Abstract Dormant disseminated tumor cells (DTCs) remain viable for years to decades before establishing a clinically overt metastatic lesion. DTCs are known to be highly resilient and able to overcome the multiple biological hurdles imposed along the metastatic cascade. However, the specific metabolic adaptations of dormant DTCs remain to be elucidated. Here, we reveal that dormant DTCs upregulate de novo lipogenesis and favor the activation and incorporation of monounsaturated fatty acids (MUFAs) to their cellular membranes through the activation of acyl-coenzyme A synthetase long-chain family member 3 (ACSL3). Pharmacologic inhibition of de novo lipogenesis or genetic knockdown of ACSL3 results in lipid peroxidation and non-apoptotic cell death through ferroptosis. Clinically, ACSL3 was found to be overexpressed in quiescent DTCs in the lymph nodes of breast cancer patients and to significantly correlate with shorter disease-free and overall survival. Our work provides new insights into the molecular mechanisms enabling the survival of dormant DTCs and supports the use of de novo lipogenesis inhibitors to prevent breast cancer metastasis. Keywords: Tumor cell dormancy, Breast cancer, Metastasis, Lipid peroxidation, Lipid metabolism, Ferroptosis, Monounsaturated fatty acids activation Graphical abstract Image 1 [43]Open in a new tab Highlights * • de novo lipogenesis is activated in dormant disseminated breast cancer cells. * • ACSL3 mediates the incorporation of MUFAs into the plasma membrane of dormant breast cancer cells. * • MUFA-enriched membranes protect dormant breast cancer cells from ferroptosis. Abbreviations AAE Acyl-activating enzyme AAPH 2,2′-Azobis (2-methylpropionamidine) dihydrochloride ACSL3 Acyl-coenzyme A synthetase long-chain family member 3 ACSL4 Acyl-coenzyme A synthetase long-chain family member 4 AT2 Alveolar type II BC Breast cancer BME Basement membrane extract BSA Bovine Serum Albumin COL Collagen 1 CPT1a Carnitine palmitoyltransferase 1a CTRP Cancer Therapeutics Response Portal D5 Day 5 D7 Day 7 DAPI 4,6-diamidino- 2-phenylindole DFS Disease-free survival DHA Docosahexaenoic acid DTCs Disseminated tumor cells ER Estrogen receptor FAMEs Fatty acids methyl esters FAO Fatty acid oxidation FAs Fatty acids FASN Fatty acid synthetase FDR False discovery rate Fer-1 Ferrostatin 1 GC-MS Gas chromatography-mass spectrometry GO Gene Ontology GPX4 Glutathione Peroxidase 4 GSEA Gene Set Enrichment Analysis HR Hazard Ratio ISA Isotopomer Spectral Analysis KAT2A Lysine acetyltransferase 2A LC-MS Liquid chromatography-mass spectrometry LN lymph node LOF Loss of function M Metastasis MFI Mean fluorescence intensity MUFAs Monounsaturated fatty acids NES Normalized enrichment score NF-κB Nuclear factor-kappaB Non-ox Non-oxidized OA Oleic acid OS Overall survival Ox Oxidized CK Cytokeratin PFTBA Perfluorotri-n-butylamine PI Propidium iodide PLs Phospholipids POA Palmitoleic acid PPARγ Peroxisome proliferator activated receptor γ PT Primary tumor PUFAs Polyunsaturated fatty acids QTOF-MS Quadrupole time-of-flight mass spectrometer RSL3 1S,3R-RSL3 SCOM Single cell organ microscopy s.e.m. Standard error of the mean SFAs Saturated fatty acids SLC40A1 Solute Carrier 296 Family 40 (Iron-Regulated Transporter), Member 1 SLC7A11 Solute Carrier Family 7 (Anionic Amino Acid Transporter Light Chain, Xc- System), Member 11 SSO Sulfosuccinimidyl Oleate TAGs Triacylglycerols TCA Tricarboxylic acid TGOLN2 Trans-Golgi network protein 2 1. Introduction Breast cancer (BC) is a biologically and clinically heterogeneous disease and the most incident and mortal cancer among women worldwide [[44]1]. Most BC-related deaths are the consequence of metastasis, but life-threatening metastatic recurrences can occur from months to decades after diagnosis and treatment of the primary tumor [[45]2]. Timing of BC recurrences significantly correlates with the molecular subtype of the primary tumor [[46]3]. On the other hand, research over the past few years has firmly established that BC dissemination beyond the primary tumor occurs very early, even before diagnosis [[47][4], [48][5], [49][6]]. This highly variable and long intermission in the course of BC progression, from early dissemination to late disease recurrence, is the consequence of the existence of disseminated tumor cells (DTCs), which prevail in a dormant state at distant sites for years before initiating clinically active metastasis [[50]7]. Tumor dormancy poses a major challenge to the achievement of long-term disease-free survival (DFS) and cure of BC patients, whose treatment is based on conventional antiproliferative therapies [[51]8]. Conversely, the development of targeted therapies based on the specific biology of dormant DTCs opens a window of opportunity to treat disseminated disease before the clinical symptoms of metastasis manifest [[52]9]. A growing body of scientific evidence indicates that both cell intrinsic and microenvironmental mechanisms regulate DTC dormancy induction, survival, and transition to a proliferative state [[53]10]. These mechanisms include interactions between the dormant DTCs and the vascular, stromal, immune and extracellular matrix niches [[54]11], as well as the modulation of molecular pathways controlling autophagy, hypoxia, adhesion, redox and stress signaling, nuclear receptor-mediated signaling and epigenomic reprogramming [[55][12], [56][13], [57][14], [58][15], [59][16]]. From a therapeutic standpoint, targeting processes that promote dormant DTCs survival would provide the safest clinical intervention to block BC progression, as it would eradicate the source of metastasis. In this regard, there is a growing scientific interest in the vulnerability of dormant DTCs to high levels of oxidative stress [[60]15,[61]17,[62]18]. Interestingly, we reported an intricate connection between the activation of mitophagy and the maintenance of redox homeostasis in the survival of dormant BC cells [[63]19]. In eukaryotic cells, mitophagy fosters metabolic fitness and redox homeostasis, ultimately promoting cell survival under stress [[64]20]. In addition, metabolic rewiring has been long recognized as a hallmark of cancer [[65]21] and metastasis formation [[66]22], although the specific signaling pathways modulated to rewire the metabolism of dormant DTCs remain largely unknown. Mounting evidence shows that lipid metabolism is intensified throughout tumor progression and that these adaptations transcend classical cellular bioenergetics, impacting tumor cell signaling and fate through different mechanisms [[67]23,[68]24]. Among these mechanisms, the inhibition of ferroptosis-mediated cell death shows the interplay between lipid metabolism and the maintenance of redox homeostasis in the survival of tumor cells. Ferroptosis is a newly discovered form of non-apoptotic cell death driven by enhanced lipid peroxidation and lethal membrane permeabilization [[69]25,[70]26]. Changes in the lipid composition of cellular membranes have been shown to modulate ferroptosis, as some lipids are more prone to oxidation. Cellular membranes enriched in polyunsaturated fatty acids (PUFAs) are known to sensitize cells to ferroptosis [[71]27]. In this regard, ferroptosis can be prevented by deletion of an acyl-coenzyme A synthetase long-chain family member 4 (ACSL4) [[72]28]. ACSL enzymes activate free fatty acids to fatty acyl-CoA, which can be then incorporated into membrane phospholipids (PLs). Of the 5 human ACSL enzymes (ACSL1, 3–6), ACSL4 exhibits a marked preference for activating PUFAs [[73]28], while ACSL3 favors the activation of monounsaturated fatty acids (MUFAs). Strikingly, ACSL3-mediated activation of MUFAs has been shown to protect cells from ferroptosis, inhibiting the accumulation of lipid peroxides at the plasma membrane and maintaining its integrity [[74]29,[75]30]. In this study, we utilized metabolomics and ^13C tracer analysis [[76]31] to uncover the metabolic adaptations of dormant BC cells. Dormant BC cells showed increased de novo lipogenesis and ACSL3 overexpression. Both, inhibiting lipogenesis and ACSL3 expression in dormant BC cells resulted in a shift from MUFA-to PUFA-enriched lipid profiles, accumulation of lipid peroxides and increased cell death. Analysis of ACSL3 expression in BC cells disseminated to human lymph nodes validated the clinical relevance of our findings, identifying the potential therapeutic value of interfering with ACSL3-mediated lipid metabolism to prevent recurrence of BC. 2. Materials and methods 2.1. Patients and tumor specimens Our study exclusively examined samples from women because breast cancer is mainly relevant in women. Female breast cancer patient specimens (Luminal A, n = 10; Luminal B, n = 10), for primary tumor, lymph node and metastasis, were obtained from the archive of the Jaen University Hospital Pathology Unit (Jaén, Spain), with previous written informed consent signed by all patients. All samples and procedures were approved by the Coordinating Committee for Ethics in Biomedical Research of Andalusia (Spain) and were conducted in accordance with the Declaration of Helsinki and International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). Demographic and clinical data are summarized in [77]Table S2. Unstained paraffin-embedded 4 μm tissue sections were obtained and stained for ACSL3 for subsequent analysis. 2.2. Mice Our study exclusively examined female mice because the disease modeled, breast cancer, is mainly relevant in females. All animal studies were performed in accordance with University of Granada (UGR) Institutional Animal Care and Use Committee guidelines (Animal protocol number #September 05, 2019/151). 5–7 weeks old female Athymic mice (Athymic nude Crl:NU(NCr)-Foxn1^nu) were purchased from Charles River (strain code: 490) and were allowed to acclimatize to housing condition in animal facility for 1 week before use for the experiments. All mice were housed in the CIBM-UGR animal facility room maintained in a 12/12 h light/dark cycle at a temperature of 20–26 °C with 30%–50 % humidity and fed with complete universal vegetal diet for rodents (Safe-lab, #150-SP). Ad libitum access to food and water was always provided to mice and environmental enrichment was applied. Animal health was monitored once daily throughout the timeline of experiments. 2.3. Cell lines and cell culture Mouse mammary tumor D2.0R and D2A1 cells were derived from spontaneous hyperplastic alveolar nodules [[78]32,[79]33] and provided by Ann Chambers (London Cancer Center, London, Ontario, Canada). MDA-MB-231 and MCF7 cells were obtained from American Type Culture Collection (ATCC). The cells were grown in 2D and 3D cultures as previously described [[80]34,[81]35]. Briefly, cells (2 × 10^3) were resuspended in 100 μl DMEM (1 g/L glucose; Invitrogen) plus 2 % FBS and 2 % basement membrane extract (BME) (Matrigel, Corning) or 2 % FBS plus BME and COL (final concentration of COL was 2 mg/ml), either with inhibitors or vehicle, and grown in 96-well plates coated with 150 μl BME or BME plus COL per cm^2. In 2D, D2.0R, D2A1 and MDA-MB-231 were grown in DMEM supplemented with 10 % FBS. MCF7 were grown in DMEM F12 supplemented with 10 % FBS, 1 mM Sodium pyruvate (Gibco), 0.1 mM non-essential amino acid (Gibco), 5 mM l-Glutamine (Gibco), 0.01 mg/ml Human recombinant Insulin (Sigma-Aldrich) and 5 mM β-Estradiol (Sigma-Aldrich). Cells were maintained in 37 °C incubator with 5 % CO[2] and split every 3 days at 1:4 dilution. 2.4. In vitro stable isotope tracing and metabolomics analysis 2.4.1. Stable isotope labeling, quenching and cell extraction from matrices 2 × 10^5 cells were cultured in 10 cm dishes in 3D on either BME or BME + COL matrices, as described before and DMEM with 2 % dialyzed FBS and 1 g/L^13C[6] glucose (Merck Sigma) was added to 3D cell culture media. After five days in culture, media was washed and cells were metabolically quenched adding 10 ml of cold quenching buffer (60 % methanol, 10 % ammonium acetate, 30 % water) at 4 °C for 5 min [[82]36]. After quenching, cells were extracted from BME or BME + COL, as described before [[83]37]. Briefly, cells on BME matrices were rocked on a tray filled with ice with 10 ml of ice-cold PBS +5 mM EDTA for 30 min. Once the BME was solubilized, the extracts were transferred to a 15-ml tube and centrifuged at maximum speed. The remaining BME was eliminated through 3 sequential washes with ice-cold PBS +5 mM EDTA, ice-cold saline +5 mM EDTA and quenching buffer, and final cell pellets were placed in dry ice and stored at −80 °C. To extract cells seeded on BME + COL matrices, an enzymatic method was performed by adding 2.5 ml of a mixture of collagenase (Sigma, 2 mg/ml) and trypsin (Sigma, 2 mg/ml) dissolved in DPBS + 5 % dialyzed FBS and incubating at 37 °C with 5 % CO[2] for 5 min. After solubilization of the BME + COL matrices, the extracts were transferred to a 15-ml tube and centrifuged at maximum speed. The cell extracts were washed twice with quenching buffer, centrifuged at maximum speed, transfer to dry ice and stored at −80 °C until metabolite extraction. 2.4.2. Metabolite extraction and mass spectrometry (MS) analysis Metabolites for MS analysis were extracted, derivatized and measured as described previously [[84]36,[85]38]. Briefly, 800 μL of ice-cold 62.5 % methanol in water with glutarate as internal standard (2.5 μg/ml) was added to each tube containing quenched cells. Then, 500 μL of ice-cold chloroform with heptadecanoate (10 μg/ml) as internal standard was added. Each sample was vortexed for 10 min at 4 °C, and phase separation was achieved by centrifugation at 4 °C for 10 min and max rpm, after which the chloroform phase (containing the total fatty acid content) and the methanol phase (containing polar metabolites) were separated and dried by vacuum centrifugation. For total fatty acid measurements, the lipid fraction (chloroform phase) was esterified with 500 μL 2 % sulfuric acid in methanol at 60 °C overnight and extracted by addition of 600 μL hexane and 100 μL saturated aqueous NaCl. Samples were centrifuged for 5 min and the hexane phase was separated and dried by vacuum centrifugation. Samples were resuspended in hexane, fatty acids were separated with gas chromatography (GC) (8860 or 7890A GC system, Agilent Technologies, CA, USA) and combined with 5977B Inert MS system (Agilent Technologies, CA, USA). 1 μL of each sample was injected in splitless or in split ratio 1 to 3 with an inlet temperature of 250 °C on a DB-FASTFAME column (30 m × 0.250 mm). Helium was used as a carrier gas with a flow rate of 1 ml per min. For the separation of fatty acids, the initial gradient temperature was set at 50 °C for 1 min and increased at the ramping rate of 12 °C/min to 180 °C, followed by a ramping rate of 1 °C/min to reach 200 °C for 1 min. The final gradient temperature was set at 230 °C with a ramping rate of 5 °C/min. The temperatures of the quadrupole and the source were set at 150 °C and 230 °C, respectively. The MS system was operated under electron impact ionization at 70 eV and a mass range of 70–700 amu was scanned. Polar metabolites were analyzed by GC or liquid chromatography (LC) coupled to MS. For LC-MS, dry samples were resuspended in 50 μL water MS grade and analyzed in a Dionex UltiMate 3000 LC System (Thermo Scientific) with a thermal autosampler set at 4 °C, coupled to a Q Exactive Orbitrap mass spectrometer (Thermo Scientific) and a volume of 10 μL of sample was injected on a C18 column (Acquity UPLC HSS T3 1.8 μm 2.1 × 100 mm). The separation of metabolites was achieved at 40 °C with a flow rate of 0.25 ml/min. A gradient was applied for 40 min (solvent A: 10 mM Tributyl-Amine, 15 mM acetic acid – solvent B: Methanol) to separate the targeted metabolites (0 min: 5 % B, 2 min: 5 % B, 7 min: 37 % B, 14 min: 41 % B, 26 min: 95 % B, 30 min: 95 % B, 31 min: 5 % B; 40 min: 5 % B. The MS operated in negative full scan mode (m/z range: 70–800 and 760–900 from 15 to 32 min) using a spray voltage of 4.9 kV, capillary temperature of 320 °C, sheath gas at 50.0, auxiliary gas at 10.0. Data was collected using the Xcalibur software (Thermo Scientific). For GC-MS, dried metabolites samples were derivatized for 90 min at 37 °C as previously described [[86]39] and measured using a 8860 GC system combined with a 5977C Inert MS System (Agilent Technologies). The inlet temperature was set at 270 °C and 1 μl of sample was injected into a DB35MS column in splitless. The carrier gas flow of helium was fixed at 1 ml min-1. After the injection, the GC oven was kept at 100 °C for 1 min, increased up to 105 °C with a gradient of 2.5 °C min-1, then ramped to 240 °C with a gradient of 3.5 °C min-1, and after that ramped up to 320 °C with a gradient of 22 °C min-1, which was followed by 4 min at 320 °C. MS was performed at 70 eV and a mass range of 150–650 atomic mass units was measured. Metabolites abundances and isotopologue distributions were extracted from raw chromatograms, corrected for naturally occurring isotopes and normalized to the internal standard and protein content. Fractional de novo synthesis of fatty acids was calculated using Isotopomer Spectral Analysis (ISA) during the exposure to the labeled substrate ^13C[6]-Glucose for 5 days. ISA provides the fractional contribution of the tracer to the lipid that is newly synthesized per cell and over the time course of the experiment, it therefore provides a parameter that is normalized by cell number and considers the variable time [[87]40]. Following ISA equations, we calculated the newly synthesized fatty acid per condition and over the time course of the experiment. Then, we estimated the fatty acid synthesis rate for proliferative and non-proliferative (dormant) cells using the following formula respectively: (g(t)∗ion counts/Δcell number)∗growth rate, proliferative cells (g(t)∗ion counts/(cell number∗Δtime), non-proliferative cells where g(t), is the fraction of newly synthesized fatty acid in the sampled pool obtained by ISA model. Metabolite abundances from the raw chromatograms and ISA equations were calculated using MATLAB. 2.5. In vitro treatments Fasnall (Merck) was used as fatty acid synthetase inhibitor at a final concentration of 40 μM. 9-octadecenoic acid, 2,5-dioxo-3-sulfo-1-pyrrolidinyl ester, monosodium salt, also known as Sulfosuccinimidyl Oleate (SSO; Cayman Chemicals), was used as a CD36 inhibitor at 200 μM. 1S,3R-RSL3 (RSL3) (Quimigen), which is a glutathione peroxidase 4 (GPX4) and indirectly activates ferroptosis, was used at 2.5 μM. 2,2′-Azobis(2-methylpropionamidine) dihydrochloride (AAPH; Merck) a specific lipid peroxidation inducer, was used at 5 mM for 2 h in vitro. Ferrostatin 1 (Fer-1; Merck), a potent and selective [88]ferroptosis inhibitor, was used at 5 μM. H[2]O[2] (Sigma-Aldrich) was used as a non-specific oxidative stress inducer at 2,7 mM for 2 h. Free fatty acids, oleic acid (OA) and palmitoleic acid (POA) (Sigma-Aldrich), were added to 3D cultures after being coupled to Bovine Serum Albumin (BSA). Briefly, prewarmed DMEM Low glucose media (Gibco) was supplemented with 10 % fatty acid-free BSA (Sigma-Aldrich). A stock solution of 100 mM OA or POA was prepared by dissolving the fatty acids in ethanol:water (1:1). DMEM low glucose-BSA media was supplemented with OA or POA while maintaining the final concentration of ethanol to <0.2 % and providing a final OA and POA concentration of 125 μM. 2.6. Cell proliferation and viability assays In order to determine cell proliferation in various conditions, CellTiter 96® AQueous One Solution Cell Proliferation Assay (Promega) or CellTiter-Blue fluorescence cell viability assay (Promega) was performed in 96-well, flat, clear-bottom microplates according to the manufacturer's protocol. 2.7. CRISPR/Cas9 edition 2.7.1. CRISPR/Cas9 nucleofection gRNAs and Cas9 protein used to edit D2.0R were purchased from Synthego (Menlo Park, CA, USA). The RNPs were complexed at a Cas9:sgRNA molar ratio of 1:9 at 25 °C for 10 min prior to electroporation. 1 × 10^5 cells were nucleofected using the SF cell line 4D kit (Amaxa Inc.), according to manufacturer's instructions. D2.0R cells were resuspended in Complete Nucleofector Solution (Lonza, Basel, Switzerland) with complexed RNPs and nucleofected using the Lonza 4D Nucleofector (program CM-113). Cells were transferred to 24-well plates following nucleofection in the regular growth media described above. Through this procedure we obtained the CRISPR-edited D2.0R control and D2.0R ACSL3·loss of function (ACSL3^LOF) cells. 2.7.2. Indel frequency analysis by ICE 2–4 days post-targeting, D2.0R cells were harvested and a QuickExtract™ DNA Extraction Soluciton (Lucigen) was used to collect gDNA. The following primers were then used to amplify respective cut sites with KAPA2G Fast Hot Start Ready Mix (Roche) according to manufacturer's instructions: KO-KIT-ACSL3, forward: 5′- ACCCCTTCCCCCACATCATA -3′, reverse: 5′- AGGCAAAACCCGGTATCAGA -3’. PCR reactions were then run on a 1 % agarose gel and appropriate bands were cut and gel-extracted using a MiniElute Gel Extration Kit (Qiagen) according to manufacturer's instructions. Gel-extracted amplicons were then Sanger sequenced using the following primer: KO-KIT-ACSL3-seq reverse: 5′- AATTTGAGGTCAGTATATATAGTGAGTCCT-3’. Resulting Sanger chromatograms were then used as input for indel frequency analysis by ICE CRSPR analysis tool (Synthego Performance Analysis, ICE Analysis. 2019. v3.0.; accessible at [89]https://www.synthego.com/products/bioinformatics/crispr-analysis). 2.8. RNA extraction and quantitative RT-PCR Total RNA was extracted using TRIZOL (Invitrogen) protocol 0.2–1 mg of RNA was subjected to reverse transcription using High Capacity cDNA Reverse Transcription kit (Applied Biosystem). To determine relative mRNA expression quantitative RT-PCR was performed using GoTaq® qPCR Master Mix (Promega). All mRNA quantification data were normalized to b-actin. Primers used for q-PCR were: ACSL3 forward: 5′- GCGAGAAGGATTCCAAGACTGG- 3′, ACSL3 reverse: 5′- GAAGAGTAGCCGATTCGGCATC- 3’ 2.9. Western blotting Cells grown in DMEM were harvested and lysed with RIPA buffer (Sigma) with protease and phosphatase inhibitor (Sigma). For immunoblotting, 20–35 mg of protein was resolved in 7.5%–15 % SDS-PAGE (Biorad), transferred onto PVDF membrane (Millipore, USA), using a semi-dry transfer apparatus and blocked in 5 % TBST (50 mM Tris pH 7.6, 150 mM NaCl -TBS-; 0.1 % Tween-20)-milk for 1 h followed by incubation with primary antibody in 5 % TBST-milk overnight at 4 °C. Membranes were then washed in TBST and incubated with secondary antibody followed by washing and developed using West Femto super signal (Thermofisher) and imaged using the ChemiDoc™ Imaging Systems (Biorad). Likewise, densitometric analysis of the developed membranes was performed using the Image Lab Software (Biorad). When necessary, membranes were stripped during 15 min with stripping solution (Thermo Fisher Scientific) at room temperature, washed in TBST and re-blocked in 5 % TBST-milk for 1 h, for subsequent incubation with primary and secondary antibodies, as above. 2.10. In vivo studies 2.10.1. Experimental metastasis assay D2.0R and D2A1 cells stably expressing GFP were generated via lentiviral infection of the pSICO construct (provided by Tyler Jacks, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA). 1 × 10^6 D2.0R-GFP cells were tail-vein injected into nude mice. 2.10.2. In vivo Fasnall treatment Endogenous fatty acid synthesis was inhibited in nude mice by injecting 10 mg per Kg of body weight of Fasnall or vehicle (sterile NaCl salt solution 0.85 %, Sigma) i.p. 2 days a week for 3 weeks. 2.10.3. In vivo AAPH treatment In vivo lipid peroxidation assay was performed according to Kou, Y.R et al. [[90]41] Treatment consisted of nasal instillation of 200 mg/kg AAPH or Saline. The drug was dissolved in 18 ml of sterile NaCl salt solution 0.85 % (Sigma) and sonicated (Frequency: 120 Hz during 60seg and 6 pulses) to fluidify the solution and ensure homogeneous in vivo distribution. 2.10.4. In vivo Fer-1 treatment Lipid peroxidation was inhibited in nude mice by injecting 5.2 mg per Kg of body weight of Fer-1 or vehicle (sterile NaCl salt solution 0.85 %, Sigma: DMSO; 1,5:1) i.p. 7 days a week for 3 weeks. 2.10.5. Lung imaging At the experimental endpoint, mice were euthanized, their lungs extracted, perfused with PBS and imaged by fluorescent single cell organ microscopy (SCOM) imaging using an EVOS fluorescence microscope (Thermo Fisher Scientific) as previously described [[91]19,[92]35]. Images of the total surface of the lungs were captured at 10x magnification and analyzed using ImageJ software (NIH). Alternatively, imaging of perfused lungs was carried out using Operetta® CLS High-Content Analysis System (Revvity®) with 5x objective and Z-stack using 14 planes with steps of 26 μm. Then, composite images covering the entire lung surface were assembled and maximum projections of Z-stack images from each lung were analyzed using HarmonyTM software (Revvity ®). Tumor cells were segmented, and morphological features and fluorescence intensity signal were quantified. 2.11. Lipid profiling Total concentration of the fatty acids Linoleic acid (C18:2n-6), Arachidonic acid (C20:4n-6), Docosahexaenoic acid (C22:6–3), Docosatetraenoic or Adrenic acid (C22:4n-6), Palmitoleic acid (C16:1n-7), Elaidic acid, methyl ester (C18:1 trans-9) and Oleic acid (C18:1, cis-9) were quantified in three independent samples of dormant CRISPR-edited D2.0R control and D2.0R ACSL3^LOF cells cultured in 3D on BME matrices and the data replicated in two independent experiments. The samples were analyzed using a suitable methodology for the separation and detection of fatty acids methyl esters (FAMEs) to measure total fatty acid content in the cells. To do that, we extracted and derivatized the samples using the method first described by LePage and Roy [[93]42]. Frozen cell pellets extracted from 3D cultures were extracted in 1 ml Methanol-Benzene (PanReac AppliChem) 4:1 (v/v) plus K[2]CO[3] (Labkem) 6 % solution and Acetyl Chloride (Sigma-Aldrich). At this point 0.125 μg/ml of pentadecanoic acid in hexane was added as an internal standard. Pentadecanoic acid is an odd chain fatty acid that does not naturally occur in mammalian cells. It was transesterified with the sample and used to quantify the total FAMEs content on a gas chromatograph. Samples were then vortexed for 30 s and incubated at 100 °C for 1 h. After centrifuging at 3000×g for 10 min, sample supernatants were transferred to GC vials. The samples were dried under a stream of nitrogen, reconstituted in 50 μl of n-hexane (Supelco) and analyzed using an Agilent 7890B gas chromatograph coupled to a 7200 quadrupole time-of-flight mass spectrometer (QTOF-MS) with electron impact ionization. The column used was an HP-88 capillary column (30 m × 0.25 mm ID, 0.20 μm phase thickness) from Agilent Technologies. The GC oven temperature program started at 80 °C, followed by a temperature ramp of 8 °C min−1 to 145 °C (26 min held), continued by a temperature ramp 2 °C min−1 to 200 °C (1 min held), and finally a temperature ramp of 8 °C min−1 to 220 °C (3 min held). Pulsed splitless injections of sample (1 μL) were carried out at 250 °C and ultrapure grade helium was used as carrier gas at 1.0 ml/min flow rate. The interface, ion source and quadrupole temperatures were set at 240, 230 and 150 °C, respectively. MS analyses were performed in full scan mode with a mass range of 50–500 m/z and ionization was carried out in electron impact mode at 70 eV. A mass calibration between samples was performed with perfluorotri-n-butylamine (PFTBA, Agilent Technologies), as recommended by the manufacturer. All data acquisition operations were controlled with MassHunter software version B.06.00 (Agilent Technologies). Then, Qualitative MassHunter (version B.07.00) software was used to process GC–TOF/MS data files. Treatment of raw data files started by deconvolution of chromatograms to obtain a list of molecular features considered as potential compounds defined by the m/z value of one representative ion for each chromatographic peak and its retention time. For this purpose, the deconvolution algorithm was applied to each sample by considering all ions exceeding 5000 counts for the absolute height parameter, with an accuracy error of 10 ppm and a window size factor of 150 units. Tentative identification of compounds was performed by searching each mass spectrum in the NIST Atomic Spectra Database 78 [[94]43]. Only those identifications with a match factor higher than 700 were considered as valid. Tentatively identified FAMEs were confirmed whenever possible by comparison with mix of 37 FAMEs (Supelco). 2.12. ACSL3 immunohistochemistry analysis and score in BC clinical samples Tumor tissue was stained for ACLS3 (Novus Biologicals; clone NBP2-15252; at 1:100 dilution), Ki67 (Agilent Dako; clone MIB-1, prediluted) and Cytokeratin (CK) (Agilent Dako; clone AE1/AE3, prediluted) using an automated stainer (Dako Autostainer Link 48) and the Envision Flex + visualization system (Agilent, Dako) following the routine procedures of the laboratory of Anatomic Pathology of the Jaén University Hospital. Staining intensity in neoplastic cells was considered by using a semiquantitative score, defined as follows: staining intensity was graded as 0: no staining; 1+: weak or low, 2+: moderate, 3+: intense or high. In addition, staining pattern indicating the subcellular location of ASCL3 (granular cytoplasmic and perinuclear) was recorded. Supplementary Data file provides details about the validation of the specificity of the ACSL3 antibody using human positive and negative control tissue as well as illustrative images exemplifying the cut-offs used to determine the different intensities for ACSL3 staining. 2.13. Survival analysis For this analysis, the samples were subdivided in ACSL3 low (staining intensities 1 and 2) and ACSL3 high (staining intensity 3). Correlation between ACSL3 staining intensity and DFS and overall survival (OS) after diagnosis in primary tumor, lymph node and metastatic tissue of patients with BC was analyzed by the Kaplan–Meier method and log-rank test using the GraphPad Prism 9 software. 2.14. Staining with fluorescent probes and live cell imaging Cells were plated in 3D cultures as described above in 96-well plates with or without inhibitors. Prior to the observation of the cultures, the cells were incubated with 240 nM Calcein (R&D Systems), 5 mM Bodipy C11 (Invitrogen), 5 μg/ml Cell Mask Deep Red PM stain (Invitrogen), Propidium Iodide (PI) (NucRed, Invitrogen) and/or Hoechst 33342 (NucBlue, Invitrogen), according to manufacturer's instructions, and subsequently imaged using a Zeiss LSM 710 confocal microscope. 2.15. Immunofluorescence Cells were plated in 3D cultures in μ-slides (Ibidi, AI-81506). Well coatings containing the cells were fixed with 4 % paraformaldehyde for 10 min, permeabilized with 0.5 % Triton-X 100 in PBS during 10 min at 4 °C and washed with a PBS-glycine (7,5 mg/ml) solution for 10 min, three times. Then, slides were blocked with 10 % goat serum in IF wash solution (0.5 mg/ml NaN[3], 1 mg/ml bovine-serum albumin, 0.05 % Tween-20 and 0.025 % of Triton-X 100) for 1h at room temperature. Next, slides were incubated with mouse anti-ACSL3 (Santa Cruz Biotechnology, sc-271246) dilution 1/50 in IF wash solution supplemented with 10 % goat serum at 4 °C overnight. The following day, the slides were washed with IF wash solution at room temperature during 20 min, three times and incubated with Alexa fluor 488 goat anti-mouse secondary antibody (Invitrogen) for 1 h. Subsequently, the slides were washed with IF wash solution at room temperature during 20 min, three times and incubated with rabbit anti-trans-Golgi network protein 2 (TGOLN2) (Cell Signaling Technology, 65969) dilution 1/500 in IF wash solution supplemented with 10 % goat serum at 4 °C overnight. Finally, the slides were incubated with 555 goat anti-rabbit secondary antibody (Invitrogen) for 1 h, stained with a 5 μg/ml 4,6-diamidino-2-phenylindole (DAPI) in PBS for 5 min, washed with PBS for 10 min, three times and imaged using a Zeiss LSM 710 confocal microscope. 2.16. Digital image acquisition and processing Digital images were acquired using: (1) an inverted microscope EVOS fluorescence microscope (Thermo Fisher Scientific); (2) confocal images of 3D cultures were taken using Zeiss LSM-710 confocal system for which acquisition was performed using Zeiss LSM software ZEN. Images were composed and edited in ZEN or Photoshop CC (Adobe) software, in which background was reduced using brightness and contrast adjustments applied to the whole image. Co-localization of oxidized (Ox) or non-oxidized (Non-ox) Bodipy C11 and Cell Mask deep red and co-localization of ACSL3 and TGOLN2 were determined using a ZEN automated macro pipeline calculating double-positive pixels. 2.17. Statistics and bioinformatic analysis In vitro experiments were repeated at least 3 times with 3 technical replicates each with representative images, blots and graphs shown. In animal experiments each group contained at least 8–10 mice. All data are represented as the mean ± standard error of the mean (s.e.m. Represented as error bars in the graphs). Shapiro-Wilks test was used to assess data normality and, according to the results, a parametric or a non-parametric test was chosen, as specified in the Figure legends. GraphPad prism 9 software was used for all statistical analysis. The values were considered statistically significant as per p values (∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 and NS: not significant). Details of the statistical analysis of metastatic lesion measurements by SCOM have been previously described [[95]35]. Briefly, pixel counts of GFP-positive metastatic lung lesions were quantified for the entire surface area of each lung. Lesions with >10 and < 1000 pixels^2 represented individual tumor cells; lesions with >1000 pixels^2 represented multicellular metastatic lesions. Total tumor burden/lung was represented by the total number of pixels^2 detected on the entire surface of each lung. Average tumor burden/lung was represented as the mean size of the metastatic lesions. Statistical analyses of differences in the distribution of tumor burden between groups utilized Mann–Whitney U test for pairwise comparisons and Kruskal-Wallis test for multiple group comparisons with Dunn's correction. Statistical significance was set at a P < 0.05. Alternatively, we used a High-Content Analysis System-Operetta® CLS (Revvity®) to image DTCs, micro- and macrometastasis in lung surface of nude mice injected with D2A1-GFP positive cells. The fluorescence intensity signal of segmented images was analyzed using HarmonyTM software (Revvity ®) and expressed in arbitrary units (A.U.). Ranked Gene Set Enrichment Analysis (GSEA) was performed with software version 4.0.3. Default setting. Data from Cancer Therapeutics Response Portal (CTRP) v2.1 dataset [[96]44] (available from [97]https://portals.broadinstitute.org/ctrp.v2.1/) were analyzed on August 1st, 2023. The data were analyzed for statistically significant Pearson correlations between basal gene expression for ACSL3 and small-molecule sensitivity, from any primary site/subtype and spheroid growth mode. 3. Results 3.1. Dormant breast cancer cells activate de novo lipogenesis BC dormancy has been modeled in vitro and in vivo. The D2.0R and D2A1 tumor cell lines were derived from murine mammary hyperplastic alveolar nodules [[98]45] and, while both of them are able to form primary tumors when injected in the mammary fat pad of mice and disseminate systemically, D2A1 cells form macrometastases in the lungs within ∼1–3 weeks, whereas D2.0R cells remain dormant at the metastatic site for about 4 months before forming relatively few lung metastases [[99]33,[100]35,[101]46]. This in vivo dormant phenotype can be reproduced in vitro using a 3D culture model system based on the seeding the BC cell lines on BME matrices, where the D2.0R cells remain quiescent for up to 12 days. In contrast, highly proliferative D2A1 cells spontaneously outbreak into a proliferative state between day 4 and 6 of culture on BME [[102]34]. In addition, it has been previously demonstrated that changes in the microenvironment, including exposure to collagen 1 (COL) or fibronectin, induce the dormant-to-proliferative switch of D2.0R cells [[103]19,[104]35]. To gain insights and explore possible metabolic differences between dormant (cells seeded on BME matrices) and proliferating (cells seeded on BME + COL matrices) D2.0R cells, we used ^13C[6]-glucose and MS analysis [[105]31]. First, we observed that proliferating D2.0R cells showed more incorporation of glucose carbons to tricarboxylic acid (TCA) cycle intermediates measured by LC-MS, indicating a more pronounced glucose metabolism compared to dormant D2.0R cells ([106]Fig. 1a). Unexpectedly, analysis of the TCA cycle intermediates abundances revealed a significant increase of citrate in dormant D2.0R cells on BME as compared to proliferating D2.0R growing on BME + COL matrices ([107]Fig. 1b). These observations were confirmed by gas GC-MS ([108]Fig. S1). Citrate can be diverted from the TCA cycle and shuttled from the mitochondria to the cytosol to participate in the first reactions of the de novo lipogenesis pathway [[109]47]. In line with this observation, we found that ^13C[6]-glucose-derived citrate was incorporated into fatty acids, which was evident from the presence of labeled carbons in palmitate and stearate measured by GC-MS, in both D2.0R cells on BME and BME + COL 3D cultures ([110]Fig. 1c). Therefore, D2.0R cells were found to divert carbons from glucose to de novo fatty acid synthesis regardless their proliferative status, and significantly increased levels of fatty acids were observed in dormant D2.0R cells as compared with their proliferating counterparts ([111]Fig. 1b, [112]Fig. S1b). Next, we estimated the rate of de novo fatty acid synthesis and found that this rate was significantly higher in dormant D2.0R cells seeded on BME matrices versus (vs.) proliferating D2.0R cells growing on BME + COL matrices ([113]Fig. 1d). Together, these data provide evidence that de novo fatty acid synthesis is upregulated in dormant breast cancer D2.0R cells. Fig. 1. [114]Fig. 1 [115]Open in a new tab Dormant breast cancer cells activate de novo lipogenesis. a.^13C[6]-glucose (1 g/L) stable isotope tracing over 5 days was used to reveal flux of carbon from glucose into TCA cycle metabolites in D2.0R cells on BME and BME + COL matrices, as measured by LC-MS. Absolute isotopolog distribution is shown (n = 3–4 biological replicates). b. Relative total abundances of polar metabolites and fatty acids in dormant D2.0R cells on BME matrices as compared to proliferating D2.0R cells on BME + COL matrices at day 5 of culture (n = 3–4). Unpaired two-tailed t-tests with Welch correction. ∗P < 0.05, ∗∗P < 0.01. c. Palmitate (left) and stearate (right) labelling from ^13C[6]-glucose in D2.0R cells on BME and BME + COL matrices at day 5 of culture. d.de novo palmitate (upper graph) and stearate (lower graph) synthesis rate in dormant D2.0R cells on BME matrices as compared to proliferating D2.0R cells on BME + COL matrices at day 5 of culture (n = 6–8 biological replicates). Unpaired two-tailed t-tests with Welch correction. ∗∗∗P < 0.001. 3.2. Inhibiting fatty acid synthesis reduces both the viability of dormant BC cells and the fraction of proliferating BC cells in 3D culture To investigate the role of lipid biosynthesis in tumor cell dormancy, dormant D2.0R cells were treated with two drugs: Fasnall, a fatty acid synthase inhibitor blocking endogenous fatty acid synthesis, and SSO, a CD36 inhibitor that prevents the uptake of exogenous fatty acids ([116]Fig. 2a). Fasnall significantly reduced the viability of dormant D2.0R cells, as compared to vehicle treated control cells, whereas SSO treatment had no significant effect ([117]Fig. 2b). Conversely, Fasnall did not affect the viability of proliferating D2.0R cells ([118]Fig. 2c and d), although we observed a significant impairment of their proliferative capabilities during the first 4 days of treatment ([119]Fig. 2d). Over time, Fasnall reduced the number of viable proliferating cells, while SSO treatment significantly decreased their viability shortly after seeding ([120]Fig. 2d). Fig. 2. [121]Fig. 2 [122]Open in a new tab Inhibiting fatty acid synthesis reduces the viability of dormant D2.0R cells and reduces the fraction of proliferating D2.0R cells in 3D culture. Figure shows data of one out of three independent experiments done in triplicate with equivalent results. a. Outline of the proliferation and viability experiments on BME matrices. b. Proliferation (left) (mean ± s.e.m, n = 3 wells. Comparisons by one-way ANOVA plus Tukey's multiple comparisons post-test for Day 0, Day 3 and Day 6 time points. ∗∗P ≤ 0.01; ∗∗∗∗P ≤ 0.0001) and viability (right) (mean ± s.e.m, n = 3 wells. Comparisons at Day 10 by one-way ANOVA plus Tukey's multiple comparisons post-test. ∗∗P ≤ 0.01 relative to Day 1 for each group) assays of D2.0R cells on BME matrices treated with Fasnall (40 μM) and SSO (200 μM). c. Outline of the proliferation and viability experiments on BME + COL matrices. d. Proliferation (left) (mean ± s.e.m, n = 3 wells. Comparisons by one-way ANOVA plus Tukey's multiple comparisons post-test for Day 0, Day 3, Day 6 and Day 9 time points. ∗∗P ≤ 0.01) and viability (right) (mean ± s.e.m, n = 3 wells. Comparisons at Day 7 by one-way ANOVA plus Tukey's multiple comparisons post-test. ∗∗P ≤ 0.01; ∗∗∗∗P ≤ 0.0001 relative to Day 1 for each group) assays of D2.0R cells on BME + COL matrices treated with Fasnall (40 μM) and SSO (200 μM). e. Representative propidium iodide (PI, red) and Calcein AM (green) staining of D2.0R cells on BME, either treated with Fasnall, SSO or vehicle. Scale bar is 20 μm. Viability and cell death indexes of D2.0R on BME were calculated as the percentage of cells positively stained for Calcein AM or PI, as shown in f. (mean ± s.e.m, n = 100–269 cells per group. Comparison by Mann-Whitney U test, two-sided. ∗P ≤ 0.05; ∗∗∗∗P ≤ 0.0001). g. Representative PI (red) and Calcein AM (green) staining of D2.0R cells on BME + COL, either treated with Fasnall, SSO or vehicle. Scale bar is 20 μm. h. Viability and cell death indexes of D2.0R on BME + COL were calculated as the percentage of cells positively stained for Calcein AM or PI (mean ± s.e.m, n = 139–287 cells per group. Comparison by Mann-Whitney U test, two-sided. ∗P ≤ 0.05; ∗∗∗∗P ≤ 0.0001). DMSO, cell treated with vehicle; Fasnall, cells treated with 40 μM Fasnall immediately after plating; SSO, cells treated with 200 [MATH: μ :MATH] M SSO immediately after plating. Viability staining of 3D cultures confirmed these findings. Dormant D2.0R cells treated with Fasnall showed a 29.67 % decrease in cell viability over 6 days of culture ([123]Fig. 2e and f upper graph), with a corresponding increase in nonviable cells ([124]Fig. 2e and f lower graph). SSO treatment, however, did not significantly alter their viability ([125]Fig. 2e and f). When Fasnall treatment was applied to proliferating D2.0R cells, we did not observe any significant change in the percentage of viable cells until day 6 of culture ([126]Fig. 2g), with a 23 % decrease in live cells ([127]Fig. 2g and h upper graph) and a 34 % increase in nonviable cells ([128]Fig. 2g and h lower graph) between days 3 and 6. SSO consistently decreased the viability of proliferating D2.0R cells from day 1 ([129]Fig. 2g and h). Mouse D2A1 cells and human MDA-MB-231 and MCF7 cell lines were used to expand these findings ([130]Figs. S2 and S3). As described above, mouse D2A1 cells spontaneously outbreak into a proliferative state between day 4 and 6 of culture on BME [[131]19,[132]34] ([133]Fig. S2a). In addition, human BC MDA-MB-231 and MCF7 cells have been previously shown to exhibit comparable proliferative behavior to D2A1 and D2.0R (proliferative vs. dormant), respectively ([134]Figs. S2b and S3), in vivo [[135]35] and in 3D culture [[136]19,[137]34]. D2.0R and MCF7 cells model ER (estrogen receptor)-positive (ER+) luminal BC subtypes [[138]48,[139]49] associated with late relapses [[140]50], while D2A1 and MDA-MB-231 cells represent aggressive, ER-negative (ER-) BC subtypes [[141]49,[142]51]. Treatment with Fasnall significantly decreased the number of viable D2A1 cells, as compared with their vehicle treated counterparts ([143]Fig. S2a). However, treatment with SSO precluded their proliferative outbreak after 4 days in culture on BME matrices without significantly affecting cell viability ([144]Fig. S2a). Treatment with Fasnall significantly reduced the proliferative rate of spontaneously outbreaking MDA-MB-231 cells on BME matrices but it did not exert a significant effect on cell survival ([145]Fig. S2b). This difference between D2A1 and MDA-MB-231 cells upon treatment with Fasnall may account for the earlier transition of MDA-MB-231 cells to a proliferative state (typically at day 3 of culture on BME), as compared to D2A1 cells on BME matrices ([146]Figs. S2a and S2b). Likewise, treatment with SSO of MDA-MB-231 cells on BME matrices decreased the number of proliferating cells over time without affecting their survival ([147]Fig. S2b). Finally, treatment with Fasnall significantly reduced the number of viable quiescent MCF7 cells, whereas SSO treatment did not affect their viability ([148]Fig. S3), similar to the effect of these drugs observed for dormant D2.0R cells cultured on BME matrices ([149]Fig. 2b). To address lipid biosynthesis during the dormant-to-proliferative switch, Fasnall and SSO treatments were delayed until D2A1 and MDA-MB-231 cells had transitioned to a proliferative state (5–7 days post-seeding) ([150]Figs. S4 and S5). Beginning treatment with Fasnall of D2A1 cells after 5 or 7 days in culture (already switched to proliferation) resulted in a significant attenuation of the decrease in the number of viable cells compared to cells treated with SSO at the same time points ([151]Figs. S4c and d). In the case of MDA-MB-231 cells, cell numbers were consistently decreased over time upon treatment with SSO compared to cells treated with Fasnall and regardless of when the drugs were added to the 3D cultures ([152]Figs. S5b and c). However, the reduction in the number of viable MDA-MB-231 cells treated with SSO vs. Fasnall was much more significant and occurred earlier upon the addition of the drugs to the culture in treatment-delayed ([153]Fig. S5c) compared to immediate treatment conditions ([154]Fig. S5b). This latter observation might owe to the fact that MDA-MB-231 cells spontaneously outbreak into a proliferative state earlier than D2A1 cells on BME matrices, as mentioned before. Visualization of viable and non-viable cells under the delayed or immediate treatment conditions confirmed the differential effect of Fasnall and SSO treatments depending on the proliferative status of the metastatic cells ([155]Figs. S4e–h for D2A1 cells and [156]Figs. S5d–g for MDA-MB-231 cells). Collectively, these results demonstrate that mouse and human metastatic BC cells in a dormant state are sensitive to the inhibition of de novo lipogenesis ([157]Table S1). 3.3. Inhibition of fatty acid synthesis reduces the lung tumor burden of D2.0R cells and inhibits the metastatic outgrowth from D2A1 DTCs in vivo Based on our in vitro results, we investigated the effect of inhibiting de novo lipogenesis on the survival and maintenance of disseminated dormant BC cells in vivo. To do that, we utilized previously developed and validated in vivo model systems to study BC dormancy [[158]19,[159]34,[160]35,[161]52]. D2.0R cells stably expressing green fluorescent protein (D2.0R-GFP) were intravenously injected into the tail-vein of athymic nu/nu mice and treated with 10 mg/kg of Fasnall, twice a week for 3 weeks ([162]Fig. 3a). Notably, Fasnall treatment was well tolerated by the mice and no signs of acute toxicity or significant weight loss were found in the Fasnall-treated compared to the vehicle-treated control mice (data not shown). Consistent with the in vitro data described above, treatment with Fasnall significantly impaired the survival of dormant BC disseminated cells in vivo, as we observed a 23-fold decrease of total metastatic lung burden in Fasnall-treated mice as compared to controls ([163]Fig. 3b). Likewise, we observed a significant decrease in the number of micrometastasis found in the lungs of mice treated with Fasnall, which ranged from 0 to 1 per mouse, compared to control mice, in which we found up to 5 lung micrometastasis per mouse ([164]Fig. 3c and d). Fig. 3. [165]Fig. 3 [166]Open in a new tab Fatty acids synthesis inhibition reduces the lung tumor burden of D2.0R cells in vivo. a. Experiment design. b. Total lung surface burden of athymic nu/nu mice receiving tail-vein injections of 1 × 10^6 D2.0R GFP cells, followed by vehicle (DMSO + Saline) or 10 mg/kg body weight of Fasnall, twice week for 3 weeks (data show mean ± s.e.m, n = 8–9 mice per group. Comparison by Mann-Whitney U test, two-sided). c. Absolute numbers of dormant DTCs and micrometastasis per lung in mice either treated with vehicle or Fasnall (data show mean ± s.e.m, n = 8–9 mice per group. Comparisons by Mann-Whitney U test, two-sided). Lesions <1000 pixels^2 correspond to dormant DTCs and lesions >1000 pixels^2 represent micrometastasis d. Representative images of dormant disseminated tumor cells (DTCs) and micrometastatic lesions in the lung from the experiment are shown in b. Scale bar is 400 μm. In contrast to D2.0 R cells, D2A1 exhibit a similar dormant phase in vivo for 1–3 weeks after which they form large metastatic lesions [[167]19,[168]35]. We injected D2A1-GFP cells into the lateral tail vein of nude mice and treated them with 10 mg/kg of Fasnall either shortly after cell injection or after one week (treatment-delay group) ([169]Fig. S6a). Fasnall treatment upon cell injection in the mice significantly reduced total metastatic tumor burden as compared to the control and treatment-delay groups. In line with previous results ([170]Fig. S4), the delay of Fasnall administration for 7 days after cell injection resulted in equivalent total metastatic tumor burden as compared to controls. On the contrary, total metastatic tumor burden was significantly higher in the treatment-delay group compared to mice treated with Fasnall shortly after cell injection ([171]Figs. S6b and c). These results confirmed previous findings using in vitro model systems and suggest that de novo lipogenesis is of higher relevance for the maintenance and survival of dormant DTCs as compared to metastatic cells which has undergone the dormant-to-proliferative switch. 3.4. Dormant breast cancer cells are protected against lipid peroxide-induced toxicity To gain insights into the molecular mechanisms regulating cell death upon de novo lipogenesis inhibition in dormant BC cells, we interrogated previous transcriptomic data comparing gene expression changes in dormant vs. proliferating D2.0R cells [[172]19]. GSEA indicated that several biological processes involved in cellular lipogenesis and iron ion homeostasis were activated in dormant vs. proliferative D2.0R cells ([173]Fig. 4a). We therefore became interested in ferroptosis, which is an iron-dependent type of programmed cell death characterized by the accumulation of lipid peroxides [[174]25]. Further exploration of our transcriptomic data revealed that several ferroptosis antagonist genes, such as Solute Carrier Family 7 (Anionic Amino Acid Transporter Light Chain, Xc- System), Member 11 (SLC7A11), GPX4 or Solute Carrier Family 40 (Iron-Regulated Transporter), Member 1 (SLC40A1), were overexpressed in dormant vs. proliferative D2.0R cells ([175]Fig. 4b), indicating that dormant BC cells might have acquired a ferroptosis resistant state. Fig. 4. [176]Fig. 4 [177]Open in a new tab Dormant breast cancer cells exert a ferroptosis-resistant state. a. Pathway enrichment analysis (upper panel) focusing on cellular metabolism for RNA expression in D2.0R cells on BME as compared with D2.0 R cells on BME + COL after 5 days in culture. The graph represents metabolic pathways which are significantly modulated in dormant as compared to proliferating D2.0R cells based on the gene hit size (adjusted P value (padj) < 0.05). The values are represented in logarithmic scale (base 10). The gene sets were obtained from the Gene Ontology (GO) Biological Process database [178]https://geneontology.org/. Enrichment plots (lower panel) of positive regulation of lipid biosynthetic process (left) and iron ion homeostasis (right) pathways in dormant D2.0R cells compared with proliferative D2.0R cells, as identified by the GSEA computational method [179]https://www.gsea-msigdb.org/gsea/. Columns indicate individual samples and rows represent each gene. Red represents a high expression level and blue indicates a low expression level. NES, normalized enrichment score; NOM P value, nominal P value; FDR value, false discovery rate value. b. Graphical summary of the expression profile of overexpressed (red) and downexpressed (green) genes involved in ferroptosis, in D2.0R cells on BME as compared with D2.0 R cells on BME + COL. Ferroptosis inducers (purple) and inhibitors (pink) used in this study were included. Representative images are shown for c. Fluorescent staining of oxidized lipid (BODIPY C11 oxidized: BODIPY C11^Ox, green) and non-oxidized lipid (BODIPY C11 non-oxidized: BODIPY C11^Non-ox, red) markers of live D2.0R cells on BME (upper panels) and BME + COL (lower panels) after treatment with either RSL3 (2,5 μM) or vehicle for 5 days, with quantification of the mean fluorescence intensity (MFI) ratio of the oxidized BODIPY C11 probe (BODIPY C11^Ox) to total BODIPY C11 signal detected (BODIPY C11^Ox + BODIPY C11^Non-ox) (mean ± s.e.m, n = 40–128 cells per group. Comparisons are relative to D2.0R cells treated with vehicle in both conditions by Kruskal-Wallis, Dunn's post-test. ∗P ≤ 0.05; ∗∗∗∗P ≤ 0.0001) (upper graph) and quantification of the relative increase in BODIPY MFI ratio in D2.0R cells on BME or BME + COL matrices upon treatment with the ferroptosis inductor RSL3 (mean ± s.e.m, n = 40–128 cells. Comparison by Mann-Whitney U test, two-sided. ∗∗∗P ≤ 0.0001) (lower graph). Scale bar is 20 μm. d. Dose-dependent reduction in D2.0R cell numbers on BME (upper left graph) and BME + COL (lower left graph) upon treatment with increasing concentrations (0–8 μM) of RSL3 as determined by MTS assay at days 0 and 3 of culture (mean ± s.e.m, n = 3 wells. Comparisons at day 3 by one-way ANOVA plus Tukey post-test. ∗∗P ≤ 0.01; ∗∗∗∗P ≤ 0.0001). On the right, quantification of the relative decrease in D2.0R cell numbers, seeded on BME and BME + COL matrices, upon treatment with 2 μM RSL3 for 3 days with respect to vehicle treated control cells. e. Representative images of live D2.0R cells stained with BODIPY C11 and treated with either Fasnall or vehicle for 5 days on BME matrices. Scale bar is 20 μm and 5 μm. On the right, quantification of the MFI ratio for BODIPY C11 of D2.0R cells on BME matrices treated with vehicle or Fasnall for 5 days (mean ± s.e.m, n = 66–84 cells per group. Comparison by Mann-Whitney U test, two-sided. ∗∗∗P ≤ 0.0001). Surprisingly, we did not find a significant change in the mRNA levels of fatty acid synthetase (FASN), the molecular target of Fasnall, but a key enzyme downstream in the lipid biosynthesis pathway, known to favor MUFAs activation and counteract lipid peroxidation and ferroptosis, ACSL3, was found to be overexpressed in dormant cells ([180]Fig. 4b). To further investigate this lack of correlation between the mRNA levels of FASN and ACSL3, we analyzed the protein levels of both enzymes in dormant and proliferative D2.0R and MCF7 cells. We found that both dormant D2.0R and MCF7 cells overexpressed ACSL3 and FASN proteins as compared to their proliferative counterparts ([181]Fig. S7), in line with the bibliography highlighting the role of FASN in tumor biology and metastasis [[182]53]. However, a recent report has demonstrated the complex chain of prost-transcriptional modifications to which FASN protein is subjected, that significantly affects its enzymatic activity and are not captured by the analysis of gene expression levels [[183]54]. Therefore, to perform further functional assays on the impact of de novo lipogenesis inhibition in tumor dormancy we specifically CRISPR-edited the acyl-activating catalytic domain of ACSL3 (see below). To evaluate the biological significance of these data, we analyzed the accumulation of lipid peroxides in dormant vs. proliferating D2.0R cells. We found a significant increase in the accumulation of cellular lipid peroxides upon treatment with the ferroptosis inducer RSL3 in both conditions, although the increase was significantly higher in the proliferating cells ([184]Fig. 4c). Indeed, treating D2.0R cells with RSL3 significantly decreased the number of viable proliferating D2.0R cells as compared to their dormant counterparts ([185]Fig. 4d). Considering this data, we sought to validate the observed dependence of a ferroptosis-resistant phenotype on the proliferative status of BC DTCs using a different model system. To do that, we seeded D2A1 cells on BME matrices and treated them with RSL3 either immediately or after 5 and 7 days in culture. As shown by the data from this experiment ([186]Fig. S8), D2A1 viable cell numbers did not significantly increase from day 0 to day 3 of culture (coinciding with the quiescent phase observed for D2A1 cells cultured on BME matrices; [187]Fig. S8a) in vehicle-treated as compared to RSL3-treated D2A1 cells in any of the conditions tested ([188]Figs. S8b–d). In contrast, after the dormant-to-proliferative switch on day 5 of culture, the numbers of viable and proliferating D2A1 cells decreased significantly, both under the immediate and delayed treatment conditions ([189]Figs. S8b–e). These data confirmed that, alike D2.0R cells, D2A1 cells exhibit maximal sensitivity to ferroptosis induction under proliferative conditions. Finally, aiming to correlate the observed effect of Fasnall and RSL3 treatments on the survival of dormant BC cells as compared to proliferative BC cells, we assessed the level of intracellular lipid peroxides in D2.0R cells seeded on BME, either treated with Fasnall or vehicle. In line with previous data showing an increase of intracellular lipid peroxides in proliferating and non-lipogenic vs. dormant and lipogenic D2.0R cells treated with RSL3, we observed a pronounced increase of lipid peroxides accumulation in dormant D2.0R cells in which de novo lipogenesis was inhibited through Fasnall treatment as compared to cells treated with vehicle ([190]Fig. 4e). Taken together, these data would indicate that dormant BC cells counteract the cytotoxic effects of lipid peroxidation through the activation of de novo lipogenesis. 3.5. ACSL3-mediated de novo fatty acid synthesis protects dormant breast cancer cells against ferroptosis Since transcriptomic data pointed out the involvement of ACSL3 in the de novo lipid biosynthetic pathway, we explored its potential role in dormant tumor cell survival through ferroptosis inhibition. In agreement with this notion, analysis of the data compiled in The CTRP, showed that low ACSL3 expression was significantly correlated with high sensitivity to GPX4 inhibitors, such as RSL3, ML162 and ML210, across all catalogued cancer cell lines (spheroid growth mode only) ([191]Fig. 5a). Therefore, decreased ACSL3 expression strongly correlated with increased ferroptosis sensitivity, in the context of tumor cell biology, as also found by others before [[192]29]. Fig. 5. [193]Fig. 5 [194]Open in a new tab ACSL3 protects dormant breast cancer cells against lipid peroxide-induced toxicity. a. Significant correlations between low expression of ACSL3 gene and compound lethality from the Cancer Therapeutics Response Portal dataset. Each dot represents one compound. Ferroptosis-inducing compounds are highlighted in green. b. Proliferation (left graph. Mean ± s.e.m, n = 3 wells. Comparisons by one-way ANOVA plus Tukey's multiple comparisons post-test for Day 0, Day 3 and Day 6 time points. ∗∗P ≤ 0.01) and viability (right graph. Mean ± s.e.m, n = 3 wells. Comparisons at Day 7 by one-way ANOVA plus Tukey's multiple comparisons post-test. ∗∗P ≤ 0.01 relative to Day 1 for each group) assays of D2.0R control and D2.0R ACSL3^LOF cells on BME. c. Representative images of live D2.0R control and D2.0R ACSL3^LOF cells stained with BODIPY C11 and cultured on BME matrices for 6 days. Scale bar is 20 μm and 5 μm. On the right, quantification of the MFI ratio of the oxidized BODIPY C11 probe (BODIPY C11^Ox) to total BODIPY C11 signal detected (BODIPY C11^Ox + BODIPY C11^Non-ox) in D2.0R control and D2.0R ACSL3^LOF cells on BME matrices for 6 days (mean ± s.e.m, n = 44–52 cells per group. Comparison by Mann-Whitney U test, two-sided. ∗∗∗P ≤ 0.0001). d. Quantification of dormant D2.0R control and D2.0R ACSL3^LOF cell numbers upon treatment with either vehicle, Fasnall and/or Fer-1 (upper graph); vehicle, AAPH and/or Fer-1 (center graph) and vehicle, hydrogen peroxide (H[2]O[2]) and/or Fer-1 (lowest graph) for 6 days (data show mean ± s.e.m, n = 3 wells. Comparisons by one-way ANOVA plus Tukey post-test at day 6 and relative to vehicle-treated control cells, with pairwise comparisons indicated by square brackets. NS, not significant; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001; ∗∗∗∗P ≤ 0.0001). e. Total lung surface burden of athymic nu/nu mice tail-vein injected with 1 × 10^6 D2.0R-GFP control or D2.0R-GFP ACSL3^LOF cells. The animals received a single-dose treatment with either saline (vehicle) or 200 mg/kg AAPH, via nasal instillation at day 18 post-injection of the tumor cells (data represent mean ± s.e.m, n = 8–10 mice per group. Comparison by Kruskal-Wallis, Dunn's post-test). f. Total lung surface burden of athymic nu/nu mice tail-vein injected with 1 × 10^6 D2.0R-GFP control or D2.0R-GFP ACSL3^LOF cells. The animals received a daily dose of Fer-1 (5,2 mg/kg) or vehicle (DMSO + Saline) intraperitoneally, for 3 weeks (data represent mean ± s.e.m, n = 8–10 mice per group. Comparison by Kruskal-Wallis, Dunn's post-test). Representative images of the data shown in e. and f. can be found in Supplementary [195]Fig. S11, a and b, respectively. Next, we CRISPR-edited D2.0R cells to both knockdown the expression of ACSL3 and specifically impair its Acyl-activating enzyme (AAE) consensus motif, which is the conserved catalytic domain responsible for the activation of newly synthesized lipids by ACSL3 ([196]Figs. S9 and S10) and observed a significant reduction over time in the number of viable dormant D2.0R cells in the edited vs. control groups ([197]Fig. 5b). Consistent with previous observations, knocking down the expression of ACSL3 resulted in a significant increase of lipid peroxides in dormant D2.0R cells compared to non-edited controls ([198]Fig. 5c). To further test the influence of fatty acid biosynthesis on lipid peroxidation and downstream ferroptosis in dormant BC cells, we performed a series of experiments aiming to rescue the viability of dormant D2.0R cells upon treatment with different oxidative stress inducers. To do that we used Fer-1, a specific ferroptosis inhibitor that prevents lipid peroxidation via radical scavenging [[199]55] and observed that the decrease in the viability of dormant D2.0R cells treated with Fasnall could be partially rescued by addition of Fer-1 to the 3D culture, regardless the expression levels of ACSL3 ([200]Fig. 5d, upper graph). This effect was also observed after treatment with AAPH, a specific lipid peroxidation inducer that generates peroxyl radicals without generating hydrogen peroxide (H[2]O[2]) as an intermediate [[201]56], confirming the involvement of lipid peroxidation in the viability decrease of dormant D2.0R cells ([202]Fig. 5d, middle graph). Moreover, we observed that the addition of Fer-1 alone to ACSL3 knocked down dormant D2.0R cells rescued their viability, further suggesting the key role of ACSL3 in protecting dormant breast cancer cells against ferroptosis ([203]Fig. 5d, middle graph). Finally, the viability of dormant D2.0R cells was not rescued using Fer-1 upon treatment with a general oxidative stress inducer, such as H[2]O[2] ([204]Fig. 5d, lower graph), indicating that both Fasnall and ACSL3 expression knockdown-mediated toxicities in dormant BC cells are mediated by lipid peroxidation and subsequent ferroptosis. To assess the role of ACSL3 in protecting dormant BC cells from lipid peroxidation and ferroptosis in vivo, we tail-vein injected ACSL3 knocked down and non-edited control D2.0R cells into mice and treated them either with vehicle or AAPH ([205]Fig. S11a). Consistent with the in vitro data described above, ex-vivo imaging showed a significant decrease in the metastatic lung burden of mice injected with ACSL3 deficient D2.0R cells as compared to control mice, which was further reduced by treatment with AAPH ([206]Fig. 5e and [207]Fig. S11a). In addition, we performed another experiment based on the same design ([208]Fig. S11b) but treated the mice with vehicle or Fer-1. Notably, ferroptosis inhibition rescued the viability of ACSL3 deficient D2.0R cells in vivo, as we observed a 173-fold increase in the metastatic tumor burden of mice injected with ACSL3 deficient D2.0R cells and treated with Fer-1 as compared to mice injected with the same cells but treated with vehicle. The raise in the metastatic tumor burden found in the lungs of mice injected with ACSL3 deficient D2.0R cells and treated with Fer-1 led to equivalent metastatic tumor burden compared to controls ([209]Fig. 5f and [210]Fig. S11b) suggesting that ACSL3 protects dormant BC cells through ferroptosis inhibition. 3.6. ACSL3 activates MUFAs to protect dormant breast cancer cells from ferroptosis We next asked if ACSL3 knock-down could influence the lipid profile in dormant BC cells. This question was relevant since it is well-known that monounsaturated fatty acids protect cells from ferroptosis while polyunsaturated fatty acids facilitate it [[211]28,[212]29]. Therefore, we subjected lipid extracts from dormant ACSL3 knockdown and non-edited control D2.0R cells to GC-MS analysis. We found that ACSL3 knockdown significantly decreased the relative abundance of MUFAs as compared to PUFAs previously reported to be prone to peroxidation and positively regulate ferroptosis, such as Docosahexaenoic acid (C22:6 n-3, DHA) and Adrenic acid (C22:4 n-6) [[213]57,[214]58] ([215]Fig. 6a, upper graph). Under the same conditions, additional treatment with Fasnall further decreased the relative abundance of all fatty acids but maintained the same lipid profile ([216]Fig. 6a, lower graph), in line with previous data suggesting that both Fasnall treatment and ACSL3 knockdown resulted in increased lipid peroxidation and decreased viability of dormant BC cells. It has been reported that treatment with exogenous MUFAs can protect cells from ferroptosis [[217]29,[218]59]. To test this observation in our experimental system, we supplemented the 3D culture media with either oleic (OA) or palmitoleic (POA) acid and treated dormant ACSL3 knockdown and control D2.0R cells with Fasnall. The data showed that the ferroptosis-protective effect of OA and POA acids is dependent on ACSL3 activity, since the addition of these two MUFAs was only able to rescue the viability of cells treated with Fasnall in the non-edited control cells but not in ACSL3 deficient dormant D2.0R cells ([219]Fig. 6b). Fig. 6. [220]Fig. 6 [221]Open in a new tab ACSL3 knockdown and fatty acid synthesis inhibition cooperatively reduce MUFAs synthesis and lipid peroxidation in the plasma membrane. a. Relative levels of eight different lipid species determined by GC-MS in D2.0R control cells and D2.0R ACSL3^LOF cells seeded on BME and treated for 5 days with either vehicle (upper panel) or Fasnall (40 μM) (lower panel). Data show the logarithm to the base 2 of the fold change for each lipid specie in the D2.0R ACSL3^LOF vs. D2.0R control cells (comparisons are relative to DHA and Adrenic acid by Kruskal-Wallis, Dunn's post-test. NS; not significant; ∗P ≤ 0.05; ∗∗P ≤ 0.01). b. Quantification of dormant D2.0R control and D2.0R ACSL3^LOF cell numbers upon treatment with either vehicle, Fasnall, exogenously added OA or/and exogenously added POA to BME-based 3D cultures for 6 days (mean ± s.e.m, n = 3 wells. Comparisons by one-way ANOVA plus Tukey post-test at day 6 and relative to vehicle-treated control cells, with pairwise comparisons indicated by square brackets. NS, not significant; ∗P ≤ 0.05; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001; ∗∗∗∗P ≤ 0.0001). c. Representative images of dormant D2.0R cells on BME matrices, either treated with vehicle or Fasnall (40 μM) and stained with BODIPY C11^581/591 and Cell Mask (plasma membrane marker). Scale bar is 20 μm and 5 μm. d. Percentage of co-localization between red (BODIPY C11^OX) and far red (Cell Mask. Colored in aqua for clarity) pixels and between green (BODIPY C11^Non-ox) and far red (Cell Mask. Colored in aqua for clarity) pixels ± Fasnall (mean ± s.e.m, n = 53–58 cells per group. Comparisons are relative to untreated control cells by Kruskal-Wallis, Dunn's post-test. ∗∗∗∗P ≤ 0.0001). DHA, Docosahexaenoic acid; MUFA, monounsaturated fatty acid; OA, oleic acid; POA, palmitoleic acid; PUFA, polyunsaturated fatty acid. Finally, we identified the plasma membrane as the subcellular location significantly affected by lipid peroxidation upon treatment with Fasnall in dormant D2.0R cells. We observed a significant increase in the co-localization of oxidized lipids and a plasma membrane marker in dormant D2.0R cells treated with Fasnall as compared with control cells treated with vehicle ([222]Fig. 6c and d). 3.7. ACSL3 expression levels are informative of DTCs metastatic fitness and disease progression in BC patients To validate the clinical impact of the findings described above, we stained BC patient samples for ACSL3 in primary tumor and overt metastatic lesions. In addition, we analyzed the expression level of ACSL3 in tumor cells disseminated to the lymph node, which also scored negative (<5 %) for the proliferation marker Ki67 and were presumably dormant ([223]Fig. S12a). Cells of epithelial origin were identified by their positive staining for cytokeratins (CK) in all samples. Quantitative image analysis was used to score CK + tumor cells as ACSL3^high, ACSL3^moderate, ACLS3^low ([224]Fig. 7a). This analysis revealed that the percentage of ACSL3^high cells dropped from 31 % to 36 % percent in metastasis and dormant cells disseminated to the lymph nodes, respectively, to 2 % in the primary tumor ([225]Fig. 7b), suggesting that ACSL3 expression is significantly upregulated in a subpopulation of tumor cells disseminated to secondary organs as compared to cells in the primary tumor. In addition, we observed a markedly different ACSL3 staining pattern in dormant DTCs in the lymph nodes, as compared to tumor cells in the primary tumor and metastatic lesions. In lymph nodes, 73 % of dormant DTCs exhibited a granular cytoplasmic ACSL3 staining. On the contrary, a significant proportion of tumor cells in the primary tumor (82 %) and metastasis (69 %) showed perinuclear ACSL3 expression ([226]Fig. 7c). These data suggested that the subcellular location of ASCL3 could be functionally relevant in dormant vs. proliferating tumor cells. In fact, it has been reported that ACSL3 localizes to membranes that belong to the trans-Golgi network [[227]60]. To test whether ACSL3 protein co-localized with the trans-Golgi network in dormant as compared to proliferative BC cells, we seeded D2.0R cells on BME (dormant) and BME + COL (proliferative) matrices and stained the cells for ACSL3 and a specific trans-Golgi network marker, such as TGOLN2. This experiment not only confirmed the increased expression of the ACSL3 protein in dormant vs. proliferative cells but, importantly, it showed a significantly higher degree of co-localization between ACSL3 and TGOLN2 proteins in dormant D2.0R cells on BME as compared to proliferative D2.0R cells on BME + COL matrices ([228]Fig. 7e and f). Fig. 7. [229]Fig. 7 [230]Open in a new tab ACSL3 expression promotes the survival of human DTCs and BC progression. a. BC patient samples of primary tumors (n = 16), lymph nodes with solitary DTCs (n = 20) and metastasis (n = 16) were immunostained for ACSL3, Ki67 and CK. Scale bar, 500 and 50 μm. The total number of cells analyzed is 102,585 cells for primary tumors, 113,390 DTCs in lymph nodes and 70,656 cells in metastatic lesions. b. Graph showing the percentage of ACSL3^high, ACSL3^moderate, or ACSL3^low cells in primary tumor (PT), DTCs in lymph node (LN), or mestastasis (M). ∗, P < 0.05 by Fisher's exact test. c. Graph showing the percentage of cells in PT, LN or M exhibiting either a granular cytoplasmic or perinuclear ACSL3 staining pattern. ∗∗∗∗, P < 0.0001 by Fisher's exact test. d. Kaplan–Meier plots of DFS (upper panel) or OS (lower panel) generated from 19 BC patients in this study stratified by low (n = 12) or high (n = 7) ACSL3 staining intensity of DTCs in the lymph node. Hazard ratio (HR) with 95 % confidence intervals and log-rank P value are shown. e. Representative ACSL3 (green) and TGOLN2 (red, trans-Golgi network marker) immunofluorescence of D2.0R cells seeded on BME matrices (dormant) or on BME + COL matrices (proliferative). Scale bar is 20 μm and 10 μm. f. Overlap coefficient between green (ACSL3) and red (TGOLN2) pixels (mean ± s.e.m, n = 111–118 cells per group. Comparison by Mann-Whitney test. ∗∗∗∗P ≤ 0.0001). To further explore the relevance of these data, we assessed how ASCL3 expression in the diverse samples analyzed (primary tumors, lymph nodes and metastasis) is correlated with the clinical outcomes of BC patients using Kaplan-Meier plot analysis. Kaplan-Meier plots showed that the probability of both DFS and OS was significantly lower only in patients who presented high ASCL3 expression in DTCs in the lymph node ([231]Fig. 7d), while the levels of ACSL3 expression in the primary tumor and metastasis did not significantly correlate with DFS and OS ([232]Figs. S12b and S12c). Therefore, the data presented herein would indicate that ACSL3 expression is clinically relevant only in the context of dormant DTCs in BC and that, in line with the pre-clinical data presented in this manuscript, it favors the survival of these cells. 4. Discussion DTCs undergo metabolic rewiring to adapt and survive in distant organs [[233]61]. Beyond cellular bioenergetics, metabolic adaptations are known to be crucial in maintaining cellular redox homeostasis [[234]22,[235]61]. Quiescent cells of diverse origin, such as memory T cells [[236]62], embryonic stem cells [[237]63,[238]64], and cancer or metastasis initiating cells, as well as dormant DTCs [[239]15,[240]19], have been shown to be highly sensitive to oxidative stress and, therefore, activate several different antioxidant mechanisms to counteract redox imbalance and survive. Here, we uncover a metabolic adaptation of dormant disseminated BC cells to endure lipid peroxidation and ferroptosis. Dormant BC cells divert a significant number of glucose-derived carbons from the TCA cycle to de novo lipogenesis, a metabolic pathway that generates saturated FAs (SFAs) and MUFAs as end products [[241]23,[242]24]. ACSL3 overexpression and activity functionally link de novo lipogenesis and ferroptosis resistance in dormant BC cells, through activation and incorporation of MUFAs to cellular membranes. Our data are consistent with previous results which reported the protective effect of incorporating exogenous MUFAs into cellular membranes in a ACSL3-dependent manner [[243]29,[244]65,[245]66]. Adding to this known mechanism, we provide the first experimental evidence that connects endogenous lipid biosynthesis to ACSL3-mediated resistance to ferroptosis. In melanoma, cancer cells metastasizing through the lymphatic system before entering the bloodstream are less sensitive to ferroptosis than those that disseminate directly through blood. The authors found that lymph node-derived melanoma DTCs incorporated significantly more OA to their cellular membranes (a process dependent on ACSL3 activity), were more resistant to erastin-induced ferroptosis and formed proliferative metastasis more efficiently, as compared to melanoma cells disseminating from the bloodstream [[246]65]. These results were aligned with previous data showing that human oral carcinoma cells scavenge exogenous lipids from the lymph node microenvironment to efficiently metastasize in a CD36-dependent manner [[247]67]. Indeed, one of the more striking differences between blood and lymph is the enrichment of the latter in OA, mainly contained in the form of triacylglycerols (TAGs) in ApoB^+ vesicles [[248]65]. It has been reported before that nutrient availability in the metastatic and even in the premetastatic niche, determine metabolic adaptations of DTCs and influences metastatic fitness. Exposure to exogenous palmitate can induce a prometastatic memory in primary oral carcinomas and melanoma via stable transcriptional and chromatin changes and stimulating tumor-associated Schwann cells and innervation [[249]68]. In addition, lung resident alveolar type II (AT2) cells have been shown to release and enrich the lung pre-metastatic niche in palmitate, as a response to soluble factors secreted by BC cells in the primary tumor. The exposure of BC DTCs to this palmitate-rich microenvironment induced fatty acid oxidation (FAO) through Carnitine palmitoyltransferase 1a (CPT1a) overexpression and the synthesis of acetyl-CoA, which is subsequently processed to acetylate nuclear factor-kappaB (NF-κB) subunit p65 through the lysine acetyltransferase 2A (KAT2A), finally resulting in enhanced metastatic capabilities of BC cells [[250]38]. On the contrary, in lipid-poor microenvironments, such as the brain, breast DTCs have been shown to upregulate fatty acid synthesis as compared to tumors growing in extracranial locations, including the primary site [[251]69,[252]70]. The authors concluded that the scarcity of lipids in the brain microenvironment drove a targetable metabolic dependency of BC cells on de novo lipogenesis. Interestingly, they found a relative enrichment in TAGs containing MUFAs in brain BC metastases, as compared to lipids analyzed in the breast tumors growing in the mammary fat fad of mice, which were highly abundant in TAGs containing PUFAs [[253]70]. Surprisingly, a recent paper described that BC latent cells disseminated to the brain uptake microenvironmental lipids secreted by activated astrocytes, to increase mitochondrial β-oxidation of FAs [[254]71]. Astrocytes have been shown to transfer PUFAs to melanoma cells in the brain, activating Peroxisome proliferator–activated receptor γ (PPARγ) and promoting the growth of advanced brain metastasis. Notably, PPARγ signaling exerted and anti-tumoral effect during carcinogenesis and early steps of the metastatic cascade [[255]72]. These findings are consistent with the pro-oxidant and deleterious effect of PUFAs enrichment in early DTCs. In addition, it has been reported that breast cancer cells switch transcriptional programs from lipid uptake to synthesis, depending on lipid availability [[256]73], which further highlights the metabolic flexibility of tumor cells in adapting to their environment. Furthermore, fatty acid synthesis and fatty acid oxidation have been shown to co-exist in BC cells and provide protection against acute increases in oxidative stress [[257]74]. Whether fatty acid synthesis and fatty acid oxidation take place simultaneously in dormant BC cells and their potential interplay to counteract oxidative stress, remain to be elucidated. This and other unresolved questions in the field of lipid metabolism and tumor progression warrant further investigation, but the data presented in this manuscript together with the above-described research allow to conclude that lipid availability dictates the metabolic adaptations of BC cells. In a context of limited microenvironmental lipid availability, such as early tumor dormancy, disseminated BC cells prioritize protection against lipid peroxidation and ferroptosis over other metabolic needs, through rerouting glucose-derived carbons to FA synthesis and MUFAs activation in a ACSL3-dependent manner. In addition, our data may explain previous observations in which DTCs never exposed to MUFA-rich environments can still form metastasis, although less efficiently [[258]65], due to de novo lipogenesis upregulation and ACSL3 activation. The opposite transition is also possible and, indeed, we have shown that exogenous supplementation of 3D cultures with either OA or POA rescued the decrease in viability observed in dormant BC cells treated with the FA synthesis inhibitor Fasnall, although this effect was restricted to ACSL3 proficient cells, further underscoring the key role of this enzyme in the survival of dormant BC cells. A recent study used single-cell RNA sequencing (scRNA-seq) to analyze gene expression differences between dormant and proliferative BC cells disseminated to the bone marrow [[259]75]. The authors identified a gene expression signature overexpressed in dormant BC cells disseminated to both the bone marrow and lungs. ACSL3 was not differentially expressed in dormant vs. proliferative DTCs in the bone marrow and, as a result, its expression in lung DTCs was not assessed. These apparent discrepancies with the data reported by Ren Q et al. may reflect the key influence of the microenvironment in shaping the genetic and metabolic adaptations of dormant BC DTCs. In addition, we have shown that BC cells significantly upregulate ACSL3 expression in secondary sites, such as the lymph node and distant metastasis, as compared with breast primary tumors. Interestingly, the main difference we observed with respect ACSL3 expression between DTCs in the lymph nodes and metastatic cells was the subcellular location of the ACSL3 protein, which was granular and cytoplasmic in the case of DTCs in the lymph nodes and perinuclear in cells within metastatic lesions. It has been reported that catalytically active ACSL3 protein localizes to membranes involved in the intracellular trans-Golgi network and endosomal trafficking system, which indeed presents a granular and cytoplasmic pattern when stained for specific proteins [[260]60]. Moreover, the trans-Golgi network and endosomal system comprise the main intracellular hub for lipid biosynthesis, sorting and transport to the plasma membrane [[261]76]. Along this line of reasoning, we found a significantly higher co-localization between ACSL3 and TGOLN2, which is a specific marker of the trans-Golgi network, in dormant vs. proliferating cells. This fact further argues in favor of the functional significance of the differential subcellular location of ACSL3 in dormant DTCs with respect tumor cells in the metastasis from BC patients, as we found that ACSL3 mediates the survival of dormant BC cells via de novo synthetized MUFAs activation and incorporation to the plasma membrane. Besides, the subcellular location of ACSL3 is strongly correlated with the proliferative status of the human tumor tissue analyzed. We found a predominant perinuclear ACSL3 staining pattern in proliferative cells from primary breast tumors and metastasis, as compared with the granular and cytoplasmic ASCL3 pattern exhibited by dormant DTCs in the lymph nodes. Accordingly, ACSL3 expression significantly impacted the survival rates of BC patients only when it was overexpressed in DTCs in the lymph node. It was intriguing to find that ACSL3 overexpression in DTCs at the lymph node was significantly associated with worse DFS and OS in BC patients. This suggests that ACSL3 expression and activity may increase the subpopulation of surviving DTCs capable of subsequent metastatic spreading and colonization at distant sites. Whether ACSL3 expression is dynamically modulated across the different stages of the metastatic cascade that involve cell quiescence, such as dissemination to the lymph nodes and tumor cell dormancy at distant metastatic sites, remains to be demonstrated. Notwithstanding, using human and mouse preclinical models of BC dormancy and BC clinical samples, we have shown that ACSL3 overexpression promotes the survival of dormant DTCs, regardless the site of dissemination. Overall, the data from this study reveal a novel molecular mechanism that identifies a metabolic dependency of dormant BC tumor cells, which might be exploited therapeutically to prevent metastatic recurrence. 5. Limitations of the study First the marked scarcity of validated BC dormancy models is a limitation for this study and the entire field. In vitro studies conducted using the 3D culture system are limited to 11–12 days by the unavoidable fact that the matrix degrades and detaches from the culture plates. This limit imposed by the 3D model system limit the follow up of cells seeded on the matrices. Second, due to the proliferation kinetics of D2A1 and MDA-MB-231 cells, which spontaneously outbreak into a proliferative state when culture in 3D matrices, it is possible to study cell behavior and phenotype upon different treatments or genetic modifications before and after the cell undergo the dormant-to-proliferative switch. However, evaluating such interventions in proliferating-to-dormant cells is not possible using the model systems reported herein. Finally, the analytical pipeline followed to process ACSL3 expression in human samples grouped the data from all the patients included in the study per each anatomical location and were presented as total percentage of cells. Likewise, analysis of ACSL3 and Ki67 expression were performed in serial cuts of the tissue microarrays where the samples were included instead of performing a double staining. CRediT authorship contribution statement Beatriz Puente-Cobacho: Writing – review & editing, Methodology, Investigation, Formal analysis, Data curation. Cintia Esteo: Methodology, Investigation. Patricia Altea-Manzano: Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Jose Luis Garcia-Perez: Investigation, Formal analysis. José L. Quiles: Investigation, Conceptualization. Pedro Sanchez-Rovira: Resources, Investigation, Conceptualization. María D. Martín-Salvago: Methodology, Formal analysis, Data curation. Lucía Molina-Jiménez: Methodology, Formal analysis, Data curation. Rafael J. Luque: Methodology, Investigation, Data curation, Conceptualization. Sarah-Maria Fendt: Writing – review & editing, Investigation, Conceptualization. Laura Vera-Ramirez: Writing – review & editing, Writing – original draft, Supervision, Resources, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Data availability * • Uncropped high-resolution scans of all the blots presented in this paper and images of ACSL3 antibody validation for immunohistochemical staining of human samples are available in the “Supplementary Data” file. * • The raw RNA-seq analyzed in this manuscript are available through Gene Expression Omnibus (GEO) under the accession number [262]GSE112094 . * • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Funding Work under the supervision of LV-R is supported by the Consejería de Salud y Familias, Junta de Andalucía (Spain), award numbers PI-0068-2021 and PI-0072-2019, and the Consejería de Universidad, Investigación e Innovación de la Junta de Andalucía (Spain), award number EMERGIA20_00313. B.P–C has received funding from the Ministerio de Universidades (FPU22/02634). P.A.-M. was supported by a Marie Sklodowska-Curie Actions individual fellowship (MSCA-IF-2018-839896) and has received funding from European Union (ERC-StG-101116912) and Beug Foundation. S.M.F. acknowledges funding from FWO Projects, Beug Foundation, Fonds Baillet Latour, KU Leuven, Interuniversity BOF (iBOF) programme and Stichting tegen Kanker. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: S.-M.F. has received funding from BlackBelt Therapeutics, Gilead and Alesta Therapeutics, is in the advisory board of Alesta Therapeutics, has consulted for Fund+ and Droia Ventures and is in the advisory board of several Cell Press Journals. All other authors have no conflicts of interest to declare. Acknowledgments