Cold Atmospheric Plasma Stimulates Clathrin-Dependent Endocytosis to Repair Oxidised Membrane and Enhance Uptake of Nanomaterial in Glioblastoma Multiforme Cells

Cold atmospheric plasma (CAP) enhances uptake and accumulation of nanoparticles and promotes synergistic cytotoxicity against cancer cells. However, the mechanisms are not well understood. In this study, we investigate the enhanced uptake of theranostic nanomaterials by CAP. Numerical modelling of the uptake of gold nanoparticle into U373MG Glioblastoma multiforme (GBM) cells predicts that CAP may introduce a new uptake route. We demonstrate that cell membrane repair pathways play the main role in this stimulated new uptake route, following non-toxic doses of dielectric barrier discharge CAP. CAP treatment induces cellular membrane damage, mainly via lipid peroxidation as a result of reactive oxygen species (ROS) generation. Membranes rich in peroxidised lipids are then trafficked into cells via membrane repairing endocytosis. We confirm that the enhanced uptake of nanomaterials is clathrin-dependent using chemical inhibitors and silencing of gene expression. Therefore, CAP-stimulated membrane repair increases endocytosis and accelerates the uptake of gold nanoparticles into U373MG cells after CAP treatment. We demonstrate the utility of CAP to model membrane oxidative damage in cells and characterise a previously unreported mechanism of membrane repair to trigger nanomaterial uptake. This knowledge will underpin the development of new delivery strategies for theranostic nanoparticles into cancer cells.


Results
Numerical modelling of the uptake of AuNPs by GBM cells. The accumulation of AuNPs inside U373MG cells was monitored using atomic absorption spectroscopy and the dose response curve of AuNPs with or without CAP treatment has been presented in previous study 26 . The experimental results, reproduced in Fig. 1 (open circles/dashed lines) were further analysed according to a simulated uptake model to better understand the possible mechanism of CAP-stimulated AuNP uptake. Uptake of nanoparticles by cell populations in vitro has previously been modelled according to a phenomenological rate equation approach [28][29][30] , which was extended here to further investigate the role of CAP in AuNP uptake.
The rate of uptake of AuNPs into a cell can be described by the equation: Figure 1. Modelling uptake of AuNPs. Numerical modelling of experimental data from our previous uptake study 26 (shown with open circles and dashed lines) was carried out for simulated AuNP uptake (green solid line), AuNP uptake quenched by incubation of the cells with NaN3 (blue solid line), AuNP uptake on application of low dose CAP (red solid line). The simulated uptake of AuNP has been normalised to the maximum calculated value. where N AuNP is the number of internalised AuNPs, D is the initial dose of AuNPs, (D-N AuNP ) allows for the depletion of the applied AuNP dose, and k doub is the doubling time of the cells. N 1 /R 1 , N 2 /R 2 and N 3 /R 3 allow for three different principle uptake pathways, with respective limiting capacities of N and rates R. The first two terms describe independent active and passive uptake mechanisms, respectively, with limiting cellular capacities N 1 (0) = N 1max and N 2 (0) = N 2max , such that: 2 1 AuNP 3 Figure 1 (solid green line) shows the simulated uptake of AuNPs, normalised to the maximum uptake observed for AuNP + CAP, for the case of R 1 = 3 ×10 −3 hr −1 , R 2 = 2.5 ×10 −5 hr −1 , R 3 = 0, which faithfully reproduces the experimentally observed behaviour. Quenching of the active uptake of AuNPs by NaN 3 is best simulated by addition of a further term in Eq. 2, such that where NaN 3 is the effective dose of sodium azide, and R 4 allows for the rapid depletion of the active uptake pathway. The experimentally observed uptake was well simulated (solid blue line) by a value of R 4 = 3 ×10 −5 hr −1 , keeping all other rates as before.
In simulating the increased uptake of AuNPs upon CAP treatment, it was noted that the enhancement of a single pathway described by Eqs. (2)(3)(4) by CAP treatment, by increasing a single uptake rate, did not faithfully reproduce the experimentally observed behaviour, as the uptakes were limited by the parameters N 1max and N 2max . Rather, faithful reproduction of the observed behaviour required the introduction of independent uptake mechanisms for untreated and CAP treated AuNP uptake, an observation which was critical to the interpretation of the effects of CAP treatment on the cells. Such a pathway can be represented by: 3 3 AuNP 3 such that N 3 (0) = N 3max . Upon the application of CAP, the enhanced uptake was well fitted (solid red line) by R 3 = 2.5 ×10 −4 hr −1 , keeping all other rates as before.
The modelling process indicates that CAP treatment increases the capacity of the U373MG cells to uptake AuNPs by introducing a third uptake route distinct from passive uptake and normal active uptake processes. The modelling parameters employed are detailed in Table 1. Note that the parameters relating to the limiting cell uptake, N nmax , were determined by the definition of the dose as 100 mg/mL. Furthermore, the process was one of simulation, rather than a mathematical fitting, so the parameters should be considered within ~10% confidence.
Reactive species generated by CAP treatment. To identify the possible uptake pathways implicated by the numerical modelling of our data, RONS generation of 30 s, 75 kV CAP treatment was first investigated. Several RONS were measured using optical emission spectroscopy (OES) and Gastec gas detector tubes, including N 2 , N 2 + , • OH, and O 3 . OES emission intensities from the N 2 second positive system (SPS) were measured at 315, 337, 357, and 377 nm, the N 2 + first negative system (FNS) at 391 nm, and • OH at 310 nm were measured. The data demonstrated a relatively constant RONS production throughout the 30 s treatment (Fig. 2a).
The electron energy distribution function (EEDF) of the plasma was also determined. As seen in Supplementary Figure S1, EEDF remained close to a ratio of 7 during CAP treatment, which indicated that the electron energies were distributed more so on the lower end of the energy scale (11-12 eV) than the higher energy levels (18.8 eV). The low variability of the EEDF indicated that the electric field was stable, and that the formation of the reactive species was in a steady state manner. Using Gastec ozone detector tubes, the concentrations of generated O 3 in the extracted gas were measured post-discharge of CAP treatment (Fig. 2b) Table S1).
The level of the lipid peroxidation indicator malondialdehyde (MDA), measured using the Thiobarbituric acid reactive substances (TBARS) assay, was significantly higher in U373MG cells 24 h after CAP treatment compared with the control group (Fig. 2e), indicating a high level of lipid peroxidation is induced by CAP treatment. In our previous study, the IC 50 of CAP treatment (75 kV) was determined to be 74.26 s (95% confidence range of 47.24-116.8 s) for U373MG cells 36 and we demonstrated that 60 s CAP treatment induces rapid permeabilisation of U373MG cell membranes 37 . Here, we confirm that no significant increase of PI uptake is evident following 30 s CAP treatment compared to the control (Fig. 2f), indicating that membrane integrity was not significantly affected by 30 s CAP treatment and CAP-induced lipid peroxidation. This is in agreement with our previous findings, where 30 s CAP treatment induces very low, non-significant levels of toxicity in U373MG cells 48 hours post treatment (18.52%, SEM = 5.41%) 26 .
CAP induced lipid peroxidation was further analysed and visualised using the lipid peroxidation fluorescent sensor C11-BODIPY (581/591). C11-BODIPY (581/591) is a lipophilic fluorescent dye that remains membrane bound, can react with various RONS and oxidation leads to a shift of the emission peak from around 590 nm (Orange) to around 510 nm (Green) 38 . To determine the level of lipid peroxidation, more than 60 cells were analysed using ImageJ software for each group. A significantly stronger emission of the oxidised green fluorescence was observed in CAP-treated cells compared to the untreated cells (****p < 0.0001, Fig. 3a, 2 nd panel and Fig. 3b), which was accompanied by a significant decrease in the non-oxidised orange fluorescence (

Effects of CAP treatment on AuNPs endocytosis.
A membrane repair mechanism has been described where cells can quickly remove damaged regions of membranes from the cell surface through rapid endocytosis 39,40 . These impaired membranes can be trafficked into endosomes and finally into lysosomes. We wished www.nature.com/scientificreports www.nature.com/scientificreports/ to determine whether rapid endocytosis may contribute to the increased uptake of AuNPs and other materials into cells following CAP treatment. To test this hypothesis, CellLight Early Endosomes-RFP, BacMam 2.0 and CellLight Late Endosomes-RFP, BacMam 2.0 were used to further visualise the route of uptake of AuNPs after CAP treatment. Rab5a and Rab7a chimeras tagged with RFP were transfected and expressed inside cells. Following overnight incubation, Rab5a-RFP and Rab7a-RFP specifically tracked early endosomes and late endosomes, respectively 41 . Cells were then incubated with AuNPs for 3 h after CAP treatment. In Supplementary Figure S4, the white arrows identify examples of co-localisation of AuNPs (red) with early (left) and late (right) endosomes (orange). Lysosomes (green) were also counterstained. We have previously demonstrated that AuNPs accumulate in lysosomes 24 hours after 30 s, 75 kV CAP treatment using confocal imaging and 3D-image construction 26 . We demonstrate here that that AuNPs enter CAP-treated U373MG cells mainly through endocytosis and colocalise with Rab5a and Rab7a chimeras corresponding to early and late endosomes respectively, eventually accumulating in lysosomes (Supplementary Figure S4).
Transferrin conjugated with Alexa Fluor 546, is used as an early endosome marker. To investigate the immediate uptake route, CAP treated U373MG cells were incubated with Alexa546-Transferrin for 5 min, then fixed with 4% PFA. As seen from confocal imaging, within 5 mins after CAP treatment, the number of transferrin-containing endosomes was greater compared to the control group (Fig. 4a). Quantification of transferrin uptake confirmed a significant increase of endosomes 5 min after CAP treatment (Fig. 4b, p < 0.0001).
Endocytosis is typically subdivided into four types, including clathrin-mediated endocytosis (CME), caveolae-mediated endocytosis, macropinocytosis and phagocytosis. We used a panel of inhibitors 42 to delineate the specific endocytic pathway activated by CAP (see Table 2). Clathrin-inhibitors and MβCD-induced cholesterol depletion decreased the number of transferrin labelled endosomes and AuNP-uptake 3 h following CAP   (Fig. 4c,d) whereas caveolae-specific inhibitors and an inhibitor of micropinocytosis did not lead to any significant inhibition of Transferrin or AuNP endocytosis in cells 3 h after CAP treatment (Fig. 4e,f). Original data for quantification can be viewed in Supplementary Table S4.
To further confirm that clathrin-mediated endocytosis played the main role in CAP-accelerated cellular uptake, MISSION Endoribonuclease-prepared siRNA (esiRNA) against human Clathrin heavy chain 1 (CLTC) was used to disrupting endocytosis mediated by clathrin coated pit formation. AuNPs uptake was visualised in clathrin-silenced cells and controls treated with or without CAP using confocal microscopy ( Fig. 5a-d). AuNP uptake was significantly lower in clathrin-silenced cells and no increase in AuNP uptake was observed 3 h after CAP treatment (p < 0.0001, Fig. 5e and see Supplementary Table S5 for original data). Together, our data confirms that clathrin-mediated endocytosis played an important role in AuNP uptake, and accelerated endocytosis following CAP treatment was clathrin dependent.

Discussion
The cytoplasmic membrane separates and protects the cellular interior from the exterior environment and provides specific and efficient exchange channels for the remaining intercellular balance and cell viability. Therefore, the integrity of the membrane is vital for all cells. Mammalian cells have developed efficient membrane repair mechanisms that can recover and reseal an injured cytoplasm membrane quickly to retain cell viability. Although  Supplementary  Table S3 for original data. (c-f) After incubation with various inhibitors as indicated, U373MG cells were treated with CAP for 0, 30 s at 75 kV and then loaded with transferrin for 5 min or 100 μg/ml AuNPs for 3 h respectively before observing using confocal microscopy, with the fluorescence integrated densities quantified using ImageJ. The statistical significance in (b-f) was assessed by one-way ANOVA with Tukey's multiple comparison post-test (*P < 0.05, **P < 0.01, ***P < 0.001, ****p < 0.0001), n ≥ 50.

Scientific RepoRtS |
(2020) 10:6985 | https://doi.org/10.1038/s41598-020-63732-y www.nature.com/scientificreports www.nature.com/scientificreports/ investigations of the precise membrane repairing mechanisms have been limited, four possible mechanisms, including patch, tension reduction and exocytosis/endocytosis and budding repair mechanisms have been proposed 40 . The study of cytoplasmic membrane repair usually employs bacterial pore-forming toxins, such as Streptolysin O, to create mechanical injuries on membranes 40 . Meanwhile, lipid peroxidation is a complex process that damages cellular membrane structure and function, which is believed to link to numerous human diseases and aging, including Alzheimer diseases, dementia, Huntington, Parkinson, and traumatic injuries, under presence of oxidative stress 43,44 . Many studies have shown that lipid peroxidation have various significant effects to cellular membranes, such as increased membrane permeability [45][46][47] , alteration of the lipid order and membrane fluidity 46,48-50 and activity change of membrane proteins [50][51][52][53] . However, there remains a paucity of literature identifying mechanisms of lipid peroxidation-related membrane damage repair, and the mechanisms of oxidised membrane repair remain unknown.
CAP is known to generate reactive species and thereby cause lipid peroxidation of cells. In this research, we explored CAP-induced lipid peroxidation using a low, relatively non-toxic dose of CAP and studied the possible mechanisms of accelerated cellular uptake of AuNPs following CAP-induced oxidative membrane damage.
As seen in Fig. 1, the uptake and accumulation of AuNPs into U373MG cells was first modelled. To faithfully reproduce the experimentally observed results in our mathematical model, a new independent uptake rate was necessary in the model (Eq. 5). This numerical model indicates that CAP treatment may introduce a new uptake route. Following a detailed experimental analysis, we can confirm this occurs and is due to a CAP-triggered, membrane repairing, clathrin-dependent, pathway of endocytosis.
As seen in Fig. 2, we observed a relatively high level of reactive species generated in air during CAP treatment such as • OH, ONOO − and O 3 . This was also evident in media, cells, and ultimately in lipid membranes with the detection of RONS including hydrogen peroxide, nitrite and nitrate and detection of peroxidated lipids and  Table 2. Inhibitors used to inhibit endocytosis in this research 42 . The fluorescence level was quantified using ImageJ and presented as integrated density. The statistical significance was assessed by one-way ANOVA with Tukey's multiple comparison post-test (*P < 0.05, **P < 0.01, ***P < 0.001, ****p < 0.0001), n ≥ 50.
CAP treatment has been shown to alter membrane structures, which may be partly due to the reactive species-caused lipid peroxidation 24,37,54,55 . However, in our study, cell membranes remains PI impermeable and therefore intact after exposing to 75 kV CAP for 30 s. Interestingly, not only did we observe a significant increase in oxidised lipids using C11-BODIPY following CAP treatment, but there was also a clear and significant increase in vesicle-like structures with high levels of oxidised C11-BODIPY, suggesting that peroxidised membrane lipids were trafficked inside cells via endocytosis (Fig. 3a). An immediate (within 5 minutes) increase in endosomes was also confirmed using transferrin conjugated with Alexa Fluor 546.
Therefore, we propose that low, sub-toxic doses of CAP can cause cytoplasmic membrane oxidation which triggers a rapid membrane repair system. The increased endocytosis induced by membrane repair is initiated within 5 minutes and accelerates the uptake of AuNPs into U373MG cells.
To further explore the route of uptake, AuNPs were tracked in cells counterstained with Rab5a-RFP, Rab7a-RFP, and LysoTracker Green DND-26. Co-localisation of AuNPs with early and late endosomes and lysosomes were observed. AuNPs uptake displayed a tendency to enter the periphery region of the cells through endocytosis, then gather at the central zone of the cells via endosome trafficking into lysosomes. This demonstrates that CAP treatment of U373MG cells triggers uptake and trafficking of AuNPs through early endosomes into late endosomes and ultimately accumulation lysosomes.
Our inhibitor studies and RNA interference studies demonstrate a primary role for clathrin-dependent endocytosis of AuNPs, and not other routes of endocytosis in CAP-triggered uptake. We compared uptake of AuNPs with Transferrin, an essential iron-binding protein that facilitate iron-uptake in cells via transferrin receptors. Uptake of Transferrin is well characterised and involves Clathrin-mediated, receptor dependent endocytosis 56 . Interestingly, Shi et al. found that CAP treatment decreased transferrin receptor-1 but could induce iron-dependent oxidative stress to stimulate significant increase in fluid-phase endocytosis, lysosome biogenesis and autophagy in mesothelioma cells 27 . In contrast to this, as seen in Fig. 4a,b, our data demonstrates that the uptake of transferrin was significantly increased just 5 min after CAP treatment, which may due to the diversity of cell lines or CAP generation devices. Similar to AuNPs, formation of transferrin-trafficking endosomes was inhibited in both groups with 0 and 30 s CAP treatment, after exposure to specific/non-specific clathrin inhibitors, including pitstop, CPZ and MβCD. Tracking the accumulation of AuNPs demonstrates the long-term CAP-triggered uptake via a single, clathrin-dependent endocytosis route are retained for at least 24 hours post CAP treatment. Jinno M et al. demonstrated in a CAP-induced gene transfection model using mouse fibroblasts that a clathrin specific inhibitor showed 60% decrease in the gene transfection efficiency and that both clathrin-dependent endocytosis and electroporation played a role in enhanced transfection efficiency 57 . The work of Jinno M et al. supports our investigation that clathrin may play key roles in CAP-induced cellular uptake mechanisms in response to different materials.
Our data indicate that oxidised lipids can be seen to colocalise with endosomes and lysosomes demonstrating the intracellular sites where the repair of peroxidised lipid membranes is taking place. Further investigation of these mechanisms and sites of membrane repair following CAP-stimulated oxidative damage is ongoing, to identify novel therapeutic targets that can augment CAP cytotoxicity and to fully understand the cellular mechanisms governing oxidative membrane repair.
In summary, we report that the enhanced uptake of AuNPs induced by CAP can be as a result of ROS-caused lipid peroxidation, leading to rapid plasma membrane repair via clathrin-dependent endocytosis. This contributes to our understanding of the cellular effects induced by CAP, especially membrane damage and endocytosis activation, which can be employed for efficient uptake of nanomaterials and pharmaceuticals into cells when combining CAP with cancer therapies. This mechanism of RONS-induced endocytosis will also be of relevance to researchers optimizing other cancer therapies that induce an increase in extracellular RONS.

Methods
Cell culture and gold nanoparticle treatment. U373MG-CD14, human brain glioblastoma cancer cells (Obtained from Dr Michael Carty, Trinity College Dublin) were cultured in DMEM-high glucose medium (Merck, Arklow, Ireland) supplemented with 10% FBS (Merck) and maintained in a 37 °C incubator within a humidified 5% (v/v) CO 2 atmosphere. Gold nanoparticles were synthesised by trisodium citrate reduction of auric acid. 20 nm sphere citrate-capped AuNPs were used to treat cells whose properties were determined in a previous study 26 . The gold colloid was concentrated to 2500 μg/ml then diluted in culture medium to 100 μg/ml. CAP configuration and treatment. The current research uses an experimental atmospheric dielectric barrier discharge (DBD) plasma reactor, which has been described and characterised in detail 36,58 . Unless otherwise stated, all U373MG cells were treated within containers, which were placed in between two electrodes, at a voltage level of 75 kV for 30 s. The culture medium was removed prior to CAP treatment then replaced with fresh culture medium afterwards.
H 2 DCFDA assay and optical emission spectroscopy (OES) and ozone measurement. H 2 DCFDA (Thermo Fisher Scientific, Ballycoolin, Ireland) was used to detect ROS induced by CAP treatment. U373MG cells were seeded into the TC dish 35 standard (Sarstedt, Belfast, UK) at a density of 2×10 5 cells/ml and incubated overnight to allow adherence. After washing twice with PBS, cells were incubated with 25 μM H 2 DCFDA in serum-free medium for 30 min at 37 °C. Cells were then washed with PBS twice, culture medium once and treated with CAP at 75 kV for 30 s. The fluorescence of H 2 DCFDA was measured using flow cytometry 30 minutes later.
Optical emission spectroscopy was carried out using an Edmund Optics CCD spectrometer with a spectral resolution of between 0.6 nm to 1.8 nm. The spectra was measured using BWSpec TM software with a spectral range Scientific RepoRtS | (2020) 10:6985 | https://doi.org/10.1038/s41598-020-63732-y www.nature.com/scientificreports www.nature.com/scientificreports/ between 200 and 850 nm and was acquired every 7.5 s with an integration time of 1500 ms. Total relative intensity of each emission line was calculated using the integral of the area under each peak. EEDF was calculated using a line ratio method (N 2 at 337 nm and N 2 + at 391 nm) 59 . O 3 was sampled using a standard Gastec sampling pump in conjunction with a Gastec detection tubes immediately after plasma discharge had ceased.

Measurement of hydrogen peroxide, nitrite and nitrate concentrations. The concentrations of
hydrogen peroxide, nitrite and nitrate were quantitatively measured in CAP-treated culture medium without phenol red. Concentrations of hydrogen peroxide, nitrite and nitrate were determined employing the TiSO 4 assay, Griess reagent and 2,6-dimethylphenol assay, respectively. The quantitative methods have been described in detail elsewhere 33 . To eliminate the effect of the culture medium on photometrical measurements, the results of CAP-treated groups were standardised with untreated culture medium.
Lipid peroxidation. For TBARS assay, thiobarbituric acid (TBA), trichloroacetic acid, MDA were purchased from Merck. U373MG cells were seeded into TC Dish 150 (Sarstedt) at a density of 1×10 5 cells/ml and incubated until confluence. After CAP treatment, cells were further incubated for 24 h, and collected by trypsinisation, centrifuged (100 g for 5 min), and homogenized by sonication. 100 μl homogenate was mixed with 200 μl ice cold 10% trichloroacetic acid and incubated on ice for 15 min to precipitate protein, then centrifuged (2200 g for 15 min at 4 °C). 200 μl of each supernatant was then mixed with 200 μl 0.67% (w/v) TBA and incubate at 100 °C for 10 min. After cooling, samples were measured at 532 nm for MDA.
Lipid peroxidation sensor, C11-BODIPY (581/591) (Thermo Fisher Scientific) was used for in-situ detection and localization of the lipid peroxidation induced by CAP treatment. Cells were incubated in fresh culture medium containing 5 μM of the probe at 37 °C for 30 min in advance. Then the cells were washed with PBS twice and culture medium once. After CAP treatment, cells were further incubated with fresh medium for 30 min at 37 °C, and observed using flow cytometry and confocal microscope as described later.
Flow cytometry. BD Accuri C6 Plus flow cytometry (BD Bioscience, Allschwil, Switzerland) was used in this study. Cells were loaded with C11-BODIPY and treated with CAP as described above (Lipid peroxidation). To prepare aliquots, all floating and attaching cells were collected by trypsinisation and then washed twice with PBS. For the measurement, a 488 nm laser was used for excitation, and 10,000 gated events were collected. Green fluorescence (oxidised dye) and red fluorescence (non-oxidised dye) was measured using an FL1 standard filter (533/30 nm) and FL2 standard filter (585/40 nm), respectively.
For Propidium iodide (PI) staining, cells were exposed to CAP 75 kV for 30 s and incubated at 37 °C for 30 minutes afterwards, then collected by trypsinisation, resuspended into 1 ml PBS. Resuspended cells were stained with 1 µg/ml PI for 5 minutes. The fluorescence of PI was then measured at FL2 (585/40 nm) standard filter.
Transferrin conjugated with Alexa Fluor 546 was used to determine the change of early endosomes induced by CAP combining various endocytosis inhibitors. After the inhibiting and CAP treatments indicated above, the cells were incubated in prewarmed fresh medium for 0 or 3 h, then incubated with 25 μg/ml transferrin in medium for 5 min. Afterwards, cells were fixed with 4% PFA and then observed using confocal microscopy. The details of the confocal microscope are described in following section.
Clathrin silencing. MISSION esiRNA (human CLTC) and MISSION siRNA transfection reagents were purchase from Merck. 50,000 U373MG cells were seeded into each 35 mm glass-bottom dishes (Greiner Bio-One, Lörrach, Germany) and incubated overnight. 1.2 ul esiRNA stock was mixed with 20 ul of transfection reagent in 400 μl serum-free medium and incubated for 15 minutes at room temperature. In glass-bottom dishes, previous medium was replaced with 1 ml of prewarmed fresh medium with serum. The siRNA/transfection reagent solution was then added onto the cells and homogenized to final volume of 1.2 ml. Afterwards, U373MG cells were incubated at 37 °C for 24 h and then incubated with fresh medium containing 100 μg/ml AuNPs for 3 h before observing using confocal microscopy.