Background
Iron is an essential component of several cellular enzymes, such as catalases, peroxidases, cytochromes, ribonucleotide reductase, desaturases, and aconitase, which are crucial for physiological functions and have been implicated in several diseases, including cancer, because of alterations in iron metabolism [
1]. Adequate iron supply is critical for various cellular processes, including DNA synthesis and cell cycle progression. Many in vitro and in vivo studies have demonstrated that compared with normal cells, cancer cells are more sensitive to iron deprivation because of their marked Fe requirements [
2]. Furthermore, their strong dependency on iron is evidenced by their increased expression of transferrin receptors compared with that of normal cells. In vitro and animal studies have also indicated the antitumour activity of several iron chelators [
3‐
5].
Clinical data support the concept that iron deficiency increases angiogenesis and causes breast cancer recurrence [
6]. Specifically, iron deficiency can contribute to the high recurrence of breast cancer in premenopausal women, whereas iron load might play a role in the metastasization of breast cancer in postmenopausal women.
Iron chelators are known to induce apoptosis in several types of proliferating cells [
7] and are therefore considered promising anti-proliferative agents in the treatment of human cancers. Iron chelators, such as deferoxamine (DFO), deferiprone, and deferasirox, have several advantages: they have been clinically approved for iron overload disorders [
8], they have a well-studied, long-term use toxicity profile, and experimental results could be readily translated into clinical trials for cancer. Nevertheless, it is crucial to consider several side effects related to their use, including myelosuppression, hypoxia, and methemoglobinaemia, as observed from clinical trials [
9].
DFO is a clinically approved non-toxic iron chelator that has been effectively used for long-term iron chelation therapy in beta-thalassemia and other iron overload disorders. DFO has also been reported to have some antitumour activity [
10‐
13]. Novel chelators based on the di-2-pyridylketone thiosemicarbazone (DpT) scaffold, such as di-2-pyridylketone 4,4-dimethyl-3-thiosemicarbazone (Dp44mT), induce iron sequestration and also form redox-active metal complexes that demonstrate potent and selective antitumour activity [
14]. Notably, Dp44mT and its analogues possess broad anti-cancer and anti-metastatic activity, in vitro and in vivo, against several aggressive solid tumours [
15‐
20].
Relevant studies have focused on the causality between iron chelation and cancer cell death, without considering that substantial regions of cancers often grow in hypoxic conditions owing to the lack of a functional vasculature. The amount of bio-available iron is often limited in poorly vascularised areas, and the iron uptake in cancer cells is inadequate to fulfil their need. Moreover, the lack of cellular iron content has an essential role in positively regulating hypoxia-inducible factor (HIF) protein stability and therefore hypoxia mechanisms, even under non-hypoxic conditions [
21]. Thus, cancer cells are more susceptible to iron depletion than non-cancer cells, a phenomenon we have termed
iron addiction; however, it is important to note that cancer cells adapt in response to low iron levels, directly affecting cancer cell metabolism [
22]. Recent findings on medulloblastoma cell lines indicate that modulation of iron-related proteins during hypoxia may increase cell proliferation as well as tumour aggression and stemness [
23]. Oxygen and iron are intimately linked in producing signals through the hypoxia response pathway, and they exert considerable influence on cancer cell metabolism [
24].
Based on these observations, we aimed to understand the morphological, proteomic, and metabolic effects of iron depletion in breast cancer cells with a deeper insight into the cellular effects and drawbacks of iron starvation. We combined biochemistry, microscopy, flow cytometry, and mass spectrometry-based methods to investigate the cellular and molecular events induced by DFO or Dp44mT in two human breast cancer cell lines, MDA-MB-231 and MDA-MB-157.
Cellular stress and organelle dysfunctions due to the disruption of cellular homeostasis were observed, and these events led to cell death. A dramatic increase in endoplasmic reticulum (ER) complexity, with the appearance of large vacuoles and the accumulation of lipid droplets (LDs), accompanied mitochondrial dysfunction and bioenergetics collapse and was followed by cell death and LD leakage. The results of this study are expected to provide important insights into the fundamental molecular mechanisms of the adaptation of breast cancer cells to iron limitation.
Methods
Cell lines and culture
MDA-MB-231 and MDA-MB-157 breast cancer cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). They are characterised as triple-negative/basal B mammary carcinoma and considered models of triple-negative breast cancer growth and progression. Cell lines were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) containing 10% heat-inactivated foetal bovine serum (FBS; HyClone), DMEM w/ 4.5 g/L glucose w/o L-glutamine at 37 °C and 5% CO2 in air.
Reagents
Dp44mT and DFO were purchased from Sigma-Aldrich. Dp44mT was dissolved in dimethyl sulfoxide (DMSO) and further diluted to a final concentration of 5 μM in the culture medium, whereas DFO was diluted to a final concentration of 250 μM in the culture medium and used at a final concentration of 100 μM. Cells were incubated with either control media containing or not containing DMSO at 0.05% (v/v) to match the concentration of the dissolved Dp44mT or DFO.
Assessment of antiproliferative activity
3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide thiazolyl blue (MTT) assay was performed in 48-multiwell plates containing MDA-MB-231 cells, seeded at a concentration of 1.8 × 104 cells per well. After 24 h, MDA-MB-231 cells were treated with 5 μM Dp44mT or with 100 μM DFO for 120 h. Cell proliferation was assessed by MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; thiazolyl blue) assay according to the manufacturer’s recommendations (Sigma). Absorbance intensity was quantified at 492 nm using a microplate reader (Infinite 200, Tecan). Data are shown as mean ± standard error of the mean (SEM) of quadruplicated wells and are representative of three independent experiments. Statistical tests were performed using GraphPad Prism version 5.0 (GraphPad Software Inc.).
Sample preparation, SDS-PAGE, and immunoblotting
MDA-MB-231 and MDA-MB-157 cell lines were treated with the iron chelators for the times indicated earlier. Cell pellets were solubilised as previously described [
25]. Protein concentrations were detected using a bicinchoninic acid (BCA) kit (Pierce) according to the manufacturer’s instructions, and SDS-PAGE and electroblotting were performed, as previously described [
26]. The antibodies used for western blotting analyses are listed in Additional file
1.
Mitochondrial membrane potential analysis
After treating cells with the iron chelators for the times indicated earlier, the mitochondrial membrane potential of the cells was detected using a MitoProbe™ JC-1 assay kit for flow cytometry (Thermo Fisher Scientific). Flow cytometry analyses were performed as previously described [
27].
Fluorescence microscopy
The subcellular location of various proteins was visualised by fluorescence microscopy (Eclipse E1000, Nikon Instruments, Inc.). To visualise different subcellular compartments, we used antibodies for RAB5, RAB7, LAMP1, NDRG1, and RTN4 as markers for the ER and LysoTracker red (50 nM; Life Technologies) as a marker for lysosomes. To visualise mitochondria in live cells, we stained the cells with the vital dye Rhodamine 123 (500 ng/mL; Sigma-Aldrich) for 15 min at 37 °C. To visualise nuclei in live cells, we stained the cells with the cell-permeable DNA dye 4′,6-diamidino-2-phenylindole (DAPI) (10 μM; Molecular Probes) for 15 min at 37 °C.
Confocal microscopy
Confocal laser scanning microscopy was performed using the Leica TCS SP8 X microscope (Leica Microsystems GmbH).
Nano-scale LC-MS/MS analysis
Cells were lysed in 50 μL of 0.1% RapidGest SF Surfactant (Waters) and diluted in 50 mM ammonium bicarbonate (pH 8.0). After reduction and alkylation, the proteins were digested with trypsin sequence grade (Roche) for 16 h at 37 °C using a protein:trypsin ratio of 20:1 [
28]. LC-ESI-MS/MS analysis was performed on a Dionex UltiMate 3000 HPLC System with a PicoFrit ProteoPrep C18 column (200 mm in length and with an internal diameter of 75 μm) (New Objective). The eluate was electrosprayed into an LTQ Orbitrap Velos (Thermo Fisher Scientific) through a Proxeon nanoelectrospray ion source (Thermo Fisher Scientific) [
29]. Four technical replicate analyses of each sample were performed. Data acquisition was controlled by Xcalibur 2.0 and Tune 2.4 software (Thermo Fisher Scientific).
Mass spectra were analysed using the MaxQuant software (version 1.3.0.5). Enzyme specificity was set to trypsin. Carbamidomethylcysteine was set as a fixed modification, and N-terminal acetylation, methionine oxidation, and asparagine/glutamine deamidation were set as variable modifications. The spectra were searched by the Andromeda search engine against the human Uniprot sequence database (release 29.06.2015). Protein identification required at least one unique or razor peptide per protein group. Quantification in MaxQuant was performed using the built-in XIC-based label-free quantification (LFQ) algorithm [
30] using fast LFQ. The required false positive rate was set to 1% at the peptide level and 1% at the protein level. Statistical analyses [
31] were performed using the Perseus software (version 1.4.0.6,
www.biochem.mpg.de/mann/tools/).
Only proteins present and quantified in at least 3 out of 4 technical repeats were considered positively identified; 2478, 2488, and 2496 proteins were identified in untreated, DFO-, and Dp44mT-treated cells, respectively; 108 proteins were exclusively expressed in untreated/control cells, 54 proteins in DFO/cells, and 137 proteins in Dp44mT/cells. An ANOVA test (false discovery rate 0.05) was carried out to identify proteins differentially expressed among the three conditions: 1682 out of 2149 common proteins differed with statistical significance and were selected for further analyses. In particular, we focused on the differential proteomics between proteins expressed in untreated cells and proteins expressed in DFO- or Dp44mT-treated cells. Differential expression was considered significant if (1) a protein was present only in untreated or treated cells or (2) its normalised (according to the LFQ algorithm) intensity resulted in a statistical difference, as calculated by the Welch’s
t-test (
t-test cut-off at
p value = 0.0167). The MS proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository [
32] with the dataset identifier PXD007595.
Gene ontology (GO)
The Search Tool for the Retrieval of INteracting Genes/proteins (STRING) database (version 10.5, Database issue: D412–416) [
33] was used for prediction of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways [
34‐
36]. A GO scatterplot was constructed in Excel.
Oil red O staining
To determine the presence of LD accumulation within MDA-MB-231 and MDA-MB-157 cells, Oil Red O (Sigma-Aldrich) staining was performed. To visualise cell nuclei, samples were stained with haematoxylin (Sigma-Aldrich). Cells were imaged on a Leica DM IRB microscope (Leica Microsystems).
Fatty acid (FA) quantification in lipid droplets
Cells were cultured in 10-cm dishes for 96 h in the presence of 100 μM DFO or Dp44mT. The presence of LDs was evaluated with Oil Red O staining. Cell debris was recovered from the plates and LDs purified by density sucrose gradient [
37].
Lipids were prepared by homogenizing the samples in ethanol containing (50 ppm) butylated hydroxy toluene (BHT) to avoid oxidation [
38]. A lipid chromatogram was obtained by gas chromatography–mass spectrometry (GC-MS) analysis using a Shimadzu gas chromatograph equipped with a quadrupole mass spectrometer for electron impact ionisation (GC-MS-QP2010). An SH Stabilwax DA column (30 m in length, 0.25 mm in diameter, and with a film thickness 0.25 μm) was used to separate the FA methyl ester at a flow rate of 1.0 mL/min. The injector temperature was set to 200 °C and the transfer line temperature to 280 °C. The GC oven was programmed as follows: after 2 min at 50 °C, the temperature was increased at 30 °C/min to 150 °C, then at 15 °C/min to 230 °C. The total run duration was 25 min. GC-MS analysis was conducted in the full scan mode (m/z 35–600). Qualitative analysis was based on the characteristic ions of the FA methyl esters and their relative retention times. Quantitative analysis was based on the ratio between the peak area of each FA and the corresponding internal standard peak area, using the respective standard curves.
Raman spectroscopy
To perform coherent anti-Stokes Raman scattering (CARS) imaging, a home-built laser scanning multi-modal microscope, described previously [
39], was utilised. To acquire CARS images, the treated cells were placed under the laser focus of the microscope, and the laser spot was galvo-scanned over the 50 × 50 μm
2 sample area at two fixed pump-Stokes frequency detuning: 2850 cm
− 1 (for lipid molecules) and 3050 cm
− 1 (for medium and vacuoles). Spectral profiles and spatially resolved chemical maps of the pure chemical components were generated using the MATLAB tool box MCR-ALS (Multivariate Curve Resolution Alternating Least Squares) algorithm [
40,
41]. We collected two images: one at 2850 cm
− 1, indicating the symmetric vibration of CH
2 chains in lipids, and the other at 3050 cm
− 1, indicating the stretching vibration of hydrogen-bonded water of saline environments. The output of the software was two matrices (representing the concentration at 2850 cm
− 1 and 3050 cm
− 1) and two spectra (representing the two identified species). For the set-up of the figure, we have plotted these two matrices from grayscale to a single-color scale (green for lipids, blue for medium) and created a composite image thereof with the FIJI image processing package.
Discussion
Iron is an essential nutrient required for many biosynthetic and metabolic processes, such as haem and sulphur–iron cluster biosynthesis, oxygen transport, mitochondrial respiration, lipid desaturation, and DNA replication [
72]. Compared with normal cells, cancer cells require a large amount of iron for several essential processes. Iron bio-availability can be affected by the lack of functional vasculature in tumours, resulting in a hypoxic/anaemic tumour microenvironment. Iron affects the HIF pathway, and iron deficiency can produce responses similar to hypoxia, as observed from the effects of iron chelation in cell culture and in humans [
73]. However, the effects of iron deficiency on hypoxia behaviour and signalling mechanisms have not been completely understood.
Our study provides novel insights on how iron deprivation modulates the cell structure and the proteomic and metabolic pathways in breast cancer cells. Prolonged iron deficiency in cultured tumour cells induces significant changes in cellular phenotypes and mesenchymal markers and an adaptive cellular response similar to that induced by hypoxic stress, prior to killing the tumour cells [
74]. HIF-1 was up-regulated in MDA-MB-231 cells (Additional files
4,
6) after treatment with iron chelators, sustaining EMT and hypoxia programs; mesenchymal marker genes, NDRG1 and vimentin, increased in parallel with the decrease in E-cadherin expression and the increase in CD44 and CD166 expression, which are well-known cancer stem cell markers [
75] (Additional files
4,
6). Remodelling of iron-starved cells showed a marked reduction in translation and the increase of redox pathways, along with perinuclear mitochondrial clustering disruption, depolarisation, and a hypoxia-like reprogramming of metabolism. The reprogramming suggests an adaptation of the cancer cells to iron deprivation, which is accompanied by the augmentation of mesenchymal characteristics.
The expanded ER formed large vacuoles and a complex network of membranes around the nucleus. Coalescent spherical units of translucent material around the nucleus were stained with lipid dyes such as Oil Red O and observed using Raman spectroscopy. These structural modifications occurred in parallel with the increased expression of ER markers (NDRG1 and RTN4) and LD-associated proteins (PLIN2/PLIN3). Using Raman spectroscopy, we could also determine that the vacuoles contained sacs of fluid-filled extracellular medium. These modifications increased with time of treatment with DFO or Dp44mT and were tolerated by the cells before they underwent cytoplasmic collapse, succumbing to a non-apoptotic and non-autophagic type of death.
Thus, under iron deprivation, breast cancer cells exhibited a large amount of fluid-filled vacuoles and lipid accumulation. A type of extensive vacuolation followed by cell death, known as methuosis, or death by macropinocytosis has been previously described as an efficient method for inducing cell death in different types of cancers [
76,
77]. Proteomics analyses of pre-death cells clearly indicated a striking metabolic plasticity, based on the scavenging of nutrients destined for short-term cell survival.
By recycling nutrients from intracellular macromolecules, autophagy represents an important cellular strategy to sustain viability during periods of limited nutrient availability [
78]. We observed a blocking of autophagy and possible loss of cell viability, which may be derived from the cytoplasmic hyper-vacuolisation associated with the loss of metabolic capacity (as indicated by the decrease in mitochondrial membrane potential) and plasma membrane integrity, until osmotic pressure permitted. Cells formed non-acidic and non-autophagic vacuoles, with massive cytoplasmic vacuolations developing from the dilation of the ER lumen. The formation of intracellular vacuoles and subsequent cell death indicated that the events occurring in the ER initiate this destructive pathway. This massive dilation of the ER finally culminates in a paraptosis-like cell death [
3], releasing LDs into the extracellular environment. Consequently, in vivo anaemic/hypoxic regions in tumours with poor iron availability may be considered to be at risk for dissemination of possible bioactive vesicles.
The association between LD metabolism and cancer cell survival and metabolism is unknown. LD accumulation probably results from the increased lipid scavenging activity in MDA-MB-231 cells, rather than augmented lipogenesis [
79]. In accordance with this view, we observed a reduction in FASN levels, which could limit the initial step in FA biosynthesis, and the KEGG indication of an increase in lipid degradation, suggesting that cancer cells scavenge lipids from the extracellular environment. Importantly, hypoxic cells exhibit increased uptake of unsaturated lipids from their environment, thus bypassing the requirement for FA desaturation [
80]. In fact, FA desaturases, mainly SCD, are among the most sensitive targets of oxygen and iron starvation [
81]. We observed only a decrease in delta-6 desaturase (FADS2) abundance in treated cells accompanied by the increase in LD accumulation. Based on this result, we expected a change in lipid composition with an increase in saturated FAs. Instead, a large amount of monounsaturated oleic acid was observed in DFO-and Dp44mT-induced LDs (29.3 and 38.4% of total FAs), which is generally regarded as cytoprotective [
82]. Although an abundance of saturated stearic acid (27 and 31.5%), possibly due to the reduction of desaturase functions, was also observed in the LDs, the unsaturated/saturated FA ratio indicated a greater content of unsaturated FAs. Some polyunsaturated FAs observed in the LD analysis have distinct and contrasting effects in cancer: arachidonic acid mostly exhibits pro-tumourigenic effects [
82]; eicosapentaenoic (EPA; 20:5, ω-3) and docosahexaenoic acids (DHA; 22:6, ω-3) possess anti-tumourigenic, anti-inflammatory, and pro-apoptotic effects in cancer cells [
83].
The mitochondrial dysfunction observed in our studies does not support an active metabolic role for LDs and further suggests that other compensatory metabolic adaptations to support cell survival under conditions of iron starvation must be considered.