*2.4. Acoustical Characterization*

To study both the acoustical behavior and the lipid phase separation of single microbubbles simultaneously, the combined confocal microscopy and Brandaris 128 ultrahigh-speed camera system was used [31]. Microbubble spectroscopy was employed to characterize the acoustic behavior of single microbubbles as described previously [11,32]. Microbubbles were washed by flotation once and counted using the Coulter Counter Multisizer 3, as described above. An acoustically compatible [32] CLINIcell (MABIO, Tourcoing,

France) with 50 µm membranes (25 µm<sup>2</sup> ) was first blocked with 12 mL of 2% (*w/v*) bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h, to avoid unspecific microbubble binding to the membranes. The CLINIcell was washed three times with PBS before inserting 12 mL of 10<sup>5</sup> microbubbles/mL in PBS. Next, the CLINIcell was placed underwater in the experimental setup and kept at room temperature for up to 2 h. To study the lipid phase separation, the custom-built confocal microscope (Nikon Instruments, Amsterdam, The Netherlands) was used with a 561 nm laser to excite rhodamine-DHPE and emitted light was detected in a 595/50 nm channel. *Z*-stacks with 0.4 µm steps were acquired with a CFI Plan 100 × W objective of single microbubbles directly before and after insonification. To perform microbubble spectroscopy, each individual microbubble was insonified over a range of transmitting frequencies (*fT*) from 1 to 4 MHz in steps of 200 kHz. The microbubbles were insonified with 8-cycle Gaussian tapered sine wave bursts either at 50 kPa or first at 20 kPa and then at 150 kPa external peak negative pressure (PNP), generated by a Tabor 8026 arbitrary waveform generator (AWG, Tabor Electronics, Tel Hanan, Israel). The signal was first attenuated by a 20-dB attenuator (Mini-Circuits, Brooklyn, New York, NY, USA), then amplified by a broadband amplifier (ENI A-500, Electronics and Innovation, Rochester, New York, NY, USA), and finally transmitted to the microbubble sample at a 45◦ incidence angle with a single-element transducer (1–9 MHz bandwidth, 25 mm focal distance, −6 dB beamwidth at 1 MHz of 1.3 mm, PA275, Precision Acoustics, Dorchester, UK), which was calibrated using a 1-mm needle hydrophone (Precision Acoustics, Dorchester, UK) in water. The Brandaris 128 ultra-high-speed camera [33], coupled with the confocal microscope [31], was used to record the microbubble oscillation behavior at approximately 17 million frames/s. First, a recording was made without ultrasound to establish the initial microbubble size. Next, 16 recordings at 50 kPa PNP, or 16 recordings at 20 kPa PNP and then 16 recordings at 150 kPa PNP were made of a single microbubble upon ultrasound insonification at the different transmit frequencies with 80 ms in between recordings. To avoid any effects from nearby microbubbles on the oscillation behavior, only microbubbles which were at least 0.7 mm from other microbubbles were investigated.

To quantify microbubble oscillation, custom-developed image analysis software in MATLAB was used to determine the change in microbubble radius as a function of time (*R*—*t* curve) [19]. As previously described, the resonance frequency and shell parameters can be obtained from the spectroscopy dataset [11,19]. Briefly, the relative oscillation amplitude (*x*0) of each microbubble was defined as the maximum of the filtered *R-t* curve (a third-order Butterworth bandpass filter centered at *f<sup>T</sup>* with a 300 kHz bandwidth) and divided by the resting size of the microbubble (*R*0; mean size of the first five frames). Next, for each *fT*, the *x*<sup>0</sup> obtained at 50 kPa were fitted to the harmonic oscillator model:

$$\chi\_0 = \frac{|P|/\left(4\pi^2\rho R\_0^2\right)}{\sqrt{\left(f\_0^2 - f\_T^2\right)^2 + \left(\delta f\_T f\_0\right)^2}}\tag{1}$$

with *P* being the acoustic pressure and *ρ* = 10<sup>3</sup> kg/m<sup>3</sup> being the density of water. The eigenfrequency (*f* <sup>0</sup>) of the microbubble is defined as:

$$f\_0 = \frac{1}{2\pi} \sqrt{\frac{1}{\rho R\_0^2} \left[ 3\gamma P\_0 + \frac{2(3\gamma - 1)\sigma\_w}{R\_0} + \frac{4\chi}{R\_0} \right]} \tag{2}$$

with *γ* = 1.07, the ratio of specific heats for C4F10, *P*<sup>0</sup> = 10<sup>5</sup> Pa the ambient pressure, *σ<sup>w</sup>* = 0.072 N/m the surface tension in water, and *χ* the microbubble shell elasticity. The damping coefficient (δ) is given by:

$$\delta = \frac{\omega\_0 R\_0}{c} + 2\frac{4\mu}{R\_0^2 \rho \omega\_0} + \frac{4\kappa\_s}{R\_0^3 \rho \omega\_0} \tag{3}$$

with *<sup>ω</sup>*<sup>0</sup> = 2π*<sup>f</sup>* <sup>0</sup>, *<sup>c</sup>* = 1500 m/s the speed of sound in water, <sup>µ</sup> = 10−<sup>3</sup> Pa·s the viscosity of water and *κ<sup>s</sup>* the microbubble shell viscosity. The resonance frequency is defined by *fres* = *f*<sup>0</sup> √ 1 − *δ* <sup>2</sup>/2.

The variability in the acoustical response of each microbubble type was quantified by determining the interquartile range (IQR) of the relative oscillation amplitude (*x*0) at each *f* <sup>T</sup> and in diameter bins of 0.5 µm (*N* > 3 per bin). Since the microbubbles deflated after insonification, the acoustic stability was evaluated by quantifying the relative diameter decrease upon insonification as (*D*<sup>0</sup> − *Dend*)/*D*0, with *D*<sup>0</sup> the mean microbubble diameter of all 128 frames of the first recording without ultrasound and *Dend* the mean microbubble diameter of the last ten frames of the last recording.

The nonlinear behavior of microbubbles was assessed by calculating the fast fourier transforms (FFTs) of the *R*-*t* curves. The noise level of each microbubble was determined by the FFT of the first recording before the ultrasound. A microbubble was categorized as exhibiting nonlinear behavior when in at least two recordings it showed a detectable peak in the FFT (using the *islocalmax* function in MATLAB) around <sup>1</sup> 2 ·*f<sup>T</sup>* for the subharmonic or around 2·*f<sup>T</sup>* for the second harmonic and the peak's amplitude was at least 6 dB above the noise level. If so, then the amplitude of the nonlinear component was defined as the maximum FFT amplitude in a 300 kHz bandwidth around <sup>1</sup> 2 ·*f<sup>T</sup>* for the subharmonic component and around 2·*f<sup>T</sup>* for the second harmonic component and normalized to the fundamental at *fT*.

Finally, the confocal microscopy recordings were scored manually for the presence of buckles (none, single, multiple, or extensive) before and after the ultrasound and for change in the microbubble coating before and after ultrasound (unchanged, buckles formed, coating material shed). Only bright spots with 1 µm diameter or larger were classified as the buckle (Supplemental Figure S2). Microbubbles between 4.5 and 6.0 µm in diameter were manually scored for the LC domain size as well (mostly large, large and small, undefined). The relationship between these classifications and the acoustical data were evaluated to determine the effect of the lipid phase distribution and buckling in the microbubble coating on the resulting acoustic response. To rule out size-dependent differences in oscillation amplitude, only microbubbles with an initial diameter in the range of 4.5–6.0 µm were included in this analysis.
