescently labeled with NR668 (modified Nile red) dye [39] to enable their tracking during the release process and for this system to act as a model drug [40].Figure 1A depicts a schematic representation of the protein-and-nanoparticle−stabilized microbubble.
The microbubbles were produced on the day of the experiments as follows.First, 0.5% (w/v) casein (Sigma-Aldrich, St. Louis, MO, USA) was mixed in 0.9% phosphate-buffered saline (Sigma-Aldrich), ultrapure water (Milli-Q, Merck, Darmstadt, Germany) and 1% (w/v) nanoparticles in a glass vial.To break up nanoparticle aggregates, the vial was placed in an ultrasonic bath (Branson Ultrasonics, Brookfield, CT, USA) for 10 min.Afterward, perfluoropropane gas (C 3 F 8 , F2 Chemicals, Bristol, UK) was added to the headspace of the vial using a glass pipet.The solution was then stirred at 24,000 rpm for 4 min using a dispersion tool (ULTRA-TURRAX, IKA), during which the microbubble shell self-assembled [29].The vial was then sealed with a rubber stopper.
Figure 1B and 1C illustrate two single bubbles, imaged using a confocal microscope (A1R, Nikon Instruments) and a scanning electron microscope (S-5500 S(T)EM, Hitachi HighTech Corp.) [26], respectively.The fluorescence of the nanoparticles in Figure 1B reveals that the shell has a non-uniform particle distribution, and Figure 1C illustrates the varying nanoparticle size.Figure 1D shows the size distribution of the bubbles selected for the optical recordings in this study (blue), compared with the size distribution measured with a Coulter counter (Multisizer 3, Beckman) (orange) and the optically measured size distribution reported before by Snipstad et al. [26] (green, dashed line).It should be noted that the minimum detectable radius from the optical recordings was 0.9 μm, limited by the optical resolution of the microscopy system.The size distribution of the microbubbles that were analyzed indicated a selection bias toward relatively large microbubbles compared with the size distributions measured with the Coulter counter and that previously measured by Snipstad et al. [26].This bias arises from the manual selection of microbubbles for positioning into the field of view and subsequent recording.

Optical setup
High-speed bright-field microscopy was used to investigate the bubble dynamics and to observe shell rupture of single microbubbles in response to a varying driving pressure field.The release of the nanoparticle payload was investigated using high-speed fluorescence microscopy.A schematic of the setups is provided in Figure 2A.
For both types of high-speed imaging, a high-speed camera (HPV-X2, Shimadzu), and a modular microscope (BX-FM, Olympus) with a 2× magnification module were used.A water-immersion microscope objective (LUMPlanFL, Olympus) with a 60× magnification and numerical aperture of 0.9 was connected to the microscope.For the bright-field microscopy experiments, an optical fiber was placed at the bottom of the water tank.The fiber was connected to either a continuous halogen light source (ACE I, SCHOTT) during the alignment of the setup and during localization of fresh microbubbles after each measurement, or to a strobe light (MVS 7010 High-Intensity Xenon Strobe, Perkin-Elmer) during the high-speed recordings.To study the release of the fluorescent nanoparticles, a continuous wave laser (wavelength 532 nm, Cohlibri-532, Lightline) and a dichroic mirror were used to fluorescently excite the nanoparticles.

Sample preparation
For both sets of experiments, the microbubble solution was diluted to image single microbubbles while minimizing the effects of neighboring microbubbles.To this end, 10 μL of the microbubble solution was  diluted in 10 mL phosphate-buffered saline (ISOTON II, Beckman Coulter) to a concentration between 2×10 5 and 6×10 5 microbubbles/mL.The diluted solution was then placed in a 10 mL cell culture cassette (CLINIcell 25, Laboratoires Mabio International), with two 125-μmthick optically transparent membranes, using a syringe.The cassette was then attached to an XYZ-translation stage and placed horizontally in a tank filled with water at room temperature, to ensure proper propagation of ultrasound.

Ultrasound setup
To acoustically drive the microbubbles, a pulse/delay generator (BNC 575, Berkeley Nucleonics) was connected to an arbitrary waveform generator (100 MS/s Dual-Channel Arbitrary Waveform Generator 8026, Tabor Electronics), which was connected to both an oscilloscope (DPO3014, Digital Phosphor Oscilloscope, Tektronix) to monitor the signal and to a 56-dB amplifier (VBA100-200, Vectawave).The amplifier was connected to an ultrasound transducer, which was placed in another entry port of the container at an angle of 35°relative to the vertical (Fig. 2A) to minimize ultrasound reflections from and within the cassette.To study the radial dynamics and shell rupture of the microbubbles during insonation, a focused immersion transducer (Panametrics-NDT C305, Olympus) was used, with a center frequency of 2.25 MHz, and a focal distance of 1 inch.To study the release of the fluorescent nanoparticles during insonation, another focused immersion transducer (Panametrics-NDT V304, Olympus) was used, with a center frequency of 2.25 MHz and a focal distance of 1.88 inches.The presence of both the cell culture cassette membrane and the microscope objective results in reflections of the incident ultrasound, caused by acoustic impedance mismatches, which changes the in situ acoustic pressure [41,42].The transmission and reflection characteristics were therefore calculated using a k-Wave simulation [43] to estimate the local pressure amplitude as described in detail in Appendix A (see Supplementary Material).

Microbubble dynamics: bright-field microscopy
The complete protocol of the bright-field microscopy experiments has been described before [44].The full parameter space of the experiments is given in Table 1.In brief, single microbubbles were brought into the field of view.The high-speed camera recorded 256 images of 400×250 pixels at a frame rate of 10 million frames per second (fps), thus resulting in recordings with a duration of 25.6 μs.After each burst of ultrasound, the cell culture cassette was moved in the x-and y-directions (Fig. 2A) by at least 2 mm to bring new bubbles into the field of view that were not insonified before.
The bubbles were insonified during the high-speed bright-field recordings by a single burst of 40 cycle ultrasound with an 8 cycle Gaussian envelope at the beginning and end, at a frequency of 1, 2 or 3 MHz.The pressure amplitude was varied per set of experiments and ranged from 131 to 1548 kPa. Figure 2B shows the timing diagram for the bright-field microscopy recordings.
Nanoparticle release: fluorescence microscopy Again, the complete protocol of the fluorescence microscopy experiments has been described before [44], and the full parameter space can be found in Table 1.In brief, the high-speed camera recorded 128 images of 400×250 pixels at a frame rate of 0.5 Mfps, thus resulting in recordings with a duration of 256 μs.
During high-speed fluorescence recordings, a single burst of 140 cycles of ultrasound with a 10 cycle Gaussian envelope at the beginning and end was used, again at a frequency of 1, 2 or 3 MHz.The driving pressure amplitude was varied per set of experiments at the same range of amplitudes as for the bright-field experiments.As before, after each burst of ultrasound, the cell culture cassette was moved in the x-and ydirections by at least 2 mm. Figure 2C shows the timing diagram for the fluorescence microscopy recordings.

Analysis
For each frame of the high-speed (fluorescence) recordings, the image was cropped around each microbubble, and the background was subtracted, and then the image was binarized to determine the centroid for each microbubble.For the high-speed bright-field recordings, the inflection point of the radial intensity profile from the centroid was then taken as the radius, over 15 equally spaced angular sections, and then averaged over these sections to obtain a singular radius value per point in time, as illustrated in Figure 3A.The initial radius of the bubble is determined by averaging over the radius of five consecutive frames, within 4−4.5 μs from the start of the recording (Fig. 2B).
For the high-speed fluorescence recordings, the point of the highest derivative of the radial intensity profiles at 28 μs from the start of the recording, i.e., 2 μs before ultrasound arrival, was taken as the initial bubble radius.
Next, the time-dependent normalized radial excursion R e (t) was determined for each bright-field microbubble recording, which is defined as where max(R(0 → t)) and min(R(0 → t)) are the maximum and minimum radii of the bubble at time t, respectively, and R 0 is the initial radius of the bubble.Finally, the maximum normalized radial excursion was determined for the microbubbles that remained intact during the recording by taking the maximum value of R e (t).Because of the limited and time-varying intensity of the strobe light, the radial excursion was determined until 15 μs from the start of each recording.Furthermore, only bubbles for which the radial excursion had a signal-to-noise ratio of at least 6 dB were included.An example of the radial excursion analysis is provided in Figure 3B and 3C.
To determine the probability of microbubble rupture, a microbubble was labeled as having "ruptured" its shell when the relative radial excursion exceeded an empirical threshold of R e > 0.25.This threshold was determined from the relative radial excursion of randomly selected microbubbles at the moment in the high-speed recordings when a sudden increase in the radial excursion occurs.In some instances, rupture is also evident by fragmentation of the microbubble resulting in multiple smaller bubbles.To determine the probability of nanoparticle release, the fluorescence microscopy recording of each microbubble was evaluated manually to establish whether detachment and deposition of the fluorescent nanoparticles from the microbubble core was observed, as illustrated in Figure 4A.The release was defined as unsuccessful when the microbubble remains intact after exposure to ultrasound (an example is provided in Fig. 4B).The high-speed fluorescence recordings corresponding to the images in Figure 4 can be found in Videos S1 and S2 (online only).

Microbubble dynamics
A total of 1522 microbubbles were analyzed from the high-speed bright-field recordings.The behavior of the microbubbles during highspeed bright-field microscopy imaging can be divided into three main categories: (i) volumetric oscillation, (ii) non-volumetric oscillation and (iii) shell rupture.Examples of this behavior can be observed in Figure 5, where images (a)−(e) from the high-speed recordings are shown with the corresponding microbubble radius over time.The microbubbles mainly exhibited volumetric oscillations, where the microbubbles expand and contract volumetrically (Fig. 5A) [23].Several microbubbles exhibited non-volumetric behavior, with a clear observation of buckling of the shell where part of the microbubble shell remained stationary, while another part of the shell was observed to radially oscillate (Fig. 5B) [45].Finally, the shells of 41% of the microbubbles were observed to rupture, dislodging shell material and a free gas bubble (Fig. 5C) [46].The latter was then observed to oscillate at higher relative radial excursions than when the microbubble was still coated (Video S3, online only).Less frequent observations include microbubble shape oscillations displaying surface modes [47] and subharmonic oscillations at half the driving frequency.The high-speed recordings corresponding to the oscillations illustrated in Figure 5 can be found in Videos S4−S6 (online only).
To investigate the effect of the shell on bubble dynamics, the relative radial excursion of each microbubble was normalized by dividing it by the driving pressure amplitude (following Eq. ( 1)), and is shown in Figure 6 as a function of the initial bubble radius.Recordings of microbubbles that were observed to rupture were not considered here, as the moment in time at which shell rupture occurred cannot be determined accurately.The lines indicate the simulated normalized radial excursion, plotted as a function of the initial bubble radius, using the linearized Rayleigh−Plesset equation (Eqs.( 1) and ( 2)).
As illustrated in Figure 6, the variation in the radial excursion for each initial bubble radius indicates highly non-uniform shell properties within the population.Nevertheless, a pronounced resonance behavior is observed.By comparison of the subfigures of Figure 6, it can be observed that, as expected, this resonant size decreases with increasing driving frequency.Moreover, the normalized radial excursion of unruptured microbubbles depends on the driving pressure amplitude.Of the microbubbles that remained intact during insonation, large normalized radial excursion amplitudes correspond predominantly to microbubbles driven at low pressure amplitudes (red), whereas lower normalized radial excursion amplitudes correspond to microbubbles driven at high pressure amplitudes (light orange to yellow).This dependence can be attributed to the increased probability of rupture for microbubbles at resonance.The two lines depicting the simulated normalized radial excursions in Figure 6 illustrate that we do not observe good agreement between the data and the linearized Rayleigh−Plesset model (Eqs.( 1) and ( 2)) for all three driving ultrasound frequencies, even for simulations with a wide range of shell parameters.Therefore, we conclude that the dynamics of the protein-and-nanoparticle−stabilized microbubbles cannot be described adequately by this linearized viscoelastic shell model.The disagreement between the experimental data and model could have been caused by the presence of the cassette membrane that supports the floating bubbles, which is known to shift the resonance frequency [48−50].This resonance shift is proportional to the resonance frequency in the absence of a membrane, and cannot be described by a frequency-independent shell elasticity.Alternatively, the oscillations observed are directly linked to the mechanical response of a complex viscoelastic shell, which may require the inclusion of an intrinsically non-linear shell bending modulus [51] in addition to the dilatational shell properties in the linearized model we have considered.These considerations also suggest that the effect of nearby cells should be considered in the future [8].

Shell rupture and nanoparticle release
In addition to the ultrasound driving frequency and pressure amplitude, the duration of insonation required to achieve microbubble rupture is a relevant ultrasound parameter for optimal drug delivery.The cavitation threshold generally shifts toward higher pressure amplitudes for increasing ultrasound frequency [2]. Figure 7 illustrates the number of cycles of ultrasound that the microbubbles experience before they rupture, for varying pressure amplitudes and at driving frequencies of 1 MHz (blue dots), 2 MHz (orange squares) and 3 MHz (green triangles).Note that we now group microbubbles of all sizes.Though the moment of rupture could not be determined with subcycle accuracy, the standard deviation in Figure 7 illustrates that the interbubble variation in the number of cycles of ultrasound required for bubble rupture is larger than a single cycle.It is evident for all driving frequencies that at higher pressure amplitudes, rupture of the microbubble shell occurs after fewer cycles of ultrasound.Moreover, for low to moderate pressure amplitudes (<800 kPa), the number of cycles necessary for rupture at a given pressure amplitude increases with increasing ultrasound frequency.This could be explained by the resonance behavior, which predicts higher radial excursions for lower ultrasound frequencies (Fig. 6).Thus, the number of cycles of ultrasound until shell rupture is typically lowest for a driving frequency of 1 MHz.For higher pressure amplitudes, the number of cycles until shell rupture plateaus for all three driving frequencies.It should, however, be noted that there is a larger standard deviation at frequencies of 2 and 3 MHz, especially for relatively high pressure amplitudes (>800 kPa).Note that, at high pressure amplitudes, the number of cycles of ultrasound to rupture falls below the duration of the Gaussian tapering in the driving pressure (dashed line in Fig. 7).Thus, most bubbles ruptured before the driving pressure had reached its peak value.
The probability of microbubble rupture within an equiprobable size range and for the varying driving pressure amplitude is illustrated in Figure 8A through 8C, as determined from the high-speed bright-field recordings.Similarly, Figure 8D through 8F illustrate the probability of release as a function of microbubble size as determined from the highspeed fluorescence recordings.Here, each column of plots corresponds to a different driving frequency, and the color indicates either the rupture or release probability.Figure 8A through 8C show that the likelihood of rupture of the microbubble shell due to ultrasound insonation is highly dependent on the initial microbubble size, as resonance behavior is apparent even from the limited number of microbubble recordings.The bubble size at maximum rupture probability decreases with increasing driving frequency.Furthermore, increasing the driving pressure amplitude is observed to typically increase the probability of rupture, for an increasing range of microbubble radii.Similarly, from Figure 8D through 8F, the probability of nanoparticle release is observed to increase with increasing driving pressure amplitude.Comparison of Figure 8A through 8C with 8D through 8F illustrates that the radii of the highest rupture probability are similar to the radii of the highest release probability, indicating that the nanoparticle release is in line with the resonance behavior.Furthermore, obtaining a similar probability of release generally requires a larger pressure amplitude as the driving frequency is increased.However, several regions can be identified in Figure 8 where the probability of nanoparticle release is higher than the rupture probability, for example, for microbubbles with a radius near 4 μm, insonified at a frequency of 3 MHz and a pressure amplitude of 1.5 MPa.These regions could potentially be explained by the shorter insonation period during the bright-field recordings (40 cycles) compared with the fluorescence recordings (140 cycles).The insonation duration was shorter during the bright-field recordings because of the higher frame rate necessary to temporally resolve the radial dynamics.Nevertheless, Figure 8 indicates that shell rupture and nanoparticle release are closely related phenomena.

Limitations
As a result of the finite recording length combined with the limited light intensity from the strobe light of the bright-field high-speed recordings, the microbubbles were already insonified for 1−1.5 μs before the initial radius could be determined.This could have influenced the estimation of the initial bubble radius and thus the relative radial excursion from these recordings.However, because of the Gaussian tapering of the acoustic driving pressure, in that time the pressure amplitude would have reached 30−46%, 60−80% and 80−93% of the maximum during experiments at driving frequencies of 1, 2 and 3 MHz, respectively.The influence of the short insonation period before the initial radius was determined was therefore neglected.
Another limitation stems from the separate high-speed recording setups and the corresponding different insonation periods.Therefore, it remains uncertain whether the occurrence of shell rupture and the release of the nanoparticle payload mutually imply one another.To this end, simultaneous high-speed bright-field imaging and high-speed fluorescence imaging such as described in [52,53] could provide a more detailed insight into the exact nanoparticle release mechanism.

Perspectives
In pre-clinical studies of blood−brain barrier disruption and ultrasound-triggered drug delivery in cancer with microbubbles, ultrasound is typically transmitted for periods on the order of 10,000 cycles [54,55].Yet in the present work, insonation lasted only 140 cycles, and release of the nanoparticle payload was observed at a mechanical index (MI) as low as 0.13.The highest probability of release was observed at MIs of at least 0.4−0.5, which are similar to the MI of 0.5 previously used for ultrasound drug delivery with 10,000 cycle bursts [26,30].
However, the effect of lowering the insonation period and the MI on the efficacy of in vivo drug delivery is not considered in this work.Due to insonation at relatively low MIs, microbubbles cavitate in a stable regime, which induces microstreaming [53,56 −58].In contrast, at high MIs, there will be relatively more inertial cavitation events, which are more violent in nature [10,59].Although cavitation is necessary for sonoporation to allow nanoparticles to enter or pass the endothelial cells [60], inertial cavitation can also lead to irreversible damage [59] and potentially cell death [61].
Overall, the results indicate that the ultrasound pressure amplitude and possibly the insonation period may potentially be reduced with respect to their typical values, to minimize adverse side effects of ultrasound-triggered drug delivery.Insonation periods longer than those considered in this work might be necessary to increase the efficacy of delivery after release of the nanoparticle payload.

Conclusion
Here, we report that protein-and-nanoparticle−stabilized microbubbles have a rigid shell, with a large variation in shell properties within the population.Although our understanding of the particle release mechanism is incomplete, the efficiency of drug delivery, quantified through the probability of shell rupture and nanoparticle release, was found to be highly dependent on microbubble size and ultrasound frequency, in a manner consistent with acoustic bubble resonance.For each acoustic driving frequency, we observed a corresponding resonant radius at which the likelihood of shell rupture and payload release was the highest.The release was further enhanced by increasing the pressure amplitude, which increased the probability of shell rupture and particle release at the resonant radius, and decreased the number of ultrasound cycles needed to rupture the microbubbles.The results suggest that ultrasound-triggered nanoparticle release of the nanoparticle-loaded microbubbles considered here can be achieved most efficiently through insonation at an ultrasound frequency of 1 MHz and at a pressure amplitude of at least 500 kPa, for a minimum duration of 10 cycles.

Figure 2 .
Figure 2. (A) Schematic representation of the setup for both high-speed bright-field imaging (top left inset) and high-speed fluorescence imaging (top right inset).High-speed imaging was performed using an HPV-X2 (Shimadzu) camera, which operated at a frame rate of 10 million fps for bright-field imaging and 0.5 million fps for fluorescence imaging.(B) Schematic representation of the timing in the experimental procedure of high-speed bright-field imaging relative to the start of the recording.The triggers are indicated by the triangles.The ultrasound was triggered 16 μs before the high-speed recording started, to ensure the arrival of the ultrasound pulse at the optical focus point after recording for 3 μs.The strobe light was triggered 1.5 μs before the start of the recording, and the intensity profile is shown in yellow.(C) Schematic representation of the timing in the experimental procedure of high-speed fluorescence imaging relative to the start of the recording.The laser was turned on 20 μs after the recording was started and turned off after recording for 170 μs.To ensure that the release of nanoparticles from the microbubble shell was captured within the time frame of the recording, the ultrasound was triggered at an earlier time such that the ultrasound arrived at the focal point after a recording time of 30 μs.

Figure 3 .
Figure 3. (A) Image processing analysis to extract the radial excursion for a single microbubble: original frame (left), after image processing (middle), where the green dot indicates the centroid, and the radius for 15 angular sections (right).Bar = 5 μm.(B) The radius after averaging over the angular sections, for each recorded frame.For the point in time indicated by the red dot, the radial excursion R e (t) is determined by the initial radius (pink, solid line), the maximum radius (blue, dashed line) and the minimum radius (orange, dotted line) until that time point.(C) Corresponding relative radial excursion based on the radius −time curve in (B).

Figure 4 .
Figure 4. (A) Image sequence of a high-speed fluorescence recording depicting successful release of fluorescently labeled nanoparticles of a 2.97-μm-radius microbubble insonified at an ultrasound frequency of 2 MHz and an acoustic pressure amplitude of 584 kPa.(B) Image sequence of a high-speed fluorescence recording depicting no release of fluorescently labeled nanoparticles of a 3.87μm-radius microbubble insonified at an ultrasound frequency of 2 MHz and an acoustic pressure amplitude of 205 kPa.The time stamps in each upper left corner indicate the time from the start of the recording.Bars = 10 μm.The laser was turned on at 20 μs and turned off 150 μs later, at 170 μs.

Figure 5 .
Figure 5.The three categories of observed microbubble behavior, with several exemplary recordings, and the corresponding radius−time curves.(A) Volumetric oscillation of a 3.48-μm-radius microbubble insonified at an ultrasound frequency of 1 MHz and an acoustic pressure amplitude of 131 kPa.(B) Non-volumetric oscillation of a 2.11-μm-radius microbubble insonified at an ultrasound frequency of 1 MHz and an acoustic pressure amplitude of 229 kPa.The arrow in image (b) indicates the oscillating side of the microbubble.(C) Shell rupture of a 1.83-μm-radius microbubble insonified at an ultrasound frequency of 1 MHz and an acoustic pressure amplitude of 459 kPa.A free gas bubble was observed to oscillate after shell rupture.Bars = 5 μm.

Figure 6 .
Figure 6.Relative radial excursion divided by the driving pressure amplitude as a function of the initial bubble radius at driving frequencies of 1 MHz (A), 2 MHz (B) and 3 MHz (C).The color of each data point indicates the driving pressure amplitude.Lines correspond to the simulated normalized radial excursion for shell parameters χ = 0.3 N m −1 and δ tot = 0.9 (blue, dashed line), and χ = 1 N m −1 and δ tot = 1 (green, solid line).For both simulated curves, the surface tension at rest is σ(R 0 ) = 0 N m −1 , the density of the medium ρ L = 1×10 3 kg m −3 , the polytropic exponent of the gas core κ = 1.1 and the ambient pressure P 0 = 100 kPa.

Figure 7 .
Figure 7. Pressure amplitude and number of ultrasound cycles after which the microbubbles were observed to rupture during ultrasound insonation at a frequency of 1 MHz (blue dots), 2 MHz (orange squares) and 3 MHz (green triangles).The markers indicate the median cycles to rupture, and the error bars indicate the standard deviation.The open markers indicate n <10, and for the closed markers n = 15−66.For each data set, the maximum pressure was reached after 8 cycles, as indicated by the gray dashed line.

Figure 8 .
Figure 8. (A−C) Probability of microbubbles that are observed to rupture during ultrasound insonation at frequencies of MHz (A), 2 MHz (B) and 3 MHz (C) at varying pressure amplitudes, within equiprobable size ranges (n = 5−9 per size range).The color indicates the probability of shell rupture.(D−F) Probability of microbubbles that are observed to release fluorescent shell material during ultrasound insonation at frequencies of 1 MHz (D), 2 MHz (E) and 3 MHz (F) at varying pressure amplitudes, within equiprobable size ranges (n = 5−10 per size range).The color indicates the probability of nanoparticle release.

Table 1
Parameter space of the high-speed bright-field recordings and high-speed fluorescence recordings