Characterization of different bubble formulations for blood-brain barrier opening using a focused ultrasound system with acoustic feedback control

Focused ultrasound combined with bubble-based agents serves as a non-invasive way to open the blood-brain barrier (BBB). Passive acoustic detection was well studied recently to monitor the acoustic emissions induced by the bubbles under ultrasound energy, but the ability to perform reliable BBB opening with a real-time feedback control algorithm has not been fully evaluated. This study focuses on characterizing the acoustic emissions of different types of bubbles: Optison, Definity, and a custom-made nanobubble. Their performance on reliable BBB opening under real-time feedback control based on acoustic detection was evaluated both in-vitro and in-vivo. The experiments were conducted using a 0.5 MHz focused ultrasound transducer with in-vivo focal pressure ranges from 0.1–0.7 MPa. Successful feedback control was achieved with all three agents when combining with infusion injection. Localized opening was confirmed with Evans blue dye leakage. Microscopic images were acquired to review the opening effects. Under similar total gas volume, nanobubble showed a more reliable opening effect compared to Optison and Definity (p < 0.05). The conclusions obtained from this study confirm the possibilities of performing stable opening using a feedback control algorithm combined with infusion injection. It also opens another potential research area of BBB opening using sub-micron bubbles.


Methods and Materials
Focused ultrasound system and acoustic feedback control algorithm. The ultrasound system was adopted from our previous report 40 and modified accordingly (see details in Supplementary Information). To evaluate the frequency response of acoustic emissions during BBB opening, a hydrophone was integrated into the center of the focused ultrasound transducer to capture the acoustic signals emitted by bubbles stimulated within the brain. The hydrophone was constructed from a flat 20-mm diameter PZT composite material (DL-54, center frequency 0.75 MHz, DeL Piezo Specialties, LLC, West Palm Beach, FL, USA), which was mounted in a 3D-printed housing and connected electrically to a 50 ohm coaxial cable. The signal acquired by the hydrophone was sampled at 20 MHz using a 14 bit PCI digitizer (ATS460, Alazar Technologies Inc, Pointe-Claire, QC, Canada) (Fig. 1a). Although the hydrophone was not focused, it was placed coaxial with the axis of the acoustic beam from the transducer, giving it a preferential sensitivity to signals coming from the focal region. No amplification of the hydrophone signal was required due to the strength of the measured emissions from bubbles.
The feedback control algorithm was implemented in LabVIEW (LabVIEW 2014, National Instruments, Austin, TX, USA). A flow chart of the algorithm is shown in Fig. 2. Each second a 10 ms burst was transmitted into the brain from the transducer in order to stimulate circulating bubbles. During each burst, a Fast-Fourier transform (FFT) was performed on the collected data, and the area under the curve (AUC) was calculated by summing the amplitude at sub-or ultra-harmonics within a ±0.05 MHz bandwidth (1,000 samples). Since the resonant frequency of the hydrophone was 0.75 MHz, this study mostly focused on the acoustic emissions at this frequency, which is the ultra-harmonic of the 0.5 MHz transmitting frequency (1.5 × F 0 ). In order to improve stability, the AUC value that was input to the feedback algorithm was the average of the last three acquisitions. The average AUC was compared with a desired AUC threshold (AUC targ ), and the pressure for the next transmission of ultrasound was determined. In addition, a dead-band, ε, defined as 0.2 × AUC targ , was applied to reduce the sensitivity to noise. If the AUC was within the range of AUC targ ± ε the pressure level was not changed. For an AUC less or greater than AUC targ ± ε, the pressure level was increased or reduced by 0.01-0.03 MPa. This process of measuring the AUC from acoustic emissions, comparing with the target threshold, and adjusting the pressure to maintain the AUC within the desired range, continued until the end of the prescribed exposure duration.
Three sonication protocols were applied in this study: (A) constant pressure sonication, (B) pressure sweep, and (C) feedback controlled sonication. The flowchart of all three sonication protocols are shown graphically in Fig. 3. In the constant pressure sonication (Fig. 3a), ultrasound was delivered at a constant pressure and for a defined duration (120 seconds in a typical BBB opening experiment). This protocol was designed to evaluate the stability of various bubbles. The pressure sweep sonication (Fig. 3b) was used to observe the acoustic emission spectrum evolution with increasing pressure and determine the range of stable and inertial cavitation. Bubbles were infused through the sample volume during sonication, and the pressure was increased from 0.21 to 1.13 MPa in steps of 0.01-0.03 MPa. Up to 5 acquisitions were collected at each pressure level and the mean AUC and standard deviation across these measurements were recorded. For feedback controlled sonications (Fig. 3c), an initial 5 acquisitions were obtained prior to bubble infusion as a baseline reference. Once the infusion of bubbles started, the ultrasound pressure was increased until the measured AUC fell within the targeting range described above. At that point the pressure was adjusted for each pulse following the rules of the feedback algorithm until the total exposure time was reached. Feedback controlled sonication could be used to evaluate the feasibility of maintaining the stimulated bubble response at a fixed level and the corresponding opening effect.
In-vitro characterization of three agents. Bubble description and preparation. Three bubble-based agents were evaluated in this study: Optison, Definity, and a custom-made nanobubble 42 . A brief description of the three agents is given in Supplementary Table S1. Optison (GE Healthcare, Milwaukee, WI, USA) consists 10 mg human albumin and 0.22 ± 0.11 mg/ml octafluoropropane in a 3 ml vial 43 . Definity (Lantheus Medical Imaging, Billerica, MA, USA) is perflutren lipid microsphere. After mechanical activation, a 2 ml Definity vial  Flow chart for the feedback control algorithm. The digitizer captures the acoustic emissions from stimulated microbubbles during sonication, and a Fast Fourier transform (FFT) is then performed on the raw data. The area under curve (AUC) is calculated for sub-/ultra-harmonics (band width ± 0.05MHz). The mean AUC across 3 separate acquisitions is then compared to a target AUC threshold. If the measurement falls within a pre-defined error dead-band (Threshold ± ε), the pressure will be maintained, otherwise, a ±0.01-0.03 MPa adjustment will be applied to the transducer prior to the next acquisition. This algorithm runs continuously during the BBB exposure until the desired treatment time is reached. contains 0.75 mg lipid blend and 1.1 mg/ml octafluoropropane 44 . The nanobubble was prepared based on the method described in previous published articles 45 (see the details in Supplementary Information).
In this study, the gas volume was kept constant across all samples of each bubble, in order to make comparisons of the frequency response of the different agents. According to the bubble size, volume of a single bubble (unit: µl) could be calculated using equation (1): where d is the mean bubble diameter in µm. Minimum and maximum gas volume can be calculated based on the range of bubble size and bubble concentration. The mean gas volume for each agent is listed in Supplementary  Table S1. To achieve a similar gas volume for all three agents, different dilutions were performed for each agent for all experiments in this study (Supplementary Table S2).
In-vitro frequency response characterization. A flexible tube with outer diameter of 1 mm (EXT-12HF, SAI Infusion Technologies, Lake Villa, IL, USA) was oriented across the ultrasound focus (Fig. 1b). For each experiment, the bubble solution was infused using an MRI compatible syringe pump (Chemyx NanoJet Stereotaxic Syringe Pump, Chemyx, Stafford, TX, USA) at a rate of 0.15 ml/min to refresh the bubble population within the focal volume of ultrasound after each pulse. All three sonication protocols were evaluated in the in-vitro study for each of the bubble formulations. Samples with five different gas volumes (listed in Supplementary Table S2) were characterized. The detailed characterization methods can be found in Supplementary Information.
In-vitro feedback control evaluation. The feasibility of maintaining the AUC of stimulated bubbles at a target level during sonication was first evaluated in-vitro prior to animal studies. Based on preliminary calibration experiments, an AUC controlling level of 5,000 was selected as it being the minimum consistent threshold with adequate signal-to-noise ratio (SNR). Bubble solutions of each agent were prepared with a gas volume of 1.1-1.2 µl/ml, and were infused into the flow tube via a syringe pump at a rate of 0.15 ml/min. The total sonication duration was 100 seconds and ultrasound power was adjusted after each pulse based on the difference between the measured AUC and the desired threshold.
Animal studies. Animal preparation. Female rats (Sprague Dawley, 230-300 g, n = 40) were used in this study. All procedures were approved by UT Southwestern Institutional Animal Care and Use Committee and followed guidelines set forth by the Guide for the Care and Use of Laboratory Animals. Nine rats were used for primary in-vivo evaluation of the system. Twenty-five rats were used for in-vivo characterization and evaluation of three different agents and 6 rats were utilized for fluorescence microscopy. Detailed procedures are described in Supplementary Information.
Ultrasound exposure. The ultrasound setup for the animal study is shown in Fig. 1c. For bolus injection, Optison was injected with the dosage of 30 µl/kg. To facilitate accurate administration, the bubble solution was diluted with saline at a ratio of 1:20. To achieve a similar total gas volume (1.1-1.2 µl/ml) per dose for comparison, Definity was given at a dosage of 6 µl/kg with 1:100 dilution of saline. In the same manner, nanobubble was administrated as 737 µl/kg with 1:1 dilution. A constant pressure sonication was performed with the focal pressure to be 0.47 MPa calculated based on the insertion loss in rodents 46 . At least 5 minutes were allowed between two sonication for the bubbles to clear from circulation. The infusion administration was controlled at 0.3 ml/min infusion rate using an MRI compatible syringe pump. The bubble solutions were prepared for three agents to reach a total gas volume of 1.1-1.2 µl/ml. For pressure sweep sonication, three acquisitions were acquired at each power level starting from 0.1 to 0.7 MPa. The duration of each power sweep sonication was approximately 125 seconds. For feedback controlled sonication, the desired threshold was set to be 5,000 for the ultra-harmonic of 0.75 MHz. This value was selected arbitrarily based on the previous characterization results as well as the consideration of signal-to-noise ration. The sonication duration for each target was approximately 100 seconds.
Microscopy. In order to assess brain tissue microstructure following BBB opening, after sacrificing the animal, brains were flash frozen and embedded in OCT compound (Sakuea Finetek, Torrence, CA) for cryosectioning. 10 µm serial sections were collected and stored at −80 °C until use. Micrographs of each whole brain slice were created by using stitching assistant software (NIS Elements, Nikon Instruments Inc, Melville, NY, USA) to combine adjacent 40x images. Details about microscopic imaging setups are included in Supplementary Information.

Statistical Methods.
Statistic studies were performed on three aspects of the results to evaluate the difference among three bubbles in terms of ultrasound persistence, characteristics under pressure sweep, and the performances under in-vivo feedback control. All analysis was performed in SAS 9.4 (SAS Institute Inc., Cary, NC). Details for statistical studies are described in Supplementary Information.

Results
In-vitro characterization of three agents. Figure 4 shows the relationship between the frequency response (AUC) and focal pressure. Using samples with gas volume of 1.1-1.2 µl/ml, the detected AUC increased with higher focal pressure (Fig. 4a). At a maximum focal pressure of 1.13 MPa, the AUC level for Optison, Definity, and nanobubble was 31000 ± 370, 24000 ± 150 and 28000 ± 160 respectively. According to statistical assessment, the corresponding focal pressure level at the first break point is 0.37 MPa for Optison, 0.47 MPa for Definity and 0.59 MPa for nanobubble. Nanobubble required a higher power to trigger the rising phase compared to both Optison and Definity (p-value < 0.0001). The frequency spectrum at various focal pressures are included in panel b-d showing the transition from stable to inertial cavitation. At a lower focal pressure of 0.21 MPa (Fig. 4b), acoustic emission at the ultra-harmonic of 0.75 MHz was detected while with higher focal pressure of 0.47 MPa (Fig. 4c), broad-band emissions were observed for Optison and Definity, but not for nanobubble. As the focal pressure was increased to 0.91 MPa (Fig. 4d), all three agents showed broad-band emissions, which is believed to be an indicator of inertial cavitation. Figure 5a,b shows the dependence of the AUC (0.75 MHz) and persistence on gas volume for all three agents. A higher AUC was observed for samples with a greater gas volume, although this reached a plateau above 0.5 µl/ ml for Definity and nanobubble (Fig. 5a). Under the maximum focal pressure of 1.13 MPa, samples for Optison, Definity and nanobubble with less gas volume (0.06 µl/ml) presented the AUC level of 13000 ± 450, 20000 ± 310, and 21000 ± 340. On the other hand, with more gas volume (1.1-1.2 µl/ml) the AUC level was 27000 ± 100, 23000 ± 190 and 24000 ± 810, respectively. Interestingly, the AUC level of Definity started decreasing when the gas volume was larger than 0.23 µl/ml. Figure 5b evaluates the persistence of acoustic emissions as a function of gas volume. A longer persisted time was measured on samples with more encapsulated gas. For nanobubble samples with gas volume higher than 0.4 µl/ml, AUC level didn't return to baseline value after 200 seconds, at which point data collection was terminated. Nanobubble had significantly longer persistence than both Optison and Definity at any gas volume (p-value < 0.0001), no significant difference was found between Optison and Definity (p-value > 0.2).
In-vitro feedback control. The result of feedback control using the tube phantom is shown in Fig. 5c-e. Panel c-e are corresponding AUC and pressure curves for three agents respectively. With AUC targ set at 5,000, the average AUC level across the control period for Optison, Definity, and nanobubble is 5490 ± 260, 5120 ± 330, and 5510 ± 240. The focal pressure required to maintain the AUC level at the AUC targ was 0.41 ± 0.00, 0.39 ± 0.00 and 0.82 ± 0.01 MPa, respectively. Effective feedback control was confirmed in-vitro. For the in-vitro setup, the AUC could be maintained at the target value with very little change in pressure once the appropriate level was reached.
In-vivo characterization of three agents. Figure 6 shows the AUC level observed during constant pressure sonication with bolus injection of three agents. Figure 6a shows the AUC over time and b-d are corresponding frequency spectrums over time shown as a spectrogram. Significant AUC increases were observed for all three agents shortly after the bolus injection, once bubbles reached the brain circulation. Definity and nanobubble had higher maximum AUC levels (39635.4 at 16 seconds for Definity, 47072.4 at 49 seconds for nanobubble) than Optison (27151.5 at 17 seconds). At the end of the sonication, the AUC level for Optison returned to baseline (AUC = 415.9) while weak frequency response still observed with Definity (AUC = 3654.9). A strong frequency response was still detected with nanobubble at the end of the sonication (AUC = 34576.4) indicating a longer in-vivo circulation time. The amplitude of acoustic emissions at ultra-harmonic of 0.75 MHz are shown in b-d. Continuous acoustic emissions across the entire sonication period were observed with nanobubble (Fig. 6d).  Figure 7 shows the in-vivo frequency response of three agents for a pressure sweep sonication. The AUC was plotted as a function of focal pressure, with a correction applied for insertion loss through the skull 46 (Fig. 7a). All three agents showed AUC increases compared to baseline acquired prior to bubble infusion (maximum AUC = 453.5 ± 28.7). Definity and nanobubble showed higher maximum AUC (62000 ± 1000 at 0.6 MPa for Definity, 80000 ± 4900 at 0.6 MPa for nanobubble) than Optison (21000 ± 1800 at 0.6 MPa). Panel b-d are frequency spectrums acquired at different focal pressure indicating the presence of either stable cavitation or inertial cavitation. Under 0.10 MPa (Fig. 7b) no acoustic emission at ultra-harmonic was detected for all three agents (frequency amplitude < 0.5). At approximately 0.16 MPa (Fig. 7c), acoustic emission at 0.75 MHz was detected with nanobubble (frequency amplitude = 9.28) while not with other two agents. At 0.47 MPa (Fig. 7d), broad band emissions were detected with all three agents.  In-vivo feedback control. A feedback-controlled sonication protocol was applied in-vivo with each of the three agents, with representative results shown in Fig. 8. The first column shows the AUC and delivered pressure during controlled treatment with each of the agents. With the controlling threshold set to be 5,000, the overall AUC level across all animals was 4690 ± 2745 for Optison, 4650 ± 3355 for Definity and 4870 ± 1853 for nanobubble. Transcranial focal pressure required to maintain the AUC level was 0.42 ± 0.04 MPa for Optison, 0.34 ± 0.03 MPa for Definity and 0.23 ± 0.02 MPa for nanobubble. Nanobubble had a significantly lower standard deviation of AUC than Optison (32% lower, p-value = 0.0019) and Definity (44% lower, p-value = 0.0002). It also required a lower focal pressure for control (0.19 MPa lower than Optison, p-value = 0.017 and 0.11 MPa lower than Definity, p-value = 0.0065). The variability of the required pressure (calculated as standard deviation of the required focal pressure) was also smaller for nanobubble compared to Optison (47% lower, p-value < 0.0001) and Definity (36% lower, p-value = 0.0147). Overall, nanobubble showed a better controlling performance compared to the other two agents.
BBB opening effect with feedback controlled sonication. The second column in Fig. 8 shows block face photos of a brain dissection slice transverse to the ultrasound beam and at the level of the focus. Localized Evans blue dye leakage indicates BBB opening, and is present for all three agents. Immunofluorescent staining was performed to observe the degree of BBB opening and the cellular morphology of FUS targeted regions. The third column in Fig. 8 shows micrographs of 10-micron brain sections indicating the location of BBB opening with fluorescent Evans blue dye (red) in the target location. Regions of BBB opening were selected (indicated by white square) for imaging of endothelial cell morphology (green) as well as overall cellular morphology (blue). Overall cellular morphology appeared normal across regions of BBB opening however, endothelial cell changes were observed in regions of BBB opening compared to non-targeted regions. All three agents presented with endothelial cells that were wider in morphology and in the Optison and nanobubble groups the endothelial cells appeared more aggregated and uneven 47 in regions of BBB opening compared to non-targeted regions. Furthermore, in the nanobubble and Optison groups, the morphological changes were more profound than in the Definity group which seemed to correspond to the degree of Evans blue dye leakage (red). Though further study is needed, these results suggest that bubble type may be sufficient to alter the effect of FUS on BBB cells and that RECA-1 maybe an attractive cellular marker for the degree of BBB opening in future studies evaluating the biological effects of BBB opening with feedback controlled sonication.

Discussion
Ultrasound contrast agents are commonly used clinically in patients with suboptimal echocardiograms. When comparing the diagnostic performance of Definity, Optison and SonoVue ® in a preclinical model, no significant differences were detected among the different agents 48 . However, when being applied for BBB opening with focused ultrasound, differences in the acoustic characteristics as well as the stimulated emissions may impact how to best use each agent for this application. The acoustic behavior of bubbles depends on the properties of the surrounding medium, applied acoustic pressure, and the physical characteristics of the agent such as size distribution, internal gas, and shell type 49 . Wu et al. 50 compared three contrast agents (Definity, SonoVue ® and USphere) with similar bubble dosage (4 × 10 7 bubbles/kg), and found that the BBB opening effect can be predicable and bubble-independent given carefully controlled bubble dosage and administration. In our study, due to the significant variations in bubble size and concentration across the agents investigated, performing the characterization test with same number of bubbles was not considered a suitable control. Instead, we selected to compare their frequency response with similar total encapsulated gas in order to take both the bubble size and bubble concentration into consideration. These results might open another pathway for using different agents for BBB opening and drug delivery into the brain.
With a single tube phantom, the measured stimulated emissions with increasing pressure demonstrated a higher focal pressure was required to detect a harmonic response from the nanobubble (Fig. 4a). This result matches the finding from a previous study showing a higher acoustic pressure threshold is required with smaller bubble to achieve BBB opening 19 . As expected, samples with larger gas volume had a higher AUC response (Fig. 5a). Interestingly, each agent exhibited its own "saturation threshold" where the AUC no longer increased with increasing gas volume. For example, the maximum AUC level of Definity started to drop when the gas volume is larger than 0.23 µl/ml, while nanobubble started saturation with even lower threshold of 0.06 µl/ml. Under higher ultrasound exposures like shown in Fig. 5b (focal pressure = 1.13 MPa), bubble destruction impacts the persistence of the stimulated emissions 51,52 . Higher bubble concentration led to longer persistence from the bubbles, however the persistence time for the nanobubble was an outlier compared to the other agents (Fig. 5b). The exact mechanism for this increased persistence in still unclear, but may be related to the much increased number of bubbles for a given gas volume (due to the small size).
The in-vivo pressure sweep results were quite different from the in-vitro results. (Fig. 7a). Compared to Optison, Definity and nanobubble exhibited a much stronger in-vivo frequency response. Furthermore, the maximum in-vivo AUC levels for Definity and nanobubble were approximately 2-3 times higher than their in-vitro response. However, Optison maintained a similar AUC level both in-vivo and in-vitro. It is worth noting that the bubble size distributions are likely to evolve immediately after injection into the body 53 , which might account for some of these changes. Temperature change, in-vivo blood pressure and other factors such as gas flux might also contribute to alter the bubble distribution. The bio-distribution of these three agents might also be different, which is not addressed or controlled in this study. From the frequency spectrums acquired during the pressure sweep, a lower acoustic pressure level was required with nanobubble to observe ultra-harmonic activity (Fig. 7c), suggesting successful BBB opening can be achieved with nanobubble under relative lower focal pressure. This observation differed from the previous study performed in mice showing a higher acoustic pressure threshold was often required with smaller bubble 19 . One possible explanation is the nanobubble has a relatively higher concentration, certain amount of small nanobubbles may have a chance to combine with each other and form a cluster of bubbles with larger diameter compared to both Optison and Definity, but confirmation of this requires further investigation.
Previous studies have successfully implemented various feedback control algorithms to achieve BBB opening with focused ultrasound 28,32,50 . These studies, however, were mostly conducted with bolus injection where the systematic bubble concentration varies significantly after injection across a 2 minutes treatment. In our study, a steady infusion of bubbles was implemented for the first time. During bubble infusion, the focal pressure was adjusted in real-time to maintain the AUC level at a certain threshold, which means acoustic emissions were present throughout the entire treatment. This differs from protocols developed by others in which the focal pressure was reduced upon detection of sub-or ultra-harmonic emission to a level where these emissions were no longer present. Further study is required to determine whether the extent of BBB opening and impact on surrounding brain tissue is different from previous approaches.
Another limitation is the challenges and considerations for clinical translation of these concepts. First, the sonication duration required to achieve BBB opening at a particular AUC threshold should be carefully characterized. Second, the microbubble concentration required to obtain adequate signal to noise ratio from the acoustic receivers should be investigated. Last but not least, the system should have the ability to cover multiple targets in the brain to create a BBB opening over a larger region. The last point could be addressed with a phased array ultrasound system 54 . The other two factors, however, have not been fully evaluated yet.
The controlling threshold selected in this study (T = 5,000) was an arbitrary selection based on the characterization results and previously determined pressure thresholds for successful BBB opening 13 . This threshold also satisfied the adequate signal-to-noise ratio. However, a big limitation exists in this current setup is the lack of link between our selection of AUC value and the corresponding biological effect. Further studies to determine the appropriate controlling threshold to achieve BBB opening without causing damage to the surrounding tissue are required. The corresponding biological effect of various controlling threshold will be evaluated as well. The robustness of feedback control algorithm was evaluated both in-vitro and in-vivo. Clear and stable control of acoustic emission levels at the ultra-harmonic was successfully achieved in the single tube phantom for all three agents (Fig. 5c-e). In the rat model, although successful BBB opening was achieved with all three agents using feedback control, the variability in acoustic emission levels during feedback controlled sonication varied a lot among three agents. One possible reason is the difference between experimental animals. The skull thickness, the size of the brain, and the vascular structure will result in an uneven spatial distribution of bubbles in the brain. Secondly, due to the possibility of different bio-distribution amongst the agents, using the same infusion rate for all three agents could result in different amount of bubble within the targeted region inside the brain. To perform reliable BBB opening with feedback control, the algorithm should be optimized for specific agents according to their acoustic properties and bio-distribution.
In BBB opening treatment with bolus injection, the circulation time plays a critical role especially during multiple spots sonications. Conventional multiple spot sonications were often performed in an alternating fashion where the transducer move between targets and return to its original location within one pulse 32,54-60 . In our study, the circulation time measured from the frequency response decay after a bolus injection (Fig. 6a) could serve as a valuable factor in optimize the sonication protocol. According to the in-vivo characterization results acquired in this study, nanobubble has a circulation time of approximately 6-8 minutes. This circulation time is much longer compared to both Optison and Definity (1-2 minutes). Previous studies from other groups have evaluated the effectiveness of multiple sonication after 1 dose of bolus injection 50 . An agent with longer circulation time will definitely simplify the treatment protocol. When combined with this specific nanobubble, FUS energy could be delivered across multiple spots without extra administration of agent. At the same time, this property will highly benefit the ability to cover a larger region without jeopardizing the treatment effect.