Distinct photooxidation-induced cell death pathways lead to selective killing of human breast cancer cells

Lack of effective treatments for aggressive breast cancer is still a major global health problem. We previously reported that Photodynamic Therapy using Methylene Blue as photosensitizer (MB-PDT) massively kills metastatic human breast cancer, marginally affecting healthy cells. In this study we aimed to unveil the molecular mechanisms behind MB-PDT effectiveness. Through lipidomic and biochemical approaches we demonstrated that MB-PDT efficiency and specificity relies on polyunsaturated fatty acids-enriched membranes and on the better capacity to deal with photooxidative damage displayed by non-tumorigenic cells. We found out that, in tumorigenic cells, lysosome membrane permeabilization is accompanied by ferroptosis and/or necroptosis. Our results broadened the understanding of MB-PDT-induced photooxidation mechanisms and specificity in breast cancer cells. Therefore, we demonstrated that efficient approaches could be designed on the basis of lipid composition and metabolic features for hard-to-treat cancers. The results further reinforce MB-PDT as a therapeutic strategy for highly aggressive human breast cancer cells.


35
Breast cancer is the most frequent malignancy in women worldwide 1,2 . In its advanced 36 stages, when distant organ metastases occur, it is considered incurable with the 37 currently available therapies 2 . The reason being that metastatic lesions are usually 38 multiple, molecular and cellular heterogenous, and resistant to conventional 39 treatments 3 . Thus, effective and safe therapies for this stage of the disease are still 40

needed. 41
Photodynamic therapy (PDT) has been the focus of several cancer centers as it might 42 represent an important advancement in treatment due to its high but also controlled 43 cytotoxic effect 4 . Additionally, the enhanced antitumor effects combining PDT and 44 chemotherapies have already been demonstrated in preclinical studies on breast 45 cancer 3 . PDT consists in the uptake of a photosensitizer (PS) molecule which, upon 46 excitation by light in a determined wavelength, reacts with oxygen and generates 47 oxidant species (radicals, singlet oxygen, triplet species) in target tissues, leading to 48 photooxidative stress (PhOxS) 5,6 , which results in photodamage of membranes and 49 organelles 7,8 . The extent of the damage, and the cell death mechanisms involved, are 50 dependent on the PS type, concentration, subcellular localization, the amount of energy 51 and fluence rate applied as well as on the intrinsic characteristics of each tumor type 9-52 12 . The bottleneck of PDT is that little is known about the complex molecular 53 mechanisms behind its cytotoxicity and even less about the factors that could improve 54 its specificity against aggressive cancer cells. In order to address these underpinnings, 55 our group has been studying PDT using methylene blue as photosensitizer (MB-PDT) in 56 human breast cell (BC) models. 57 In previous studies we have already reported that there were differences in MB-PDT 58 sensitivity regarding MB concentration, time to achieve maximal cell death and the 59 effect of fluence rate 9,13 . Moreover, our results have shown that non-tumorigenic 60 breast cells are more resistant to MB-PDT, whereas the very aggressive triple negative 61 breast cancer cells (TNBC) displayed the highest susceptibility 13 . However, the 62 mechanisms behind these effects are still not well understood. In the present study, we 63 set out to unveil the molecular mechanisms triggered by this PhOxS therapy that are 64 responsible for its selectivity in the elimination of human breast cancer cells.  Collectively, our results showed that while MB-PDT is efficient in inducing multiple 71 regulated necrosis mechanisms only in tumor cells, non-tumorigenic breast cell were 72 able to mount an antioxidant response that led to impairment of the extensive 73 photooxidative damage. We believe that these data highlight MB-PDT potential to be 74 safe, accessible and an efficient adjunct to surgery for breast cancer treatment. 75 Furthermore, this study contributes to the cancer and cell biology fields, providing 76 further molecular mechanisms explaining why breast cells displaying distinct molecular 77 makeups are able to undergo different regulated cell death pathways upon the same 78 trigger. 79

Human breast cells (BC) presenting variations in PDT sensitivity displayed differential 81 cellular lipid composition 82
As a first step in our study, we confirmed previous results from our group 13 by showing 83 that cell death kinetics after MB-PDT exerted a higher impact in the malignant cell lines, 84 being TNBC cells the most susceptible ( Figure 1A). We then set out to investigate the 85 molecular basis of these cell type specific differential responses to the therapy. Since 86 MB-PDT relies on a massive intracellular generation of oxidant species 9,13,17 , with a 87 widespread impact in membranes, we evaluated whether there was a link between the 88 cellular lipid profile and the sensitivity to MB-PDT. For this purpose, we performed a 89 comparative lipidomic profiling of BC using reversed-phase ultra-high-performance 90   iron available in the labile iron pools (LIP) has already been reported for being more 134 prone to participate in ferroptosis, LIPs were analysed in the three cell lines before MB 135 photooxidation. The highest LIP levels were found in MDA-MB-231 cells ( Figure 2C). 136 Remarkably, these data revealed that these cell lines constitute three distinct cell 137 models to explore the role of ferroptosis in the context of photooxidation: one that 138 exhibits PUFA but also disposes high lipid detoxification capacity, like GXP4 and CoQ 10 , 139 (MCF-10A); another that besides displaying high levels of GPX4, does not posses high 140 abundance of PUFA nor CoQ 10 (MCF-7); and finally, the most agressive cell line 141 containing a higher proportion of PUFA and LIP and a very low capacity to deal with lipid 142 peroxidation due to the low levels of reduced gluthatione (GSH) 13 , GPX4 and CoQ 10 143 In order to investigate the role of ferroptosis in these cells submitted to photooxidation, 145 protective role in MCF-7 cells submitted to photooxidation, highlighting that a 189 differential cellular response mechanism against MB-PDT has been triggered. 190 Altogether, these results allowed us to conclude that the non-tumorigenic cells were 191 able to activate a proficient antioxidant response through the increase of G6PD, which 192 in turn would lead to the higher production of NADPH that could then be used to 193 regenerate GSH, as well as reduced CoQ 10 , contributing to maximize the detoxification 194 process and thus minimize cell death. 195

MB-PDT can also trigger necroptosis 196
In order to investigate whether necroptosis was a cell death mechanism activated by

MB-PDT induces lysosome damage 215
Based on the fact that we had previously reported that MB was mainly accumulated in 216 lysosomes 13 , the hypothesis that photoactive MB could be capable to damage lysosome 217 membrane and thus induce lysosome-dependent cell death (LDCD) was raised and 218 tested. The first evidence was obtained by assessing the cytosolic activity of the  We demonstrated that MB-PDT trigger multiple RCD in BC by inducing modifications in 228 lipid membranes. In malignant cells, our data pointed that LMP was followed by MLKL 229 phosphorylation and necroptosis. In case of high PUFA-lipid content, in addition to 230 LDCD and necroptosis, cells were also able to undergo ferroptosis, which was triggered However, this enzyme is absent or inactive during ferroptosis, resulting into toxic lipid 290 accumulation 39,40 . As a consequence, the inhibition of GPX4 or depletion of GSH has 291 emerged as therapeutic strategies to induce cancer cell death 41 . We have previously 292 shown that MB-PDT was able to deplete GSH levels in breast tumor cells 9 . Additionally, 293 in the present study the levels of GPX4 were decreased in all BC tested upon 294 photooxidation, underscoring MB-PDT as a potential ferroptosis inducer. This 295 observation is consistent with the fact that sensitivity to GPX4 inhibitors varies greatly 296 across cell lines and ferroptosis may have additional regulation mechanisms 42 . Indeed, a 297 glutathione-independent regulation axis involving the ferroptosis suppressor protein 1 298 (FSP1), previously called apoptosis-inducing factor mitochondrial 2 (AIFM2), was 299 recently uncovered 18,43 . FSP1 acts as a NAD(P)H-oxidoreductase that reduces coenzyme formation, also contributing to their resistance to photooxidation-induced cell death. 373 It has been previously shown by others that RIPK1 and RIPK3 can be degraded in 374 lysosomes and that inhibition of lysosomal function, with LMP, leads to this kinases 375 accumulation and necroptosis induction 59,60 , strengthening the possibility of a link 376 between LMP and necroptosis. Indeed, we have shown that, in breast cancer cells 377 submitted to MB-PDT, lysosomal cathepsin inhibition was able to suppress cell death 378 and MLKL phosphorylation, providing a clear evidence of the existence of cross-talks 379 between LDCD and necroptosis, triggered by MB-PDT. We suggest that, because as MB 380 localizes mainly in lysosomes of these cells, LMP is an inevitable consequence of MB 381 photoactivation, and thus may be driving the RCD events. The activation of different cell 382 death subroutines will then depend on the availability of the required components of 383 each pathway. We have described that different human breast epithelial cells, from 384 non-tumorigenic to very aggressive malignant cells, display distinct structural and 385 metabolic traits within different signaling pathways were activated upon MB-PDT. In 386 this study, LMP appears as the common event, which was then followed by an 387 antioxidant response (in non-tumorigenic cells), necroptosis (in non-invasive tumor 388

cells) or both necroptosis and ferroptosis (in highly aggressive tumor cells). 389
Collectively, our data have provided molecular mechanisms behind a hitherto 390  Proteins were transferred by tank blot onto a PVDF membrane that was subsequently 454 blocked in a solution containing 5% Blocking Buffer (Life Technologies) and 5% BSA (1:1) 455 at 4°C overnight. Primary antibodies were diluted in a solution of 5% BSA in PBS (Table  456 1) and were incubated overnight at 4°C. Membranes were washed three times in PBS-457 Tween (0.1%) and then incubated at RT for 1h with HRP-labeled secondary antibodies, 458 diluted in a solution of 1% BSA in PBS. The protein-antibody complex was visualized by 459 using enhanced chemiluminescence (Millipore Corporation, Billerica, MA, USA). Images 460 were acquired using Uvitec Image System (Cleaver Scientific Limited, Cambridge, UK). 461 Quantitative densitometry was carried out using the ImageJ software (National 462 Institutes of Health). The volume density of the chemiluminescent bands was calculated 463 as integrated optical density × mm 2 using ImageJ Fiji. 464

Cathepsin activity assay 465
Cells were washed with phosphate buffer (PBSA: NaCl 137 mM, KCl 2.7 mM, Na 2 HPO 4 466 10 mM, KH 2 PO 4 1.8 mM, pH 7.2) and detached from the plate using trypsin solution 467 (0.5% p/v). The cells were then centrifuged at 800 x g for 2 minutes. Cell pellets were 468 washed with PBSA and resuspended in 2 mL PBSA. The number of cells was counting in 469 a hemocytometer. Samples were then homogenized in syringes with insulin needle 10 470 times, and centrifuged at 4°C, 700 x g for 10 minutes. The supernatants were collected 471 and centrifuged at 4°C, 25000 x g for 2 hours for cytosol and organelles fractionation. 472 The supernatants (cytosolic fraction) were used in cathepsin-B/L kinetics assays using Z-473 FR-MCA as substrate (10 µM) in 100 mM citrate phosphate buffer, pH 6. Protease

Lipid peroxidation analysis 489
After irradiation, the cells were incubated with 1 µM BODIPY C11 at 37°C during 20 min. 490 The probe was then removed and 1 mL of fresh medium without serum and phenol red 491 was added. Cells were imaged using a fully motorized Leica DMi8 widefield microscope 492 (from Leica Microsystems) using the FITC and Texas Red filter sets and a 20x objective. 493 Imaging was performed on two independent biological replicates. In each independent 494 experiment at least 4 different images (100 cells) per condition were analysed. All 495 imaging acquisition parameters were kept constant for each experiment. Images were 496 quantified using ImageJ Fiji and quantified as follows. Cell outlines were free-handed 497 drawn on the bright field channel to generate a cell selection mask for quantifying the 498 fluorescence intensity in the green and red channels. Oxidation of BODIPY C11 581/591 499 was calculated as the ratio of the green (fluorescence emission of the oxidized probe)/ 500 red fluorescence mean intensity (fluorescence emission of reduced probe) within the 501 cell outlines. 502

Lipidomic analysis 503
Non-targeted lipidomic analysis of major lipids was performed by reversed-phase ultra-504 high-performance liquid chromatography (RP-UHPLC) coupled to electrospray 505 ionization time-of-flight mass spectrometry (ESI-TOFMS). Prior to lipid extraction, a 506 mixture of lipid internal standards (Table 2) was added to the samples for semi-507 quantification of reported lipid molecular species. Lipid extraction was performed 508 according to a method adapted from 62 . 500.000 cells were homogenized in 500 μL of 50 509 mM phosphate buffer (pH 7.4) containing 100 μM deferoxamine mesylate. This 510 homogenate was mixed with 400 μL of ice-cold methanol, containing 100 μM of 511 butylated hydroxytoluene (BHT), and 100 μL of internal standards (10 µg/mL). 2 mL of 512 chloroform: ethyl acetate (4:1) were added to the mixture, followed by vortexing during 513 30 s. After centrifugation at 1,500 x g for 2 min at 4°C, the lower phase containing total 514 lipid extracts (TLE) was transferred to a new tube and dried under N 2 gas 63 . Dried TLE 515 were dissolved in 100 µL of isopropanol and the UHPLC injection volume was set at 2 516 µL. The separation conditions of mass spectrometry analysis were performed as 517 previously described. The MS/MS data were analysed with PeakView®, and lipid 518 molecular species were identified by an in-house manufactured Excel-based macro.

Statistical analysis 554
All results were analysed for Gaussian distribution and passed the normality test. The 555 statistical differences between group means were tested by One-way ANOVA followed 556 by Tukey´s post-test for multiple comparisons or by Two-way ANOVA followed by 557     In this cell line, no antioxidant response was mounted upon PhOxS. In addition, low levels of GPX4 and CoQ 10 , combined with high amount of iron (Fe 2+ ) and PUFA-phospholipid content (PL-PUFA), resulted in ferroptosis activation by MB-PDT (purple arrows and letters). This cell death was inhibited by Fer-1 pre-treatment. Lysosomal damage was observed in all cell lines, evidenced by the release of cathepsin B through lysosomal membrane permeabilization, LMP, (green arrows and letters). Pre-treatment with CA-074, a cathepsin B inhibitor, alleviated cell death. In both tumorigenic cells, MDA-MB-231 and MCF-7, necroptosis activation with MLKL plasma membrane pore formation (blue arrows and letters) was observed. Inhibition of RIPK1, RIPK3 or MLKL phosphorylation, by gene silencing or pre-treatment with Nec-1 or NSA, rescue tumorigenic cells from death. A possible link between LMP and necroptosis was found in tumorigenic cells (green dotted arrows). Because MCF-7 cells lack significant amounts of oxidizable phospholipids, lipid peroxidation was not observed and therefore ferroptosis did not contribute to death. However, a complete antioxidant response was not sustained in these cells, making them also highly affected by MB-PDT. The scenario after PhOxS was quite different for MCF-10A cells. Even occurring LMP and lipid peroxidation, they were significantly more resistant to MB-PDT than the other cells. Neither ferroptosis nor MLKL phosphorylation and necroptosis were observed.