Selenium nanoparticles trigger alterations in ovarian cancer cell biomechanics

High dose selenium acts as a cytotoxic agent, with potential applications in cancer treatment. However, clinical trials have failed to show any chemotherapeutic value of selenium at safe and tolerated doses (b90 μg/day). To enable the successful exploitation of selenium for cancer treatment, we evaluated inorganic selenium nanoparticles (SeNP), and found them effective in inhibiting ovarian cancer cell growth. In both SKOV-3 and OVCAR-3 ovarian cancer cell types SeNP treatment resulted in significant cytotoxicity. The two cell types displayed contrasting nanomechanical responses to SeNPs, with decreased surface roughness and membrane stiffness, characteristics of OVCAR-3 cell death. In SKOV-3, cell membrane surface roughness and stiffness increased, both properties associated with decreased metastatic potential. The beneficial effects of SeNPs on ovarian cancer cell death appear cell type dependent, and due to their low in vivo toxicity offer an exciting opportunity for future cancer treatment. © 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons. org/licenses/by-nc-nd/4.0/).

antioxidant supplements has been correlated with a decreased risk of cancer development 13 , possible due to the form of selenite used in supplements forming endogenous SeNPs that inhibit glycolysis, causing mitochondrial dysfunction, autophagy and cytoskeletal depolymerisation. 14 Tumor metastasis in advanced ovarian disease is the leading cause of death and is an inherently mechanical process where these properties are known to be altered. 16 , 17 The acquisition of invasiveness by tumor initiating cells is accompanied by the loss of the epithelial features and the gain of a mesenchymal phenotype, termed epithelial to mesenchymal transition (EMT). 15 Chemotherapeutic drugs have been shown to modify such cellular biomechanical features, through architectural changes to the cell cytoskeleton. 18 Similarly cellular cytoskeletal components such as actin microfilaments, intermediate filaments, and microtubule polymer networks play determinant roles in cellular mechanical properties, locomotion, while regulating cellular integrity during differentiation. 18 Modifications to those networks influence cytoadherence, migration, invasion and tumor metastasis. 19 , 20 Here, we assessed the anticancer activity of protein (BSA) and carbohydrate (chitosan) surface coated SeNPs in two distinct high grade serous ovarian cancer cell lines, OVCAR-3 and SKOV-3. Both SeNPs were significantly more cytotoxic than soluble Se in OVCAR-3 cells at high doses (N40μg/mL), but similar to soluble Se with SKOV-3 cells, which were more sensitive to Se treatment than OVCAR-3 cells, highlighting the differences between cell types. Further analysis revealed SKOV-3 cells exhibited stable EMT markers and decreased motility, and interestingly, an increase cell surface roughness and cellular stiffness. In contrast OVCAR-3 cells displayed a decrease in cellular stiffness indicative of altered cytoskeletal dynamics that, alongside decreased vimentin expression levels and autophagy, can be interpreted as sensitization toward apoptosis. 5 , 6 , 21 It appears that reduction in cell viability following SeNP exposure occurs through different mechanisms that result in contrasting perturbations in cellular mechanics in serous ovarian cancer subtypes. SeNPs may therefore offer the potential for pan-cancer treatments, not least in ovarian cancer, that is a complex and multifaceted disease with a very poor prognostic outcome.

SeNP characterization
SeNPs were purchased from NANOCS (New York, USA) with two different coatings, BSA and chitosan. Manufacturer specifications stated 25 -50 nm of diameter for both nanoparticles. Size shape and charge analysis was conducted by Dynamic Light Scattering and Zeta Potential measurement using a ZetaSizer Nano (Malvern Instruments, Malvern, UK) with a 173°scattering angle using SeNPs at 1μg/mL in water (reflexive index of 1.33) at 25°C.

Cell growth and Se treatment
Both SKOV-3 and OVCAR-3 cells were cultured in growth medium until 80% confluency, trypsinized per regular passage and seeded to sterile petri dishes (Corning, UK) at 37°C and 5% CO 2 . After 48 h incubation culture medium was removed and 2 mL of fresh medium containing the expected concentration of aqueous Se (selenite Se 4+ ) or SeNPs added for a further 48 h prior to analysis, with a minimum of three biological repeats.

Cell viability assay
Ovarian cancer cell viability in the presence and absence of Se treatments was determined using the Real-Time Glo assay (RTGlo, PROMEGA, Southampton, UK). 1 × 10 3 ovarian cancer cells/well were plated within 96-well white plates (Corning, UK). After 24 h of growth, culture media was aspirated and 100 μL of fresh medium containing aqueous Se (selenite Se 4+ ) or SeNPs (BSA or chitosan) were added. An increasing dose range (0.06 μg/mL to 40 μg/mL) was applied by dilution in appropriate medium for 48 h. The RTGlo with RPMI-1640 medium was added 1:1 with treatment medium. A BMG Labtech Fluostar Omega was used to measure luminescence every 24 h and presented as absolute values. IC20/IC50 doses were determined as the concentration of treatment that reduced by 20%/50% the luminescence signal compared to untreated control. The IC20/IC50 values shown (average ± standard deviation) are from a minimum of four independent experiments performed with 6 technical repeats.
Migration assay 2 × 10 5 SKOV-3 cells/per well were seeded in 12 well plates and cultured in complete culture medium until the cells reached confluence. A scratch was then introduced to the monolayers, using a sterile (20 μL pipette) tip. The media was then aspirated and a fresh, fetal bovine serum (FBS, Gibco, UK) free medium containing IC20 selenium treatment was added to each well and cultured for 48 h. The scratch was imaged at 0, 24, 48 h using a Primo Vert inverted light microscope (Zeiss, Cambridge, UK). Average scratch width measurements were measured at three random areas, in three wells for each condition, as previously reported. 22 The migration rate was calculated according to the following equation: cell motilities (%) = [1 − (distance of scratched area at 24 h/distance of scratched area at 0 h)] ×100%.

Protein blotting
Total cellular protein was extracted using TRIzol™ reagent (Sigma, Gillingham, UK) and quantified using the DC™ Protein Assay (BioRad, Deeside, UK). Protein from each sample was mixed with Laemmli sample buffer containing βmercaptoethanol (5%) and boiled at 95°C for 5 min. Equal amounts of protein (30 μg) were separated by SDS-PAGE (4-20% gels) and subsequently transferred onto a PVDF membrane (Biorad, Deeside, UK). The membranes were blocked for 1 h in 5% BSA prepared in 0.1% Tris-buffered saline-Tween20® (TBS-T). Blots were then incubated with the corresponding primary antibody (E-cadherin: mouse monoclonal (Abcam ab1416, Cambridge, UK), vimentin: mouse monoclonal (Santa Cruz sc-6260, Wembley, UK), N-cadherin: rabbit polyclonal (Abcam ab18203, Cambridge, Wembley, UK) or GAPDH: mouse monoclonal (Santa Cruz sc-47724, Wembley, UK)) at a concentration of 200 μg/mL overnight at 4°C. Blots were then washed 3 times with TBS-T and incubated at room temperature for 1 h with the appropriate secondary antibody (goat antimouse Abcam ab150113 or goat anti-rabbit Abcam ab6721 HRP secondary, Cambridge, UK) at a concentration of 400 μg/ mL. For signal detection, membranes were processed using the Clarity™ Western ECL Substrate kit (BioRad, Deeside, UK) according to the manufacturer's recommendations and visualized using a ChemiDoc XRS system (BioRad, Deeside, UK). Analysis of the intensity of the bands was done using Image Lab (BioRad, Deeside, UK) software tracing fixed size-limited rectangle around the bands of interest and reporting the Adjusted Volume (Intensity corrected by the background noise). Protein expression was normalized GAPDH and relative expression expressed as the mean fold induction ± standard deviation.

qRT-PCR
Following RNA extraction and quantification, qPCR was carried out in accordance with the manufacturers' recommendations, using the RETROscript® kit two-step method (Invitrogen Ltd., UK). Following cDNA synthesis from 100 ng of RNA, each sample was analyzed by qPCR in triplicate using iQ SYBR Green supermix (BioRad, Deeside, UK) and gene specific primers (Sigma-Aldrich, Gillingham, UK) to evaluate different gene expression GAPDH (GAPDH F o r w a r d : G T C C A C T G G C G T C T T C A C , R e v e r s e : CTTGAGGCTGTTGTCATACTTC) and GPX1 (GPX1 Forward: GTGCTCGGCTTCCCGTGCAAC, Reverse: CTCGAA-GAGCATGAAGTTGGGC). Serial dilutions of cDNA were used to plot a calibration curve, and gene expression quantified by plotting threshold cycle values. Expression levels were normalized to values obtained for the reference gene (GAPDH) and relative expression expressed as the mean fold induction ± standard deviation. Statistical differences between the treatment groups and the control were determined by analysis of variance (ANOVA) (where P b 0.05 was considered significant).

AFM analysis Young modulus, indentation and adhesion
Force-indentation curves were obtained using a Nanowizard II AFM (JPK, Berlin, Germany) mounted on a ZEISS 510 confocal microscope (Zeiss, Cambridge, UK) as described in 23 . During AFM, cells were kept alive in serum free, pH indicator free culture media at 37°C in a petri dish using a standard stage heater and analyzed for a maximum of 3 h. The inverted optical microscope was used to position the tip on the cell and force volume conducted using borosilicate colloidal (Novascan, UK) cantilevers, with a nominal spring constant of 0.35 N/m with a radius of 2.5 μm. Prior to measurements, deflection sensitivity and spring constant were experimentally determined, the latter using the subroutine of the JPK software. Three individual force curves (ramp size of 6 μm) were taken on a total of 25 cells, across 3 independent biological experiments, using a maximum force indentation of 6 nN was used. JPK Data Processing program was used to process the acquired force curves. For each force curve the baseline was corrected to 0 and the approach curve in the contact regime of each force curve was fitted using the classical Hertz model according to Eq. 1. In both cases, the fitting module in the JPK software was used and only curves with a goodness of fit between 0.85 and 1 were considered for statistical analysis.
In this equation, F is the force applied by the cantilever tip to the cell, E is the Young's modulus (fit parameter), υ is the Poisson's ratio (0.5), R the radius of the indenter, δ is the indentation depth and α is the half-angle of the indenter (18°for the used sharp probes).

Topography and surface roughness
In order to resolve the membrane architecture, cells were fixed for imaging according to the protocols outlined in Francis et al. 24 Briefly cells were washed 2 times with PBS then fixed for 30 min in 4% PFA (Merck, UK) diluted in PBS at RT. PFA was subsequently removed and replaced by phosphate buffered saline (PBS), at RT. SKOV-3 and OVCAR-3 cell morphology and topography were analyzed using a BioScope Catalyst (Bruker Instruments, USA) mounted on a Nikon Eclipse Ti-S inverted optical microscope (Nikon Instruments, Netherlands). The inverted optical microscope was used to carefully position the tip on the desired cell and tapping mode imaging undertaken using MLCT-E silicon nitride cantilevers (Bruker-Nano, UK). Offline processing for AFM height data consisted of first-order flattening and plane fitting. The membrane roughness was measured using the subroutine in the Nanoscope Analysis software v1.50, on areas of 25 μm 2 each on five cells for control and treated, from a minimum of 3 biological repeats. Membrane roughness was calculated using Eq. 2, where N is the number of height points in the analyzed area and Zi is the vertical distance of data point i from the mean image data plane. Sixteen roughness measurements were calculated per image, with 1 μm 2 areas of measurement.

Statistical analysis
All data presented are calculated from a minimum of three biological repeats, with technical repeats included per sample, as denoted. Data normality was analyzed using the Kolmogorov Smirnov test, with normally distributed data analyzed with the one-way and two-way analysis of variance (ANOVA) or the Mann-Whitney pairwise test for non-parametric data. In all cases in which ANOVA was significant, multiple comparison methods were used. Differences were considered significant for P ≤ 0.05 (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001). All data were analyzed in MiniTab 14.

Physicochemical and biological characterization of SeNPs
Based on previous observations showing Se has anti-proliferative effects on ovarian cancer cells, 14,25,26 and the good tolerability of  (Figure 1, A and B). SeNP-chitosan had a positive charge of 16.4 ± 4.4 mV, an average size of 320 ± 221 nm and PDI of 0.220 ± 0.012 and was considered as polydisperse. 29 SeNP-chitosan size values with the ZetaSizer measurement were higher than the supplier's specification of 50nm average size however TEM images confirmed those specifications (Figure 1, A  and B) To demonstrate that SKOV3 and OVCAR3 cells were responding to Se treatments, and therefore confirming that Se had been effectively taken up by cells, GPX1 mRNA levels were measured. 30 GPX mRNA levels increased 2-2.5-fold following Se treatment in both cell types (Figure 1, C).
IC50 values confirmed SKOV-3 cells were more sensitive than OVCAR-3 to both selenite and SeNP-BSA treatments (8 vs 56 μg/mL and 19 vs 42 μg/mL respectively, Pb 0.05), whereas the IC50 for SeNP-chitosan the same (40 μg/mL) for both SKOV-3 and OVCAR-3 (Figure 2, B and C). It appears that selenite has a greater cytotoxic effect on SKOV-3 cells than SeNPs, whereas the opposite was observed for OVCAR-3 cells, which are more sensitive to SeNPs than selenite. The concentrations of SeNPs required to decrease cell viability are all significantly higher in OVCAR-3 than SKOV-3 suggesting that the two cell types are responding differently to Se treatment despite the GPX stimulation being similar.
Epithelial-mesenchymal transition (EMT) is a cellular mechanism linked to differentiation during cancer progression, 31 and includes alterations in E-and N-cadherin and vimentin expression, and increased cell migration. 32 Efforts to reverse these transitions are an important consideration in treating aggressive cancers. The effect of Se on cell motility was assessed using a scratch assay, treating cells with IC20 levels of selenite and SeNPs for 48 h. Motility rates were decreased by 25% following SeNP-BSA treatment, 17% after SeNP-chitosan treatment and 19% with selenite (P b 0.05, Figure 3, A), suggesting that whilst selenite was move cytotoxic the three treatments has a similar extent in reducing cell motility. Scratch assays were not possible with OVCAR-3 cells as these cells do not survive when depleted with FBS. In SKOV-3 cells assessment of EMT markers revealed that selenite and SeNPs had no effect with E-cadherin expression remaining at very low levels, and N-cadherin and vimentin remaining highly expressed consistent with an unaltered mesenchymal phenotype. (Figure 3, B and C. In OVCAR-3 cells the E-cadherin:Ncadherin was the opposite to that observed in SKOV-3 cells suggestive of a more epithelial phenotype (Figure 3, B and D), however neither marker responded to selenium treatments. In contrast, vimentin levels significantly (P b 0.05) decreased following treatment with SeNPs, an effect that has previously been correlated with a decrease in cancer cell mechanical integrity. 18 Figure 7. Effect of SeNP on SKOV-3 and OVCAR -3 cell stiffness and nano-indentation. AFM probe with a colloidal tip was used as a nano-indentor to monitor changes in cell elasticity following 48 h of treatment. Using Hertz mechanics, elasticity was calculated from the observed changes in the contact regime of the force curve. The 0 point in the x-axis indicates when the cantilever tip makes contact with the cell surface. Total cell elasticity values are drawn in frequency curve for control and treated cells at 48 h. Highly significant alterations in median values were detected between control and treated cells. Indentation data reported that were applying the same force results in treated OVCAR-3 (B) cells deforming more than control cells. SKOV-3 (A) cells seem to deform less after treatment. Statistical significance was determined using the Mann-Whitney test with the following used symbols NS=P N 0.05, *P b 0.05 **P b 0.01, ***P b 0.001.

Effect of SeNP treatment on surface roughness and biomechanics
Previously ovarian cancer cell biomechanics have been shown to differ depending on the invasive potential of cells, 16 and perturbations in gene expression patterns linked to enhanced cellular movement, migration, and invasion. 17 Given the decreases in motility rates described above we reasoned that this may be linked to altered biomechanical properties following SeNP exposure. Using high-resolution tapping mode AFM we initially resolved gross cellular morphology and nanoscale surface topography.
Membrane topography is sensitive to changes of both physical or chemical factors, and NP exposure has been shown to modulate cell membranes. 33 To investigate any SeNP effects on membrane roughness, 1μm 2 areas of the cell surface selected at random from a 25 μm 2 image area and analyzed for roughness. Surface roughness increased in SKOV-3 cell treated with SeNP-BSA (37.31 ± 1.58 nm) and SeNP-chitosan (42.89 ± 2.37 nm), compared to untreated cells (33.08 ± 1.56 nm), whereas no change was detected following selenite treatment (26.41 ± 1.15 nm) (P N 0.05, Figure 6). For OVCAR-3 cells the surface roughness of SeNP treated was significantly decreased from 39.39 ± 2.48 nm in untreated controls to 28.41 ± 1.82 nm with selenite, 27.92 ± 1.75 nm with SeNP-BSA, and 30.05 ± 1.11 nm with SeNP-chitosan treatment (P b 0.05). There was no significant difference in OVCAR-3 surface roughness was observed between selenite and SeNP treatments (P N 0.05, Figure 6).
The mechanical properties of cells change during cancer progression, with metastatic cells becoming more elastic 17 after initial transformation to enable basement membrane penetration. [34][35][36] To investigate whether Se treatment resulted in any changes in cell elasticity, both SKOV-3 and OVCAR-3 cells were characterized using AFM based nanoindentation, with a colloidal probe (Figure 7).
Cellular adhesion to the AFM probe was also analyzed, with no significant alterations in SKOV-3 or OVCAR-3 cells respectively, as shown in Supplementary Figure 4.

Discussion
Despite the anticancer properties associated with selenium, its use as a cancer therapy has not yet been realized due to systemic toxicity when administered in an aqueous form. Here we demonstrated that, like selenite, SeNPs are effective at inducing a cytotoxic effect on two different serous ovarian cancer derived cell lines and observed distinct responses for each cell type. Both selenite and SeNPs were cytotoxic to SKOV-3 cells, with selenite having a 2.5-and 5-fold lower IC50 than SeNP-BSA and SeNP-chitosan respectively. However, given that aqueous selenium is significantly more toxic when administered systemically in murine models, 5 , 27 , 28 the use of SeNP-BSA is likely to provide an opportunity to deliver cytotoxic doses of Se in vivo. In addition, tumor specific delivery could occur through the enhanced permeability and retention effect, 37 further reducing toxicity. For OVCAR-3 cells SeNP-BSA and SeNP-chitosan had an approximately 2-fold greater cytotoxic effect than selenite. Given the above considerations, and that the IC50 for SeNPs in OVCAR-3 cells was almost five times greater than for SKOV-3 cells, both types of SeNP are likely to be the only safe and tolerable route for Se administration to target cancer cell growth in vivo.
Changes in cell mobility in SKOV-3 led us to investigate cell membrane dynamics in response to selenium treatment as we and others have previously observed differences in ovarian cancer cell biomechanics which were dependent on the invasive potential of cells, 16 and associated with genes involved in enhanced cellular movement, migration, and invasion. 17 Detailed AFM analysis revealed that SKOV-3 cells become less elastic following SeNPs treatment but remained unchanged when treated with selenite. This suggests that SeNPs are triggering a mechanism that results in a lower metastatic potential, and in agreement with the reduced migratory capacity of these cells following treatment. 20 , 22 In contrast, AFM analysis revealed that OVCAR-3 cells became more elastic following treatment with the higher IC20 concentrations of SeNP and selenite required to induce a cytotoxic effect with these cells. These results highlight specific differences in membrane architecture between SKOV-3 and OVCAR-3 and indicate that SeNP treatment can cause differential cell type specific cytoskeletal effects.
Further contrasting effects were seen with surface roughness, which increased following SeNP treatment in SKOV-3 cells, possibly as a result of nanoparticle internalization as previous studies have shown that BSA coated SeNPs, similar to those used in the present study, were internalized through endocytosis or clathrin mediated vesicles. [38][39][40] As selenite would be expected to be taken up by anion transporters 41 this difference in surface roughness was anticipated. Surprisingly, the surface roughness of OVCAR-3 cells decreased following SeNPs and selenite treatment and was accompanied by the appearance of intracellular vesicles which are likely to be autophagosomes ( Figure 5, B-D). 42 , 43 This suggests that, uniquely, Se forms may induce autophagy which would act as a resistance mechanism in OVCAR-3 cells 41 and offers an explanation as to why the levels of SeNPs required to illicit a response is higher for these cells. This detailed nanomechanical assessment of two distinct ovarian cancer cell types has established that changes in the mechanical properties of ovarian cancer cells, which are likely to result from reorganization of the actin cytoskeleton induced by selenium. SeNPs are effective at preventing cell proliferation in high grade serous ovarian cancer cells, apparently through different biological pathways. High grade serous ovarian cancers continues to present obstacles to currently available treatment through the presentation of advanced and highly aggressive forms of the disease, resulting in a high rate of mortality. 8 The observations made here pave the way for further investigations into evaluating the utility of SeNPs as a pan-cancer treatment, at least in ovarian cancer, although Se is known to be effective in many other cancer types.