ESTABLISHMENT AND CHARACTERIZATION OF PRIMARY tumor cellS FROM LUNG ADENOCARCINOMA CANCER PATIENTS
Low passage primary cultures of tumor cells represent an excellent ex vivo model for translational research because they are believed to maintain the same features of resident cancer cells present in tumor lesions. Primary cell cultures of lung adenocarcinoma were obtained by isolating tumor cells from MPEs of patients, according to a specific isolation protocol previously described (8, 9). In order to characterize primary cell cultures, they were subjected both to phenotypic and genotypic investigations. For the immunophenotypic characterization of the primary cultures, the expression levels of some specific markers were analysed by flow cytometry. In particular: Epithelial Cell Adhesion Molecule (EpCAM), expressed in epithelial cells and correlated with tumorigenesis (17); Cytokeratin 7 (CK7), normally produced by epithelial cells in the lung and commonly used as marker for diagnosis of lung tumors (18); CD90 (alias Thymus cell antigen 1), a well-known marker of fibroblast (19, 20), and CD45, a ubiquitous cell surface marker of nucleated hematopoietic cells (21). The LUAD primary cell cultures isolated exhibited a marked positivity for EpCAM and CK7/8 markers, assuring that they were constituted by cancer cells of epithelial origin with the same phenotypic features of the original tumour, as well as negligible expression of CD90 and CD45 markers, thus excluding the presence of fibroblasts and tumor infiltrating immune cells (Table 1). Mutational profiling was performed by Next Generation Sequencing (NGS), enabling the simultaneous detection of variants in 52 genes relevant for solid tumors. For this study we selected four primary cultures (Table 1), all harboring mutations in KRAS gene, accompanied or not by alterations in other genes, such as Keap1 and TP53, known to be associated with ferroptosis protection (14, 22).
lung adenocarcinoma CSCs ARE resistant to ferroptosis
Given the role of CSCs in cancer biology and resistance to therapy, targeting CSCs metabolism may serve as an important approach for future cancer treatments. Moreover, since lipid metabolism undergoes significant changes in CSCs (8, 11, 23), we investigated whether ferroptosis, a type of cell death strictly dependent upon the level of lipid peroxidation, is somehow impaired in CSCs and is associated with drug resistance in the 3D culture model. The four primary cell cultures selected for the study were thus grown in a serum-free medium supplemented with growth factors and cultured in non-adherent conditions as three-dimensional spheroids to enrich for CSCs; in parallel, they were grown in adherence conditions. Figure 1A shows the morphological aspect of the four primary cultures grown in 2D and 3D contexts. In order to verify the enrichment of CSCs in 3D cultures, we evaluated by Real Time PCR the expression of stem markers such as nanog, oct4, and sox2 (Fig. 1B). Although not all markers show statistically significant up-regulation in the spheroids, a general pattern of overexpression is observed. To evaluate the sensitivity of the two culture systems to ferroptotic cell death, we treated the four lung cancer primary cultures (BBIRE-T248, PUC30, PUC36 and PUC37) with RSL3, a ferroptosis activator with selectivity for tumor cells bearing oncogenic RAS, which acts by inhibiting the glutathione peroxidase 4. The data clearly show that cells grown in adherence were sensitive to the treatment with RSL3, while a high degree of resistance to this agent of CSCs enriched cultures was observed (Fig. 1C). Based on the evidence that STK11 and KEAP1 co-mutations cooperate to promote protection from ferroptosis, independent from the KRAS status (13), we also performed viability assays following RSL3 treatment in the NCI-H460 stable cell line that, having triple mutations in KRAS-STK11-KEAP1 genes, could represent a positive control of ferroptosis resistance. As shown in Fig. 1D, NCI-H460 cell line confirmed to be highly resistant to RSL3, both in 2D and in 3D conditions. To note, the baseline sensitivity to RSL3 (relative to growth in adhesion) varied greatly across cancer cell lines, probably due to their mutational status (KRAS, TP53 and KEAP1 mutations). In particular, PUC37 cell culture, that harbors KRAS-KEAP1-TP53 mutations, has a similar sensitivity to NCI-H460 cell line, while PUC30 cell culture, that harbors only KRAS mutation, is similar to BBIRE-T248. Finally, the dose-response curve of PUC36 is in the middle of the graph, probably due to the additional mutation in TP53 (Fig. 1E). This led us to hypothesize that resistance to ferroptosis in lung CSCs may be linked to dynamic molecular and epigenetic changes occurring in cells when grown in 3D conditions. To verify this hypothesis, we disrupted the 3D spheroids of the BBIRE-T248 cell culture and grew cells in adherence for 6 days to then re-evaluate their sensitivity to RSL3 (Fig. 1F). 3D disrupted cells showed a sensitivity similar to 2D cells, supporting the concept of CSC plasticity in which cancer cells harbor the dynamic capacity to switch from a non-CSC state to a CSC state and then back to non-CSC, in parallel with their susceptibility to ferroptotic cell death (Fig. 1G).
LUNG ADENOCARCINOMA CSCs resist to RSL3-induced ferroptotic CELL death by COUNTERACTIng lipid and mitochondrial ROS production
To understand the effect of RSL3 on the cell cultures described in Fig. 1, we investigated the phenomenon of lipid peroxidation, i.e. the core reaction of ferroptosis, in cells treated for 2 hours with RSL3 and grown in 2D and in 3D conditions respectively for 72 hours. Lipid peroxidation is caused by free radicals that attack unsaturated fatty acids of membranes producing lipid peroxides; in turn, lipid peroxides, as highly reactive compounds, propagate further generation of ROS and alter the assembly, composition, structure, and dynamics of lipid membranes, leading to membrane rupture and cell death (5). We performed a flow cytometric analysis using C11-BODIPY (581/591) as reporter of lipid peroxidation and assessed the levels of lipid peroxides in samples treated with 1 µM RSL3 alone or in combination with 2 µM of Ferrostatin-1, a ferroptosis inhibitor with antioxidant capacity (Fig. 2A). In the same experimental set-up, we also evaluated the percentage of dead cells (Fig. 2A). In Fig. 2B are reported the density curves of lipid peroxidation of a representative experiment for each cell line, while the histograms in Fig. 2C and Fig. 2F show the results of three independent experiments of measure of lipid peroxidation and cell death respectively. It is evident that 2D cell lines treated with RSL3 show increased levels of lipid peroxidation compared to the control but to a different extent, in line with their different sensitivity to cell death shown in the previous paragraph. Moreover, lipid peroxidation was rescued, at least in part, in cells co-treated with the antioxidant Ferrostatin-1, that acts via a reductive mechanism to prevent damage to membrane lipids, thereby inhibiting lipid peroxidation and cell death. Interestingly, the same cells grown in 3D conditions were completely resistant to lipid peroxidation, in perfect agreement with lack of sensitivity to cell death induced by RSL3 (Fig. 1 and Fig. 2F).
Another class of ROS involved in ferroptosis is represented by the mitochondrial reactive oxygen species (24). In order to investigate the oxidative stress by mitochondrial superoxides, we decided to performed flow cytometric analyses using the MitoSOX Red reagent, a fluorogenic dye specifically targeting mitochondria in live cells. In particular, oxidation of the MitoSOX Red reagent by mitochondrial superoxide produces red fluorescence indicative of cellular oxidative stress. We set up the experimental conditions culturing all cell lines previously investigated in 2D and in 3D conditions and assessed the percentage of mitochondrial superoxides after 2 hours of treatment with 1 µM RSL3 alone or in combination with 2 µM Fer-1 (Fig. 2A). As reported in density curves of representative experiments in Fig. 2D, we observed that RSL3 treatment causes mitochondrial superoxide production in all 2D samples except for NCI-H460 cell line. Interestingly, at the basal level, cells grown in 3D conditions have a higher level of mitochondrial oxidative stress than cells grown in 2D conditions but treatment with RSL3 does not produce further mitochondrial superoxides production (Fig. 2E). For each cell line, three biological replicates were performed and the percentages of positive cells for MitoSOX Red dye are reported in the histograms in Fig. 2E. In all cases, the co-treatment with Ferrostatin-1 rescues the percentage of mitochondrial superoxides. As a consequence, these data suggest that mitochondria are crucial players in ferroptosis and that in our cells the level of ferroptosis corelates with the degree of mitochondrial disfunction.
lung adenocarcinoma CSCs show an increased expression of a set antioxidant genes
Given the importance of antioxidant defence systems in resistance of cancer stem cells to ferroptotic cell death, we investigated at the molecular level the mechanisms implemented by 3D cells to protect themselves from stress. Therefore, we carried out a molecular characterization of primary cell cultures, by evaluating the expression of antioxidant genes, namely NRF2, HO-1, NQO1 and AKR1C2, and genes related to glutathione metabolism such as SLC7A11, GCLC and GPX4. In Fig. 3A are reported the graphs of the relative mRNA expression as fold changes observed in the primary cell lines in the transition from 2D to 3D growth condition. We noticed that each cell line grown in 3D conditions showed a statistically significant over-expression of genes involved in antioxidant defence systems but with a different pattern of increased genes. Notably, this up-regulation trend was particularly pronounced in BBIRE-T248 and PUC30 cell lines, which harbor only mutations in KRAS gene (Table 1) and show the greatest difference in terms of sensitivity to RSL3 between 2D and 3D (Fig. 1C). In contrast, in PUC36 e PUC37 cell lines the fold changes of antioxidant genes analysed were less marked. However, it is evident, in general, that 3D cultures trigger a mechanism to detoxify cells and to escape from cell death.
Glutathione peroxidase 4 (GPX4) is a core enzyme that regulates lipid peroxidation and therefore plays a critical role in maintaining lipid homeostasis. GPX4 is a selenoprotein with a dual role: converts its cofactor GSH (glutathione reduced form) to GSSG (glutathione disulfide oxidized form) and reduces phospholipid hydroperoxides (PLOOHs) to their corresponding alcohol (PLOHs) to avoid cell damage (25). Since GPX4 is a key regulator of ferroptosis, we investigated, by western blotting, the expression levels of this enzyme in the four primary cell cultures analysed in this study. As shown in Fig. 3B, a marked overexpression of GPX4 is common to all primary cell lines, although mRNA levels did not show consistent modulations (Fig. 3A). We then investigated the intracellular content of GSH and GSSG through a luminescence assay and we observed that the levels of GSH decreased in all cell lines in the passage from 2D to 3D grown condition, while the GSSG content showed an opposite trend (Fig. 3C). We hypothesize that in 3D grown condition, the decreased levels of GSH are due to high expression of GPX4 that use it as cofactor in enzymatic reaction, converting it into GSSG.
To understand the molecular mechanisms involved in the induction of ferroptosis in 2D vs 3D cells and to explore the role of GPX4 in this context, we treated BBIRE-T248 cell line with 1 µM RSL3 alone or in combination with 2 µM Ferrostatin-1 for 2 hours and evaluated the expression of GPX4. Results obtained by western blotting confirmed that GPX4 is strongly upregulated in 3D spheroids compared to the 2D cells, without substantial modulation following treatment. At a closer inspection, however, we observed that in the untreated cells, GPX4 is present with two isoforms: mitochondrial (the higher band) and cytosolic (the lower band). After treatment with RSL3, the cytosolic isoform seems to disappear, and we know that in 2D cells this is sufficient to induce ferroptotic death (Fig. 3D). The same phenomenon is observed in the 3D spheroids, where, however, the mitochondrial isoform of GPX4 seems to increase following RSL3 treatment, to compensate for the absence of the cytoplasmic isoform. This suggests that 3D cells, in addition to having basally higher levels of GPX4 compared to their 2D counterpart, which confers lower sensitivity to RSL3, undergo also GPX4 isoform switching, which further enhances the anti-ferroptotic effect.
lung adenocarcinoma CSCs MODULATE OTHER MOLECULAR PATHWAYS potentially involved in ferroptosis
Another mechanism to escape ferroptotic cell death is to maintain iron homeostasis in the cell (26). Storage, outflow or inflow of iron makes cells more or less susceptible to ferroptosis (27). For this reason, we assessed the different expression of iron metabolism-related genes in cells grown in 3D conditions versus 2D conditions by RT-PCR. In Fig. 4A is shown that the expression of ferritin heavy chain 1 (FTH1) and ferroportin-1 (also known as SLC40A1), responsible for iron storage and export respectively, increased in 3D condition. Similarly to the antioxidant genes, this up-regulation is more accentuated in BBIRE-T248 and PUC30, which show the greatest degree of sensitivity shift between 2D and 3D. Accordingly with these data, our analysis suggests that in 3D conditions potentially toxic effects of iron are under control. TRFC instead is responsible for transferrin bound iron (Tf-Fe3+) uptake concurring for cell iron accumulation and does not show significant differences in gene expression.
It is known that lipid metabolism plays a crucial role in CSCs, providing the energy needed for their survival (28, 29) and affording protective mechanisms against peroxidation. The production of lipid droplets, for example, keeps PUFAs away from lipid oxidative damage (30), while lipid desaturation, regulated by the enzyme stearoyl-CoA desaturase 1 (SCD1), decreases levels of PUFAs (16). On the contrary, the enzyme Acyl-CoA Synthetase Long Chain Family Member 4 (ACSL4), that catalyses the biosynthesis of polyunsaturated fatty acid-containing lipids, promotes the accumulation of lipid peroxidation products, leading to ferroptosis (31). We then evaluated the expression of lipid metabolism genes in our 2D and 3D cellular models by quantitative RT-PCR. SCD1 and FADS2 resulted to be over-expressed in BBIRE-T248, PUC30 and PUC37 cell lines when cultured in 3D conditions, while ACSL4 expression was quite stable in all cell cultures comparing the two growth conditions (Fig. 4B). This suggests that lipid metabolism plays an important role in controlling CSCs maintenance, making cells less sensitive to peroxidation and cell death.
In order to identify other possible mechanisms involved in the protection against ferroptosis in 3D cells, we relied on microscopy analyses. In particular, we performed an ultrastructural investigation of organelles and subcellular compartments by means of a Transmission Electron Microscope (TEM). The images in Fig. 4C represent BBIRE-T248 cells grown in 2D and 3D conditions, treated according to the previously described experimental scheme (Fig. 2A) and analysed by TEM. RSL3-treated 2D samples have clear characteristic mark of ferroptotic cell death. In particular, electron-bright nuclei (red asterisks) with damage to the outer membrane (red arrow number 1), endoplasmic reticulum stress (green asterisks) and mitochondria with flattened cristae are found (red arrow number 2). The phenomenon of ferroptosis in the 2D samples progressed to damage the cytoplasm as well (yellow asterisk). Numerous non-homogeneous lipid droplets are noted, probably because they are composed of different lipid species (red arrow number 3). Cells grown as spheroids and treated with RSL3 show evident activation of autophagy for the presence of many autophagosomes (red arrow number 4). Probably, massive activation of autophagy could be a defence mechanism of the cells against ferroptotic cell death. Mitochondria are in contact with the endoplasmic reticulum cisternae indicating intense cross-talk among these organelles that can imply different processes such as lipid synthesis and/or lipid metabolism, modulation of mitochondrial morphology (fission and fusion), endoplasmic reticulum (ER) stress, autophagy, and Ca2+ handling (red arrow number 5). In some places, the inner and outer membrane of the nucleus are separated (red arrow number 6). Finally, we noticed that in the internal part of the spheroid there is partial loss of cell-cell junctions (red arrow number 7) which is partially restored in the samples treated with Fer-1. The sample treated with Fer-1 maintains a high number of autophagosomes compared to the control (red arrow number 4).
We were able to appreciate this different sensitivity to ferroptosis activator also by looking at morphological changes of the cells through an inverted microscopy. As shown in Fig. 4D, BBIRE-T248 cells showed loss of adherent and occlusive junctions between cells after treatment when grown in 2D conditions. This phenomenon was reverted when cells were cotreated with antioxidant Fer-1 because they showed a similar level of attachment as in the control. Contrariwise, the same cells cultured in 3D conditions showed no morphological alterations compared to the control.
LUNG ADENOCARCINOMA CSCs overexpress GGT1, SEPW1 and MUC1 genes that POTENTIALLY confer ferroptosis protection
To finely characterize our established collection of patient-derived CSCs and identify common putative genes involved in driving the phenomenon of ferroptosis resistance, the four selected LUAD primary cell lines (BBIRE-T248, PUC30, PUC36, PUC37) were profiled by RNA-sequencing, comparing the 2D and 3D in vitro culture systems (Table S1-4). As shown in Fig. 5A, 48 genes were found to be commonly up-regulated in the four cell cultures grown in 3D condition compared to 2D counterpart, while the commonly down-regulated genes were 17. Interestingly, among the down-regulated genes, AXL, AJUBA, CYR61 and CTGF are downstream genes of YAP/TAZ, two important transcription factors that positively regulate ferroptosis (32). Among the common up-regulated genes in 3D samples, we identified three genes which could contribute to ferroptosis protection, namely SEPW1, MUC1 and GGT1. SEPW1 encodes for Selenoprotein W that plays a role as a glutathione-dependent antioxidant and may be involved in redox-related processes (33); Mucin 1 (MUC1), instead, might inhibit ferroptosis through Keap1-Nrf2-GPX4 pathway (34) and, finally, Gamma-Glutamyltransferase 1 (GGT1), an enzyme involved in the metabolism of glutathione (cofactor of GPX4), could confer resistance to ferroptosis regulating the activity of GPX4 (Fig. 5B) (35). To validate this finding, we measured the expression levels of GGT1, SEPW1 and MUC1 by RT-PCR analyses. In Fig. 5C are reported the graphs of the relative mRNA expression as fold changes observed in all primary cell lines in the transition from 2D to 3D growth condition. Data confirmed the up-regulation of GGT1, SEPW1 and MUC1 genes in 3D spheroids and suggested that, since GGT1 provides substrates for glutathione synthesis, which is utilized as cofactor of both GPX4 and SEPW1, lung CSCs may be protected from ferroptosis by the activation of a GSH-dependent antioxidant mechanism. Based on the commonly up-regulated genes in 3D spheroids (GPX4, GGT1, SEPW1 and MUC1) and their known function in the literature, we hypothesized a molecular model consisting of a limited number of genes, in which each of them may contribute partially in conferring resistance to ferroptotic cell death (Fig. 5D).