Time-intensity-curve Analysis and Tumor Extravasation of Nanobubble Ultrasound Contrast Agents

Abstract

Our group recently presented a simple strategy using the nonionic surfactant, Pluronic, as a size control excipient to produce nanobubbles in the 100 nm range, which exhibited stability and echogenicity on par with clinically available microbubbles. The objective of the current study was to evaluate biodistribution and extravasation of the Pluronic-stabilized lipid nanobubbles compared to microbubbles in two experimental tumor models in mice. Standard lipid-stabilized perfluoropropane bubbles or Pluronic L10 and lipid-stabilized perfluoropropane nanobubbles were intravenously injected into mice bearing either an orthotopic mouse breast cancer (BC4T1) or subcutaneous mouse ovarian cancer (OVCAR-3) through the tail vein to perform perfusion dynamic studies. No significant differences between the nanobubble and microbubble groups were observed in the peak enhancement of the three tested regions (tumor, liver, and kidney). However, the decay rates of nanobubble in tumor and kidney of BC4T1 bearing mice as well as in OVRCAR-3 tumor were significantly slower than those of microbubble. To quantify extravasation, fluorescently-labeled bubbles were intravenously injected into mice bearing the same tumors. Histological analysis showed that nanobubbles were retained in tumor tissue to a greater extent compared to microbubbles in both tumor models at 3-hour time point. Our results demonstrate unique nanobubble behavior compared to microbubbles and support augmented application of these agents in ultrasound molecular imaging and drug delivery beyond the tumor vasculature.

Introduction

Breast cancer is by far the most frequently diagnosed cancer and cause of cancer death among women. There were an estimated 1.7 million new cases (25% of all cancers in women) and 0.5 million cancer deaths (15% of all cancer deaths in women) in 2012. Nearly 43% of the estimated new cases and 34% of the cancer deaths occurred in Europe and North America (Stewart and Wild 2014). Ovarian cancer ranks second among gynecologic tumor after cervical cancer worldwide. It is estimated that ovarian cancer affects 238,719 women worldwide and causes more than 150,000 deaths annually (Ferlay, et al. 2015). Early detection affects the long-term survival rate of the tumor. The reported sensitivity and specificity of conventional ultrasound in distinguishing benign and malignant adnexal masses or breast tumor are 93%– 96.7% and 88%–98%, respectively (Cohen, et al. 2001, Menon, et al. 2009, Stavros, et al. 1995, Topuz, et al. 2005). However, 2D ultrasound has low positive predictive value (PPV) for cancer diagnosis. For ovarian cancer screening, according to the largest screening trail, UKCTOCS, the PPV of transvaginal ultrasound was 2.8% (Menon, et al. 2009). Stavros AT, et al reported that the PPV of ultrasound for the solid breast mass was 38% (Stavros, et al. 1995). Doppler and power Doppler US techniques, which enable examination of blood flow in a region of interest, can provide more specific information than the standard anatomical (grayscale) US (Rao A 2011). However, they are sensitive only to vessels > 2mm in diameter (Cha, et al. 2005). Moreover, using color flow Doppler to analyze blood flow to suspicious areas or masses were not found to significantly add to the assessment of malignant lesions (Valentin, et al. 2006).

During the past two decades, the advent of microbubble ultrasound contrast agents (UCA) has enhanced the capabilities of ultrasound as a medical imaging modality and stimulated innovative strategies for cancer detection, therapy and post-therapy monitoring. UCA increase vascular signal and improve the delineation of microvascular architecture. It has been previously reported that contrast-enhanced ultrasound (CEUS) has higher sensitivety, specificity and PPV in the diagnosis of breast and ovarian tumors compared to conventional US (Cassano, et al. 2006, Hu, et al. 2015, Xiang, et al. 2013). Further expanding the field of CEUS, microbubbles which target specific biomarkers have shown potential for detection and quantification of tumour angiogenesis at the molecular level. For example, in a recent preclinical study, microbubbles targeted with endothelial cell markers (anti-integrin, anti-endoglin, or anti-vascular endothelial growth factor receptor 2) successfully attached to tumor vessels and were detected by ultrasound in breast, ovarian, or pancreatic tumors. (Deshpande, et al. 2011). However, the diameters of commercial UCA are in the range of 1–8 microns. Due to their larger size, microbubbles remain in the vasculature. This limits its application in molecular imaging and drug delivery, especially in oncology. The ability to formulate robust sub-micron contrast agents able to extravasate hyper-permeable tumor vasculature and penetrate deep into tissue is attractive as it can facilitate detection of cell surface markers or molecules in the tumor tissue.

A number of groups have recently explored the use of nanobubbles in contrast-enhanced ultrasound (Guvener, et al. 2017, VanOsdol, et al. 2017, Yang, et al. 2016). Our previous work demonstrated a simple strategy using Pluronic as a size control excipient to produce stable and echogenic nanobubbles in the 100–300 nm range (Hernandez, et al. 2017, Krupka, et al. 2010, Perera, et al. 2017, Wu, et al. 2013) which have been shown to accumulate in tumors (Gao, et al. 2017). We hypothesize that nanobubbles are likely accumulate in tumor due to extravasation. The purpose of this study was to explore the perfusion dynamics and extravasation of this nanobubble as compared to microbubbles in a breast and ovarian cancer tumor model.

Materials and Methods

Materials

The lipids DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine), DPPA (1,2 dipalmitoyl-sn-glycero-3-phosphate), and DPPE (1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine) were obtained from Avanti Polar Lipids (Pelham, AL, USA), and mPEG-DSPE (1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt)) was obtained from Laysan Lipids (Arab, AL, USA). The molar pluronic:lipid ratio in the formulation was 0.02 (Hernandez, et al. 2018). 1,1’-Dioctadecyl-3,3,3’,3’-Tetramethylindocarbocyanine perchlorate (DiI) and Lycopersicon esculentum lectin (FL-1171) were purchased from Sigma Aldrich (Milwaukee, WI, USA) and Vector Laboratories (Burlingame, CA, USA), respectively. Pluronic L10 was donated by BASF (Shreveport, LA, USA).

Preparation of contrast agents

Fluorescently-labeled lipid microbubbles or lipid-Pluronic nanobubbles were prepared as previously described (Krupka, et al. 2010, Wu, et al. 2014). Briefly, the lipids DPPC, DPPE, DPPA and mPEG-DSPE (7 mg total) were dissolved in 1 mL of chloroform at a mass ratio of 4:1:1:1. For fluorescent bubbles, 50 μg of DiI was added to lipid solution in chlorofom. The solvent was then removed by evaporation, which resulted in the formation of a lipid film. Lipid films were then hydrated in a solution of PBS or Pluronic L10 (0.6 mg/mL) in the presence of glycerol to produce microbubbles and nanobubbles, respectively. Hydration of the lipid films took place at 75 °C in a water bath for 60 min. Next, the vials were sealed, and the air in the vial was replaced with octafluoropropane (C₃F₈) gas. Finally, the vial was placed on a VialMix shaker (Bristol-Myers Squibb Medical Imaging, Inc., N. Billerica, MA, USA) for 45 sec to form the bubbles. The nanobubbles were allowed to separate at 4°C for 2 hours in the sealed vial, and then the milky portion was withdrawn and used for this study. To determine the size and polydispersity of the bubble populations, nanobubble and microbubble samples were first analyzed with a bright field microscope (Zeiss AxioScope) using a 40x objective. The bubbles were then characterized using resonant mass measurement (RMM) (Archimedes, Malvern Panalytical Inc., Westborough, MA, USA) which measures particle size and concentration (Burg, et al. 2007, Godin, et al. 2007). A micro sensor was used to measure the microbubbles, which can provide size measurements from 250 nm to 5 µm, and a nanosensor was used to characterize the nanobubbles, which provides measurement from 100 nm to 2 µm. Sensors were calibrated using NIST traceable 565 nm and 994 nm polystyrene bead standards (ThermoFisher 4010S, Waltham MA, USA). Microbubbles were diluted 1:10000 and nanobubbles were diluted 1:1000 with phosphate buffered saline (pH 7.4) resulting in acceptable limit of detection (<0.02 Hz) and coincidence (<10%). A total of 1000 particles were measured for each trial performed. The sensor and microfluidic tubing were cleaned with deionized water in between each run. Measurements were repeated in triplicate. The sharp cut-off in size is the result of the limit of detection (LOD) of the nanosensor that was determined for each measured sample at the start of each measurement and accounts for the baseline noise (Hernandez, et al. 2019).

Determination of experimental and theoretical perfluorocarbon volume of buoyant and non-buoyant particles

Headspace gas chromatography/mass spectrometry (GC/MS) was performed to quantify the amount of perfluoropropane in the nanobubble sample. (Yang et al. 2016) Measurements were made using an Agilent 5977B-MSD equipped mass spectrometer with an Agilent 7890B gas chromatograph (GC/MS) system. A DB5-MS capillary column (30 m × 0.25 mm × 0.25 μm) was used with a helium flow of 1.5 ml/min. Headspace samples of 1 µl were injected at 1:10 split. Gas chromatography conditions used were as follows: oven temperature was at 60°C, held for 1 min, ramp 40°C/min until 120°C and held for 3.5 minutes. Perfluoropropane eluted at 1.2 minutes. Samples were analyzed in Selected Ion Monitoring (SIM) mode using electron impact ionization (EI). M/z of 169 (M-19) was used in analyses. Ion dwell time was set to 10 msec. Perfluoropropane was verified by NIST MS spectra database. Briefly, a standard calibration curve was prepared by diluting perfluoropropane with air (0 – 1 % v/v). The linear regression plot generated from the curve was used to determine the experimental perfluoropropane volume of the bubble samples. Nanobubble samples (50 uL) were added to a headspace vial containing 0.45 mL PBS and sealed. The bubbles were destroyed via sonication at 50^oC for 20 min, which consequently released the gas in the bubble to the headspace of the vial. The samples were then analyzed using headspace GC/MS.

Resonant mass measurement was used to determine the buoyant (bubble) and non-buoyant particle concentration of the sample (Fig. S1). The total theoretical amount of perfluorocarbon gas expected in the nanobubble sample is shown in Table S1. The amount was calculated by assuming that the thickness of the particle shell is 2.5 nm (since we expect that the shell is a lipid monolayer and the typical lipid bilayer thickness is 5 nm).(Shim et al. 2012) The lipid layer thickness was subtracted from the particle radius in calculating the volume of the perfluoropropane gas for each bubble, using the density of perfluoropropane (0.008 g/mL). The concentration used was that of the buoyant (bubble) particle component. The theoretical perfluoropropane gas that may arise from the non-buoyant particle population of the sample (which may contain liquid perfluoropropane forming nanodroplets that may appear as non-buoyant in Archimedes) was calculated by subtracting the lipid layer thickness from the particle volume, using the density of liquid perfluoropropane (1.6 g/mL), and the concentration of the non-buoyant population of the sample. (Table S1 and Fig. S1)

In vitro Ultrasound Stability Characterization

In vitro measurements were performed using two conditions via (a) similar gas volume concentration (0.044 nL/mL, NBs 3.6 × 10⁷ particles / mL, MBs 4.1×10⁵ particles / mL) and (b) similar particle concentration (bubbles/mL at 4 × 10⁶ particles / mL). Bubble solutions were placed in a custom-made 1.5% (w/v) agarose phantom with a thin channel (L × W × H = 22 × 1 × 10 mm) (Hernandez, et al. 2017). The agarose phantom was fixed over a 12 MHz linear array transducer and imaged using a clinical ultrasound scanner (AplioXG SSA-790A, Toshiba Medical Imaging Systems, Otawara-Shi, Japan). Contrast Harmonic Imaging (CHI) was used to image changes in the ultrasound signal as function of time using the following imaging settings: CHI frequency, harmonic frequency 8.0 MHz; MI, 0.08; dynamic range, 65 dB; gain, 80 dB; imaging frame rate, 12 frames/min. The change in ultrasound signal versus time for each concentration was determined using the pre-loaded quantification software (CHI-Q).

In addition, we also performed ultrasound scans of the nanobubble sample before and after exposure to high intensity ultrasound. The destruction procedure was carried out as described in our recent paper (Hernandez et al 2018). Briefly, the flash-replenishment feature on board the clinical ultrasound unit was repeatedly applied to bubbles in PBS solution placed in a 1.5 wt.% agarose phantom which was fixed above the transducer. We previously demonstrated that this procedure breaks bubbles. This results in the escape of the gas into the air, as the experiment is performed in an open system. We then analyzed the remaining sample using RMM and GC/MS for residual C₃F₈ content.

Animal models

Animals were handled according to a protocol approved by the Institutional Animal Care and Use Committee at Case Western Reserve University and were in accordance with all applicable protocols and guidelines with regards to animal use. Two mouse tumor models, the 4T1 mammary adenocarcinoma and OVCAR-3 ovarian tumor models were used. A suspension of 5×10⁵ 4T1 cells in 50 µL was orthotopically injected to the #9 mammary fat pad of female BALBc/4j mice (Jackson Laboratories, ME, USA) mice, or 1×10⁶ OVCAR-3 cells in 50 µL was subcutaneously injected into right flank of NCRnu/nu nude mice (Athymic Animal and Xenograft Core Facility of Case Western Reserve University). Before all procedures, the animals were anesthetized with inhalation of 3% isoflurane with 1 L/min oxygen. Two weeks after inoculation, the tumor diameter reached 1.5 cm or above and the animals were used in the study.

Perfusion dynamic study

To acquire ultrasound images, a US probe (PLT-1204BT) was placed longitudinally to the axis of the animal body. The probe was immobilized using a clamp and was adjusted so that the field of view included the liver, kidney and tumor. Contrast Harmonic Imaging (CHI) was used to image changes in the tissue contrast density as function of time using the following imaging settings: CHI frequency, harmonic frequency 8.0 MHz; MI, 0.08; dynamic range, 65 dB; gain, 80 dB; imaging frame rate, 12 frames/min. The images were acquired in the raw data format. For contrast agent administration, 100 µL of DiI-labeled microbubbles or nanobubbles were diluted to 1 mL with normal saline, and 100 µL of the diluted bubble solution was administrated through the tail vein 10 sec after the start of the raw data acquisition. The total image acquisition time was 12 minutes. For imaging studies, a total of 5 mice were used for each tumor model. On the first day, each mouse was randomly selected to be imaged with either nanobubbles or microbubbles. The bubbles were allowed to clear for 24 hours before each mouse was re-imaged using the alternative bubble group.

Data analysis

The acquired linear raw data images were processed with CHI-Q quantification software (Toshiba Medical Imaging Systems, Otawara-Shi, Japan). Regions of interest (ROIs) were drawn outlining the liver, kidney, tumor and unenhanced area. The mean echo-power value (P) in each ROI as a function of time, also called time-intensity-curve (TIC), was exported into MATLAB

$$f_{(t)}=P_0+\frac{A}{(t-t_0)\sigma \sqrt{2\pi }}{e}^{\frac{-{\left[ln\frac{t-t_0}{m}\right]}^2}{2{\sigma }^2}}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}(x\ge \theta ;m,\sigma >0)$$ (1)

R2013a (MathWorks Inc., Natick, MA). The TIC data were fitted with lognormal function, Where P₀ is the baseline intensity, A is the area under curve, t₀ is the time to start, m is the scale parameter, and σ is the shape parameters. The baseline (P₀) was then subtracted from the TIC. The peak enhancement was defined as the maximum of the f_(t)-P₀. The data post peak (P) was normalized to the peak value (P_peak) and log-compressed in decibels (P_dB) using the following function (Payen, et al. 2013)

$$P_{dB}=10{\log }_{10}\frac{P}{P_{peak}}$$ (2)

The P_dB data was fit using simple linear regression to derive the decay slope.

Bubble extravasation

A total of 13 mice per tumor model were used for this study. For each tumor model, the mice were divided into 3 groups, nanobubble (n=5), microbubble (n=5), and control (n=3). Mice were injected with 300 µL DiI-labeled microbubbles (microbubble group), DiI-labelled nanobubbles (nanobubble group), or PBS (control group) through tail vein. Three hours later, 0.1 mL of 1 mg/mL fluorescein labeled Lycopersicon esculentum lectin (FL-1171, Vector Laboratories, Burlingame, CA, USA) was intravenously injected. Five minutes later, the animals were perfused with 50 mL PBS through left ventricle. Liver, tumor and muscle tissue samples were harvested and frozen on dry ice after being embedded in OCT (optimal cutting temperature compound, Sakura Finetek USA, Inc., Torrance, CA, USA). Frozen tissue samples were cut into 8-µm thick slices using a Leica CM1850 cryostat (Leica, Germany) and mounted with DAPI mounting medium (Vectashield, Vector Laboratories). The fluorescence images of the slice were acquired by an inverted microscope using a Rhodamine filter for DiI-labeled bubble imaging, and FITC filter for lectin imaging. The images were then analyzed with ImageJ. The regions of DiI or lectin (vessels) in tumor were first segmented. The percentage of area of DiI or lectin in tumor tissue was then calculated.

Statistics

All values were expressed as mean ± standard error (SD). Student t-test was used to compare the difference in means between two groups. A p value of less than 0.05 was considered statistically significant.

Results

Physical properties of bubbles

Under optical microscopy, the size of visible microbubbles measured was 1.53 ± 0.39 µm (Figure 1a). No larger bubbles were visible in the nanobubble sample given the limit of resolution of the optical microscope (Figure 1b). Resonant mass measurements of the size, concentration, and gas volume of the bubbles particles are shown in Table 1. The mean nanobubble diameter was 138 ± 43 nm (range: 97 – 347 nm), while the mean microbubble diameter was 596 ± 327 nm (range: 182 – 1824 nm). The concentrations of nanobubbles and microbubbles were similar, approximately 4 × 10¹⁰ /mL. The size and distribution of the microbubbles and nanobubbles are shown in Figure 1c and 1d, respectively. The size and concentration for non-buoyant particles present in nanobubble and microbubble samples are also reported in Table 1. The nanobubble sample has non-buoyant particles with an average size of 183 ± 39 nm and a concentration of 7 × 10¹⁰ /mL, while the microbubble sample has non-buoyant particles with an average size of 457 ± 124 nm and a concentration of 1 × 10¹⁰ /mL. Figure 2 shows representative RMM plots of concentration vs buoyant mass for all both buoyant and non-buoyant particles detected in the total nanobubble sample (a and b) and microbubble sample (c and d) with respect to PBS solution. A large population of buoyant particles (negative buoyant mass) is seen in both samples, supporting the notion that C₃F₈ is present in the gas form in both bubble types.

Figure 1.

Optic microscopic images of microbubble (a) and nanobubble (b). The size and distribution of the nanobubble (c) and microbubble (d) analyzed with resonant mass measurement (RMM).

Table1.

Resonant mass measurement of the nanobubbles and microbubbles.

	Nanobubbles (buoyant)	Nanobubbles (Non-buoyant)	Microbubbles (buoyant)	Microbubbles (Non-buoyant)
Size (nm)	138 ± 43	183 ± 39	596 ± 327	457 ± 124
Concentration (× 1010 /mL)	3.99	7.11	4.07	1.13
Gas volume (× 109 um3/mL)	2.88	N/A	12.7	N/A

Figure 2.

RMM plots of concentration vs buoyant mass for both buoyant and non-buoyant particles of (a and b) nanobubbles and (c and d) microbubbles with respect to PBS solution.

To evaluate whether C₃F₈ nanodroplets were formed in addition to nanobubbles, the total gas volume generated by the buoyant and non-buoyant fractions of the sample were quantified experimentally and compared to theoretical calculations. Results (Table S1 and Figure S1) show that the experimentally determined volume of C₃F₈ matches that of the theoretically expected gas volume of the bubble population of the sample when assuming that none of the buoyant particle population is comprised of liquid nanodroplets. Thus, the formation of liquid nanodroplets is very unlikely. If all of the non-buoyant particle population of the sample forms liquid nanodroplets, this should give rise to an additional 3854 nL of C₃F₈ gas. This value is 329x that of the experimentally determined C₃F₈ gas volume. These two calculations using the buoyant particle population and the non-buoyant particle population indicate that it unlikely contains any significant quantity of liquid nanodroplets.

In addition, we also performed in vitro ultrasound scans of the sample before and after exposure to high intensity ultrasound. Here, we demonstrate that high intensity ultrasound (1.52 MI at 12 MHz) leads to destruction of bubbles releasing the perfluoropropane gas in the process. It is shown in Fig. S2 (a & b) via headspace GC/MS that no perfluoropropane gas is observed following high intensity exposure of the bubbles. In addition, Archimedes measurement shows that exposure to high intensity ultrasound converted the sample population from a mixture of buoyant and non-buoyant particles (Fig. S2c) to one that is predominantly non-buoyant (Fig. S2d). This is expected since all the bubbles were destroyed after high intensity ultrasound exposure.

In vitro bubble stability

Using ultrasound parameters similar to the perfusion dynamic study, at similar total gas volume concentrations, nanobubbles showed higher signal enhancement compared to microbubbles (Fig. 3a and 3b). Nanobubbles were also more stable than microbubbles over time (Fig. 3c) with a decay slope of 0.041 for nanobubbles and 0.048 for microbubbles. For similar concentrations (bubbles/mL), microbubbles showed higher US signal as expected (Fig. 3d and e) but comparable stability as the nanobubbles (Fig. 3f) with a decay slope of 0.034 for nanobubbles and 0.035 for microbubbles.

Figure 3.

In vitro bubble stability study. CHI images of nanobubble and microbubble at 0 sec and 12min at similar total gas volume concentrations (0.044 nL/mL) (a) and at similar particle concentration (4×106 bubbles / mL) (d). Time-intensity curves of in vitro nanobubbles and microbubbles acquired by CHI-Q quantification software by Toshiba showed as signal intensity in dB at similar total gas volume concentration (b) and at same particle concentration (e). (c) and (f) show the corresponding normalized time-intensity curves.

In vivo characterization of bubble distribution

In vivo distribution of bubbles was examined using ultrasound. Contrast enhancement was rapid in the wash-in phase followed by a rapid wash-out in the liver and kidney in both tumor models. In tumors, slower wash in and washout phases were noted in both models compared to those in liver and kidney. Figure 4 shows the ROIs outlining liver, kidney tumor and unenhanced area in 2D mode image (a), CHI image (b), and their corresponding time intensity curves of a typical 4T1 tumor bearing mouse after administration of nanobubble (c). In this case, the signal intensities (dB) in kidney and liver reached their peaks at 13 sec and 30 sec after contrast injection, respectively, then rapidly washed out. In contrast, the signal intensity in tumor reached its peak enhancement at 90 sec and gradually washed out. There were no significant differences of time to peak in all ROIs between microbubble and nanobubble groups (data not shown). The linear TICs were fitted by lognormal function (Figure 5c). No significant differences in peak enhancement were observed between the nanobubble and microbubble groups in the three tested regions (tumor, liver and kidney) of both models, as shown in Figure 6a and 6c. Linear correlations were noted in log-compressed signal intensity (dB) with time in wash out phase. The decay slopes of nanobubbles in all 3 ROIs were slower than those of microbubbles. Significant differences in the decay rate between the nanobubble and microbubble groups were observed in tumor and kidney of BC4T1 bearing mice (P < 0.05, 0.79 ± 0.40 dB/min compared to 1.13 ± 0.24 dB/min in tumor, 0.63 ± 0.17 dB/sec compared to 1.34 ± 0.47 dB/min in kidney), as well as in OVACAR 3 tumor (P < 0.05, 1.66 ± 0.76 dB/min compared to 2.64 ± 0.46 dB/min) (Figure 6b, 6d).

Figure 4.

Regions of interest (ROIs) outlining liver (L), kidney (K), tumor (T) and unenhanced focus (N) in representative 2D image (a) and CHI image (b). Time-intensity curves in liver (ROI 1), kidney (ROI 2), tumor (ROI 3) and unenhanced focus (ROI 4) acquired by CHI-Q quantification software by Toshiba showed as signal intensity in dB (c).

Figure 5.

Time intensity curves of a typical 4T1 tumor after microbubble (MB) and nanobubble (NB) iv injection. (a) 2D US and CHI images before contrast administration. (b) CHI images at 30 sec, 5 min, 10 min, and 20 min after NB or MB administration. Dash line, region of interest in tumor. (c) shows the TICs in tumors of MB and NB as well as the curve fitting using lognormal function.

Figure 6.

Comparisons of peak enhancement and decay rate in the ROI of tumor, liver, kidney between the microbubble (MB) and nanobubble (NB) groups.

An enhancement deficit at the tumor center was seen in all tumors. Figure 5a and 5b show the 2D US image and CHI images of a typical breast cancer tumor at different time points after an intravenous administration of microbubbles and nanobubbles. Prominent enhancements were observed in tumor rims with both bubble groups; however, nanobubbles penetrate deeper into the tumor compared to microbubbles at 5, 10 and 20 min time points.

Bubble extravasation

Figure 7 shows typical fluorescence images of tumors after receiving an intravenous injection of either PBS (a, d), DiI-microbubbles (b, e), or DiI-nanobubbles (c, f). Blood vessels were labeled with Lectin (green signal), and appeared to be more densely populated around the tumor rim than tumor core (Figure 7a, 7d). In the DiI-nanobubble group, DiI signal (red color) was localized around vessels (7c, 7f), and this was not observed at the same frequency in the microbubble group (7b, 7e). The signal from tumor vasculature comprised 0.14 ± 0.05% (0.01% – 0.33%) of the total area in the BC4T1 tumor and 0.43 ± 0.30 % (0.01% – 1.19%) of the total area in the OVCAR3 tumor without significant difference between microbubble and nanobubble groups. However, the nanobubble signal intensities were 4.5 and 22.7 times higher than those of microbubbles in the BC4T1 and OVCAR3 tumors, respectively. The ratio of nanobubble to vascular signal was significantly higher compared to microbubbles (P<0.05, 6.03 ± 3.51 vs. 0.33 ± 0.61 in OVCAR3 tumor, 3.47 ± 2.33 vs. 0.88 ± 0.64 in BC4T1 tumor) (Figure 8).

Figure 7.

Representative fluorescence images of BC4T1 tumor (a, b, c) and OVCAR3 tumor (d, e, f) in groups control (a, d), microbubble (b, e) and nanobubble (c, f). Blue color: DAPI nuclei staining; Red color, DiI; Green color: lectin endothelial staining.

Figure 8.

The average percent of total slide area showing fluorescent signal from bubbles (either microbubble (MB) or nanobubble (NB)) and vessels in both the BC4T1 (a) and OVCAR3 (b) tumor models.

Discussion

Pluronic is a triblock copolymer. An extensive body of work has demonstrated that Pluronic can interact with lipid membranes, change lipid membrane fluidity, and control particle size (Batrakova, et al. 2001, Chang, et al. 2005, Demina, et al. 2005). Our previous studies have shown that nanobubbles can be formulated by incorporating Pluronic into monomolecular layer of lipid-shelled bubbles filled with perfluoropropane gas (C₃F₈) (Krupka, et al. 2010). In this work, to further characterize the nanobubbles, we used resonant mass measurement to measure the particle size, size distribution, concentration, and ratio of bubbles (buoyant) and non-buoyant particles in the total population. To confirm whether the perfluorocarbon in the bubbles is in gas or liquid state, we also measured the buoyant mass compared to PBS using Archimedes. The Archimedes can distinguish between buoyant and non-buoyant particles in a bubble sample, which typically are composed of both bubbles (buoyant) and aggregated lipids (non-buoyant) (Hernandez, et al. 2019). The density of gaseous C₃F₈ is 0.008 g/mL and that of liquid C₃F₈ is 1.6 g/mL (density of PBS is 1 g/mL). Based on these numbers, if C₃F₈ is gaseous then the buoyant mass of the particle would be negative (vs. PBS) and the buoyant mass of the particle would be positive if it is liquid (vs. PBS). The results (Figure 2) showed that the buoyant mass of the bubbles are negative for both nano- and microbubbles, indicating that the perfluorocarbon C₃F₈ is gaseous in the bubbles. Thep non-buoyant particle signal most likely results from aggregated lipids that are unavoidably produced during the self-assembly process. Based on the RMM results, the amount of perfluorocarbon gas per particle can be calculated by taking into consideration that the thickness of the particle shell is 2.5 nm (since we expect that the shell is a lipid monolayer and the typical lipid bilayer thickness is 5 nm) (Shim, et al. 2012). The bubble volume and gas volume of 138 nm nanobubbles with the concentration of 3.99 × 10¹⁰/mL is 0.014 × 10⁻⁷ nL and 49.2 nL/mL, respectively. The in vitro bubble stability study showed that in our test ultrasound setting, microbubbles have higher echogenicity than nanobubbles. However, at similar total gas volume concentrations, nanobubbles showed higher signal enhancement compared to microbubbles. This may be due to the higher nanobubble concentration (3.6 × 10⁷ particles / mL vs. 4.1 × 10⁵ particles / mL) and yet undetermined population interactions, which will require further in depth study.

Time-intensity-curve analysis can provide valuable insight into the behavior of contrast agents. Peak enhancement can reflect bubble echogenicity and bubble concentration at the site, while the rates of contrast during wash in and wash out phases from the region of interest can be indicative of tissue perfusion and vascular permeability. Since US images are inherently noisy, many models are used to analyze the TIC in tissue. Indicator dilution models are used for quantification of tissue blood flow with bolus administration of ultrasound contrast agents. The models that best fit the experimental data are the lognormal function and the diffusion with drift (Krix, et al. 2004, Strouthos, et al. 2010). In this study, we used the lognormal function to fit the TIC of ROIs (tumor, liver and kidney).

The overarching goal of this study was to explore the feasibility of using a nano-scale ultrasound contrast agent for tumor molecular imaging and to compare nanobubble behavior to bubbles closer to the clinical agent size range. In this study, we used an 8.0 MHz clinical ultrasound probe to image the breast and ovarian tumor models. This frequency is within the frequencies used for clinical breast imaging (8–15 MHz) and transvaginal pelvis ultrasound imaging (6–10 MHz). At this frequency, we found that there were no significant differences in peak enhancement of the liver, kidney, and tumor between the nanobubble and microbubble groups in both tumor models. However, nanobubbles penetrate deeper into the tumor core compared to microbubbles. The reasons for this are not clear, but in general the differences may be due to the smaller size of nanobubbles penetrating into smaller tumor vasculature, which in turn would enable detection of the acoustic activity in those regions.

The clearance parameters are related to several factors, including the stability of contrast agents, clearance by the reticuloendothelial system, as well as potential extravasation in regions of high vascular permeability(Discher and Eisenberg 2002, He, et al. 2007, Ishida, et al. 2002, Wu, et al. 2013). In this study, the post-peak data was normalized to the peak. We found that the log-compressed normalized signal intensity post peak was linear with time. Since the maximal enhancement was affected by many parameters, such as the number of bubbles administrated, animal heart function, and heart rate, we opted to normalize the curves to the peak enhancement to eliminate these factors. This data indicates that the decay slopes of nanobubbles were significantly less than those of microbubble in both tumor models. However, there were no differences observed between the two groups in liver and kidney models. This suggests that nanobubbles are more stable than microbubbles and can leak out of the tumor vasculature into the tumor parenchyma.

Lipid-based microbubbles are rapidly phagocytosed by activated neutrophils and monocytes, a process that depends on rapid surface opsonization. Adding PEG (polyethylene glycol) to the shell of particles can inhibit particle uptake by the reticuloendothelial system and this increases the half-life of particles in vivo (Discher and Eisenberg 2002, He, et al. 2007). In our previous study, in a mouse colon cancer model, nanobubbles without PEG functionalization had 2 phases of decay, fast initial phase (about 200 sec) and slow latter phase, especially in liver (Wu, et al. 2013). In the current study, the PEGylated bubbles decayed at a constant speed in all ROIs. This suggests that PEGylation affects in vivo bubble clearance.

Extravasation into the extravascular space in tumor tissue is a potential advantage of nanobubbles, and a parameter that will be crucial to potential future application of these agents in tumor detection and drug delivery. In this study, the nanobubbles had a narrower size distribution (97–347 nm) while the microbubbles had a much broader range, ranging from 182 to 1824 nm and included a considerable population of bubbles in the sub-micron range. However, as seen in the size distribution data, the majority of bubbles in the “nanobubble” group were nonetheless smaller than the smallest bubbles in the “microbubble” group. We used DiI-labeled bubbles to observe bubble extravasation after 3 hours (Yin, et al. 2012). The contrast enhanced ultrasound images showed more central enhancement within the core (as opposed to rim enhancement) at 10 and 20 min (Figure 5b). It is based on this data that we make the statement that the smaller bubbles have better penetration. However, in addition to bubble size, this could also be a result of higher nanobubble concentration visible in the smaller, more tortuous vasculature of the tumor core, which cannot be achieved with the microbubbles. Our histology results showed that more fluorescence was detected outside of the tumor vessels in the nanobubble group than in the microbubble group, which may indicate that the smaller nanobubbles passively extravasate out of tumor vessels, especially the nanobubbles, which are smaller than 180 nm. Acoustic driving was reported as one of the mechanism of the bubble extravasation (Kauer, et al. 2018). There was no ultrasound present in the histology experiments in this study, so it is unlikely in this case. In the perfusion dynamic study, we imaged the animal continuously for 12 minutes after contrast administration with low MI (0.08). Both passive contrast extravasation and acoustic driving may contribute to the contrast extravasation. Further studies will be performed to verify the role of acoustic driving on the nanobubble extravasation.

Our study has some limitations. First, it would be ideal to use a commercial microbubble contrast agent as a control. This would enable the comparison between the nanobubble and a well-known standard. However, it is technically challenging to label the commercial microbubbles contrast agent with DiI. Furthermore, it is also difficult to remove all of the sub-micron bubbles from the population, as their presence is likely even in commercial formulations. Second, we used subcutaneous mouse breast and ovarian tumor models in which the vascular architecture may not be clinically relevant. In future studies, orthotopic tumors, instead of flank ones described above, may lead to acquisition of more directly clinically relevant information. Third, we performed the histological analysis in perfused animals and indeed this experiment showed more fluorescent signal outside of tumor vessels. However, the presence of fluorescent signal does not indicate the presence of intact nanobubbles, and the signal could be from lipid fragments or DiI dissociate from the bubble itself. However, we can assume that these issues would appear in both groups, and since DiI fluorescence in the nanobubble group was higher, the effect could indeed be a result of greater accumulation of nanobubbles in tumor tissue. Furthermore, the stability and echogenicity of the extravasated bubbles while quite important, is outside of the scope of this study and will be explored in the future experiments. Finally, while there was a 6-fold difference in bubble diameters, the microbubbles were not in the clinical agent range. Thus, one can expect that the differences would be further magnified if larger microbubbles with fewer nanobubbles present were utilized.

Conclusions

In this study, we explored the acoustic properties and in vivo performance of surfactant-stabilized lipid nanobubbles. Our results suggest that the acoustic properties of the nanoscale contrast agents are distinct from larger microbubbles. Nanobubbles exhibit a unique perfusion dynamic behavior which can be clearly distinguished from microbubbles by the higher peak intensity and slower washout rate compared to those of microbubbles. The same differences were observed in the two tumor models examined, but no differences were seen in normal tissues, suggesting that tumor physiology and potentially the irregular and highly permeable tumor vasculature plays a role in the distinct observed kinetics for different size bubbles. Histological analysis demonstrated enhanced extravasation and enhanced retention of particles from the nanobubble group within tumor tissue. Overall this study demonstrates potential augmented utility of nanobubble agents in ultrasound molecular imaging and drug delivery beyond the tumor vasculature.