Lethal effects of mitochondria via microfluidics

Abstract Tumor cells can respond to therapeutic agents by morphologic alternations including formation of tunneling nanotubes. Using tomographic microscope, which can detect the internal structure of cells, we found that mitochondria within breast tumor cells migrate to an adjacent tumor cell through a tunneling nanotube. To investigate the relationship between mitochondria and tunneling nanotubes, mitochondria were passed through a microfluidic device that mimick tunneling nanotubes. Mitochondria, through the microfluidic device, released endonuclease G (Endo G) into adjacent tumor cells, which we referred to herein as unsealed mitochondria. Although unsealed mitochondria did not induce cell death by themselves, they induced apoptosis of tumor cells in response to caspase‐3. Importantly, Endo G‐depleted mitochondria were ineffective as lethal agents. Moreover, unsealed mitochondria had synergistic apoptotic effects with doxorubicin in further increasing tumor cell death. Thus, we show that the mitochondria of microfluidics can provide novel strategies toward tumor cell death.


| INTRODUCTION
Tumor cell can undergo various morphological changes in response to different therapeutic agents. For example, mammalian target of rapamycin (mTOR) can convert tumor cells into fat-like cells. [1][2][3][4] Various morphological alterations in lamellipodia, filopodia, or tunneling nanotubes have been shown to occur in these cells. 5,6 However, the functional significance of tunneling nanotubes during tumor cell reprogramming has not been fully elucidated.
Tunneling nanotubes are thin membrane elongations that are formed between cells that can mediate trafficking of subcellular vesicles, proteins, or organelles. 7 They consist of filamentous-actin and are about 0.05-1 μm in width and 100 μm in length. 7 In addition, they have been implicated as one mechanism by which mitochondria can be transferred from one cell to another. 8 The speed of mitochondrial migration through tunneling nanotubes is about 80-1400 nm/s. 9 Several studies have been conducted on mitochondrial migration between cells. [10][11][12][13] An in vitro study showed that oxidative stressinduced apoptosis of endothelial cells or H9c2 cardiomyocytes was abolished by transferring of mitochondria from mesenchymal stem cells. 14,15 An in vivo study showed that mitochondrial transfer inhibited hypoxia-induced apoptosis in cardiomyocytes. 16 Moreover, mitochondria can be transferred to support the survival of metabolically compromised cells. 17,18 Therefore, mitochondrial transfer through tunneling nanotubes can provide important insight in our understanding of tumor cell fate.
We describe here the application of optical tomography whereby an imaging technique is used to obtain cross sections of cells remotely. 19 In optical tomography, projected images are obtained by waves passing through the cells at various angles, and digital images are subsequently reconstructed to obtain the internal structure of cells in a cross-sectional manner. 19,20 Since the tomographic microscope can render reconstructed cell images according to varying refractive index, transparent objects such as lipid droplets and mitochondria can be detected without staining. 21,22 Here, we examined breast tumor cells and their response to antitumor agents, rho-associated protein kinase (ROCK) and mTOR inhibitors by tomographic microscopy. Intriguingly, we found that mitochondria within tumor cells transferred into other nearby tumor cells via tunneling nanotubes and induced apoptosis.
Therefore, we hypothesized that mitochondria of tunneling nanotubes can be an effective novel strategy to achieve tumor cell death.
To this end, we investigated the mechanisms by which mitochondria can instigate tumor apoptosis. Specifically, we examined the physical processing of mitochondria that occurs during transfer through tunneling nanotubes and recapitulation of this process.

| Animal treatment
All mice were in C57BL/6 background and purchased from Charles River. All experiments using animals were performed and approved by the NIH guidelines (Guide for the care and use of laboratory animals) and the Ewha Womans University Animal Care Committee respectively. Forty female mice were each given 6 weekly 1 mg doses of 7,12-dimethylbenzathracene (DMBA) in 0.2 ml of sesame oil by oral gavage and implanted 30 mg of subcutaneous pellets of compressed medroxyprogesterone acetate (MPA), beginning at 5 weeks of age. Doxorubicin was administered weekly by intraperitoneal injection (5 mg/kg body weight, 300 μl) for 4 weeks. Mice were then maintained for another week.

| Quantitation of fat-like cells
We defined fat-like cells when the sum of the area of the lipid droplets accounted for about 30% of the total cell area. At the start of the experiment, the number of MDA-MB-453 was approximately 10 4 in a 2 cm-diameter dish. Following treatment with ROCK-mTOR inhibitors, approximately 3 Â 10 3 (30%) of cells were dead, 2.5 Â 10 3 (25%) were nondifferentiated, and 4.5 Â 10 3 (45%) of cells were differentiated into fat-like cells. For each experiment, three triplicate plates were used to obtain an average value, and the experiments were repeated six different times.

| Experimental procedures using microfluidic device
Microfluidic device with a diameter of 300 μm was purchased from Relative mRNA levels were quantified using the comparative 2 ÀΔΔCT method.  Blots were stripped and re-exposed to detect β-actin (1:1000;

| ELISA for detection of apoptosis
sc-47778, Santacruz Biotechnology, USA) as housekeeping protein.

| Transmission electron microscopy
We fixed the mold of CC-microfluidic with 3% buffered glutaraldehyde (G5882; Sigma-Aldrich) for 2 h and processed into resin (02334;

| Echocardiographic assessment
For echocardiography, Vevo 2100 was used at Cardiovascular Research Center in Seoul, Korea. Mice were anesthetized with 2% isoflurane and maintained with 1.5% isoflurane followed by application of depilatory cream to the chest and wiped clean to remove all hair in the area of interest. The scanning probe (20 MHz) was used to obtain 2D images of the parasternal long axis. These 2D images were converted to M-mode.

| Statistical analysis
Values were means ± SE. The significance of differences was determined by a two-way analysis of variance (ANOVA), or a one way ANOVA followed by a Bonferroni post hoc analysis where appropriate. Tukey's hsd (honestly significant difference) test was used to confirm synergistic effect, supported by department of statistics at Ewha Womans University. Differences were considered significant when p < 0.05. were not formed and mitochondria were not able to migrate even to cells in close proximity ( Figure S1A). In addition, L-778123 decreased apoptosis and Endo G levels in response to rapamycin and Y27632

| Reprogramming of breast tumor cells into fat-like cells
( Figure S1B,C). However, L-778123 did not affect caspase-3 activity following rapamycin and Y27632 treatment ( Figure S1D). Therefore, mitochondrial tunneling preceded apoptosis in this setting.

| Physical mimicking of tunneling nanotubes in breast tumor cells
In order to mimic the nanotubes through which mitochondria are transferred, we designed a microfluidic system to recapitulate these tunnels. Microfluidic devices were sealed with polydimethylsiloxane

| Endo G is essential for apoptosis through cellular transport of mitochondria
We found that mitochondrial migration through tunneling nanotubes increased the amount of Endo G in the mitochondria-recipient cells  (Figure 5g,h). In caspase-3-depleted cells, rapamycin and Y27632 did not induce in apoptosis nor alterations of Endo G levels ( Figure S2C-F). Therefore, both caspase-3 and Endo G are required for apoptosis in this setting.
Mitotracker itself did not affect cytochrome C, caspase-3, Endo G, or apoptosis ( Figure S3A-C). The stained and internalized mitochondria were maintained until 4 days, and the loss of mitochondria began to appear after 6 days ( Figure S3D). Overall, mitochondrial transport promoted recipient cell apoptosis with release of Endo G through tunneling nanotubes and moderately increased caspase-3 activity.
Mitochondria released Endo G through a narrow space, which we referred to herein as unsealed mitochondria are depicted in a diagram ( Figure 5i). doxorubicin but did not appear to be dependent on unsealed mitochondria. On the other hand, the release of Endo G was increased by unsealed mitochondria and this did not appear to be affected by doxorubicin. Importantly, unsealed mitochondria induced significant apoptosis combined with low concentration of doxorubicin (0.01 μM), a dose that does not by itself induce apoptosis. Overall, treatment of unsealed mitochondria has a synergistic effect on apoptosis with doxorubicin. Doxorubicin is a widely used chemotherapeutic agent, but it is known to cause cardiotoxicity. Therefore, we investigated the role of unsealed mitochondria on cardiac cells. Human neonatal cardiomyocytes were treated with unsealed mitochondria and doxorubicin;

| Unsealed mitochondria have a synergistic effect with doxorubicin
and unsealed mitochondria did not affect caspase-3 activity in these cells (Figure 6e). Similar to cancer cells, the increased Endo G was dependent on unsealed mitochondria in neonatal cardiomyocyte.
However, unlike tumor cells, unsealed mitochondria or unsealed mitochondria with 0.01 μM of doxorubicin did not induce apoptosis in cardiomyocytes. Consistent with these findings, Endo G has been shown to promote cellular survival rather than apoptosis in cardiomyocytes. 23 In summary, unsealed mitochondria can potentiate lethal effects of doxorubicin in requiring lesser amounts to effectively kill tumor cells in vitro.

| DISCUSSION
During the reprogramming of breast tumor cells into fat-like cells, formation of tunneling nanotubes was detected. 3,5,6 In general, intercellular migration of mitochondria through tunneling nanotubes has been known to impede apoptosis and optimize metabolism. 9,12,15 However, detection of tunneling nanotubes and mitochondria with accuracy using an optical microscope has had limitations due to suboptimal resolution. Although these components can be visualized with a fluorescence microscope after staining, the difficulties around processing using chemicals and fixatives in addition to the limitations of twodimensionality can all contribute to the overall limitations of accurate assessment of tunneling nanotubes and mitochondria. Since tomographic microscopy can provide the internal structure of cell images from multiple angles and render digitally reconstructed images according to the difference in refractive index, the tunneling nanotubes and mitochondria can be distinguished and visualized without the confounding elements that are associated with the processes of special staining. 19 Given that doxorubicin is a chemotherapeutic agent known for cardiotoxic effects, the unsealed mitochondria may reduce the off target effects by enabling the use of doxorubicin in low concentration.

CONFLICT OF INTEREST
The author declares that there is no conflict of interest.

DATA AVAILABILITY STATEMENT
All data supporting the findings of this study are available within the paper.