Unveiling the Biotoxicity Mechanisms of Cancer-Selective Thulium Oxide Nanoparticles

Abstract

High-Z nanoparticles (NPs) have the potential to revolutionize cancer radiotherapy by radiosensitising tumours. This is particularly important for radioresistant cancers such as glioblastoma. A newer NP candidate in this area is thulium oxide nanoparticles (TmNPs). However, prior to clinical assessment, ideal NP characteristics, including biocompatibility, biosafety, and preferential uptake in cancer, should be assessed. This in vitro study compares the effects of TmNP treatment, without radiation, on 9L gliosarcoma (9LGS), a well-established glioblastoma cell model, with exposure to Madin Darby Canine Kidney (MDCK) cells, a widely used non-cancerous cell model. The findings demonstrated selective uptake of TmNPs in 9LGS over MDCK following treatment. A biological assessment of toxicity confirmed minimal long-term effects on MDCK, whilst TmNPs were observed to induce some notable cell death in 9LGS. Excessive TmNP uptake in 9LGS over time was observed to induce cell vacuolisation, which resulted in cell death via necrosis. It was concluded that this was the explanation for the underlying mechanisms of TmNP toxicity in cancer cells. This study was therefore able to demonstrate not only that TmNPs are a biocompatible, cancer-selective candidate for radiosensitiser usage, but further provided a theory to explain its mechanisms of cancer cell toxicity.

1. Introduction

Cancer is a leading cause of mortality worldwide, responsible for millions of deaths annually [1]. Glioblastoma multiforme (GBM) is the most common malignant tumour of the central nervous system (CNS) with one the lowest 5-year survival rates in CNS cancers [2,3]. 9L gliosarcoma (9LGS), a biphasic variant of GBM with both sarcomatous and gliomatous elements, is a widely used animal cell model for in vitro and in vivo studies of glioblastoma [1,4,5]. 9LGS cells are derived from Fischer 344 rats and have similarities to high-grade human glioma cells, including both an astrocytic morphology and notable resistance to conventional treatments [6,7].

Madin Darby Canine Kidney (MDCK) cells are an epithelial tissue model suitable for a variety of uses [8]. Well-established MDCK cells provide a widely used and flexible model for normal, soft tissue [8]. While these cells are derived from canine kidneys [8], in contrast with the rodent origins of 9LGS, both cell lines represent soft tissues. Moreover, MDCK comparisons can provide an insight into the effects of treatments for cancers on non-cancerous cells, as well on the kidneys [9,10].

Surgical resection, chemotherapy, and radiotherapy (RT) remain the common and conventional methods of cancer treatment [11,12,13]. However, each modality faces limits. For brain tumours, invasive cancers may not be possible to remove safely, while chemotherapeutic drugs may harm normal tissues in the body [13,14]. RT utilizes ionizing radiation (IR) such as photons to treat cancers locally, and more directly target tumours, the success of which depends on the total radiation dose delivered [11,14,15]. The cellular damage induced via radiation occurs either directly or indirectly through the production of ionized electrons and free radicals, which can in turn result in DNA damage. This damage can result in the death of the cancer cell via apoptosis (programmed cell death) or necrosis (death due to injury) pathways (among others) [16,17].

The two primary DNA damage mechanisms are single-strand DNA breaks (SSB) and double-strand DNA breaks (DSB). With SSB, which comprise most of the breaks, only one strand of the DNA double helix is discontinued. By contrast, DSBs occur where both strands are severed, which is the most lethal lesion for a cell [18,19,20,21]. However, RT also faces the challenge of maintaining tumour control whilst minimising exposure to adjacent healthy tissues, given normal tissues can also experience DNA damage [7,15]. As radiation doses are then restricted by the tolerance of normal tissues to IR exposure, new and improved methods of targeted radiation treatment are needed.

A newer option is nanotherapy, which involves the use of nanoscale materials to act as targeted delivery agents for multifunctional nanoceramics [22,23,24]. Novel nanoparticles (NPs), typically measured in size from 1 to 100 nm, have become a popular new means of nanotherapy due to their nanoscale size making passage into living cells easier and more effective [23,24]. Biocompatible NPs would ideally be rapidly taken up and selectively accumulate in targeted cancer cells, while also maintaining the properties necessary for theranostic applications. They would further have negligible toxic effects whilst being safely cleared from the body (typically via the kidneys) [23,24], thereby making MDCK a useful cell model for assessment.

As NPs, such as gold [25,26,27,28], can deliver targeted materials to cancers such as 9LGS [7], this newer modality then compliments both theranostic RT and diagnostic medical imaging [13], where high-Z materials are preferred due to the well-established energy dependence of photon interactions [29,30]. Oxide NPs in particular are commonly used [7,9,31,32,33,34] and can deliver high-Z materials directly to cancer cells to induce tumour radiosensitisation after IR exposure. This occurs through photoelectric interactions, where secondary electron emissions increase the likelihood of inducing lethal genetic lesions in the cell [29,30].

Several studies have demonstrated that the cellular uptake of NPs, which is typically via the endocytosis pathway [35,36], and distribution within the cell are important factors to consider when characterising NPs’ potential in cancer radiosensitisation. Non-cancerous MDCK has been rarely studied to assess NP internalisation and distribution within normal tissue lines. However, a study by Engels et al. was conducted with both 9LGS and MDCK using tantalum oxide (Ta₂O₅) NPs (TaNPs), which were observed to have negligible cytotoxicity in both cell lines [37].

NP internalisation and distribution in 9LGS cancer has been studied more frequently, however. Brown et al. found that some TaNPs aggregate clusters were too large to be internalised within 9LGS, even following sonication, with a mean crystallite size of 56 nm found by X-ray diffraction measurements in two independent studies [38,39]. Due to this, TaNPs would distribute throughout the cytoplasm following cellular uptake, later confirmed in a study by McDonald et al. for the same TaNPs concentration [33]. Further study demonstrated a “shell effect” when these NPs were internalised by 9LGS (noticeably present at high concentrations) in which the NPs aggregate around the nucleus in a shell-like layer [38]. It was reasoned that this effect could act to absorb secondary electrons that may otherwise contribute to dose enhancement [33,38]. Hence, assessing the size, uptake, and distribution of NPs within a target cell are then critical to successful use of these types of cancer therapies.

Engels et al. assessed an alternative NP candidate and found a smaller mean crystallite size of 42.52 nm for thulium oxide (Tm₂O₃) NPs (TmNPs), suggesting greater internalisation potential given the optimal NP core size range for maximum cellular uptake is 30–50 nm [36]. This was observed in flow cytometric experiments performed by both Brown et al. and Engels et al. [7,38], in which a larger mean side scatter (SSC) value was found for TmNPs compared to the untreated 9LGS control at the same 50 µg/mL concentration. TmNPs exhibited some cytotoxicity after cellular uptake, with a clonogenic survival of 75% for 9LGS without any IR exposure, compared to nearly 100% for Ta₂O₅ NPs for the same concertation [7,38]. This was confirmed by McDonald et al. with 80–90% survival and Engels et al. with 80–100% for both 9LGS and MDCK [33,37]. This also suggested that the smaller size of TmNPs may have contributed to greater cellular uptake and cytoplasmic internalisation, and from this, cell cytotoxicity. This was also verified by Engels et al. via flow cytometry for TmNPs internalised by 9LGS [7].

Cytoplasmic distribution of NPs was further observed by Engels et al. through DSB assessment using γH2AX immunofluorescence via fluorescent microscopy. TmNPs were found to be internalised into the cytoplasm of 9LGS cells following uptake, with some observations of larger NP clusters localised near the nucleus, similar to the “shell effect” observed in previous studies [7,38]. Engels et al. also observed that NP cluster proximity to the cell nucleus corresponded to clustered and more numerous occurrence of DSBs [7]. This indicated that a site-dependent increase in DNA damage was responsible for the radiosensitisation of 9LGS cells treated with TmNPs [7].

These findings in previous studies clearly highlighted that use of NP candidates for successful radiosensitisation is dependent on cellular uptake, internalisation, distribution, and localisation. Cytotoxicity in targeted cancer cells, even without IR exposure during RT, was often present, depending on the NP material used. While some NPs demonstrated negligible toxicity to 9LGS cancer cells [9,31,32,33], TmNPs demonstrated some ability to kill the cancer cells on their own [7]. This was partly due to their smaller size permitting greater cellular uptake (although this also depends upon NP reactivity) [7,36,40], making TmNPs a good candidate for radiosensitisation once confirmed for cancer selectivity and biocompatibility. However, as prior assessment with non-cancerous cells such as MDCK is lacking, further work to confirm the effects of TmNPs on normal cells is critical. Assessment of cancer selectivity in the uptake of the NPs is also needed to ensure the targeted capability of these high-Z treatment agents is maintained [22,23,24].

As such, this study focused on a comparative assessment of the biological effects of treating cancerous 9LGS cells and non-cancerous MDCK with TmNPs without radiation. It was reasoned that a comprehensive study would be significant to confirm that TmNPs could selectively target cancer cells over non-cancerous cells. This would be key to assuring NP targeting is retained so safe future use in clinical patients is assured. Additionally, the uptake, internalisation, distribution, and localisation of TmNPs within cells was assessed, as well as the underlying mechanisms of any toxic effects on cells. For this purpose, short, medium, and long-term effects on 9LGS metabolism, growth, and survival were studied, as were biological impacts affecting the cell cycle, cellular DNA, and cell death. These results would demonstrate the potential of TmNPs for us as a NP candidate for RT enhancement, by first assessing the viability of using these NPs without radiation and explaining their mechanisms behind any observed effects.

2. Materials and Methods

2.1. Nanoparticle Preparation

Thulium(III) oxide (Tm₂O₃) nanoparticles (NPs) (99.9% trace metals basis) were obtained from Sigma Aldrich (Merck, Bayswater, VIC, Australia #289167). Following the protocols by Engels et al. [7], the TmNPs were used for physicochemical analysis as powder or in a 70% ethanol/water (v/v) solution that was sonicated for 40 min to separate particles using an ultrasonic water bath (Branson, Danbury, CT, USA).

For in vitro assays, TmNPs were sonicated sterile for 40 min in DPBS (Ca²⁺/Mg²⁺ free, Gibco, Melbourne, VIC, Australia, #14190144) at a concentration of 1 mg/mL (w/v). NPs were added to cells for an optimal concentration of 50 μg/mL in media and then incubated for 24 h of exposure prior to the cells reaching 100% confluence.

2.2. X-Ray Diffraction for Materials Analysis of Nanoparticle Size and Properties

Powder X-ray diffraction (XRD) data for TmNPs were collected using a PANalytical Aeris diffractometer (Malvern PANalytical Sydney, Chipping Norton, NSW, Australia) equipped with a Cu Kα radiation source (λ = 1.5406 Å), operating at 35 kV and 28.4 mA. Diffraction patterns were recorded over a 2θ range of 10–90°, with a step size of 0.02°.

Rietveld refinement was performed using the Materials Analysis Using Diffraction (MAUD) software package [41]. The initial structural model corresponded to the standard cubic phase of thulium oxide (space group Ia-3, No. 206), with atomic coordinates obtained from the Crystallography Open Database (COD) [42]. The background was modelled using a polynomial function, and peak shapes were fitted using a pseudo-Voigt profile. Refinement parameters included the scale factor, background coefficients, zero-point correction, lattice parameter (a), peak shape parameters (U, V, and W), atomic positions, and isotropic atomic displacement parameters for both Tm and O.

2.3. Transmission Electron Microscopy for Nanoparticle Physiochemical Analysis

TmNPs were visualised by transmission electron microscopy (TEM) for a physiochemical analysis of particle size and properties following 40 min of sonication. Imaging was conducted using a JEOL F-200 TEM (JOEL (Australasia) Pty. Ltd., Frenchs Forest, NSW, Australia) at the Australian Institute for Innovative Materials (AIIM), University of Wollongong Innovation Campus (Wollongong, Australia).

TEM employs an electron beam generated by a tungsten filament emission gun and a series of electromagnetic lenses to achieve high-resolution imaging down to the nanometre scale. This allowed a detailed characterization of morphology, lattice spacing, and average particle size, as well as an estimate of crystallite size.

In this study, the TEM operated at an accelerating voltage of 200 kV. After proper instrument alignment, a series of high-resolution images were taken at varying magnifications to examine grain fringes, single-crystal diffraction patterns, and overall surface morphology. Average particle size was determined by measuring particle dimensions using a known pixel-to-nanometre calibration.

2.4. Energy Dispersive X-Ray Spectroscopy for Nanoparticle Elemental Analysis

The TEM used in Section 2.3 is equipped with an energy dispersive X-ray spectroscopy (EDS) detector, permitting an elemental distribution analysis of the nanoparticles. Prior to measurement, a beam shower was applied to decontaminate the region of interest. The electron beam was then focused on the sample, producing both an image and characteristic X-ray signals. These X-rays were detected and matched against a built-in elemental database to determine the elemental composition and spatial distribution. This provided insight into the stoichiometric ratio of thulium (Tm) to oxygen (O) in the NPs.

2.5. Subculture of Adherent Cells

Both 9LGS and MDCK cell lines were obtained from the European Collection of Authenticated Cell Cultures (ECACC). Both cell lines were cultured in T75 cm² flasks (Greiner Bio-One via Interpath, Melbourne, VIC, Australia, #658175) containing complete Dulbecco’s Modified Eagle Medium (c-DMEM) (Gibco via ThermoFisher Scientific, Brisbane QLD, Australia, #11965118), with added 10% foetal bovine serum (FBS) (Gibco, AUS, #10099141) and 1% PenStrep (10,000 units/mL penicillin, 10,000 μg/mL streptomycin) (Gibco, AUS, #15140122). All subcultures were incubated at 37 °C and 5% CO₂ (v/v), with a doubling time of 36 h for 9LGS cells and 16 h for MDCK cells.

When passaged or harvested, both cell lines were washed with DPBS (Dulbecco’s Phosphate Buffered Saline) (Ca²⁺/Mg²⁺ free, Gibco via ThermoFisher Scientific, Brisbane, QLD, Australia #14190144) before being suspended with 0.05% Trypsin EDTA (Gibco, 25300054). Cells were harvested via this passaging method and counted and seeded for monolayers at 100% confluence into 1 cm² (well area) micro-chamber slide (Ibidi via DKSH Australia, Sydney NSW, Australia, #80827) wells for imaging or T12.5 cm² flasks (Corning Incorporated, Corning, NY, USA, #353107), T25 cm² flasks (Greiner Bio-One via Interpath, Melbourne, VIC, Australia, #690175), and 96-well microplates (Corning^® via Sigma (via Merck), Melbourne VIC, Australia, # CLS3599), or 24-well microplates (Greiner Bio-One via Interpath, Melbourne, VIC, Australia, #662160), for other experiments.

2.6. Short-Term Cell Viability via MTT Assay

Assessment of short-term cell viability was performed using a MTT assay for each cell line, with or without TmNPs. A monolayer of 9LGS was cultured for confluence of 10,000 cells, or MDCK monolayers at 6000 cells, at the time of assay in a 96-well plate. A total of 6 wells were seeded for each treatment type, with 6 wells for a cells-only control and another 6 wells of c-DMEM media as a plate blank. MTT tetrazole (reduced to formazan in cells) reagent stock powder (Invitrogen, Melbourne, VIC, Australia, #M6494) was diluted in DPBS (Ca²⁺/Mg²⁺ free, Gibco via ThermoFisher Scientific, Brisbane, QLD, Australia, #14190144) for a final 5 mg/mL (w/v) final concentration (i.e., 12 mM).

Following treatment, 10 μL of 12 mM MTT was added to each plate well in darkness in aseptic conditions. The plate was wrapped in aluminium foil and incubated in accordance for 4 h in darkness at 37 °C and 5% CO₂ (v/v). Following incubation, the MTT reagent was discarded, and 200 μL of DMSO (≥99.5%) (Sigma Aldrich via Merck Life Science, Melbourne, VIC, Australia, #276855) was added to each well. The plate was then analysed by a SpectraMax Plus 384 Microplate Reader (Molecular Devices, San Jose, CA, USA) at a wavelength of 540 nm.

2.7. Medium-Term Growth via Live Cell Imaging

Medium-term growth and efficacy were assessed using live-cell imaging time trials. An IncuCyte ZOOM live cell imaging system (Essen BioScience via Sartorius AG, Göttingen, Germany) was used to image cells at regular intervals for 9LGS or MDCK monolayers. Each cell line was first cultured in T25 cm² flasks and treated for 24 h TmNPs (while cells-only controls were left untreated).

Following NP treatment, flasks were washed, and cells were harvested and counted to seed into the wells of 24-well microplates (Greiner Bio-One via Interpath, Melbourne, VIC, Australia, #662160) for imaging over time. 9LGS cells were seeded for 250,000 cells at confluence after 10 days, or 150,000 cells after 5 days for MDCK. Once seeded, 3 µL of a 1 mg/mL stock of propidium iodide (PI) (Sigma via Merck, Melbourne, VIC, Australia, #P4170) powder diluted in DBPS was added to each well. The PI was detected using a red excitation wavelength band of 565–605 nm and an emission detection range of 625–705 nm, sufficient to detect PI signals.

Images were taken at 4 h intervals across the time trial for both cell lines with a 10× dry resolution objective. A total of 9 images were taken at each time point in each well in a 3 × 3 grid pattern. After the time courses were completed, the IncuCyte ZOOM software (version 2016A) provided the analysis. All images were processed and quantified to obtain the percentage confluence for the sample at each time point and the PI signal red dot count at each time point.

These were plotted on graphs over time as percentage confluence measures, and the area under the curve (AUC) for each graph was obtained to provide an overall measure of population growth inhibition in response to treatment. Cytotoxic damage represented by increased cell death marked by PI signal increases (via staining of released genetic content from damaged or dying cells with ruptured membranes as PI cannot permeate cell membranes without fixation) was also obtained over time, and the AUC was obtained to calculate an overall measure. The PI red dot measures were divided by the corresponding confluence measure for each image, at each time point, to obtain a measure of damage per cell (thereby accounting for increased PI measurement that may occur naturally with more cells in a population). The overall measures are represented as an enhancement ratio by comparing the ratio of the AUC of each treated sample (for both grown inhibition and cell damage) to that of the cells-only control for that cell line.

2.8. Long-Term Survival via Clonogenic Assay

Clonogenic cell survival assays were performed with and without TmNPs for both cell lines to assess long-term cell survival and treatment efficacy. Following the treatment of cells with 50 µg/mL of TmNPs seeded into T12.5 cm² flasks, the cells were plated for a clonogenic assay following the protocol of Engels et al. 2018 [7].

Cells were washed and harvested following Section 2.5. Cells were then counted and seeded at the desired density in triplicates of 3 Corning Primaria™ 100 mm Cell Culture (petri) Dishes (Corning Incorporated, Corning, NY, USA, #353803) for each sample. At least two seeding numbers (each with a triplicate of plates) was used for each sample. Cells were then mixed and incubated at 37 °C and 5% CO₂ (v/v) for up to 15 doubling times. Following incubation, plates were washed with DPBS (with Ca₂+/Mg₂+ salts, Gibco, Australia, #14040) before being stained with a crystal violet solution (Sigma Aldrich via Merck Life Science, Melbourne, VIC, Australia, #HT90132) diluted 1:3 in 70% ethanol (v/v).

Colonies with less than 50 cells, and plates with less than 50 colonies or more than 300, were discounted. The plating efficiency (PE) was determined as the ratio of surviving colonies to the initial seeding number plated. For each treated sample, the surviving fractions (SF) was calculated by taking the ratio of the PE of the treated cells over the cells-only control.

2.9. Nanoparticle Uptake and Toxicity Imaging via Confocal Microscopy

For cytotoxicity assessment and visualisation, including cell membrane damage and DNA leakage in response to treatment, confocal microscopy was performed. A monolayer of cells was cultured for confluence of 100,000 cells for 9LGS, or 60,000 for MDCK, in the wells of an 8-well micro-chamber slide. Following TmNP treatment (Section 2.1), at confluence of 90–100%, fluorescent stains Hoechst 33,342 (H) (Sigma Aldrich via Merck, AUS, #14533) and propidium iodide (PI) (Sigma via Merck Life Science, Melbourne, VIC, Australia, #P4170) were used to stain genetic content.

H stock powder was dissolved in distilled water for a concentration of 1 mg/mL. PI was dissolved in DPBS (Ca²⁺/Mg²⁺ free, Gibco via ThermoFisher Scientific, Brisbane, QLD, Australia, #14190144) for a concentration of 1 mg/mL. Each dilution was left to dissolve for 24 h while refrigerated at 2–8 °C and then heat-treated and filtered.

A total of 2 µL of each stain was added and incubated for 1 h prior to imaging. Cells are then washed once with 300 µL of DBPS, and a Leica TCS SP8 confocal microscope (Leica Microsystems, Lane Cove West, NSW, Australia) with a 93× glycerol immersion objective resolution was used to image the cells live. Live incubation at 37 °C and with 5% CO₂ (v/v) was used.

DAPI (for H), FITC, and Texas Red (TXR) spectra (for PI) were applied using a 405 nm excitation wavelength for the H (which served as a DAPI spectra nuclear counterstain) and 561 nm for PI. A 488 nm excitation was used for the FITC spectra, which was used for light scattering off the TmNPs, thereby allowing the NPs to be better detected in the image. All detection ranges began at 10 nm above the excitation wavelength.

An optical bright field was used for the background to visualise the cell membranes. Voltage gains were optimised using the images which had the most intense signals to avoid saturation. Sequential imaging of separate channels (bright field, DAPI, FITC, and TXR) was used to acquire 512 × 512 pixel images and avoid crosstalk of signal detection from different wavelengths. A 2 × 2 tile scan with a z-stack of 10 slices was taken per image, resulting in images greater than 200 µm × 200 µm at 93× resolution to include sufficient cells (around 50–100 cells per image). At least 3 images were taken per well.

The images were analysed on the Leica Application Suite X (LASX) software (v. 3.0.11.20652, Leica Microsystems, Wetzlar, Germany) and ImageJ (v 1.53k; NIH, Bethesda, MD, USA) [43]. The 3D Object Counter plugin for ImageJ was then used with the NP light scatter channel [44], allowing for the NP clusters imaged to be analysed and quantities measured. The plugin provided data on the NP volume, surface area, and mean radius, as well as its 3D position (via the object centroid) in the z-stack. Use of a mask (by using the H channel to identify the cells) via the plugin allowed these NP size and location measurements to be targeted specifically to NPs internalised within the cells. Using the known density of thulium oxide of 8.6 g/cm³, the total mass of NPs taken up by either cell line was measured.

2.10. Nanoparticle Uptake Quantification Using Flow Cytometry

Nanoparticle internalisation within 9LGS or MDCK cells was performed with flow cytometric quantification to verify the proportion of the cell population taking up NPs. Following treatment in T25 cm² flasks, 1 × 10⁶ cells were harvested and centrifuged at 380× g for 5 min at 22 °C and washed with 1 mL of DBPS before being transferred to the flow cytometer. A BD LSRFortessa^TM X-20 (BD Biosciences via Becton Dickinson Pty. Ltd. (Australia & New Zealand), Macquarie Park, NSW, Australia) flow cytometer analysed the sample to measure forward scatter (FSC) and side scatter (SCC), and the NP light scatter signal response is detected using a 488 nm excitation wavelength with a (525 ± 50) detection range.

The mean FSC and SSC signals increased when cells were treated with TmNPs detected due to light scattering off the internalised NPs. Comparing the increase in the mean SSC signal demonstrated the presence of nanoparticles and characterised cellular uptake of the TmNPs. The ratio of SCC values for NP-treated cells was compared to the cells-only control for each cell line. The percentage of cells within the NP gating compared to the total population measured indicated the proportion of the population taking up NPs (which differed from the microscopy experiment that instead measured what quantity of NPs was taken up per cell).

2.11. Cell Cycle Analysis via Flow Cytometry

Flow cytometric analysis of cell cycle phases was conducted using PI (Sigma via Merck Life Science, Melbourne, VIC, Australia, #P4170) staining of 9LGS and MDCK cells. Following treatment of cells in T25 cm² flasks, 1 × 10⁶ cells were harvested and centrifuged at 380× g for 5 min at 22 °C and washed with DBPS (Ca²⁺/Mg²⁺ free, Gibco via ThermoFisher Scientific, Brisbane, QLD, Australia). The cells are then resuspended and fixed with 1 mL of ice-cold 100% methanol (Sigma via Merck Life Science, Melbourne, VIC, Australia, #37860) for 30 min. Cells are then centrifuged and washed twice with DPBS.

Following this, a PI dilution of 40 µg/mL PI and 100 µg/mL RNase A (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #R6513) in DPBS is added at a cell density of 1 × 10⁶ cells mL⁻¹ in darkness and incubated for 1 h. After incubation, a BD LSRFortessa^TM X-20 (BD Biosciences via Becton Dickinson Pty. Ltd. (Australia & New Zealand), Macquarie Park, NSW, Australia) flow cytometer analysed the sample to measure the stained DNA (PI) signal response using a 488 nm excitation with a (575 ± 25) detection range. After FSC and SCC gating and filtration, each of the phases of the cell cycle are identified and gated, with the number of events in each compared to the total gated population to obtain the percentage distribution of cells in each phase.

2.12. Cell Death Analysis via Flow Cytometry

Cell death pathway analysis was conducted using the Annexin V reagent (Invitrogen, Melbourne, VIC, Australia, #C10841) with flow cytometry [45]. An Annexin binding buffer is pre-prepared at pH 7.4, comprising a mix of 10 mM of HEPES (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #RDD002), 140 mM of sodium chloride (NaCl) salt (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #71376), and 2.5 mM of calcium chloride (CaCl₂) salt (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #764495). The Annexin V reagent is then diluted by adding 5 µL of the stock reagent into 1 mL of binding buffer.

Following treatment of 9LGS or MDCK cell monolayers in T25 cm² flasks, 1 × 10⁶ cells are harvested and centrifuged at 380× g for 5 min at 22 °C and then washed with DBPS (Ca²⁺/Mg²⁺ free, Gibco via ThermoFisher Scientific, Brisbane, QLD, Australia). Cells are then washed with Annexin binding buffer before 1 mL of the diluted Annexin V reagent (in buffer) is added. Samples are incubated in the Annexin V dilution for 20 min in darkness at room temperature. Following incubation, the Annexin V supernatant is discarded, and cells are washed with binding buffer once more.

A total of 100 µL of PI/RNase mix is added to cells, consisting of 40 µg/mL PI (Sigma via Merck Life Science, Melbourne, VIC, Australia, #P4170) and 100 µg/mL RNase A (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #R4875) in DPBS. Cells are then incubated in PI/RNase in darkness at room temperature for 20 min. Following this, cells are transferred to a BD LSRFortessa^TM X-20 (BD Biosciences via Becton Dickinson Pty. Ltd. (Australia & New Zealand), Macquarie Park, NSW, Australia) flow cytometer for analysis. Annexin V signals are detected using a 488 nm excitation with a (525 ± 50) detection range, while PI signals are detected using a 488 nm excitation with a (575 ± 25) detection range.

Following FSC and SCC gating and filtration, a quadrant gating is used on a scatter plot of Annexin V vs. PI to obtain the number of live cells, early apoptotic, late apoptotic, and necrotic cells. The number of cells in each quadrant is compared to the total gated population to obtain the percentage found in each state. The ratio of each percentage for each NP-treated sample (for both 9LGS and MDCK cells separately) is compared to the untreated cells-only controls for each cell line to obtain an enhancement ratio for any increase in cell death.

2.13. γH2AX Immunofluorescent Imaging and Quantification via Confocal Microscopy

DSBs were imaged by confocal microscopy using well-established biomarker γ-H2AX [46,47]. Microscopy was performed for a monolayer of cells cultured in slide wells for confluence of 100,00 cells for 9LGS or 60,000 cells for MDCK. Both treated and untreated (cells-only) control samples were imaged.

At 20 min following the 24 h exposure and incubation with TmNPs, cells were washed twice with 300 µL of ice-cold DPBS per well before being fixed with 300 µL of ice-cold 100% methanol per well for 20 min on ice. Wells are then each washed three times with 300 µL of cold DBPS, where for each wash the chambers are rocked for 5 min at room temperature. Following this, wells are treated twice with a blocking solution of 3% bovine serum albumin (BSA) (Sigma via Merck Life Science, Melbourne, VIC, Australia, #A9418) in DPBS (BSA-DPBS), with 15 min of rocking at room temperature for each wash. A primary antibody (Mouse anti-phospho-Histone H2A.X (Ser139), clone JBW301, supplied by Merck Millipore via Merck Life Science, Melbourne, VIC, Australia, #05-636) was then added 1:500 in 1% BSA-DPBS mix for a concentration of 2 µg/mL in the cells. Cells were then incubated for 2 h at room temperature in darkness.

Following incubation, the cells were washed three times with BSA-DPBS and 5 min of washing at room temperature per wash. A secondary antibody (goat anti-Mouse IgG1 Cross-Absorbed, Alexa Fluor 488, supplied by Invitrogen via Merck Life Science, Melbourne, VIC, Australia, #A21121) was added 1:500 in 1% BSA-DPBS for a concentration of 4 µg/mL to the cells and incubated for 1 h at room temperature in darkness. Finally, cells were again washed twice with 300 µL of DBPS before 100 µL of DBPS was added to each well. A total of 2 µL of 1 mg/mL Hoechst 33342 (Sigma-Aldrich via Merck Life Science, Melbourne, VIC, Australia, #14533) was then added to each well for 20 min at room temperature. Cells were then imaged with a Leica TCS SP8 confocal microscope (Leica Microsystems, Lane Cove West, NSW, Australia) with a 93× glycerol objective at room temperature.

The confocal microscope utilised a wavelength providing a 488 nm excitation with a detection range for the Alex Fluor 488 fluorophore (FITC) and another 405 nm excitation providing with the range for the H nuclear counterstain (DAPI). Detection ranges were set to a minimum 10 nm above the excitation wavelengths for each channel and higher. A 2 × 2 tile scan with a z-stack of 10 slices was taken per image. These images were then analysed via the Lecia LasX Application Suite (v. 3.0.11.20652, Leica Microsystems, Wetzlar, Germany) and ImageJ.

ImageJ (v 1.53k; NIH, Bethesda, MD, USA) [43] was used to process images to quantify DSBs observed in γH2AX images (represented by foci). A quantitative analysis of γH2AX foci was used as the key indicator of DNA damage due to the high sensitivity of this method [46,47]. Following our previous work in Valceski et al., the foci factor (FF) method was used to account for variations in individual γH2AX foci [48].

FF values provide a measure representing the average number of DSBs per cell nucleus. The FF value is determined for each individual image as the raw integrated density (the total sum of pixel intensity values in a foci) summed up across all foci in that image, divided by the number of cells counted in the image. The DSB Enhancement Ratio (DSBER) was determined as the ratio of FF values of a treatment sample to the untreated, cells-only control. DSBER is used as the final quantification of all confocal images using the γH2AX assay in this work to represent enhancement in DSBs following treatment. The DSBER values of at least six images are averaged for each sample assessed.

2.14. Statistical Analyses

All error bars were calculated as standard error using 2 standard deviations (95% confidence interval) of the mean divided by square root of the number of samples or images used. For all samples tested, at least 4 biological and technical replicates across independent repeats were averaged for each sample.

A Student’s t-test was used to compare samples for statistical significance, with the unpaired heteroscedastic t-test for all independent samples. One-tailed t-tests were used when comparing to untreated controls as the increase was the primary interest, while all other cases used a two-tailed t-test. The p values for each statistical test are presented in the corresponding figure legend.

3. Results

3.1. Physiochemical Analysis Demonstrates Optimal TmNP Size for Nanoparticle Uptake

To complete our analysis of TmNP selectivity for cancer, a physiochemical analysis was performed using XRD and TEM along with EDS. This provided a valuable perspective at the molecular level on the properties of the TmNPs and their potential for cell uptake. Figure 1 confirmed the crystallinity and particle sizes and produced an XRD spectrum that very reliably matched the previous analysis of these TmNPs conducted in our previous work [7]. This included a match, within experimental uncertainty, with the crystalline parameters and particle sizes, as well as full alignment with nearly all major peaks of reflection and corresponding miller indices. This provided assurance of our NP properties.

The refinement converged with agreement indices of R_wp = 11.1%, R_p = 3.2%, and χ = 3.46. The refined lattice constant was a = (10.49410 ± 0.00003) Å, consistent with reported values for cubic TmNPs and well in agreement (and within error) of our previous analysis [7]. The refinement yielded a single-phase model with a 100% weight fraction of crystalline TmNPs, and no secondary phases or amorphous content were detected within the sensitivity of the measurement.

Crystallinity was further confirmed by the sharpness of the dominant reflections and the narrow FWHM values across the pattern. The degree of crystallinity was calculated from the relative integrated intensity of the crystalline peaks with respect to the total scattered signal, including the background. The calculated d-spacing of the (211) reflection was 4.2842 Å, matching the spacing observed in TEM measurements and confirming phase identification.

The crystallite size was evaluated using the Debye–Scherrer equation, and instrumental broadening was determined using refined Caglioti parameters (U = 0.00231, V = 0.00272, and W = 0.00253) and subtracted from the measured FWHM values at each Bragg angle. The resulting average crystallite size was calculated to be (22.5 ± 1.4) nm, assuming a shape factor of 0.9.

This value is consistent with, and within experimental error of, our TEM observations in Figure 2, which shows an estimated mean crystallite size of 26.5 nm (Figure 2f), while the average particle size was 44.5 nm (Figure 2f,g). This further agrees with the average NP sizes observed in our previous work [7]. It is worth noting that TEM provides only an estimated value with this method and is more commonly applied to particle size distribution measurements, whilst XRD more typically provides a reliable calculation of crystallite size via Scherrer’s formula [49,50]. Despite this, the values obtained in Figure 1 and Figure 2 still agree, within error. The slightly smaller crystallite size value obtained from XRD in Figure 1 also reflects the diffracting domain size, which can be reduced by internal defects or grain boundary strain not visible in TEM (Figure 2).

TEM imaging in Figure 2 revealed that the Tm₂O₃ nanoparticles exhibit a rounded polygonal morphology (Figure 2c–e), with some particles displaying sharp edges, although most are smooth. On a larger scale (Figure 2a–c), the particles tend to form loosely aggregated clusters, extending up to the micrometre range.

EDS mapping in Figure 2h–j demonstrated a homogeneous signal distribution of thulium and oxygen across the particle. The atomic ratio of O to Tm was measured to be 1.42, which approximates the theoretical value of 1.5 for TmNPs (which are chemically represented by the formula Tm₂O₃). Given the sample size and measurement limitations, this value falls within an acceptable range. In conjunction with the TEM analysis in Figure 2a–g and XRD in Figure 1, this result in Figure 2h–j verifies the 100% purity of the TmNPs used throughout this study. This is important given the NPs are prepared in solution and sonicated for 40 min for all in vitro work in Figures 3–8 as well (albeit in water-based DPBS rather than ethanol), notably given the TmNPs have been found in Figure 1 and Figure 2 to be at the optimal size for NP uptake into tumour cells, well within the 30–50 nm range [36].

3.2. Cell Viability Indicates Short-Term Toxicity Is Selective for Cancer Cells

A preliminary assessment in vitro of short-term effects of uptaken TmNPs via MTT assay is shown in Figure 3. TmNPs are observed to be selectively toxic for 9LGS cancer cells regardless of the concentration of NPs used. For the 24 h exposure, 9LGS cell populations are left more than 50% non-viable whilst non-cancerous MDCK is generally unaffected by the presence of the TmNPs.

The TmNPs concentrations studied in the cell populations in Figure 3 are not observed to induce a significant difference in cell viability. Increasing concentration does not induce further toxicity in 9LGS nor is any threshold for MDCK toxicity observed. It is clearly indicated that TmNPs selectively impact short-term cell viability in cancer cells over non-cancerous cells. This is observed as potentially being due to selective NP uptake in Figure 4.

3.3. TmNPs Selectively Uptake into Cancer Cells over Non-Cancerous Cells

Confocal microscopy is used in Figure 4 to bring to light preferential uptake of TmNPs by 9LGS cancer cells over non-cancerous MDCK. The light scatter of the heavy metal NPs allows the size and locations of the NPs to be revealed, indicating some of their internalisation mechanisms.

Figure 4a demonstrates many TmNPs internalised within the cytoplasm of 9LGS cells. MDCK cells have noticeably less NPs present within the cells, highlighting the selective uptake of the NPs for cancer cells. This correlates well with the short-term viability results in Figure 3, where only the 9LGS cells are significantly affected by TmNPs, reinforcing our assumption of cell selectivity. Additionally, some examples of fragmented nuclear DNA and apoptotic bodies are visible in Figure 4a, suggesting some potential cell death via the apoptosis pathway (given these observations are hallmarks of apoptosis) [16].

Figure 4b shows significant vacuolisation within 9LGS cells, where a large number of vesicles contain TmNPs clusters in the attempt to process the NPs. The rounded and relatively smooth polygonal morphology of these NP clusters correlates with Figure 2, where the TmNP clusters exhibit some of the ideal NP shape and surface characteristics associated with increased cellular uptake [51,52,53,54,55]. The excessive fractionation and vacuolisation of the cytoplasm, and consequent lysosomal degradation, are also a key features of endocytosis, which has been observed to be a common uptake pathway for NPs [35,55,56,57]. Additionally, cytoplasmic vacuolisation has previously been observed to contribute to cell death [58], which may explain the selective toxicity observed in Figure 3.

This attempted metabolic processing of TmNPs was also noticeable in Figure 4c, where NPs are observed to replicate the “shell effect” previously observed by Brown et al. with tantalum oxide NPs [38]. However, TmNPs are still significantly spread throughout the 9LGS cell cytoplasm, resulting in only ‘partial’ shells forming in Figure 4c, the opposite of what transpires with MDCK cells.

TmNPs do not appear to internalize within non-cancerous MDCK at all. Using a 3D z-stack via confocal microscopy, Figure 5b indicates that the NPs are above or below the cells rather than inside, and at most attached to the membrane exterior. This is also shown in Figure 4 and further highlights the selective uptake of the NPs for cancerous 9LGS cells.

Figure 5 uses a methodical image analysis via ImageJ to quantify the mass of NPs uptake per cell and the 3D locations of the NPs around cell nuclei, before verifying uptake with flow cytometry. Figure 5a quantifies the images in Figure 4a demonstrating a significant difference between the mass of TmNPs uptaken in 9LGS and MDCK. Less NPs are internalised within MDCK, which Figure 4c reveals to be potentially attached to the cell exterior.

Figure 5b demonstrates TmNPs in 9LGS cells are located at the same location and plane in 3D as the cell nuclei, whilst there is a significant difference between NP locations around MDCK. Instead, the NPs are distinctly quantified on average as being above or below the MDCK nucleus, correlating with Figure 4c. Accordingly, TmNPs are larger in MDCK cells (Figure 5c) and much smaller in 9LGS, indicating potential breakdown following the attempted metabolic processing of NPs observed in Figure 4b, and only when they are internalised.

This further supports the selective uptake of TmNPs for cancerous 9LGS over non-cancerous MDCK cells. Flow cytometry verifies this using side scatter off the NPs in both cell lines, where Figure 5d shows a much higher signal for NP internalisation within 9LGS cells. Likewise, Figure 5e verifies that a larger proportion of the 9LGS population internalises NPs compared with MDCK. This is also verified over time for 9LGS (to see the effect of increasing NP exposure time beyond 24 h), where it is observed that the cancer cells continuously uptake an ever-greater quantity of NPs (Supplementary Figure S1).

Finally, where TmNPs are successfully internalised within the cancer cells, there is relatively uniform uptake observed across all cell cycle phases (Figure 5f). While there are also some minor yet significant increases in TmNPs-induced side scatter observed via flow cytometry in Figure 5f with MDCK, this is significantly less than the uptake observed for 9LGS and more likely correlates with observations of some minor TmNPs potentially adhered to MDCK cell exteriors in Figure 4 and Figure 5a–c.

3.4. TmNPs Selectively Shift Cell Cycle and Induce Cell Death via Necrosis in Cancer Cells

A cell cycle analysis via flow cytometry in Figure 6 re-confirms the selective uptake of NPs, such that only 9LGS cells have a noticeable phase shift. The TmNPs are observed to increase the proportion of 9LGS cells in the G1 phase in Figure 6a, doing so at the expense of both S and G2/M. This is particularly significant for G1 and S (with p < 0.001 for each shift). This indicates potential cell cycle arrest at the G1-S checkpoint, which NPs have been previously shown to induce, notably following oxidative stress, or DNA or microtubule damage [59].

By comparison, some studies have previously observed NPs resulting in G2/M cell cycle arrest, with the largest concentration of NP uptake in the G2/M phase (followed by S) [60,61]. Internalised NPs have also been previously noted to carry over from parent cells to daughter cells in mitosis [60]. This indicated that NPs may gather in all cell cycle phases, correlating with Figure 5f, yet also result in G1 population increases observable in Figure 6a, notably if G1-S checkpoint arrest occurs.

By contrast, TmNPs demonstrate no significant impact on cell cycle distribution for MDCK in Figure 6b (with p > 0.35 or higher when comparing shifts in all phases), thereby correlating with TmNPs selectivity for 9LGS. MDCK also present a significant Sub-G1 pre-apoptotic peak, even without NP exposure. This Sub-G1 peak is not present in 9LGS, and further investigation into this result is beyond the scope of this study.

However, indications of pre-apoptosis are further demonstrated in the short-term via microscopy and flow cytometry. Figure 6c demonstrates cell damage via PI staining and membrane rupture following TmNPs internalisation within 9LGS cell cytoplasm. This is not observed in MDCK populations when imaged.

Annexin V flow cytometry then permits quantification of enhanced cell death following TmNPs exposure. While MDCK cells present a minor increase in early apoptosis in Figure 6d, this correlates with the natural pre-apoptosis observable even in untreated MDCK populations in Figure 6b via cell cycle analysis. Despite the early apoptosis and necrosis increases with TmNPs in Figure 6d, this is still reinforced by the minor statistical significance and the large error bars. This may also be linked to observations of NP adherence to MDCK cell exteriors in Figure 4 and Figure 5. However, it is also worth noting that MDCK has an intact p53 tumour suppressor gene compared with the p53-mutant 9LGS, [62,63,64,65,66] which may leave MDCK more predisposed to an apoptotic response compared with 9LGS.

Accordingly, 9LGS cells did present a significant increase in the number of cells undergoing necrosis following TmNP treatment (and also some late apoptosis, correlating with some of the observed apoptosis hallmarks [16] in Figure 4 and Figure 6). This differential response of the cancer cells (compared with non-cancerous MDCK) further correlates with the cell swelling and membrane ruptures observable in Figure 6c, which are in turn signs of the necrosis pathway (among others) [16,17].

It is then clear that TmNPs induce some cytotoxicity in the cancer cells selectively, likely resulting in necrotic cell death. By contrast, non-cancerous MDCK cells are relatively unphased by the NPs, likely due to the selectivity observed in Figure 3, Figure 4, Figure 5 and Figure 6. These cancer-selective effects of the NPs are further demonstrated over time via live cell imaging in Figure 7.

3.5. TmNPs Inhibit Medium-Term Cell Growth and Recovery After Treatment in Cancer Cells Only

Figure 7a further highlights the selectivity of TmNPs as well as the medium-term efficacy of the treatment. 9LGS cells are observed to grow significantly less over time when treated with TmNPs (and less in general with or without NPs compared with MDCK, as 9LGS has a slower doubling time). By contrast, MDCK cells grow to confluence over the medium-term, regardless of treatment, which we propose to be linked to the TmNPs’ uptake selectivity observed in Figure 4 and Figure 5. This is quantified in Figure 7c, where 9LGS cells have 2.1 times slower growth due to the toxic effects (Figure 3 and Figure 6) of the TmNPs on the cancer line.

Figure 7b demonstrates that the reduced 9LGS growth and recovery over time following TmNPs’ treatment results from NP toxicity and resulting cell death. The significantly increased PI signal from 9LGS cell populations over the medium-term correlate with Figure 6c–d, highlighting that TmNPs are killing some of the cancer cells over time. By contrast, MDCK is unaffected, as quantified in Figure 7d (where TmNPs result in 2.9 times more PI damage indicators for 9LGS cells over the medium term). This indicates that cell death is preferentially occurring in 9LGS cells over time due to TmNPs.

3.6. Initial DNA Damage Results in Long-Term Cell Death Following TmNP Exposure

To determine the cause of 9LGS cell death following TmNP exposure, double-strand DNA damage is assessed and quantified using high-resolution γH2AX immunofluorescent imaging (Figure 8a–b). Following this, long-term efficacy is assessed via clonogenic assay to determine cell survival as the long-term endpoint, for both 9LGS and MDCK in Figure 8c.

Figure 8a demonstrates an observable increase in the number of γH2AX foci (representing DSBs) in both 9LGS and MDCK. This takes place after the immediate 20 min following 24 h of TmNPs exposure. This represents the short-term damage caused by the treatment after 24 h. Figure 8b confirms that TmNPs induce three times more DSBs in treated cells compared with the untreated control.

This damage is observed for both cell lines, such that each see a reduction in cell survival over time. Figure 8c demonstrates a minor reduction in MDCK survival with TmNPs relative to the control, indicating some 24 h of exposure does produce some minor toxicity to the non-cancerous cells. This does further correlate with the possible NP adhesion to MDCK cell exteriors (Figure 4c). Sub G1 was observed via cell cycle analysis (Figure 6b), and some early apoptosis in MDCK (Figure 6d) was observed following TmNP exposure.

However, a larger and more significant killing of 9LGS cancer cells is observed in Figure 8c, with long-term survival reduced by a third. This greater killing of 9LGS cells correlates with the TmNP selectivity for the cancer cells observed throughout this study. Although, it is worth noting that p53-mutant 9LGS may also exhibit a differential response compared with p53-intact MDCK cells [62,63,64,65,66]. This may also explain why MDCK killing in Figure 8c is only minor compared to its control, where simple NP exposure, despite cancer-selectivity, is sufficient to induce an effect in the more sensitive MDCK cells (hence the minor early apoptosis suggested by Figure 6d) and seemingly within error of more resistant 9LGS results. Despite this, the MDCK cell killing is still non-significant and clearly recovers over time (Figure 7a and Figure 8c), while TmNP-treated 9LGS does not (Figure 7) and demonstrates a significant killing all its own (Figure 8).

Accordingly, we propose that the selective cancer cell killing observed throughout this study still demonstrates TmNPs to be a potentially viable candidate for the cancer-selective treatment of 9LGS whilst limiting damage to non-cancerous MDCK. Moreover, the data collected in this study permits a theory to explain the TmNP toxicity within the cancer cells, allowing a complete discussion of the mechanisms of toxicity for TmNPs for the first time.

4. Discussion

The results of the study demonstrate minor-moderate toxicity for TmNPs at the tested concentrations but selectivity for 9LGS cancer cells over non-cancerous MDCK. This is notably true in the short-term via the MTT assay (Figure 3), likely due to the TmNPs having relatively smooth and rounded polygonal morphology with a mean particle size near 45 nm (Figure 1 and Figure 2), well within the optimal 30–50 nm range and consistent with our previous work [7,36].

An assessment of TmNP uptake mechanisms followed, which verifies preferential uptake of the NPs into 9LGS. By contrast, MDCK does not appear to uptake the TmNPs, with the NPs rather sticking to the exterior (Figure 4 and Figure 5). The “shell effect” previously observed by Brown et al. with tantalum oxide NPs is also observed for TmNPs [38], although only partial shell structures appear to form (Figure 4), further indicating potential implications for TmNPs use as a high-Z radiosensitiser [7]. This was verified using flow cytometry and image quantification of confocal images, which further reveal that NPs are selectively being internalised and processed by 9LGS (Figure 4, Figure 5 and Figure 6). Accordingly, while TmNP uptake is relatively uniform across all cell cycle phases (Figure 6), NP uptake does increase steadily over time with increasing treatment exposure time for 9LGS cancer cells (Supplementary Figure S1).

TmNPs are also observed to induce minor shift to the G1 phase of the cell cycle after 24 h of NP exposure and uptake (Figure 6). Accordingly, internalisation and metabolic processing (Figure 4) of TmNPs eventually results in cytotoxicity in 9LGS cells, culminating primarily in necrosis, and some late apoptosis being triggered (Figure 6). This is notably true in 9LGS cells, which demonstrate rather excessive uptake of NPs and significant vacuolisation as a result (Figure 4 and Figure 6). Over the medium-term, cell death is observed via these pathways in Figure 7 (resulting in leaked DNA content from damaged cells identified via PI staining), culminating in reduced 9LGS cell growth over time (Figure 8).

This is assessed to be further triggered by DSB DNA damage following the 24 h of TmNPs exposure (Figure 8). While MDCK suffers some DNA damage, it is largely able to recover and only suffers minor long-term damage in the end (Figure 8), attributable to the tendency of this more sensitive, non-cancerous cell line to enter the apoptotic pathway. This is correlated to the increased apoptosis observed in Figure 6 and Figure 7 with MDCK and may be related to cell cycle regulation in the more treatment-resistant 9LGS cancer cells [62,63,64,65,66,67].

The p53 tumour suppressor gene is mutated in 9LGS but remains intact in MDCK. This gene may affect the results due to its role in regulating the G1 cell cycle phase predominantly (in which Figure 6 revealed 9LGS cells with TmNPs gathered), including G1 arrest and G2-S checkpoint progress [62,63,64,65,66]. The p53 mutation also suggests that 9LGS could continue with cell cycle progression due to mis-regulation, notably given its high mitotic index as a rapidly cycling cancer [62,63,64]. The p53 gene also plays a role in triggering cell death pathways such as apoptosis in place of cell cycle arrest [64,65,66]. When arrest is chosen, it may be because the cell is attempting to repair damage (which can result in γH2AX signalling [46,47], as observed in Figure 8 for both 9LGS and MDCK). When this cannot occur, cell death (or cellular senescence) may take place instead [64,67].

Cells may then need to manage the metabolic processing of internalised TmNPs (Figure 4) when progressing to S phase (Figure 6), resulting in DNA damage signalling (Figure 8) and G1-S check-point arrest (via p53) that may see cells gather in G1 (Figure 6) [64,65,66]. This makes sense given that cells uptake NPs predominantly via the endocytosis pathway (Figure 4) [55], which would be more active in cell growth phases of the cycle like G1 (Figure 6). For 9LGS, a mutant p53 may then prevent early apoptosis from occurring after some form of TmNPs-induced damage following uptake (Figure 6) [62]. This may result in potential oncogenesis [65,66], possibly by lysosomal degradation that NP exposure has been observed to induce [57] or 9LGS cells continuing to progress through the cell cycle without significant phase distribution shifts (Figure 6). By contrast, MDCK may engage apoptosis normally with its intact p53, resulting in the substantial Sub-G1 pre-apoptotic cell populations (Figure 6). Accordingly, this may also explain why 9LGS does not demonstrate any detectable Sub-G1 peak in Figure 6.

G1 cell cycle arrest may also prevent cells from progressing to the other phases [65,66]. This can potentially leave viable cells for NP uptake only in the G1 phase, resulting in an increased G1 population following uptake of TmNPs in 9LGS (Figure 6). These also may have been 9LGS cells which internalised NPs later in the 24 h treatment exposure period, whilst cells taking up NPs earlier had already arrested in their cell cycle. This allows increased G1 gathering relative to other cells (Figure 6), although not necessarily stopping all cells from progressing to other phases. This may be because only 24 h had passed, and a full cell cycle for the 9LGS cell line used was greater than this, hence resulting in Figure 5 showing uniform uptake across phases. A future cell cycle analysis of TmNP effects over time would be needed to verify this.

However, this does provide some explanation as to why MDCK shows some initial DNA damage (Figure 8) in the presence of TmNPs outside the cell (Figure 4), despite selective TmNP uptake for 9LGS (Figure 4 and Figure 5). Accordingly, the intact genes and pathways in MDCK cells may also be the reason why the non-cancerous cells recover over time (Figure 7) and only suffer a minor and non-significant long-term survival reduction (Figure 8). This may be the result of differential cell line responses from p53-intact MDCK [62,63,64,65,66]. The more sensitive, non-cancerous model may simply be more predisposed to undergo natural apoptosis (Figure 6), elevated slightly—but not significantly—by the presence of a possible irritant like the TmNPs observed to potentially attach externally rather than uptake into the cells (Figure 4 and Figure 5). By contrast, p53-mutant 9LGS faces some notable toxic effects following uptake and internalisation within 9LGS cells, which impacts significantly on both medium-term and long-term survival (Figure 7 and Figure 8). This further correlates with previous findings by Engels et al. [7].

A theory for the mechanisms underpinning these observations is now presented to explain the results of this study, and by extension, the toxicity of TmNPs. To begin with, Figure 4 indicated possible endocytosis as the cellular uptake pathway [55]. Endocytic vacuoles are visible (Figure 4), each containing NP clusters for metabolic processing within defined organelles, highlighting endocytosis to be the mechanism of cellular uptake of NPs within 9LGS cells. As this has been previously demonstrated in NP other studies [35,56], we conclude this as the likely uptake pathway for TmNPs as well in this study via Figure 4.

Additionally, the results of this study appear to show significant and continuous uptake of TmNPs (Figure 4) in 9LGS (Figure 5) over time (Supplementary Figure S1), even to the point that the cell death pathways are triggered (Figure 6). As apoptosis presents morphological changes in cells, including tightly packed organelles and membrane blebbing [16,17], this explains the many tightly packed vesicles in Figure 4 and Figure 6. However, necrosis also presents cytoplasmic blebbing yet contrasts with apoptosis via observations of swollen cells, formation of cytoplasmic vacuoles, disrupted organelle membranes, and even disrupted cell membranes, which can result in the release of internal contents to the extracellular environment [16,17].

Figure 4 and Figure 6 also demonstrate possible apoptosis [16], as indicated by nuclear DNA fragmentation and apoptotic bodies, as well as the tightly packed vacuoles filled with NP clusters. However, necrosis is more significantly demonstrated throughout the 9LGS population (Figure 6) by large cell swelling, and possible membrane and vacuole damage (indicated by the (red) PI bursts from a digestive vacuole with NPs (Figure 6c) due to possible provocation of vacuolic lysis) [16,17]. This also correlates with Figure 7 where TmNP-treated 9LGS cells show significant PI signalling over time. It is possible that the excessive uptake of NPs observed in Figure 4, and over time in Supplementary Figure S1, may cause severe stress in the cancer cells that can in turn trigger cell death pathways [17,58,68]. These observations of metabolic stress explain the short, medium, and long-term effects observed in Figure 3, Figure 7, and Figure 8.

Moreover, larger NP cluster sizes observed with MDCK may play a role. Figure 5 demonstrates NP size (as measured in terms of NP cluster radius, surface, and volume) to be notably smaller in 9LGS than MDCK, indicating that likely metabolic processing of NPs is selectively occurring within 9LGS (Figure 4, Figure 5 and Figure 6). This may be explained as the smaller NP clusters within the cancer cells resulting from breakdown through metabolic processing in the endocytic vacuoles over the 24 h exposure period, which is not observed in MDCK (Figure 4). This may also be interpreted as smaller TmNPs being more likely to be internalised within 9LGS, notably compared to the larger NP clusters that might attach to the exterior of MDCK cell membranes (Figure 4 and Figure 5). This suggests that TmNPs, being smaller than other oxide NP counterparts such as tantalum oxide and in the most optimal size range (30–50 nm) for uptake [7,36,38], may have a greater potential for cellular uptake and effect due to the smaller size of the NPs.

This is also indicated in Figure 1 and Figure 2, where the mean TmNP particle size is found to be approximately 44.5 nm and 100% pure Tm₂O₃ under these specific experimental conditions used in this study. The lack of impurity (Figure 2) suggests that all in vitro effects in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 are likely due to the intrinsic nature of the TmNPs and not any other factors, whilst the optimal size may explain the apparent ease of internalisation into 9LGS cells (Figure 4, Figure 5 and Figure 6).

Figure 2 also revealed that most of the TmNPs had smooth surfaces, with only some sharp edges observed on occasion. Some studies have noted the importance of surface roughness, where smoother NPs have been found to uptake more easily, and greater quantities, into cancer cells than rougher particles [51,52]. Other studies have recognised that NP shape plays a significant role [53,54,55]. NPs with round shapes have been found to uptake more easily, whilst some particles with sharper angles and protrusions have been observed to more easily adhere to cell membranes to promote internalisation [53,54,55]. This may explain the TmNPs uptaken in Figure 4, where round NP clusters are observable in NPs being processed in endocytic vacuoles. Figure 2 also shows some NPs with sharper angular protrusions (although most were round and smooth).

This may also provide some explanation as to why 9LGS cells had preferential uptake (Figure 5) whilst MDCK cells only appeared to have NPs adhered to the cell exterior (Figure 4). Some of the rounded NPs with sharper protrusions (Figure 2) may have simply attached to the MDCK cell membrane (Figure 4 and Figure 5). Additionally, mammalian epithelial cells like MDCK have been observed to continuously transport NPs out of the cell via exocytosis, with smaller NPs more likely to be removed [69,70]. This may also explain why Figure 5 demonstrated larger NPs attached to MDCK cells, as the smaller NPs may have already been removed. As exosome secretion may also be regulated by the p53 gene [71], it is possible that p53-mutant, fast-cycling 9LGS may be less likely to remove NPs, whilst p53-intact MDCK may be more predisposed to NP removal (Figure 4, Figure 5 and Figure 6 and Supplementary Figure S1). Ultimately, future study is needed to empirically verify potential differential NP uptake mechanisms and cell transport pathways [53,55,72,73]. Nonetheless, this study still found a clear preferential uptake of TmNPs into 9LGS cancer cells over MDCK (Figure 4, Figure 5 and Figure 6) and selective toxic effects (Figure 3, Figure 7 and Figure 8), likely due to the favourable TmNP size and properties observed in Figure 1 and Figure 2.

Indeed, Haume et al. further emphasise the importance of NP size, both for radiosensitisation potential and toxicity purposes. This is especially true of gold NPs, where conflicting studies have reported that inert and non-reactive nature of gold renders such NPs non-toxic, whilst other studies have observed some toxicity depending on the NP size and tissues affected [26,27,29,74,75]. Some studies have indicated that reactive oxygen species (ROS) activity can induce toxicity through oxidative stress, which in turn can produce DSBs in 9LGS cells (Figure 8) [27,30,76]. Khochaiche et al. observed this with silver-doped lanthanum manganite NPs but also observed this was preferential for 9LGS cancer cells over MDCK [76], correlating with the selectivity of TmNP toxicity in this study in vitro (Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8). ROS assessments were also conducted in this study for TmNPs, but no significant changes were observed. By contrast, NPs such as gadolinium have been observed not to induce DSBs, and some studies have concluded these NPs (without radiation) as non-toxic to tumour cells [9,75,77]. Ultimately, further studies are required across NP types to better understand and control the toxic mechanisms and biocompatibility of NPs prior to irradiation, with this study offering a possible novel explanation for the case of TmNPs (Figure 9).

Overall, these findings explain all the results across this study, where a significant number of TmNPs with optimal shapes and sizes continuously internalise in 9LGS over time through the endocytosis pathway (Figure 1, Figure 2 and Figure 4 and Supplementary Figure S1) [36]. Extreme levels of vacuolisation follow, selectively within 9LGS cancer cells, to metabolise so many NPs that it can eventually result in cell death, primarily via necrosis (Figure 3, Figure 4, Figure 5 and Figure 6) [17,58]. 9LGS cells then die over time due to necrosis induced by this TmNP cytotoxicity (Figure 3, Figure 6, Figure 7 and Figure 8). This theory is presented in Figure 9 as a novel proposal for the underlying mechanism behind TmNP toxicity.

Despite some limitations (no human cell model, p53 analysis, lysosomal studies, NP dosage study, oxidative stress testing, cellular radiosensitisation, or in vivo testing or NP clearance), the results provided in Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 still use a significant volume of data to comprehensively underpin our theory proposed in Figure 9. Additionally, many of these limitations have been thoroughly investigated in previous studies for both TmNPs and other high-Z NP materials, including successful TmNP radiosensitisation, material and uptake analysis, physiochemical characterisation, and TmNP clearance in vivo from 9LGS-tumour-bearing rats [7,9,28,38,61,74,76,78,79,80]. Future studies may still be required on human cell models or in vivo pre-clinical trials or further physiochemical or biological analysis of cell death mechanisms or p53 gene expression. Nonetheless, our results comprehensively demonstrate the selective impact of TmNPs on short, medium, and long-term efficacy and preferential treatment of cancerous 9LGS cells over healthy MDCK.

5. Conclusions

This study successfully assessed and affirms the biocompatibility of TmNPs as a potential high-Z radiosensitiser candidate. The results of this work demonstrate TmNPs to be selective in uptake for cancerous 9LGS cells over non-cancerous MDCK cells. This is demonstrated repeatedly throughout this study, not only in the more significant toxic effects observed preferentially in 9LGS, but through multiple methods of uptake quantification. Indeed, while TmNPs are not observed to harm non-cancerous MDCK cells, some notable cytotoxic effects are observed in 9LGS cells, resulting in some moderate cancer cell death in the long-term once internalised selectively into the cancer cell line. Through a thorough examination of the underlying biological mechanisms, a viable theory has been formulated that 9LGS cells continuously uptake TmNPs excessively over the time that they are exposed to this treatment. The excessive uptake eventually results in necrosis over the long-term, killing the cancer cells whilst leaving the normal tissue model relatively unharmed. This study therefore provides not only a valuable assessment of the safe use of TmNPs as a high-Z candidate for potential use as a cancer-selective radiosensitiser, but also an explanation of the underlying biological mechanisms behind the selective toxicity of TmNPs. As such, we believe TmNPs to be a viable candidate for continued study of NP radiosensitisers to enhance cancer radiotherapy in future.