Enhanced Detection of Cancer Biomarkers in Blood-Borne Extracellular Vesicles Using Nanodroplets and Focused Ultrasound Free

Circulating tumor cells, extracellular vesicles (EV), and free-circulating DNA and RNA are promising sources of tumor biomarkers that could provide diagnostic and prognostic information important for personalized medicine (1–6). Despite progress in detecting circulating tumor cells (CTC), existing assays still have low sensitivity due to the reduced number of CTCs often found in patients' blood (1–10 CTC/mL).

EVs, which include exosomes, apoptotic bodies, and other vesicles, are typically between 30 and 2,000 nm in diameter and are naturally released from normal and cancerous cells (7). EV release is involved in various physiological processes including cell–cell communication and apoptosis (7). Cancer cells have been shown to shed large numbers of these vesicles compared with noncancerous cells (8). Cell surface proteins can be found on their surface, and cytoplasmic molecules including proteins, mRNA, and miRNA can be found inside EVs, suggesting that they contain many of the potential biomarkers found in intact cells (9). EVs in blood have thus been under intense investigation as a source of biomarkers for various diseases including cancer (10, 11). As a source of biomarkers, EVs may have some advantages compared to non-encapsulated nucleic acid biomarkers as vesicle-encapsulated biomarkers are likely to be protected against degradation. Furthermore, tumor-shed vesicles are far more abundant than circulating tumor cells, which should allow for more sensitive detection of circulating biomarkers. Recently, circulating vesicles containing high levels of glypican-1 identified pancreatic ductal adenocarcinoma in 190 cancer patients with 100% sensitivity and specificity demonstrating the diagnostic potential of EVs (12). Prostate cancer–specific mRNA biomarkers PCA-3 and TMPRSS2:ERG have recently been shown to be detected in exosomes in urine and blood (13).

Free-circulating and vesicle-encapsulated DNA and RNA offer considerable potential for genetic diagnostics of tumors. In particular, detection of tumor-specific mutations have yet to be widely explored, yet could open up new opportunities for genetic profiling. Unfortunately, such DNA and RNA fragments from host tumors are not always abundant in blood.

Ultrasound has been demonstrated to enhance the release of various biomarkers from cancer cells (14–16). D'Souza and colleagues showed that sonication of LS174T xenograft tumors in mice increased blood levels of carcinoembryonic antigen protein compared with pre-ultrasound levels (14). Furthermore, sonication of breast cancer ZR-75-1 cells in tissue culture enhanced the release of nucleic acids including mammary-specific mammaglobin mRNA and miR21-5p, a miRNA found overexpressed in many cancers (15). This suggests that many different cell-derived biomarkers can be liberated from cells using ultrasound, yet the principle challenge for this approach is that ultrasound alone releases relatively low levels of biomarkers from cancer cells (15, 16). Ultrasound contrast agents such as microbubbles and nanodroplets have been demonstrated to enhance the biomarker-releasing effects of ultrasound (15, 16). Microbubbles, which typically are 1 to 4 μm perfluorocarbon gas bubbles with a phospholipid or polymer shells, can undergo cavitation and highly energetic destruction when exposed to high pressure ultrasound (>1 MPa peak negative pressure), which can alter membrane permeability of nearby cells (17, 18). Phase-change nanodroplets are similar to microbubbles but typically have a liquid perfluorocarbon core and are less than a micrometer in diameter (19). High pressure ultrasound can induce nanodroplets to phase change into microbubbles, which can subsequently be destroyed with ultrasound (20). Nanodroplets and microbubbles perform similarly in cell culture for the enhancement of ultrasound-mediated biomarker release (16).

Cancer cells exposed to ultrasound in the presence of microbubbles display substantial deformation of the plasma membrane (21). This suggests that the enhanced release of biomarkers may be due in part to the enhanced release of EVs. The impact of nanodroplets/microbubbles and ultrasound on the release of biomarker-bearing EVs and free-circulating DNA/RNA in vivo has yet to be investigated.

We hypothesized that exposing tumors to high intensity focused ultrasound in the presence of nanodroplets would amplify the release of EVs carrying relevant tumor-specific biomarkers into adjacent blood vessels (Fig. 1). To test this, we exposed HT1080-GFP tumors in chicken embryos to nanodroplets/ultrasound and analyzed serum EVs using micro flow cytometry, real-time PCR, and deep sequencing. We demonstrate for the first time that nanodroplets and focused ultrasound stimulate the release of large numbers of EVs from tumor cells into the bloodstream, allowing enhanced detection of relevant protein, mRNA, miRNA, and DNA biomarkers containing tumor-specific mutations that may be of great diagnostic value.

The electronic signal from a Tektronix AFG 3021B function generator was amplified by an Electronics & Innovation 2100L power amplifier, which powered a Sonic Concepts H134 MR-006 focused transducer (43-mm diameter). The H134 transducer has a center frequency of 1.15 MHz and a focal depth of 36.5 mm from the transducer housing. The ultrasound beam focal width and length are 1.2 and 8.1 mm, respectively. A custom acrylic water coupling cone was created to allow efficient transmission of ultrasound from the transducer to target cells. Ultrasound pressures were verified using an Onda HN needle hydrophone.

A phospholipid stock containing 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (0.4 mg/mL), 1,2-dipalmitoyl-sn-glycero-3-phosphate (0.045 mg/mL), and 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (0.145 mg/mL; all phospholipids were from Avanti Polar Lipids) were dissolved in PBS containing glycerol (10%, v/v) and propylene glycol (10%, v/v), heated to 85°C for 1 hour, and stored at 4°C. For each batch of nanodroplets, phospholipid stock (1 mL) was aliquoted in a 2 mL vial with septum and the headspace was filled with perfluorobutane (FluoroMed L.P.). The vial was shaken in a VialMix shaker (Bristol Myers Squibb Medical Imaging, Inc.) for 45 seconds to create microbubbles. Microbubbles were centrifuged at 400 × g for 4 minutes and the floating microbubbles were washed once with 1 mL PBS with 10% (v/v) glycerol and 10% (v/v) propylene glycol. These washed microbubbles were used as ultrasound contrast agents for some experiments. To create nanodroplets, the vial containing washed microbubbles had approximately 10 mL of room air injected through the septum using a 10 mL syringe and 27 gauge needle. The pressurized vial caused the microbubbles to phase-change into nanodroplets, which were used for subsequent experiments. Microbubble conversion efficiency was determined using a Z2 Coulter Counter. For characterization experiments (NanoSight and cryoTEM imaging), 100 μL nanodroplets were layered on top of 200 μL 25% (w/v) sucrose (density 1.1 g/cm³) and samples were centrifuged at 16,160 × g for 5 minutes. Lipid particles without liquid perfluorobutane cores remained above the sucrose whereas nanodroplets formed a pellet below the sucrose. The nanodroplet pellets were resuspended in PBS and nanodroplets size and concentration were characterized using a NanoSight LM10 system. Cryo-transmission electron microscopy was also performed for visualization of the sucrose-purified nanodroplets (see Supplementary Materials and Methods).

The cell lines used in this study were generously provided by Dr. James Quigley (The Scripps Research Institute, La Jolla, CA). HT1080 and HT1080-GFP cells were cultured in high-glucose DMEM with 10% (v/v) FBS. Ninety-six well plates were inoculated with 7,000 cells/well. Two days after cell inoculation, medium in wells were removed and 100 μL fresh medium ± 2% (v/v) nanodroplets was added to each well. Some experiments replaced nanodroplets with 2% sucrose-purified nanodroplets or 2% diluted (0.51%, v/v) microbubbles with a final concentration of approximately 0.01% (v/v). For some wells, focused 1.15 MHz ultrasound (30 MPa pk-pk pressure, 10–10,000 cycles/burst, 1 burst per well) was applied upward from the bottom of the plate through the center of each well using the transducer coupling cone for ultrasound alignment. An Onda HNP needle hydrophone was used to measure pressures inside a well of a 96-well plate. After cell sonication, 60 μL medium from each well was acquired, centrifuged at 2,000 × g for 10 minutes to pellet large debris/cells and 40 μL of supernatant was diluted in 200 μL PBS and analyzed with the Apogee A50 micro flow cytometer. For all flow cytometry experiments, photomultiplier voltage settings for small angle light scatter, large angle light scatter, and 488 nm laser/green filter were 208, 300, and 362, respectively. For all flow cytometry cytograms, data are presented in log₁₀ scales on all axes. Nanosphere polystyrene beads (Thermo Scientific) were also run to approximate sizes of EVs.

The toxicity of ultrasound/nanodroplets was assessed in cultured HT1080-GFP cells as described in the Supplementary Materials and Methods.

All animal experiments were approved by the University of Alberta's Animal Care and Use Committee Livestock group. Shell-less chicken embryos were prepared as described previously (22). Briefly, fertilized chicken eggs were placed in a humidified 38°C rocking chamber for 3 days followed by cracking of eggs into sterilized plastic weigh boats. The embryos were placed in a 38°C humidified incubator for five days and HT1080 or HT1080-GFP cells were microinjected in the chorioallantoic membrane (CAM; approximately 50,000 cells per injection site; 2–3 injections per chicken embryo). Chicken embryos were placed back in the incubator for 6 days before embryos were used for experiments.

For each chicken embryo, 50 μL nanodroplets, sucrose-purified nanodroplets, microbubbles (0.51%, v/v), or PBS containing glycerol (10%, v/v) and propylene glycol (10%, v/v) was injected intravenously in the CAM. Unless otherwise stated, ultrasound was applied on or away from the tumor approximately 3 minutes after nanodroplet injection and 50 μL blood was sampled thereafter near the sonication site. When no ultrasound was applied to the chicken embryo, blood was sampled near the tumor. Blood was acquired from the chicken embryo by physically disrupting the larger blood vessels in the CAM using a glass needle and collecting blood that pooled on top of the CAM. If ultrasound caused blood to pool on top of the CAM, this blood was taken for analysis. Blood samples were left at room temperature for at least 20 minutes and clotted blood was centrifuged at 2,000 × g for 10 minutes. The supernatant (5 μL serum) was diluted in 295 μL PBS and analyzed with the A50 micro flow cytometer. The gate used to determine GFP⁺ vesicles was manually drawn to not include events based on cytograms from serum of embryos with sonicated HT1080 tumors (no GFP in these tumors so all vesicles in this sample should not have GFP). This same GFP⁺ gate was used for in vitro experiments described above.

Various experimental parameters were manipulated to identify the optimal conditions for vesicle release from tumors into blood. Unless otherwise stated, ultrasound parameters for tumor sonication were 1.15 MHz, 30 MPa (pk-pk) pressure, 1,000 cycles/burst, 1 burst/seconds, 30 seconds duration. If one parameter was changed, all other parameters remained as described above.

Chicken embryos bearing HT1080-GFP tumors were injected with nanodroplets and, for some embryos, had tumors sonicated as described above. Blood was collected (50 μL) and was diluted with 50 μL washing buffer, which was PBS containing 3 mg/mL ethylenediaminetetraacetic acid and 2% (v/v) FBS. Some blood samples from embryos without ultrasound exposure had 100,000 HT1080-GFP cells added to the samples to verify HT1080-GFP cells could be isolated and detected by flow cytometry. To separate red blood cells from mono-nuclear cells, diluted blood was layered on top of 200 μL Lymphoprep (STEMCELL Technologies Inc.) in 1.5 mL microcentrifuge tubes and tubes were centrifuged at 800 ×g for 20 minutes at room temperature. The mono-nuclear cells at the interface of the solutions were collected, washed with 750 μL washing buffer, and resuspended in 500 μL washing buffer. Cells were analyzed with a BD FACSCanto II flow cytometer and, based on data from samples containing additional HT1080-GFP cells, green fluorescent signal above 10,000 was used as the threshold for identification of HT1080-GFP cells.

Chicken embryos bearing a single HT1080-luciferase-GFP tumor were injected intravenously with nanodroplets and had tumors sonicated as described above. Some tumor-bearing embryos did not have nanodroplets injected or tumors sonicated. Four days after ultrasound exposure, embryos were given an intravenous injection of 50 μL luciferin (15 mg/mL) and tissues were dissected from the embryos approximately 3 minutes post-luciferin injection. Dissected tissues included a piece of CAM distant from the primary tumor, liver, lung, and brain, which were rinsed with PBS and stored in 250 μL luciferin (0.3 mg/mL). Tissues from individual chicken embryos were imaged for bioluminescence and green fluorescence using a Bruker In-Vivo Xtreme system equipped with a back-illuminated 4 Megapixel camera. Bruker molecular imaging software (version 7.5.2.22464) was used for signal analysis. Bioluminescence signal within tissues was normalized by summing tissue signal, which was divided by tissue weight.

Chicken embryos bearing HT1080-GFP tumors were injected with nanodroplets and, for some embryos, had tumors sonicated as described above. Serum GFP mRNA and miR21-5p levels were determined as described in the Supplementary Materials and Methods.

Chicken embryos bearing HT1080-GFP tumors were injected with nanodroplets and, for some embryos, had tumors sonicated as described above. Serum was collected and processed for RAC1 mRNA sequencing as described in the Supplementary Materials and Methods. The serum RAC1 mRNA sequence was compared with that of cultured HT1080-GFP cell lysates to verify that both sequences matched.

Chicken embryos bearing HT1080-GFP or HEp3-GFP tumors were injected with nanodroplets and, for some embryos, tumors were sonicated as described above. Serum was collected and was processed for whole genome sequencing and analyzed as described in the Supplementary Materials and Methods.

Data were analyzed using GraphPad Prism software (version 5.00). For experimental groups where Bartlett test of equal variance provided P values less than 0.05 (i.e., variances were not equal), data were log transformed, which caused Bartlett test P values to be greater than 0.05 (i.e., variances were not different between groups and ANOVA analysis was appropriate). The Student t-test was used when comparing two groups and one-way ANOVA was used when comparing three or more groups. Bonferroni posttest correction was used to compare all pairs of experimental groups and P values ≤0.05 were required for statistical significance. For each experiment, the number of samples analyzed per group is described in the figure legends.

To determine whether focused ultrasound can stimulate the release of EVs from cancer cells in culture, HT1080-GFP fibrosarcoma cells in 96-well plates were treated with high pressure 1.15 MHz ultrasound and the media was analyzed for EVs using the Apogee A50 micro flow cytometer, which is designed to detect particles as small as 100 nm. The bottom of the plate attenuated approximately 8% of the ultrasound pressure thus cells within the well were exposed to approximately 27.6 MPa pk-pk pressure. Media from HT1080 and HT1080-GFP cells contained very low numbers of green fluorescent EVs (1.7 and 32 vesicles per μL, respectively). When 10 to 10,000 cycles of focused ultrasound were applied to HT1080-GFP cells, the level of green fluorescent EVs in the media increased modestly (Fig. 2B, C, E), whereas sonication of HT1080 cells resulted in no significant increase in green fluorescent EVs in the media (Fig. 2A). Comparing the EVs with polystyrene beads of known sizes indicated that the majority of GFP⁺ EVs were between 180 and 240 nm in diameter (Fig. 2D).

We then investigated the impact of applying focused ultrasound on cultured cells in the presence of perfluorocarbon nanodroplets, which can enhance the effect of ultrasound (Supplementary Fig. S1A). Cryo-electron microscopy imaging of nanodroplets revealed particles between approximately 100–500 nm in diameter (Supplementary Fig. S1B). The nanoparticle tracking analysis method revealed that nanodroplets had a mean diameter of 156 ± 5 nm and a mean concentration of 4.1 ± 1.0 × 10¹⁰ particles per mL (Supplementary Fig. S1C). Microscopy imaging of nanodroplet and 4% (v/v) microbubble solutions suggested that nanodroplet solutions without or with sucrose purification had 0.51% and 0.13% of microbubbles remaining in the nanodroplet solutions, respectively (Supplementary Fig. S2). When HT1080-GFP cells were sonicated with 10,000 cycles of ultrasound in the presence of 2% nanodroplets, a significant increase in the number of GFP⁺ EVs was observed (P < 0.01, Fig. 2E) whereas cell viability did not significantly decrease even at 10,000 cycles of ultrasound (Fig. 2F). Sonication of HT1080-GFP cells with 10,000 cycles of ultrasound with 2% nanodroplets present caused 330-fold greater green fluorescent EV release compared with nonfluorescent HT1080 cells under similar conditions (Fig. 2G). We did observe some cell detachment at the higher ultrasound exposure levels, which likely reflects a limitation in utilizing this technique with cells in two-dimensional (2D) cultures. Nevertheless, these experiments demonstrate that exposing cancer cells to focused ultrasound in the presence of perfluorocarbon nanodroplets significantly increases the release of EVs, which prompted us to investigate whether this could be replicated in an in vivo model and whether released EVs could be detected in the bloodstream.

We utilized an ex ovo chicken embryo xenograft model where HT1080 or HT1080-GFP tumors are established in the CAM (22–26). Embryos bearing HT1080 and HT1080-GFP tumors were injected intravenously with PBS with or without nanodroplets and were exposed to focused ultrasound on the tumor or on another region of the CAM not containing cancer cells (Fig. 3A). Sonication of HT1080 tumors with no GFP expression or sonication of the CAM without HT1080-GFP tumor cells resulted in no increase in green fluorescent EVs in the serum (Fig. 3B and C). The only conditions that caused a significant increase of green fluorescent EVs in the serum was sonication of HT1080-GFP tumors in embryos previously injected with nanodroplets (Fig. 3D and F), suggesting that both nanodroplets and ultrasound were necessary for EV release from tumors. Interestingly, some of the ultrasound-mediated released green fluorescent EVs displayed diameters >1,300 nm, which was much greater than the released particles observed from sonicated cultured cells (Figs. 2D and 3E). This may be due to the larger amount of ultrasound energy HT1080-GFP tumors were exposed to in the in vivo experiments. When performing similar in vitro (Supplementary Fig. S3A) and in vivo (Supplementary Fig. S3B) experiments described above but replacing nanodroplets with 0.51% microbubbles, ultrasound-mediated cancer-derived EV release significantly decreased and was similar to levels without ultrasound contrast agent. Unpurified and purified nanodroplets had similar EV release enhancing effects, suggesting that nanodroplets, and not the remaining microbubbles, were responsible for the observed ultrasound enhancing effects (Supplementary Fig. S3).

Various ultrasound and experimental parameters were individually altered to determine the optimal conditions for vesicle release from tumors in vivo. The lowest ultrasound pressure (pk-pk) tested causing significant EV release from HT1080-GFP tumors was 10 MPa (Fig. 4A). Higher pressures did not significantly increase EV release from tumors and had no effect on size of released EVs (Supplementary Fig. S4A). Increasing the number of transmitted ultrasound cycles per interval (burst) increased EV release up to 1,000 cycles, which displayed the maximum level of EV release (Fig. 4B). Varying ultrasound cycles per interval had no effect on size of released EVs (Supplementary Fig. S4B). The total duration of ultrasound exposure to the tumor affected EV release with the maximum effect observed at 30 seconds exposure (Fig. 4C). Interestingly, higher ultrasound durations caused the mean particle large angle light scatter to decrease, suggesting that longer sonication times increased the release of smaller particles (Supplementary Fig. S4C). Surprisingly, nanodroplets could be diluted 128-fold and still displayed relatively high levels of tumor EV release (Fig. 4D). These doses are comparable with clinically relevant dose recommendations for Definity microbubble contrast agents (27). Dilution of nanodroplets at 256-fold eliminated any observable tumor EV release. The time between nanodroplet intravenous injection and tumor sonication inversely affected tumor EV release (Fig. 4E) with the shortest time interval of 3 minutes displaying significantly greater tumor EV release than 3 or 6 hours (P < 0.05).

To determine the effect of ultrasound/nanodroplets on tumor cell viability, some HT1080-GFP tumor-bearing chicken embryos were intravenously injected with nanodroplets and had tumors exposed to ultrasound. Tumors were excised 24 hours post-sonication and TUNEL staining demonstrated that tumor regions outside the ultrasound focus remained viable, whereas apoptosis was detectable within the ultrasound focal zones (Supplementary Fig. S5).

Using flow cytometry, we determined if sonication of tumors in embryos preinjected with nanodroplets caused increased circulating tumor cells in the blood. Compared with chicken embryos without ultrasound, embryos with tumors exposed to ultrasound had no significant increase in detectable HT1080-GFP cells in the blood (Fig. 5A–D). To determine if application of ultrasound/nanodroplets to tumors increased the tumors' metastatic potential, chicken embryos with HT1080-GFP tumors were given an intravenous injection of nanodroplets and tumors were sonicated. Four days after tumor sonication, tissues including the CAM (away from the tumor), liver, lung, and brain were dissected and imaged for green fluorescence and luciferase (Fig. 5E and F). In general, nanodroplet injection and tumor sonication caused decreased observed luciferase signal in tissues, although no significant differences were observed between groups (Fig. 5G). The results suggest that ultrasound/nanodroplets do not enhance the risk of tumor metastasis.

Using the optimal ultrasound conditions for tumor EV release, HT1080-GFP tumors were sonicated and the blood was analyzed for levels of miR21-5p and GFP mRNA. miR21-5p is an oncomir known to target a variety of tumor suppressor genes and to be up-regulated in several metastatic cancers (28, 29). Serum EVs were isolated by centrifugation, which allowed discrimination of free and EV-associated miRNAs. miR21-5p from EVs was present in serum of nonsonicated embryos at 880,000 copies per μL of serum (Fig. 6A and C). Ultrasound off or on tumors increased EV miR21-5p levels in serum three- and seven-fold, respectively, compared with nonsonicated tumors. miR21-5p is present in chicken cells, which likely explains the increase in miR21-5p when sonicating off of tumors. Compared with ultrasound off of tumors, ultrasound on tumors significantly increased EV miR21-5p levels in serum by 2.3-fold (P < 0.05, Fig. 6B). Serum miR21-5p levels were approximately 95% free and 5% in EVs regardless of ultrasound application. GFP mRNA was undetectable in the blood without tumor sonication but was detectable at 480 copies per μL of serum after tumor sonication (Fig. 6B and D). EV-associated GFP mRNA, determined by incubating samples in RNaseA, which degraded mRNA outside of EVs, represented 82% of detectable GFP mRNA (Fig. 6D).

To assess whether EV mRNA released from tumors could be used for identification of cancer genetic signatures, we sought to detect the N92I RAC1 mutation in HT1080-GFP cells, which conveys increased tumorigenicity (30). EV-associated RAC1 mRNA was only detectable in serum samples from embryos bearing HT1080-GFP tumors that were exposed to nanodroplets and ultrasound. EV-derived RAC1 mRNA contained the N92I mutation, which was detectable in cultured HT1080-GFP cells (Fig. 6E). In addition, PCR amplification of fragments encompassing heterozygous mutations reported in the catalogue of somatic mutations in cancer (COSMIC), followed by low-pass deep sequencing, allowed the identification of 13 mutations in the HT1080 genome (Supplementary Table S1). The data suggest that ultrasound released EVs contain genetic information capable of profiling tumor-specific mutations, which can predict tumor aggressiveness.

We then determined whether the application of focused ultrasound in the presence of nanodroplets could enhance the detection of tumor genomic DNA circulating in the blood. Next-generation sequencing libraries were constructed from DNA isolated from serum of chicken embryos bearing larger HT1080-GFP tumors (2–5 mm) or smaller HEp3-GFP tumors (1–2 mm), with or without ultrasound exposure and injection of nanodroplets. Nanodroplets/ultrasound exposure to the HEp3 and HT1080-GFP tumors increased the number of human DNA sequence reads in serum by 16- and 102-fold, respectively, compared with embryos without tumor sonication (Fig. 7A and B) with apparent uniform coverage of the entire tumor genome for HT1080 tumors (Fig. 7C). These encouraging results indicate that nanodroplets and ultrasound can significantly enhance the levels of tumor-derived DNA in the blood, even in very small lesions.

This study provides evidence that focused ultrasound in the presence of nanodroplets can enhance the release and detection of EVs and their associated biomarkers from tumors into the blood. Given that high-intensity focused ultrasound can be focused in regions approximately 5 × 1 × 1 mm, target tissues can receive ultrasound exposure in precise regions to stimulate localized EV release into the bloodstream. Blood samples can be acquired and compared pre-/postsonication to interrogate biomarker levels in precise regions of interest with baseline biomarker levels determined for each patient. This would be ideal for biomarkers with significant interpatient variability including PSA for prostate cancer (31).

Although our in vitro data suggested that ultrasound and nanodroplets caused minimal cell death, our in vivo data demonstrated apoptosis within small tumor regions corresponding to cells within ultrasound focal zones. This discrepancy is likely due to differences in (i) total ultrasound energy applied to cells, and (ii) the cell environment (i.e., 2D cultures with abundant nutrients vs. 3D tissues requiring blood vessels for nutrients). High pressure ultrasound, especially with ultrasound contrast agents, can damage blood vessels and impair their function, potentially increasing cell death of nearby cells (32, 33). The side effect of tumor cell death may be considered beneficial, although this may restrict our diagnostic technique to known tumor locations. Focused ultrasound will be necessary to minimize the toxicity of this technique to small regions of interest.

Deeply focused ultrasound minimizes skin exposure to ultrasound pressures necessary for microbubble destruction and likely eliminating the risk of infection. By comparison, biopsy needles physically disrupt all tissues from the skin to the target biopsy site, significantly increasing the risk of infection. Such wide-scale physical disruption of tissue may explain why biopsies have been observed to increase the risk of metastasis (34). Interestingly, ultrasound exposure to cultured PC3 prostate cancer cells incubated with microbubbles have demonstrated reduced cell migration and invasion with reduced MMP-2 and MMP-9 expression (35). In another study, metastasis was reduced in mice bearing breast 4T1 tumors when mice were injected with microbubbles and had tumors sonicated (36). These results as well as our own suggest that tumor sonication with ultrasound contrast agents is more likely to inhibit rather than promote tumor metastasis.

Although physical biopsies have some risks to patients, they have proven effective for characterizing different malignancies and provide valuable clinical information that influences clinical decisions. Biopsies provide tumor samples with intact physical architecture that cannot be acquired with ultrasound and blood samples. We do not propose that nanodroplets and ultrasound should necessarily supplant traditional biopsy methods. Instead, we believe nanodroplets and ultrasound may provide additional tumor-derived material when traditional biopsy methods may not be appropriate. Sonication of tumors for biomarker release may also be useful for research studies if longitudinal repeat tumor biopsies are too invasive. In addition, biomarkers could be liberated from different focal regions of tumors for analyzing tumor heterogeneity.

Intravenous injection of nanodroplets was required for detection of tumor-derived vesicles after tumor sonication. Given that vesicle release was greatest with the shortest time interval between nanodroplet injection and tumor sonication (3 minutes), the vast majority of the nanodroplets would have been present within the blood vessels during this time. This suggests that vesicle release from tumors is maximized when nanodroplets are destroyed near endothelial cells. Ultrasound-mediated microbubble destruction in the blood vessels of the chorioallantoic membrane has previously been shown to increase blood vessel permeabilization, suggesting that this may be an important mechanism for how tumor vesicles entered the vasculature (33).

One major limitation of our study is the difference in cancer vesicle/biomarker detection in our experiments and in the clinic. We use GFP expressing tumors to easily identify cancer vesicles in the serum, which does not translate in the clinic. Additionally, we used human-specific primers for detection and sequencing of HT1080-GFP tumor RAC1 mRNA. In the clinic, tumor-specific gene primers cannot necessarily be designed. To identify and characterize ultrasound-mediated mRNA release from tumors, tumor-derived vesicles could be purified using affinity columns or immuno-magnetic methods for cell surface markers present specifically on tumor cells. These purified tumor vesicles from the serum could then be analyzed to determine tumor genetics as previously demonstrated (37).

Analyzing circulating tumor DNA (ctDNA) is a potentially powerful diagnostic technique for profiling tumor genetics (38, 39). Unfortunately, ctDNA levels vary greatly between patients with some patients exhibiting no detectable ctDNA. Much of the ctDNA is believed to be derived from circulating tumor cells, suggesting that analysis of ctDNA would primarily be useful in patients with advanced metastatic disease. Given that chicken embryos bearing HT1080-GFP tumors had minimal ctDNA before ultrasound but >100-fold increased levels after ultrasound, our nanodroplet/ultrasound-based technique may allow genomic characterization of localized tumors that is ideal, because tumor aggressiveness should be determined before cancers metastasize while they are still typically curable.

To our knowledge, this is the first study to determine that nanodroplets and ultrasound can stimulate vesicle release from tumor cells that can be detected in the serum. The released vesicles contained tumor-derived protein/nucleic acid biomarkers with tumor-specific mutations allowing characterization of tumor phenotype including aggressiveness. Future studies optimizing this procedure in larger animal models with longer longitudinal safety studies are warranted. This diagnostic technique may provide clinically relevant tumor information for situations where (repeat) needle biopsies are not appropriate due to health risks to patients. Our findings suggest a new paradigm and path forward for biomarker research with the vision that radiographically identified lesions could be sonicated with image-guidance to stimulate release of biomarker-loaded vesicles for molecular profiling of tumors.

No potential conflicts of interest were disclosed.

Conception and design: R.J. Paproski, J.D. Lewis, R.J. Zemp

Development of methodology: R.J. Paproski, J.D. Lewis, R.J. Zemp

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.J. Paproski, J. Jovel, G.K.-S. Wong, J.D. Lewis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.J. Paproski, J. Jovel, J.D. Lewis, R.J. Zemp

Writing, review, and/or revision of the manuscript: R.J. Paproski, J. Jovel, G.K.-S. Wong, J.D. Lewis, R.J. Zemp

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Jovel, R.J. Zemp

Study supervision: J.D. Lewis, R.J. Zemp

We thank Sandra O'Keefe for assistance with whole genome sequencing, Dr. Katia Carmine Simmen for her assistance with chicken embryo tissue collections, and Shalawny Miller for performing tumor sectioning and staining.

We gratefully acknowledge funding from the following agencies. R.J. Zemp is funded by Prostate Cancer Canada (MVBRDG D2013-40), Terry- Fox Foundation and the Canadian Cancer Society (TFF 019237, TFF 019240, CCS 2011-700718, CCSRI IG 702032), NSERC (355544-2008, 375340-2009, STPGP 396444), Canadian Institutes for Health Research (CIHR CPG 134739), the Canada Foundation for Innovation, Leaders Opportunity Fund (18472), Alberta Advanced Education & Technology, Small Equipment Grants Program (URSI09007SEG). G.K.S. Wong is funded by the Alberta Innovates Technology Futures (AITF) through an Innovates Centres of Research Excellence (iCORE) Strategic Chair. J.D. Lewis is funded by Alberta Innovates – Health Solutions (AIHS 201201259), Alberta Cancer Foundation (ACF 26001), and Prostate Cancer Canada (PCC MTA TAG2014-03).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.