Lipid microbubbles as a vehicle for targeted drug delivery using focused ultrasound-induced blood–brain barrier opening

Synthesis of fluorescent microbubbles

The fluorescent microbubbles were manufactured based on our previously published protocol for in-house and size-isolated lipid-coated microbubbles⁴² (Supplementary methods).

Briefly, they were formulated by dissolving 90% mol 1,2-distearyol-sn-glycero-3-phosphocoline (DSPC) and 10% mol 1,2-distearyol-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)2000] (DSPE-PEG2000) (Avanti Polar Lipids, Alabaster, AL) in a solution consisting of phosphate-buffered saline (PBS) 1×/glycerol (10% volume)/propylene glycol (10% volume). After dissolving, the gas core, perfluorobutane (PFB) at 99% wt. purity (FluoroMed, Round Rock, TX), was introduced in order to activate the microbubbles. Immediately after activation, the 4–5 µm microbubbles were isolated and their concentration and size distribution were measured.

The microbubbles were converted into fluorescent by the addition of the lipophilic fluorescein probe, C-12 (Life technologies, Eugene, OR) (molecular weight 529.63 g/mol, maximum emission and excitation wavelengths 497 ± 3 and 518 ± 4 nm, respectively) using a post-labeling technique. C-12 contains a C₁₂ alkyl chain (Figure 1(a)) that will bind to membranes with the fluorophore at the aqueous interface and the alkyl tail protruding into the lipid interior (Figure 1(b)). C-12 was dissolved in pure ethanol at a concentration of 10 mM and put in direct contact with the microbubbles. Due to its amphiphilic nature, direct contact with the C-12 solution led to microbubbles binding to it. To allow the diffusion of C-12 to all the bubbles, the 1-mL syringe containing the suspension of microbubbles plus C-12 was incubated during 2 h at room temperature while it was rotating at 40 r/min. The size distribution and concentration of the microbubbles after making them fluorescent were determined again with the Coulter Counter Multisizer III. Figure 1(c) illustrates the appearance of the fluorescent microbubbles. It is clear that all the fluorescence is concentrated on the lipid shell.

Preparation of animals

All experiments were performed in strict accordance with the National Institutes of Health Guidelines for animal research; all animal procedures for these experiments were reviewed and approved by the Institutional Care and Use Committee of the Columbia University. A total of 47 male mice (C57BL/6, Harlan, Indianapolis, IN) (age: 3–4 months old, weight: 20–25 g) were used in this study. The animals were divided into seven cohorts depending on the applied acoustic pressure and time of sacrifice (Table 1). Randomization had not been performed because all the mice were males from the same strain with similar age and weight.

All the mice were handled as in previous studies conducted by our group^21,27 (Supplementary methods).

In vivo experimental procedure and MRI

The experimental FUS setup, as shown in Figure 2(a), was used as previously described^24,29 (Supplementary methods).

The targeting procedure was carried out with a grid system to locate the sutures of the skull; the procedure has been described in great detail elsewhere.^6,51 Briefly, the FUS focus was placed 3 mm below the skull and 18% attenuation was accounted for acoustic pressure loss through the skull.¹⁰

In this study, a pulsed FUS (pulse length 6.56 ms; pulse repetition frequency 5 Hz; duration 5 min) beam at acoustic pressures between 450 and 750 kPa was targeted transcranially to the left caudate-putamen (Cau-Pu), while the right side served as a control. Hence, each animal served as its own control, thereby reducing the variability caused by physiologic differences among animals.

Prior to fluorescent microbubbles administration, a 30-s sonication using the same acoustic parameters described above was applied in order to measure the baseline background signal needed in the acoustic emission analysis. The injected microbubbles were freshly diluted before each injection. A bolus of 50 µl diluted fluorescent microbubbles in saline solution (8 × 10⁸ microbubbles/ml) was injected intravenously through the tail vein immediately before sonication (Figure 2(b)). The microbubbles reached the targeted region in under 15 s (time window between the injection of the fluorescent microbubbles and the beginning of sonication) according to the spectrograms obtained from the PCD data (Supplementary Figure 1). Here, comparing the spectrogram corresponding to the control (Supplementary Figure 1(a), FUS without microbubbles), where only the reflected signal (1.5 MHz) and the effect from the skull (3.0 MHz, the second harmonics) are observed, with the spectrogram corresponding to the first FUS pulse after injecting the microbubbles (Supplementary Figure 1(b)). The harmonic, ultraharmonic, and broad-band emissions associated with the oscillation of the microbubbles are present since the beginning of sonication. One control cohort was used, where ultrasound was applied at 450, 600, and 750 kPa, and C-12 (2 µl at 10 mM) dissolved in 50 µl of saline solution (final concentration of C-12 0.4 mM) was co-administered 30 s after in the presence of non-fluorescent microbubbles to serve as the basis for comparison in the fluorescent imaging analysis (Supplementary methods).

After sonication, post-contrast T₁-weighted (T₁-w) MRI was conducted in order to confirm the BBB opening by a 9.4-T MRI system (Bruker Medical, Boston, MA) (Supplementary methods).

Four additional mice sonicated at 750 kPa were survived two weeks in order to analyze when the BBB integrity was restored. In these mice, T₁-w post-Gd injection MRI was repeated on a daily basis starting from the day of sonication and lasting up to 14 days after sonication.

In some mice, T₂-weighted (T₂-w) MRI was acquired for safety evaluation (presence or absence of edema at the sonicated region) (Supplementary methods).

The same four mice used for studying the closing timeline of the BBB were also imaged with T₂-w MRI on a daily basis to determine edema progression.

The mice were sacrificed at different time-points to assess the fluorescence delivery and microscopic damage 2 h after sonication in order to evaluate the fluorescence delivery and short-term microscopic damage. This time-point was selected to enable the detached C-12 to diffuse into the brain parenchyma in the region of BBB opening; and one week to evaluate long-term microscopic damage and potential microglial activation (Supplementary methods).

Cavitation emission quantification

The acoustic emissions from the ultrasound-activated fluorescent microbubbles were detected and quantified, as described previously,^33,35,44 using stable cavitation dose (SCD) and ICD to assess the two types of cavitation activity (Supplementary methods).

Fluorescence imaging analysis

The efficiency of the C-12 delivery was assessed based on the quantification of the fluorescence enhancement in the targeted regions.

A spinning-disk confocal (CSU10; Yokogawa, Tokyo, Japan), inverted microscope (Eclipse Ti; Nikon, Melville, NY) with an electron-multiplying charge-coupled device camera (Hamamatsu Photonics, Hamamatsu, Japan) and 100×/1.4 numerical aperture objective with a 1.5 × magnifier was used for image acquisitions of the fluorescent microbubbles. Image acquisition was performed using Micromanager 1.4.⁵²

Epi-fluorescence images of the brain sections were captured using an Olympus DP30BW digital camera mounted on an upright Olympus BX61 microscope. The fluorescence enhancement was quantified in a similar manner following the previously published protocol⁵³ (Supplementary methods).

In order to evaluate the spatial location of the fluorescence distribution in the brains, the slides were mounted with ProLong® Diamond Antifade Mountant (Thermo Fisher, Waltham, MA) with the nuclear marker DAPI.

Histological analysis

Whole brain histological examinations were performed using H&E stain for general histology (Supplementary methods).

Thirteen mice were used for histological observations: Four mice sacrificed 2 h after sonication at 600 (n=2) and 750 kPa (n=2); nine additional mice survived during one week to evaluate the length and severity of the damage, it means, for intensity quantification of the damage and its recovery.

Both the sonicated and nonsonicated sides were evaluated.

Immunohistochemistry

In order to detect potential microglial activation as consequence of the treatment with FUS, Iba1 bright-field immunohistochemistry was performed in three mice sonicated at 750 kPa and survived for one week (Supplementary methods).

Statistical analysis

Statistical analysis was performed using GraphPad Prism (Version 5.01, La Jolla, CA). For each study group, the mean ± standard deviation of the fluorescence enhancement was calculated. Successful C-12 delivery was concluded if the fluorescence enhancement was statistically higher than zero (one-tailed, one-sample Student's t-test). Unpaired two-tailed Student's t-tests were used to determine whether the ICD, SCD_h, and SCD_u, at 30 s, 1 min and 5 min, were significantly different between the cases, where C-12 was delivered and the cases where it was not. A p-value < 0.05 was considered to represent a significant difference in the analysis.

Fluorescent microbubble generation

The size distribution of microbubbles before and after tagging with fluorescence is showed in Figure 3. After rendering them fluorescent, the mean size decreased by approximately 12% (from 4.56 ± 0.76 µm to 4.00 ± 1.13 µm) while maintaining monodisperse distribution.

However, while non-fluorescent microbubbles are stable for, at least, three days, fluorescent microbubbles become quickly unstable in a matter of hours (data no shown) and they had to be used on the same day of the preparation.

BBB opening and targeted fluorophore delivery

Table 1 summarizes all the results obtained in this study. The BBB opening was revealed by post-Gd T₁-w MRI (Figure 4(a) to (c), while fluorescence delivery was assessed by epi-fluorescence microscopy (Figure 4(d) to (e)). Not all the mouse brains were imaged by fluorescence microscopy since some of them were used for H&E staining, gross pathology (images not shown), and Iba1 immunohistochemistry in order to evaluate the safety of our method. In all cases, fluorescence delivery was achieved at the site of the BBB opening, co-localized with fluorescence enhancement regions.

Figure 4(d) and (e) shows representative fluorescence images obtained from mouse brains sonicated at their left Cau-Pu at pressures of 600 and 750 kPa, respectively (at 450 kPa fluorescence was not observed in any case). Images of the left Cau-Pu exhibited successful delivery of C-12 in the sonicated region, while images of the right Cau-Pu depict the contralateral (unsonicated) side, where fluorescence enhancement was never observed. At 600 kPa, the BBB was opened in all the cases and the C-12 was successfully delivered in three out of the nine cases, while at 750 kPa, the BBB was also opened in all the cases and FUS treatment resulted in a successful delivery of C-12 throughout the left Cau-Pu in three out of the five cases. In order to reflect correctly, the spatial location of C-12 throughout the sonicated region the nuclear marker DAPI was added to the slices of interest.

Therefore, the likelihood of a successful delivery in the sonicated region increases with the FUS pressure. Under the control conditions (450, 600, and 750 kPa with co-administration of non-fluorescent microbubbles and free C-12) no mouse showed significant fluorescent enhancement despite achieving 100% opening efficiency. There was no statistically significant difference in fluorescence enhancement between the 600 and 750 kPa (p = 0.36) cases, but the findings were statistically significant when comparing 600 and 750 kPa against the 450 kPa and control cases (Figure 4(f)).

Microbubble cavitation emissions

In Figure 5, two cases with typical cavitation activities (ICD, SCD_u, and SCD_h) are shown: a case where the C-12 was successfully delivered (Figure 5(a)) and a case where it was not delivered (Figure 5(b)).

ICD, SCD_u, and SCD_h were calculated at three different time-points (30 s, 1 min, and 5 min) in order to establish thresholds that could explain why the C-12 delivery was not successful in all cases. Figure 5(c) to (e) shows the comparison among the ICD at (c) 30 s, (d) 1 min, and (e) 5 min. For 30 s and 1 min, the quantified ICD showed a significant increase between the cases where the fluorophore was successfully and unsuccessfully delivered to the target area. However, no statistical difference was determined at 5 min. Figure 5(f) to (h) shows the equivalent case for the SCD_u at (f) 30 s, (g) 1 min and (h) 5 min. Similar to ICD, SCD_u showed significant increase between the cases where C-12 was successfully and unsuccessfully delivered to the target area for 30 s and 1 min. No statistical difference was observed at 5 min. Figure 5(i) to (k) depicts smaller findings for the SCD_h at (i) 30 s, (j) 1 min, and (k) 5 min. SCD_h only showed significant increase between the cases where C-12 was successfully and unsuccessfully delivered to the targeted area for 30 s. No statistical difference was observed at 1 and 5 min.

According to the results shown in Figure 5, ICD, SCD_u, and SCD_h thresholds have been established (Supplementary Table 1). These thresholds were defined as two times the standard deviation above the mean cavitation dose of the non-fluorescent delivery. Applying higher cavitation doses than these thresholds may indicate a higher probability that C-12 would be delivered to the targeted (sonicated) region.

Safety

In order to evaluate the safety of this study, i.e potential edema, possible erythrocyte extravasations, dark or necrotic neurons, gross hemorrhage and/or microvacuolations tissue, BBB opening closing time, and likely microglial activation were evaluated by T₂-w MRI (edema), H&E stain (damage), T₁-w MRI (BBB closing), and Iba-1 immunohistochemistry (microglial activation) in the sonicated brains.

In some mice sacrificed about 2 h after sonication, T₂-w images were acquired 20 min after sonication and no edema was observed in most of the cases (data not shown), but in mice survived for one week after sonication, T₂-w was repeated and edema was detected in all of them at 600 (Figure 6(a)) and 750 kPa (Figure 6(b)). In order to check whether the edema is restored, four mice were sonicated at 750 kPa and survived for two weeks. These mice were scanned on a daily basis to determinate the temporal extension of the edema. It was observed that in three out of the four cases, edema was repaired between days 3 and 4 (Supplementary Figure 2(b)), while in the other mice the edema remained after two weeks.

These four mice were also scanned on a daily basis to demonstrate that the BBB integrity was restored. The BBB returned to its original condition between days 4 and 5 (Supplementary Figure 2(a)) in three out of the four cases. The mouse that showed permanent edema also showed permanent opening.

H&E was performed in brains sonicated at 600 and 750 kPa. At these pressures, which are above the IC threshold, some of the brains showed microscopic damage. In Figure 6, microscopic images showing more obvious damage at 600 kPa are shown (at higher pressures virtually all the brains showed microscopic damage). Within the FUS-sonicated Cau-Pu (Figure 6(c)), microhemorrhages, few necrotic cells, and microvacuolations were found.

Twelve additional mice were sonicated at the same pressures with FUS in the presence of fluorescent microbubbles and survived one week to test if damage remained or disappeared. Nine were used for H&E staining and three for Iba-1 immunohistochemistry. According to H&E staining, damage was repaired in four out the five cases at 600 kPa, but in most (three out of the four) cases at 750 kPa, by comparing the treated with the untreated side, a higher number of cell nuclei (blue/purple color) was observed in the sonicated region compared to the unsonicated side. Figure 6 depicts the case with potential damage at 600 kPa (Figure 6(e)) and one representative case at 750 kPa (Figure 6(g)). Three additional mice sonicated at 750 kPa and survived for one week were studied with Iba-1 bright-field immunohistochemistry in order to determine if these cell nuclei represented microglial activation (Figure 6(k)). Comparing Figure 6(i) and (k) shows that in the regions where the BBB was disrupted with FUS, although the BBB was restored (Figure 6(j)), there is microglial activation one week after treatment.

Fluorescent microbubbles generation

As a consequence of making the monodisperse 4–5 µm microbubbles fluorescent, a decrease in their mean size of about 12% occurred. This means that we might need to apply higher pressures to achieve the same cavitation (stable and inertial) levels that in the non-fluorescent microbubbles, since the smaller the microbubbles are, the higher acoustic pressure is needed to apply them to open the BBB.⁵⁴ However, this decrease is not deemed big enough to generate a significant effect over the cavitation levels.

The fluorescent microbubbles maintained their monodisperse distribution. There are some very large microbubbles (more than 10 µm) (Figure 1(c)), but they did not represent any safety issues. Commercial acoustic contrast agents approved by the FDA for intravascular injection, such as Sonovue, Optison, etc. contain microbubbles higher than 10 µm; for example, Optison, although its mean diameter is 3.0–4.5 µm, contains microspheres up to 32 µm. Moreover, in our case, as it is shown in Figure 3(b), the > 10 µm concentration is negligible comparing with the microbubbles concentration in the range 2–6 µm.

BBB opening and targeted fluorophore delivery

Our fluorescent microbubbles have proven their utility as carriers of therapeutic agents in the sonicated region of the brain, concretely the left Cau-Pu in this study (successful delivery rates of 0% at 450 kPa and control conditions; 33% at 600 kPa and 60% at 750 kPa) with a successful BBB opening rate of 100% (Table 1).

Possible explanations of these successful rates are as follows. Acoustic pressures higher than 450 kPa are needed to achieve targeted delivery, so this pressure may denote the lower threshold.

Under control conditions, it means, when C-12 was co-administered freely with the microbubbles, i.e. not anchored onto their lipid shell, the delivery was unsuccessful, even at high pressures, to the sonicated region. This may be explained in terms of the amphiphilic nature of C-12. As with all the amphiphilic molecules, the C-12 presents a free lipophilic terminal, which is where the bond between the lipid shell and the fluorophore occurs. Therefore, if the C-12 is co-administered via tail vein injection, the lipophilic terminal is free and it will be trapped by lipid membranes of the endothelial cells in blood vessels and C-12 will not be able to reach the mouse brain.

Regarding 600 and 750 kPa, the successful rates may be explained by the fact that fluorescent microbubbles become quickly unstable, so the mice sonicated first, when the fluorescent microbubbles are fresher, have higher probability of showing fluorescent enhancement than mice sonicated later, when the fluorescent microbubbles are less fresh. Future studies should be performed in order to characterize the relationship between successful fluorescence delivery and freshness of the fluorescent microbubbles.

There is a relationship between the amount of C-12 delivery and the acoustic pressure; it was found that there was a relation between 450 and 600 kPa and 450 and 750 kPa, but more studies should be done to assay if this relationship exists between 600 and 750 kPa (Figure 4(f)).

Microbubble cavitation emissions and safety evaluation

Transcranial PCD from oscillating fluorescent microbubbles has been conducted to characterize the types of cavitation during sonication in order to predict, in vivo, if the C-12 targeted delivery was successful or not. PCD has been used previously as a tool for predicting BBB opening, both in mice^{21,30,35,50,55,56} and in NHP.³³ In this study, transcranial PCD was utilized to define, for the first time, SCD_u, SCD_h, and ICD thresholds for successful targeted drug delivery. These thresholds are displayed in Supplementary Table 1 and were defined as twice the standard deviation above the mean of non-fluorescence (unsuccessful) delivery cavitation doses. Interestingly, these thresholds indicate that sonicating more than 2 min is not necessary to deliver C-12. This may be explained by the fact that, due to the high ICD, a significant amount of microbubbles collapsed during the first 2 min of the sonication. Therefore, during the rest of the sonication, less cavitation nuclei were available for opening the BBB and delivery the C-12. Pressures higher than 750 kPa were not tested because previous studies conducted in our laboratory showed that at higher pressures (810 kPa, 900 kPa) inertial cavitation occurs always, which might denote a successful delivery rate of 100%, but with a higher probability of microhemorrhage.^21,54 However, other parameters, like microbubble size, composition of the lipid shell and/or gas core, pulse length, etc. will be included in future studies to evaluate and clarify the effects of SC and IC activities.

In Figure 5(a) and (b), it is observed that the success rate was probably determined by the ICD, because when it was significantly higher than zero, part of the microbubbles collapsed and C-12 was able to be delivered to the target (Figure 5(a)), but when the ICD was not significantly higher than zero, the microbubbles remained unaltered and therefore C-12 remained attached to the lipid shell, unable to cross the BBB (Figure 5(b)).

Cavitation analysis has not been carried out in previous studies of FUS-stimulated loaded microbubbles for localized drug delivery into the brain.^16,40 These studies also may have operated in the IC regime and potentially induced microhemorrhages, although they do not report histological examinations. In this study, initial histological examinations showed microscopic damage, such as microhemorrhages, microvacuolations, and few necrotic cells (Figure 6(c), (e) and (g)), coinciding with the high ICD needed to achieve successful C-12 delivery (Figure 5). These results were consistent with the prior findings obtained by our group,^21,35 where the BBB opening at pressures higher than 600 kPa resulted in tissue damage at the sonicated regions. The presence of this kind of damage may compromise the safety of our technique. For this reason, nine mice were survived one week after sonication to determine whether damage was reversible. At 600 kPa, the microhemorrhage and microvacuolation were cleared, while a higher number of cell nuclei in the sonicated region compared to the same region in the contralateral side has only been observed in one out of the five cases. The same pattern appeared in three out of the four cases at 750 kPa, thereby indicating that our technique might not be safe at pressures higher than 600 kPa, at least when maintaining the rest of the sonication parameters fixed (pulse length, duration of the sonication, repetition pulse frequency) and/or the lipid and gas core composition, microbubbles size, etc. To better characterize this observation, Iba1 bright-field immunohistochemistry was performed in three additional mice sonicated at 750 kPa and survived for one week. All of them showed more microglia cells (Figure 6(k)) in the sonicated region compared to the contralateral (untreated side), it means, FUS induces, as it was shown in previous works,⁵⁷ microglial activation. Microglial activation can be really beneficial since, for example, it can reduce the burden of amyloid plaques in the brain,⁵⁷ but, it may also be harmful when it is involved in chronic neuroinflammation, it means, when the responses of microglial cells contribute to and expand the neurodestructive effects, worsening the disease process.

Therefore, another parameter to investigate is the survival duration and to increase it to two/three weeks in order to study how the microgliosis evolves.

Also, T₂-w MRI was performed on day 1 (the day after sonication). At both pressures, edema was observed. Previous studies reported by our group have shown that edema, at acoustic pressures until 600 kPa and pulse length of 100 cycles, is cleared out over the course of a week.⁵⁴ However, in the current study some mice are sonicated at higher pressures (750 kPa), and all of them at higher pulse length (10,000 cycles). Therefore, four mice survived for two weeks after sonication at 750 kPa were scanned, on a daily basis, by T₂-w MRI to assess whether, under the higher pressures, brain edema has been effectively repaired. In three out of the four cases, edema was repaired between days 3 and 4 (Supplementary Figure 2(b)).

These mice were also scanned, on a daily basis, by T₁-w post-Gd injection MRI to evaluate how long it took for the BBB opening to be restored. Samiotaki and Konofagou ⁵⁸ reported that when mice were sonicated in the presence of 4–5 µm microbubbles, at a pressure of 600 kPa and pulse length of 10,000 cycles (6.7 ms), the BBB remained opened for 48–72 h, but they did not reach pressures up to 750 kPa. At this pressure, the BBB was restored between days 4 and 5 in three out of the four cases (Supplementary Figure 2(a)).

This study does present indeed certain limitations based on the success on the delivery and side-effects provoked to the high pressures used and, therefore, future work is needed to fully study the safety of our FUS technique before being applied to the clinic. These studies will be focused on evaluating the effects of using different microbubble sizes and shells (different lipids and/or different proportions), different gas cores, lower pulse lengths, lower repetition pulse frequency, reducing the sonication duration, etc. With the suitable combination of all the optimal parameters, the C-12 delivery may be achieved and enhanced at lower pressures, diminishing the side-effects provoked by the IC.

However, our findings indicate potential advantages for employing our in-house manufactured lipid microbubbles as a vector for localized drug delivery since they cause less damage than the invasive alternative methods currently used clinically for brain-targeted drug delivery, such as intracerebral-ventricular infusion,⁵⁹ convection-enhanced delivery, or implementation of delivery systems.⁶⁰ This approach, after all the parameters (acoustic pressure, pulse length, pulse repetition frequency, sonication time, lipid shell composition, microbubbles size, and gas core composition) would be optimized, might be used in humans, taking into account the differences between the thickness of the mouse (≈0.2 mm)⁵¹ and human skulls (≈4.65 mm).³³ According to our previous in vitro studies using human skulls, the FUS frequency had to be lowered to 500 kHz to decrease the aberration effects, so that SCD_h, SCD_u, and ICD were measurable.³³ Therefore, making the appropriate corrections, considering that the attenuation through the human skull is 7.33 dB/mm,³³ the method proposed herein can in principle be translated to the clinic.