In vitro potentiation of doxorubicin by unseeded controlled non-inertial ultrasound cavitation

Ultrasound-generated non-inertial cavitation has the ability to potentiate the therapeutic effects of cytotoxic drugs. We report a novel strategy to induce and regulate unseeded (without nucleation agents) non-inertial cavitation, where cavitation is initiated, monitored and regulated using a confocal ultrasound setup controlled by an instrumentation platform and a PC programmed feedback control loop. We demonstrate, using 4T1 murine mammary carcinoma as model cell line, that unseeded non-inertial cavitation potentiates the cytotoxicity of doxorubicin, one of the most potent drugs used in the treatment of solid tumors including breast cancer. Combined treatment with doxorubicin and unseeded non-inertial cavitation significantly reduced cell viability and proliferation at 72 h. A mechanistic study of the potential mechanisms of action of the combined treatment identified the presence of cavitation as required to enhance doxorubicin efficacy, but ruled out the influence of changes in doxorubicin uptake, temperature increase, hydroxyl radical production and nuclear membrane modifications on the treatment outcome. The developed strategy for the reproducible generation and maintenance of unseeded cavitation makes it an attractive method as potential preclinical and clinical treatment modality to locally potentiate doxorubicin.

remain in the blood stream 11,12 with intended mechanical effects resulting in bioeffects such as sonoporation or enhancement of vascular permeability. The later mechanism is the subject of current preclinical and clinical studies aiming at transiently opening the blood brain barrier (BBB) for drug delivery in a safe and reproducible way [13][14][15][16] . Echocontrast agents on the gaseous form are limited to the micrometric scale to remain thermodynamically stable 17 and thus do not permit to nucleate cavitation in areas remote from the tumor vasculature. Alternative for such agents in the sub-micron scale as nanobubbles or nanocups are being investigated [18][19][20][21] .
Cavitation is one of the ultrasound-associated mechanism that creates numerous biological effects and might affect drug activity 22,23 . Non-inertial cavitation can permeabilize transiently the plasma membrane by creating resealable nano-sized pores, allowing transportation of molecules across the membrane [24][25][26] , a process called sonoporation 27 . Cavitation can cause intracellular effects and localize macromolecules directly into the nucleus 28,29 , through the augmentation of nuclear pores and a change in their distribution throughout the nuclear membrane, which could potentiate the action of intercalant DNA agents such as DXR. US can affect DNA structure as well as its function by inducing single-and double strand breaks (DSBs) 30 . These alterations could lead to enhanced cell apoptosis in the presence of DXR, as well as changes in cell morphology (such as membrane poration or organelle destruction) and in cell nuclei 31,32 . Ultrasound interaction with cells and tissue can also lead to US-induced mild hyperthermia (37 °C to 41 °C), which can also increase drug penetration in tumor cell suspensions 33 . US cavitation can generate free radicals, being another potential mechanism for an enhanced DXR cytotoxicity 34,35 .
We hypothesize that combining DXR with unseeded (without nucleation agents) non-inertial cavitation would enhance its therapeutic efficacy. As cavitation induced without the need of UCA can potentially be generated in all tissues, and not only the vasculature space reached by UCA, this method can provide a higher versatility compared to UCA-based cavitation. Moreover, developing a method to generate unseeded cavitation with high reproducibility over long period of time can overcome another hurdle associated with the use of UCA, which is the requirement to have UCA infusion in order to generate constant level of cavitation 36 . Alternating between high amplitudes US pulses to generate bubble cloud and low amplitude pulses to sustain its oscillation has been reported to show benefits on cavitation activity 37 . In our experimental setup, the method is used to generate and maintain cavitation activity, using monitoring of cavitation activity to dynamically dictate the alternation between the two power modes.
We report in this study the generation and maintenance of non-inertial cavitation in vitro with high reproducibility, and without the use of UCA, by using two US power modes adequately alternated by a feedback loop. This unseeded non-inertial cavitation can act synergistically with DXR to affect 4T1 cells viability and proliferation in vitro. We investigated the underlying mechanisms enhancing the therapeutic effect: changes in cell morphology such as membrane poration, organelle destruction or formation of nuclear pore complex; increased DXR internalization due to sonoporation; heating and free radicals formation.

Results
Reproducibility of non-inertial cavitation. To quantify the level of non-inertial and inertial cavitation, two cavitation indexes (CI) were defined based on specific acoustic features: the non-inertial cavitation index -niCI (dB), calculated as the difference between the maximum power level in the 540-560 kHz sub-harmonic range and the mean base level obtained in the 560-600 kHz range -and the inertial cavitation index -iCI (dB), calculated as the difference between the average of the entire frequency spectrum in dB in the 0.1-7.1 MHz range and its base level. On average over the different sets of experiments, the niCI was found well above the non-inertial cavitation threshold of 3 dB (Fig. 1). Similar niCI were obtained in the presence or absence of DXR (niCI of 10.32 ± 0.93 in the absence of DXR, 10.16 ± 0.63 in the presence of DXR). These values were also reproducible between experiments, confirming the generation and the control of non-inertial cavitation activity (Fig. 1). Noninertial cavitation was generated in exposed samples with approximately 2.7% of bursting pulses to maintain cavitation all along the exposure duration. The iCI was 14.55 ± 1.69 in the absence of DXR, and 14.23 ± 1.54 in the presence of DXR. These values are below the iCI threshold for inertial cavitation initiation (Fig. 11). The inertial cavitation activity is residual from the high amplitude bursting pulses.
In vitro enhanced cytotoxicity effect. Cell viability. The cytotoxicity of the combined treatment with doxorubicin and non-inertial cavitation was assessed in vitro 1 h and 72 h after treatment. Sonication significantly decreased the cell viability at 1 h of approximately 15% (p value < 0.0001, post-hoc test p-value indicated on Fig. 2). At 72 h, there was a statistically significant difference between the treatment groups (p value < 0.0001, post-hoc test p-value indicated on Fig. 2). DXR decreased the cell viability compared to controls and US. These results are consistent with the mortality rate and kinetics expected after a treatment with doxorubicin at 400 ng/mL, as cell death caused by DXR is expected to occur only after 24 to 48 h 38 . In the US + DXR group, cell viability was significantly lower than in the DXR group at 72 h with a drop of cell viability of 50% compared to DXR alone.
Cell proliferation. Cell proliferation was quantified over time with a cell growth index (CGI), the ratio of cell number at 72 h over cell number at the time of plating. At 72 h, the CGI was close to 6 for controls and US groups, likely limited by having reached confluence in the culture wells. In DXR treated samples, CGI was 0.82 ± 0.29 after 72 h. In US + DXR samples, the CGI was 0.32 ± 0.23 after 72 h. The CGI at 72 h was significantly lower in the US + DXR group compared to DXR alone (p value < 0.0001, post-hoc test p-value indicated on Fig. 3).
Mechanistic study. Effect of the presence of cavitation. Our system uses high pressure bursting pulses to generate cavitation, and then lower pressure pulses to maintain non-inertial cavitation activity. When the bursting pulses were not applied, no cavitation could be detected during the treatment as quantified by the very low average levels of the cavitation indexes (niCI = 1.81 ± 2.81, iCI = 9.87 ± 0.22). When boosting pulses were applied, the levels of cavitation indexes reached values associated with non-inertial cavitation activity (niCI = 8.07 ± 3.19; iCI = 12.06 ± 1.21).
In DXR groups, the cell viability with and without sonication was similar in the absence of cavitation at any time point (Fig. 4).
DXR + US-Cav treatment decreased the cell viability compared to DXR alone at 1 h (p value = 0.0048, post-hoc test p-value indicated on Fig. 4) and to DXR alone or DXR + US-NoCav at 72 h (p value = 0.009, post-hoc test p-value indicated on Fig. 4), demonstrating that cavitation is required for inducing the therapeutic effect observed.
DXR internalization. The fluorescence intensity of DXR was similar between DXR and US + DXR groups at the two time points, suggesting that the drug uptake was similar with and without unseeded cavitation (Fig. 5).
Effect of temperature. The US-induced hyperthermia was mimicked using a water bath heated at 29 °C or 37 °C. No significant difference in cell death after 72 h (p value = 0.51) was observed between the heated and non-heated groups (Fig. 6). This suggests that the transient US-induced temperature elevation did not enhance the combined DXR efficacy in our study.
Effect of reactive oxygen species. The mean quantity of ROS in each sample was assessed using the specific ROS marker H2DCFDA. At 1 h, the mean quantity of ROS was significantly higher in sonicated groups than in DXR group and tend to be higher compared to non-treated group (p value = 0.0015, post-hoc test p-value indicated on Fig. 7). At 72 h, the mean quantity of ROS was significantly larger for groups with DXR than for groups without DXR (p value = 0.0009, post-hoc test p-value indicated on Fig. 7). The data may suggest two different mechanisms of ROS production, one short term due to acoustic cavitation, and one long term due to DXR. The results concerning mean intracellular quantity of doxorubicin, cell viability and growth index in each group were similar to those obtained in the previous experiments.
To assess whether ROS production contributes to cytotoxicity, 4T1 cells were cultured in the presence of a hydroxyl scavenger (L-Histidine), and cell viability assessed at 1 h and 72 h. For all the treatment conditions tested, the viability was not significantly affected by ROS scavenging and were similar to those obtained in the previous set of experiments. ROS scavenging did not affect cell proliferation at 72 h in any of the condition tested (p value < 0.0001, post-hoc test p-value indicated on Fig. 8), nor did it affect changes in viability induced by ultrasound at 1 h, nor the increased cytotoxicity of the combined treatment US + DXR compared to DXR alone at 72 h.
Bioeffects of the US on cell membranes. Confocal microscopy observations of cells did not show any morphological differences between sonicated and non-sonicated cells, either at the levels of the entire cell, cell membrane, the nuclear membrane or the cell organelles. The quantity of nuclear pore complexes was similar in all samples (24 ± 2 per cell). Typical example images of these pores are shown on Fig. 9. Reproducibility of non-inertial cavitation. niCI and iCI values obtained for US and US + DXR groups when each experiment was considered as an independent measure. Figure shows that non-inertial cavitation was generated in a controlled and reproducible manner.

Discussion
The aim of this study was to evaluate DXR potentiation by unseeded non-inertial cavitation, to induce cell death in murine 4T1 mammary tumor cells in vitro. Using a confocal device, we first designed a feedback loop that alternates conditionally between two exposure levels that either generate bubbles or maintain the non-inertial cavitation activity. The parameters were optimized so that reproducible non-inertial cavitation exposures could be applied. In a pilot study on 4T1 tumors in mice (unpublished data), we were also able to generate and maintain  www.nature.com/scientificreports www.nature.com/scientificreports/ non-inertial cavitation in tumor area using a different set of parameters (DC 15%, PRF 100 HZ). This experiment provides encouraging leads for our regulation system to efficiently generate and maintain non-inertial cavitation in vivo.
We studied a potential synergetic effect of DXR and US. The action of US combined to that of DXR had both a higher and a faster therapeutic effect on tumor cells, therefore supporting a potential synergy. The experiments showed that ultrasound did not induce delayed cell mortality, as the immediate mortality assessed at 1 h did not have any influence on the results at 72 h. Furthermore, cells that did not die immediately after US treatment proliferated normally; cell growth in the US group was similar to that in the control group. No therapeutic effect could be observed in the absence of cavitation at 72 h, for the same amount of energy deposition in the exposed samples. At 1 h, the viability in the sonicated group with non-inertial cavitation was significantly lower than in the  www.nature.com/scientificreports www.nature.com/scientificreports/ control group and lower than the sonicated group without non-inertial cavitation, although this last difference was not statistically significant probably due to low number of samples and low sensitivity of non-parametric test. This further highlights the key role of cavitation in mechanism of action and bioeffects of US.
To investigate the mechanisms by which cavitation can enhance cytotoxicity of DXR in this in vitro experimental settings, several factors were independently investigated. One of the most important and intriguing point is highlighted by the absence of increased DXR internalization after US treatment. Indeed, one of the initial rational for this work was that US-induced cavitation would enhance the intracellular delivery of drugs by transiently increasing the cell membrane permeability 39 . The increased permeability temporal window has been studied in 4T1 cells and estimated to be 3 h 40,41 . In the current experimental settings, there was no observable difference in DXR internalization between DXR and US + DXR group at short-term (1 h) or long term (72 h) post-treatment.  www.nature.com/scientificreports www.nature.com/scientificreports/ One of the possible explanations for the absence of enhanced internalization in our specific experimental settings lies in the post-treatment flow cytometry protocol that happened after cells have been exposed for 1.5 to 2 h to DXR. Because doxorubicin transport across cell membranes takes place by simple Fickian diffusion 42 and the equilibrium between intracellular and extracellular concentration of DXR is rapidly achieved, all cells, sonicated or not, may have reached intracellular peak concentration at the time of flow cytometry analysis.
The effect of temperature elevation caused in our in vitro experiments by US exposure was investigated. Although the kinetics of the temperature elevation caused by ultrasound exposure was not exactly reproduced inside a heated water bath, both US treated and warmed samples were allowed to reach similar maximum temperature of 29 °C. In a separate condition, samples were even heated by a 37 °C water bath for 30 s. None of the heating experiments resulted in any significant therapeutic effects of DXR compared to non-heated conditions, and temperature elevation was ruled out as a causative mechanism of the observed synergetic effect between US and DXR. It may be assumed that a 30 s temperature elevation of 2 °C as observed in our experiment is too short to significantly influence cell metabolism. Another possible explanation lies in our methodology as cells were always in presence of free DXR. Temperature may influence DXR activity and cytotoxicity, as mentioned in several studies 43,44 . However, in these studies, DXR was removed from the cell medium after exposure to investigate the impact of temperature on its kinetics. In our experiments, because cells were exposed to DXR after each treatment, the change in temperature did not impact cytotoxicity at 72 h. A recent clinical study reported increased intratumoral drug delivery after ultrasound-mediated hyperthermia and injection of lyso-thermosensitive liposomal DXR 45 . However, in our in vitro experimental setting, and as reported in vivo by Lyon et al., it is unlikely that addition of ultrasound-mediated hyperthermia to free doxorubicin would significantly increase cytotoxicity.
Cavitation triggers the generation of reactive oxygen species (ROS)-derived radicals, which can initiate a chain peroxidation of cell membrane lipids 46 . ROS generation is one of the DXR ways of action. Thus, the ROS produced by cavitation may add up and increase the overall cytotoxicity of the combined treatment. This strategy is known as sonodynamic therapy 47 . ROS production was quantified using H2DCFDA. Sonication tend to increase ROS formation at 1 h, mostly likely due to stressed cells after sonication. ROS quantities were increased for DXR-treated groups at 72 h. This suggests that DXR was the major regulator of ROS production in our treatment scheme 48 . However, their inhibition with histidine did not affect cell proliferation at 72 h in any of the condition tested, and the role of ROS production in our settings remains to be clarified.
Possible opening of nuclear pore complexes upon ultrasound exposure was also investigated. Such opening could lead to increased uptake of DXR or changes and/or alterations of the cell homeostasis, such as endoplasmic reticulum stress or different mitochondria potential membrane. These effects have been reported following ultrasound cavitation 23,49-52 , but no difference could be observed between sonicated and non-sonicated cells in our study. Despite the previous uncertainty on whether the cytometry analysis is biased or not by a potential DXR peak concentration reached in both DXR and US + DXR groups, these morphological observations reinforce the hypothesis of sonoporation not being a responsible mechanism of US-enhancement of DXR in our study.
Our mechanistic investigation did not provide a clear understanding of what mechanism is responsible for the observed synergy between DXR and US non-inertial cavitation. We propose that DXR-US treatment affecting DNA damage could be a possible mechanism, as ultrasound has been reported to be able to induce DNA double-strand breaks (DSBs) 28 , and DXR intercalation with the mitochondrial 53 or the nuclear DNA double helix causes inhibition of topoisomerase II leading to DSBs 28,54 . www.nature.com/scientificreports www.nature.com/scientificreports/ In conclusion, we designed an efficient strategy to produce controlled and reproducible non-inertial cavitation. We showed that unseeded non-inertial cavitation potentiates the cytotoxic effect of doxorubicin on cell viability and proliferation in 4T1 tumor cells in vitro. Cavitation was identified as necessary to enhance DXR efficacy. Potential underlying mechanisms were investigated independently, including increase in intracellular doxorubicin, temperature elevation, production of reactive oxygen species, changes in cell morphology, but none could explain the observed enhanced efficacy. Further work will be required to understand how non-inertial cavitation potentiates doxorubicin.

Materials and Methods
Reproducible generation of non-inertial cavitation. The cavitation device used in this study is composed of two confocal spherical PZ28 piezoceramic transducers (Ferroperm, Kvistgaard, Denmark) with curvature radius of 50 mm, separated by an angle of 90° and operated at 1.1 MHz (Fig. 10). This configuration favors the occurrence of cavitation 55 . The transducers were immersed in a tank filled with Ablasonic (EDAP-TMS, Vaulx-en-Velin, France), a cavitation-inhibitor liquid used to prevent acoustic cavitation outside the sonicated sample 56,57 .
For treatment, a 2 mL safe-lock micro test tube (Eppendorf, Hamburg, Germany), containing 650 µL of cell suspension, was positioned at the focal zone of the confocal device (Fig. 10A). The emission signal was generated with a National Instruments PXI-1031 device (National Instruments, Nanterre, France) driven by a LabVIEW software and powered by a 200 W amplifier (LA-200H EMV, Kalmus, Hong Kong, China). All ultrasound treatments were performed at room temperature.
Cavitation was monitored using an in-house hydrophone made with a polyvinylidene fluoride (PVDF) film (25 µm thickness, 10 mm in diameter) embedded in resin (AY103 Araldite+ 10% HY956) placed between the two focused transducers and directed towards the exposed sample. Signals were recorded at a sampling frequency of 33.3 MHz over segments of 16384 points.
Non-inertial cavitation activity can be maintained by reducing the acoustic pressure following creation of a bubble cavitation cloud by higher pressure 58 . In our experimental settings, preliminary experiments (internal data) have shown that driving the transducers in pulse mode with a pulse repetition frequency (PRF) of 25 Hz, and number of cycles of 6600, corresponding to a duty cycle (DC) = 15%, during an exposure time of 30 seconds, could generate non-inertial cavitation with a cell death limited to less than 30%. These settings were applied in the present study. www.nature.com/scientificreports www.nature.com/scientificreports/ A non-inertial cavitation regulation system was implemented, using the recorded cavitation noise as input for a feedback loop amplitude control of the emitted signal. With this control system, two distinct modes can be emitted: a high-power mode or "bursting pulses" aimed at creating a population of bubbles, with a peak negative pressures (PNP) of 6.7 MPa (Pmax) at the focus, and a low-power mode or "regulation pulses" aimed at maintaining their oscillations, with a peak negative pressures (PNP) of 2.9 MPa (Pmin) at the focus.
The inertial and non-inertial cavitation contributions were separated based on the specific features in their frequency content for cavitation analysis (Fig. 10B). To quantify the level of inertial and non-inertial cavitation, two cavitation indexes (CI) were defined. The non-inertial cavitation index (niCI (dB)) value is monitored in real time and used to select the emission mode: if the niCI falls below the threshold associated with a lack of non-inertial cavitation, the system emits a bursting pulse to generate a new bubble cloud; if the niCI is above the threshold, the system emits low power mode to maintain bubbles oscillations. The niCI was calculated as the difference between the maximum power level in the 540-560 kHz sub-harmonic range and the mean base level obtained in the 560-600 kHz range. To calculate the inertial cavitation index, the average of the entire frequency spectrum in the 0.1-7.1 MHz range was first calculated in dB (re1mV) and then normalized by the base level, over the same frequency range, measured when the amplifier is turned on with no input signal 59 . The resulting cavitation index is therefore given in dB. The dB scale reduces the possible contribution of harmonic signal to the cavitation spectrum integration. Moreover, the length of the acquired signal is large enough to provide harmonics that are narrow enough in the frequency domain to be neglected. However, iCI is only slightly sensitive to harmonic peaks due to the presence of bubbles in medium and reflects the broadband noise due to inertial cavitation for iCI values greater than 6 dB [59][60][61] .
To validate the non-inertial cavitation regulation system, cavitation activity was also assessed by chemical dosimetry with terephthalate acid (Sigma Aldrich, Saint Quentin Fallavier, France). The terephthalate dosimeter is based on the formation of fluorescent hydroxyl-terephthalate (HTA) anions when the non-fluorescent terephthalate (TA) anions enter in contact with hydroxyl radicals (OH). Because inertial cavitation produces OH radicals, the dosimetry consists in exposing a TA solution to ultrasound-induced inertial cavitation, and measuring the HTA fluorescence being generated. HTA production, and therefore fluorescence intensity of the sonicated solution, will depend on sonication time and ultrasound intensity [62][63][64] Test tubes containing 650 µL of TA at 20 mM were placed at the focus of the confocal device and sonicated for 30 seconds at different levels of iCI. After sonication, fluorescence of the samples was read using a spectrofluorometer (FluoroMax, Horiba Scientific, Kyoto, Japan), and HTA concentration in the samples was derived by normalization using the fluorescence level of a reference 2 µM HTA solution. As seen on Fig. 11, the cavitation threshold, above which a departure from baseline of HTA concentration can be attributable only to production of OH radicals, correspond to iCI typically above 18 dB. HTA concentration only depicts the sonochemical effect of inertial cavitation. However, for iCI below 18 dB, no rise of broadband noise (associated with inertial cavitation) or presence of higher subharmonics in the spectrum (associated with bubbles cloud oscillations) have been observed (Fig. 10).
Previous experiments have shown that inertial cavitation occurs for iCI typically higher than 18 dB (Fig. 11), and non-inertial cavitation for niCI typically higher than 3 dB, conservatives estimates to classify these acoustics emissions as significantly above noise levels.
In vitro enhanced anti-tumor effect. Cells line and culture conditions. The 4T1 adherent murine mammary carcinoma cells, with a rapid doubling time of 23 h 65 , were cultured as a monolayer in T75 culture flasks in supplemented DMEM (10%FCS, 1% L-glutamine, 1% Penicillin-Streptomycine) and incubated at 37 °C in an humidified atmosphere (5% CO2, 95% O2). Passages were done every 3 days to prevent cells to exceed 90% confluence as overgrowth decreases their viability. All reagents were purchased from Invitrogen Life Technologies (Cergy-Pontoise, France). 4T1 cells were tested mycoplasma-free using Mycoalert TM kit (Lonza, Cologne, Germany).
Chemicals. Doxorubicin hydrochloride 2 mg/mL (Accord Healthcare, Lille, France) was used at a concentration of 400 ng/mL. This dose, inducing 20% cell death at 72 h, was derived from estimation of the half maximal inhibitory concentration (IC50) from modelisation of the dose-response curve using MOSAIC macro 66 . For experiments involving hydroxyl radical scavengers, L-Histidine (Fluka, Sigma-Aldrich, Saint Louis MO, USA) at 20 mM and Mannitol (Fluka, Sigma-Aldrich, Saint Louis MO, USA) at 100 mM were used.
Combined treatment by doxorubicin and non-inertial cavitation. 72 hours prior to the experiments, cells were plated at a density of 2. 10 6 cells per T75 flask to reach 90% confluence on the days of the experiments. Cells were trypsinized with 0.25% trypsin/ 1 mM EDTA (Life Technologies, Cergy-Pontoise, France), centrifuged at 300 G during 5 min, washed once with 5 mL of PBS (Life Technologies, Cergy-Pontoise, France) to eliminate traces of Fetal Calf Serum (FCS) to avoid any association between DNA of dead cells and FCS under ultrasound exposure, and then resuspended in Opti-MEM Reduced-Serum Medium (Life Technologies, Cergy-Pontoise, France) at a concentration of 2 million cells per mL. The exact volume of Opti-MEM was determined after counting cells using an automated cell counter (Cellometer, Nexcelom Bioscience, Lawrence, MA).
The study design comprised four experimental conditions: (1) Non-Treated (NT) (2) DXR (DXR) (3) Ultrasound (US) (4) Combined ultrasound and DXR (US + DXR). For treatment, a volume of 650 µL of cells solution was pipetted in 2 mL Eppendorf tube using electronic pipettes. The tubes of DXR and US + DXR groups were supplemented with DXR at 400 ng/mL. Each tube of the US and US + DXR groups was positioned at the focus of the confocal system and sonication was applied using the parameters described earlier at room temperature. As differences in dissolved gas concentration in our culture medium (Opti-MEM) over the days may lead to differences in cavitation activity, we ensured the reproducibility of the non-inertial cavitation dose delivered. NT and DXR groups were manipulated the same way except for the ultrasound exposure. After treatment, a volume www.nature.com/scientificreports www.nature.com/scientificreports/ The tank is filled with Ablasonic (a). The hydrophone (c) is placed in the solution between the two confocal transducers (b) and is directed towards the exposed sample contained in a 2 mL Eppendorf tube (d). (B) Schematic of the non-inertial cavitation control strategy. The control system is based on the alternation of high power pulses (Pmax) and low power pulses (Pmin). The level and type of cavitation are characterized by several indicators on the spectrum of the signal recorded by the hydrophone: occurrence of sub-harmonics (half the emission frequency) indicates non-inertial cavitation and high level of broadband noise indicates inertial cavitation. If the niCI (index of non-inertial cavitation) falls below a userdefined threshold (empirically defined at 3 in our case), a high power pulse is triggered to revive the cavitation cloud. The number of boosting pulses is recorded. (2019) 9:15581 | https://doi.org/10.1038/s41598-019-51785-7 www.nature.com/scientificreports www.nature.com/scientificreports/ of 20 μL of cell suspension was collected for cell counting to assess immediate mortality. Then, for each sample, a volume containing 100 000 cells was collected. Those cells were plated in a 24-well plate, supplemented with 1 mL of culture medium, DXR at 400 ng/mL was added for the DXR and US + DXR groups and cells were incubated for 72 h at 37 °C.
To account for biological variability, this experiment was repeated six times, with 4 to 10 samples per group. Cell counting. Cell counting was performed 1 h and 72 h after sonication using a Cellometer (Nexcelom Bioscience, Lawrence, MA). Automated cell counting was always checked manually to prevent potential device inaccuracies that could be caused by cellular aggregates. Special attention was placed on samples treated with DXR or US + DXR at 72 h, as the cell concentration in these samples were on the lower end of the device's sensitivity. To ensure that the immediate mortality was inferior to the limit value of 30%, the dead cells percentage was measured immediately after sonication. After 72 h of culture, a cell growth index was calculated as the absolute number of cells divided by the initial number of cells reseeded after treatment (100 000 cells). An index less than 1 indicates inhibition of cell proliferation and/or cell death, while an index greater than 1 indicates continuous cell proliferation.
Mechanistic study. A set of experiments was designed to investigate the role of cavitation, DXR internalization (sonoporation), temperature increase, generation of free radicals, and changes in cell morphology, in the cell viability and proliferation changes after the combined treatment.

Influence of cavitation.
To investigate whether cavitation is required to induce the observed biological response, an experiment was designed to deliver the same acoustic energy, with or without cavitation. In the first setting, ultrasound treatment parameters were as previously described. In the second setting, the bursting pulses were omitted to prevent onset of cavitation and cells treated only with the regulation pulses. Because the bursting pulses are responsible for a very small fraction of the total delivered acoustic energy to the cells (see results section), it was considered that cells were treated with equivalent acoustic energy dose in both settings. The presence or absence of cavitation was monitored in both settings via quantification of the sub-harmonics component in the spectrum of the hydrophone-recorded signals. The experimental design was made of three conditions with 5 samples for each: (1)   To account for biological variability, this experiment was repeated five times, with 4 to 10 samples per group.
Influence of temperature increase. To measure temperature changes induced by the US treatment, a thermocouple was placed in the Eppendorf tube during sonication. To avoid perturbations in temperature measurements due to cavitation and to artifacts from direct sonication on the thermocouple, the thermocouple was placed a few mm away from the focal zone, assuming the temperature in the solution to be homogeneous due to the strong acoustic streaming. The temperature increased of 2 °C during the 30 s sonication, rising from 24 °C (room temperature (RT)) to 26 °C.
To study specifically the potential impact of this temperature increase on cell viability, tubes containing cells solution were placed for 30 s in water bath heated either at 29 °C, (that induced 2 °C temperature increase over the 30 s course), or at 37 °C to simulate physiological conditions (Fig. 12). Tubes were removed from the water bath after a time interval equivalent to US exposure time and allowed to cool at RT. Three experimental conditions were studied, with 4 samples per condition: (1) DXR at RT of 24 °C (DXR-RT), (2) DXR heated at 29 °C (DXR-29), (3) DXR heated at 37 °C (DXR-37). Cell viability was assessed 72 h post-heating as previously described (subsection Cell viability).
Effect of reactive oxygen species (ROS). The production of ROS was assessed with flow cytometry by using H2DCFDA (Fischer Scientific, Illkirch, France), a fluorescent marker that becomes fluorescent when linked to ROS 68 . Because the emission wavelength of H2DCFDA is closed to that of DXR (intensity measured in the 488 nm blue laser), compensation was performed prior to FACS analysis. The experimental design was made of four conditions of 6 samples each: (1) Non-treated (NT) (2) DXR (DXR) (3) US-treated (US) (4) Combination of US and DXR (US + DXR). Cell viability was assessed similarly as previously described. 1 mL of H2DCFDA diluted at 1:1000 (volume: 50 μL/sample) was added between adding the DAPI and Annexin markers and washing the cells with PBS. Cells were then incubated for 30 minutes at 37 °C, washed twice with 1 mL PBS, incubated with 1 mL of DMEM at 37 °C for 10 minutes (step of fixation of the reactive) and washed twice with 1 mL PBS.
In ROS experiments, cells were treated in the presence of a ROS scavenger, L-Histidine, to specifically mitigate the effects of hydroxyl 69 . Before treatment, scavenger L-Histidine at 20 mmol/L was added to the cell suspension. Cell microstructure. Modifications in cell microstructure were observed by transmission electron microscopy (Jeol 1400 JEM, Tokyo, Japan). Ultrasound cavitation and activation of apoptotic pathways are expected to induce membrane pore complexes, nuclear pore complexes and DNA damages as a result of compaction and marginalization of nuclear chromatin, convolution of cytoplasmic and nuclear membranes and condensation of cytoplasm 32