Glucose is a key driver for GLUT1-mediated nanoparticles internalization in breast cancer cells

The mesenchymal state in cancer is usually associated with poor prognosis due to the metastatic predisposition and the hyper-activated metabolism. Exploiting cell glucose metabolism we propose a new method to detect mesenchymal-like cancer cells. We demonstrate that the uptake of glucose-coated magnetic nanoparticles (MNPs) by mesenchymal-like cells remains constant when the glucose in the medium is increased from low (5.5 mM) to high (25 mM) concentration, while the MNPs uptake by epithelial-like cells is significantly reduced. These findings reveal that the glucose-shell of MNPs plays a major role in recognition of cells with high-metabolic activity. By selectively blocking the glucose transporter 1 channels we showed its involvement in the internalization process of glucose-coated MNPs. Our results suggest that glucose-coated MNPs can be used for metabolic-based assays aimed at detecting cancer cells and that can be used to selectively target cancer cells taking advantage, for instance, of the magnetic-thermotherapy.


Glucose coated CoFe 2 O 4 NPs characterization and biocompatibility validation.
To have the control on the MNPs properties and their chemical functionalization we synthesized CoFe 2 O 4 NPs in the laboratory, following the protocol described in the Methods section. The covalent binding of glucose and its fluorescent analogue, the 2-2-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)Amino)-2-Deoxyglucose (2-NBDG), to the CoFe 2 O 4 NPs has been chemically characterized using Attenuated Total Reflection-Fourier Transform Infra-Red (ATR-FTIR) spectroscopy. The absorbance spectra of citrate stabilized MNPs (blue line), 2-NBDG functionalized MNPs (black line) and glucose MNPs (red line) are reported in Fig. 1a. The ATR spectrum of the citrate stabilized MNPs is characterized by few broad bands, the most intense of which are centered at 1396 cm −1 (symmetric stretching of COO − ) and 1579 cm −1 (asymmetric stretching of COO − ). This vibrational pattern reveals the presence of free carboxylate moieties, while the shifts of the carbonyl group of carboxylic moieties from 1710 cm −1 , in free solution, to 1679 cm −1 proved the partial single bond character of the C= O group and consequently the chemisorptions of carboxylate ions onto the MNPs surface 23,24 . The coordination band was further shifted as a consequence of the functionalization with the glucose derivative while the carboxylate spectral features were greatly suppressed due to the esterification reaction involving the -OH group of 2-NBDG. A further proof of the covalent ester bonding formation was given by the absorbance peak at 1737 cm −1 in MNPs -2NBDG sample as for the glucose one. The absorbance peaks in the spectral region between 1200 and 900 cm −1 are correlated to specific vibrational characteristics of glucose molecule.
The size of MNPs has been measured by Scanning Electron Microscopy (SEM), finding a mean diameter of 27 ± 3 nm for all the three samples (citrate, 2-NBDG and glucose coated NPs) (Fig. 1b,c). Neither the functionalization with 2-NBDG nor the one with D-glucose affected the mean diameter of NPs due to the under-nanometer size of functionalizing agents.
Before proceeding with the MNPs administration to breast cancer cell lines, both MDA-MB-231 and MCF7 cell lines have been evaluated for their specific marker expression confirmation (data provided as Supplementary Fig. 1 and Supplementary Methods). To exclude possible side effects, cell lines have been firstly tested for toxicity related to the use of MNPs functionalized with the 2-NBDG. A 3-(4,5-dimethylthiazol-2-yl)-2,5-diph enyltetrazolium bromide (MTT) assay viability assay has been employed to assess the low cytotoxicity of the 2-NBDG-MNPs administration at 2.5 μ g/mL, equivalent to 4.6*10 10 NPs/mL in accordance to literature 25 (see Supplementary Fig. 4).
Glucose concentration in culture medium drives the uptake of glucose coated CoFe 2 O 4 NPs. To assess differences in the glucose uptake between mesenchymal-and epithelial-like tumor cells, we investigated the ability of MDA-MB-231 and MCF7 cells in internalizing glucose-modified MNPs. The shift of glucose concentration from low to high was already proven to promote the proliferation of mammary cells 26 . Here, we analyzed MNPs uptake to prove salient disparities between epithelial-like and mesenchymal-like cells. Three different techniques have been employed: Perls' iron staining ( Fig. 2 and Supplementary Fig. 5), fluorescence confocal microscopy ( Fig. 3) and FIB cell milling (Fig. 4). Perl's iron staining was firstly used to assess the presence of MNPs inside the cells through the recognition of the iron. Confocal microscopy was used to quantify the fluorescent MNPs uptake avoiding underestimation and FIB cell milling was finally employed to further demonstrate the presence of internalized glucose-MNPs both as aggregates and single particles.
At normal glucose concentration, both epithelial-like and mesenchymal-like cell lines were found positive to MNPs uptake, although the MNPs internalization by MDA-MB-231 cells was significantly higher with respect to that of MCF7. This difference further increased at high glucose concentration, since the uptake ability of MCF7 cells resulted to be significantly reduced whereas that one of MDA-MB-231 remained unchanged. In fact, as it is shown in Fig. 2b (black bars), the uptake for MCF7 cells decreased by 79.3% (from 58 to 12%) from normal to high glucose concentration, while it remained unchanged for MDA-MB-231 cells (Fig. 2d, white bars).
Data recorded from co-culture samples are reported in Fig. 2e,f. Although the subtype discernment was not achievable, due to the absence of a cell specific marker, counting the Prussian-blue positive cells provided valuable results. The percentage of positive cells at low glucose concentration (77 ± 12%) was lower than that registered for the MDA-MB-231 cells alone, result explicable considering the contribution of MCF7 cells. Moreover, this effect was confirmed at high glucose concentration where a percentage of 47% positive cells were found. In fact, as expected, the number of positive cells in co-culture, due to the contribution of MCF7 cells, was more evident at high than at low glucose concentration (Fig. 2f).
To establish that MNPs were indeed internalized by cells and not simply adherent to the cell surface, their distribution within the cells (previously treated with CoFe 2 O 4 -2-NBDG NPs) were confirmed by collecting z-stack confocal images of cells labeled by DAPI and phalloidin to identify nuclei and cytoplasmic actin filaments, respectively (see Supplementary Movies 1 and 2). Figure 3a,b show representative fluorescence images of MCF7 and MDA-MB-231 cells in which internalized fluorescent CoFe 2 O 4 -2-NBDG NPs (red color) are illustrated to exemplify the distinct behavior of the two cell lines. Specifically, at normal glucose concentration both epithelial-like and mesenchymal-like cell lines were found positive to MNPs uptake, with pronounced particle internalization by MDA-MB-231. On the contrary, at high glucose concentration, the uptake ability of MCF7 cells was reduced whereas that of MDA-MB-231 remained almost unchanged (Fig. 3d). MNPs uptake has been tested also in co-cultures of the two breast cancer cell lines using an anti E-cadherin antibody to specifically detect, within the co-culture, MCF7 cells (see Methods). MNPs-fluorescence quantification by confocal microscopy of the E-cadherin -positive and -negative cells revealed comparable uptake values to those observed in single cultures ( Fig. 3c and Supplementary Movie 3). Specifically, there was a significant increased uptake of MNPs in the E-Cadherin negative cells with respect to those positive for the epithelial surface markers at both glucose concentrations. Moreover, as for the MCF7 in single culture, the uptake of MNPs in E-cadherin positive cells significantly decreased when cultured at higher glucose concentration (Fig. 3c,d).
Since the resolution of the confocal microscope is unable to resolve single MNPs we decided to quantitatively evaluate the uptake difference between the two cell lines by measuring the total fluorescence emitted by the cells rather than counting the fluorescent particles. Specifically, analyzing samples treated or not by MNPs, we defined the uptake ratio, R u , (see methods) as a metrics to evaluate MNPs uptake. The quantitative analysis confirmed that at low glucose concentration, MDA-MB-231 cells showed a significantly higher glucose uptake than MCF7 cells. Nevertheless, increasing the glucose concentration the uptake did not change significantly for MDA-MB-231 cells (from R u = 1.10 to R u = 1.05, p = n.s., Fig. 3d) while it decreased by 85.4% in MCF7 cells (from R u = 0.48 to R u = 0.07, p < 0.01, Fig. 3d).
Scientific RepoRts | 6:21629 | DOI: 10.1038/srep21629 The increased uptake of MNPs by MDA-MB-231 cells, with respect to MCF7 cells, could be explained by the different amount of GLUT1 protein expressed by these two breast cancer cell lines ( Supplementary Figs 2  and 3). In fact, at 5 mM glucose concentration the relative amount of GLUT1 was about 3.8 times higher in the mesenchymal-like representative cell line respect to the epithelial-like one, while at high glucose concentration (25 mM) this ratio increased to 10 after 48 hours (see Supplementary Figs 2 and 3). Consistently, the glucose modified MNPs uptake paralleled the GLUT1 expression levels. At high glucose concentration, the uptake ability of MCF7 cells was significantly reduced whereas that one of MDA-MB-231 remained unchanged.
The further confirmation of MNPs internalization by breast cancer cell lines was achieved by the FIB precise milling technique coupled with SEM inspection (see Methods). Two representative images of MCF7 and MDA-MB-231 cell lines, treated with CoFe 2 O 4 -2-NBDG NPs and that subsequently underwent to the cytoplasm precise FIB milling, are provided in Fig. 4. In the MCF7 case (Fig. 4a) an image of the inner part of the cytoplasm is reported, where several white objects can be appreciated (Fig. 4b). These white particles showed an average size of 30 nm, in accordance with the measured mean diameter of functionalized CoFe 2 O 4 NPs. The white color of these is due to the electronic contrast that belongs to the different atomic mass of metals (cobalt and iron in this case), usually completely absent into cells. In support, also not-treated cells underwent the milling process and such evidence was not registered at all. In Fig. 4c an example of MDA-MB-231 cells showing few big MNPs clusters (around 500 nm) on the membrane can be seen. When the cells were investigated by the FIB-SEM technique, beside the aggregate, it was also possible to assess the presence of several internalized single-particles, pointed out by the measuring bars, with a size in accordance with that of MNPs, in Fig. 4d.
The glucose functionalization is necessary for MNPs uptake. The data obtained on CoFe 2 O 4 -2-NBDG NPs uptake provided the insights for further investigation about the internalization process and its reliance on glucose shell. To exclude that MNPs internalization was a process independent from the glucose functionalization, citrate MNPs have been used as controls. When the fraction of cells internalizing MNPs was evaluated by the Perls' iron staining, with respect to citrate MNPs, a significantly increased fraction of cells resulted to internalize CoFe 2 O 4 -2-NBDG and CoFe 2 O 4 -D-glucose NPs, with no differences between these two latter groups (see Supplementary Fig. 6).
In order to better appreciate the MNPs uptake process, a time dependent analysis was carried out revealing a saturation value at 2 hours for both breast cancer cell lines. In more detail, the MCF7 samples evidenced an increasing uptake of 2-NBDG and D-glucose coated MNPs at low glucose concentrations with about 60% of positive cells after 120 minutes ( Supplementary Information Fig. 6a). The glucose-independent uptake of Data are presented as mean ± standard deviation. *, **, ***p < 0.01 vs. the first, the second and the third column, respectively (One way Anova followed by post-hoc Bonferroni test). Scale bars: 50 μ m. citrate-coated MNPs by MCF7 cells ranged between 10 and 20% after 2 hours of incubation. Hence, the 2-NBDG (or glucose) vs. citrate coated MNPs intake at 5.5 mM concentration was significantly different in MCF7. Conversely, at high glucose concentration, a quite-similar number of positive cells has been registered regardless of the functionalization, ranging from 10 to 18%, leading to the hypothesis of a completely glucose-independent uptake mechanism at this concentration in MCF-7 cells ( Supplementary Information Fig. 6b).
Considering MDA-MB-231 samples, the estimation of total Prussian-blue positive cells confirmed a time-dependent internalization trend at both glucose medium concentrations, with a peak, as for the epithelial-like counterpart, at 2 hours (Supplementary Information Fig. 6c,d). In the case of citrate coated MNPs, the total number of positive cells was ranging between 5 and 20%, depending on the incubation interval. Differently from MCF7 cells, MDA-MB-231 cells maintained a high percentage of 2-NBDG/D-glucose MNPs-positive cells at both low and high glucose concentration, being the internalization always significantly superior to that of the citrate counterpart. Comparing 2-NBDG with D-glucose functionalization, no remarkable differences were appreciated among low and high glucose-concentrations: 87% and 83% at 5.5 mM; 77% and 81% at 2 mM ( Supplementary Information Fig. 6c,d).

Glucose transporter 1 guides the uptake of CoFe 2 O 4 -2-NBDG NPs.
To confirm the role played by GLUT1 and explain the differences in the observed high glucose uptake, GLUT1 specific inhibition was carried out using both a chemical compound (STF-31) and a biological inhibition (small interference RNA, siRNA, transfection).
The STF-31 cytotoxicity has been evaluated by MTT assay, executed at different concentrations and time points, before proceeding with the MNPs uptake inhibition studies (see Supplementary Methods and Supplementary Fig. 7).
Drug concentrations of 5 μ m for MCF7 and 10 μ m for MDA-MB-231 have been chosen as the lowest concentration inducing a time-dependent MNPs uptake reduction without observing a marked cytotoxicity for each cell type. Indeed, at double concentrations (10 μ m for the epithelial-like subtype and 20 μ m for the mesenchymal-like one) an earlier uptake inhibition was observed together with an increased cytotoxicity. The MNPs uptake inhibition was quantified by fluorescence assay for the CoFe 2 O 4 -2-NBDG NPs and by Perls' staining for the D-glucose. STF-31-mediated inhibition revealed an essential involvement of the GLUT1 transporters in the glucose modified MNPs internalization in both breast cancer cell lines. In the case of MFC7, the first statistically significant uptake inhibition effect was achieved after 4 hours with the maximum reached after 8 hours (Fig. 5a,c). In contrast, MDA-MB-231 cells revealed a significant CoFe 2 O 4 -2-NBDG NPs uptake diminution after 16 hours of 10 μ M STF-31 incubation, reaching the maximum effect at 32 hours (Fig. 5b,d). The fluorescence CoFe 2 O 4 -2-NBDG NPs evaluation at the last incubation time points furnished, for both cell lines, a really low "Uptake Ratio" value: 0.1 for MCF7 and 0.2 for MDA-MB-231 (Fig. 5c,d). In order to confirm these data, Perls' iron staining was performed as well, finding about 20% of Prussian-blue positive cells after 8 and 32 hours for MCF7 and MDA-MB-231, respectively (see Fig. 5e,f). Again, no differences between the kinetic of inhibition of MNPs internalization was observed for MDA-MB-231-7 at 5.5 and 25 mM. The biological inhibition of GLUT1 protein expression has been obtained by transfecting a specific siRNA against GLUT1 mRNA and the inhibition of MNPs uptake has been evaluated 72 hours after transfection. As shown in Fig. 6, the specific inhibition of GLUT1 protein synthesis determines a consequent lower amount of GLUT1 protein presented on membranes. This diminished protein level, in both breast cancer cell lines, is associated with a significant decrease in the 2-NBDG modified MNPs internalization. Specifically, after GLUT1 siRNA administration, MCF7 and MDA-MB-231 cells showed about 13% and 15% of MNPs positive cells, respectively (Fig. 6d,h), corresponding to a 78% and 81% fold decrease in the corresponding uptake capacity, respectively.

Discussion
In the last decade, several types of NPs have been developed and proposed in the medical research field. Although diverse size and core-shell properties have been suggested and deeply investigated for in vivo purposes, several limitations slowed down their clinical applications 27 . In particular, their pronounced tropism for filter organs, with the consequent accumulation and cytotoxicity, enormously delayed the translation into therapy. On the contrary, from a diagnostic point of view, MNPs received very much attention for the development of new and efficient rare cancer-cells enrichment techniques 28 . In this field the only FDA-approved method to detect and count rare circulating tumor cells (CTCs) in metastatic breast, colon and prostate cancer is CellSearch (Veridex) 29 . This method is based on the selection of circulating cells expressing the epithelial specific EpCam antigen. Since this latter is quite low expressed by White Blood Cells (WBCs), it received an extraordinary attention to detect tumor cells of epithelial origin both for diagnostics and in targeted treatment approaches 30,31 . However, since the expression of EpCam can be lost by cells in EMT 32 , this method can fail in detecting mesenchymal CTCs. To detect mesenchymal CTC it has been so far proposed a multi-parametric approach based on the combined expression of markers useful for the recognition of mesenchymal cancer cells: CD15 33 , HER-2 34 , CD34 35 , CD44 35 , CD45 36 , CD47 36 , Plastin-3 37 , Vimentin 38 are the most diffused biomarkers used to enrich and identify CTCs of mesenchymal-origin. These aggressive and metastatic-prone cells own a higher glucose metabolism exhibiting the Warburg effect (as described in the introduction), with a consequently higher demand of energy and glucose uptake needs. The main glucose uptake way is via the glucose transporter channels (Glut family) and their expression is particularly sustained in the mesenchymal cancer subtype.
Here we introduced a MNPs-based method that in principle could provide a simple and low-cost implementation to identify GLUT1-overexpressing cells. Exploiting the glucose cell metabolism, the glucose modified MNPs developed in this work allowed targeting the highly metabolic active breast cancer cells. Epithelial-like and mesenchymal-like cell models have been investigated and compared from a metabolic point of view, leading to the conclusion that at low glucose condition they both internalized glucose-modified MNPs, even though with an appreciable disparity due to their relative GLUT1 protein expression. Increasing the medium glucose level to a non-physiological concentration (25 mM) reflected into an even higher difference between the two cell models tested. Indeed the demonstration of a very well distinguished MNPs uptake ability at 25 mM concentrated glucose could lead in the next future to new feasible solutions in CTCs diagnostics. Moreover metabolism-based methods could implement the current diagnostic techniques by simplifying and improving the mesenchymal-cancer state definition, as demonstrated by GLUT1 immunostaining.
In this work we provided advancements to the current knowledge by the development of a new metabolic assessment of cancer cells, based on the use of MNPs. In particular, we presented a successful functionalization of CoFe 2 O 4 NPs with glucose molecules and its fluorescent analogue, the 2-NBDG, where the covalent linkage was confirmed by FTIR spectroscopy and fluorescence optical microscopy. The cobalt ferrite nanoparticles due to their higher magnetization characteristics, with respect to normal iron oxide nanoparticles 39 , and the low cytotoxicity exhibited may allow to decrease the concentrations and hence becoming suitable also for future in vivo applications. The added value of glucose shell could strongly boost the cancer-targeted therapy allowing to selectively penetrating the tumor and hence reducing the accumulation in non-pathological tissues.
The glucose functionalization of MNPs allowed us to target the more metabolic active breast cancer cell line, in vitro, discerning it from the less metabolic active one. Indeed, the GLUT1 overexpressing cell line (MDA-MB-231) internalized a statistically significant higher amount of glucose-coated MNPs, with respect to the MCF7 counterpart, at both glucose medium concentrations assessed. Actually, the most remarkable result was obtained at 25 mM concentration, since at this concentration, with respect to 5.5 mM glucose, the epithelial-like cells lost three/fourth of their ability to uptake glucose coated MNPs as a consequence of the reduced GLUT1 protein expression. Considering the GLUT1 binding affinity for glucose similar in the two cancer cell lines, results can be interpreted as a consequence of a different modulation of GLUT1 expression in the two cell lines as a consequence of changes in the glucose concentration. Even after GLUT1 inhibition, both with chemical and biological methods, a variable percentage in the range of 15 to 20% of cells still continued to internalize glucose-modified-MNPs. This behavior could be interpreted as to be associated with a non-GLUT1 dependent mechanism, introducing the possibility that also other Glut-family members could be important for the MNPs uptake mechanism. For instance GLUT3, GLUT14 and GLUT12 could be involved by observing their relative gene expression (see GEO dataset, GEO accession number: GSE41445) even if GLUT3 is known to be mainly involved in neurons and there are just few evidences about GLUT12 involvement in mammary tumors 40 .
By the use of SEM the MNPs uptake has been confirmed, at nanometer resolution, for both cell types. Focused Ion Beam (FIB) milling, as reported in Fig. 4, made precise investigation of cell cytoplasm and nucleus possible. The recurring occurrence observed, with each technique and confirmed by SEM analysis, was the presence of bigger MNPs aggregates on MDA-MB-231 cells and smaller ones on MCF7 at 5.5 mM glucose concentration. Considering this latter aspect, it could be hypothesized that at a glucose concentration of 25 mM, there is a low interaction of the membrane proteins (in particular glucose transporters) with the MNPs shell due to the glucose-demand saturation. The low interaction strength exerted by MCF7 cells towards MNPs confirms also the already described lower amount of GLUT1 protein expression, with respect to mesenchymal-like counterpart 41 .
Although our evidences strongly suggest that the internalization mechanism is Glut-protein dependent, it is however still unclear how MNPs cross the membrane. In some preliminary experiments we observed, by an immuno-staining study, a slight co-localization between MNPs and lysosome markers. Indeed endocytosis might be a more feasible way for metal nanoparticles to be internalized after interacting with glucose transporters. Nevertheless, further experiments are required to adequately address this hypothesis.
The GLUT1 selective inhibition obtained by STF-31 and gene silencing definitely helped in correlating the MNPs-specific uptake to the activity of this glucose transporter channel in both the tested cancer cell lines. However, the inhibition of MNPs uptake in the mesenchymal-like subtype required, with respect to that in MCF7 cells, an increased STF-31 concentration (about twice) and incubation time (about 4 times), while by siRNA no concentration difference were required. Anyway, at the used concentration, STF-31 revealed a low cytotoxicity in both cell lines until 32 hours of incubation, leading to a marked cytotoxicity only after 72 hours of treatment ( Supplementary Information Fig. 7).
In addition to their possible use to detect mesenchymal-like tumor cells, the peculiar tropism of glucose-coated MNPs for mesenchymal-like cells, suggest their employment in therapeutic applications such as in targeted hyperthermia treatments 42 . Based on this hypothesis, here we presented an IR-laser-guided way for single-cell localized overheating. MNPs absorbance in IR allowed transferring these effects to the cell cytoplasm and consequently observing the cell reaction. Cell shape contractions, contour modifications, intra-cytoplasmic bubbles formation, cytoplasm rearrangements and cell death were the main observed occurrences (Supplementary Fig. 8 and Supplementary Movies 4-7 for details). Control cells (not previously treated by MNPs), subjected to IR-laser beam irradiation both on cytoplasm and nucleus, did not show any damaging evidence during and after this procedure (Supplementary Movie 8).
A classic approach in the field of thermotherapy is the use of oscillating magnetic field with precise frequency (100 kHz-1 MHz) and intensity (10-50 mT). Effectively the combination of the low administered MNPs concentration and the related heating-caused effects, coupled with cancer live imaging techniques could convey a significant improvement in cancer field. In this work via the exploitation of glucose coated MNPs, we selectively induced a localized heating on more metabolic active cells, the mesenchymal-like breast cancer cell line MDA-MB-231, providing interesting insights for future applications, especially because selectively targeting the population of cells considered more difficult to remove by conventional chemotherapy strategy because of an intrinsically enhanced drug-resistance. By consent, metal-based NPs are nowadays investigated for targeted heating therapy by the irradiation of precise areas with X-rays 43 . Due to the substantial metal-NPs to cell absorption ratio, in those wavelengths, several applications are now in development in order to face this challenge. The CoFe 2 O 4 -2-NBDG NPs developed in this work are perfectly in accordance with this front thanks to the low cytotoxicity exerted and the high tropism for the metabolic active cancer cells.
With the purpose to develop an implementation method for cancer diagnostic applications, here we also tested the glucose coated MNPs on WBCs obtained from healthy donors. WBC samples were incubated for 2 hours in the presence of CoFe 2 O 4 -2-NBDG NPs in order to observe their uptake ability. Remarkably, WBCs did not show any significant MNPs internalization ( Supplementary Fig. 9), supporting the possible application of glucose-coated MNPs in in vitro assays aimed at specifically recognizing GLUT1-overexpressing CTCs. This possibility could further improve the current CTCs detection methods, since it focuses the attention on the mesenchymal population, considered to be characterized by cancer stem cell properties and thus truly responsible of tumor maintenance, recurrence and metastasis 44 .
In conclusion, we think we have developed MNPs able to specifically recognize tumor cell subpopulations taking advantage of their distinctive deregulated metabolic pathways (Warburg effect), confirming the link between a more aggressive mesenchymal-like phenotype and a higher demand of energy. This will open the way to develop inexpensive in vitro assays and in vivo therapeutic approaches aimed at specifically recognizing/targeting the population of tumor cells considered to be endowed with the highest metastatic and drug-resistance potential.

Synthesis and Characterization of CoFe 2 O 4 NPs. CoFe 2 O 4 NPs were synthesized accordingly to
Massart method 45 . The protocol was partially modified, as we previous described 46 , in order to have citrate-coated MNPs with improved stability in physiological solution, which are more suitable for cellular environment. The citrate coordination of MNPs was used for covalent functionalization of MNPs with glucose molecule and its fluorescent analogue, the 2-NBDG (Life Technologies). The esterification reaction between the free carboxylic group of citrate and the more reactive hydroxyl group on 2-NBDG molecule has been achieved using HCl as catalyzing agent. The normal D-glucose molecule was also used as control sample to confirm the successful functionalization of the MNPs. The esterification was carried out by mixing 0. Immunofluorescence analysis. For immunofluorescence analyses, MDA-MB-231 and MCF7 cells were fixed in 4% buffered paraformaldehyde and permeabilized with 0.1% Triton X-100. Expression of vimentin (clone V9, Dako), cytokeratins 8-18-19 (clone 5D3, BioGenex), and the estrogen receptor alpha (clone SP1, ACZON) was evaluated by indirect immunofluorescence staining, using Alexa A555-labeled secondary antibodies. The Glucose Transporter GLUT1 was detected using an anti-GLUT-1 rabbit monoclonal antibody labeled with Alexa 647 dye (clone EPR3915, Abcam). DAPI was used to detect nuclei. All images were collected using a Leica DMI6000 B (Leica Microsystems, Germany) utilizing a 40X/63X oil immersion objective (numerical aperture: 1.25 and 1.40 respectively), keeping constant the acquisition parameters. GLUT1 immunostaining was quantified by the ImageJ software. For each specified condition, all time points were evaluated in triplicate, calculating the mean fluorescence intensity of at least 300 cells. phalloidin (Rhodamine conjugated, λ ex = 555 nm, Life Technologies) to highlight actin filaments, and by DAPI to identify nuclei. On the contrary, co-culture samples were stained overnight at 4 °C with anti-E-cadherin antibody (1:100 rabbit, Dako) to specifically recognize epithelial cell membranes 48 . After incubation, samples were stained with a secondary anti-rabbit antibody, Cy-7 conjugated (donkey, Abcam). Nuclei were labeled by DAPI.

Microscopy Evaluation of CoFe 2 O 4 NPs Uptake. The fluorescence of CoFe 2 O 4 -2-NBDG NPs uptake
by MDA-MB-231 and MCF7 cells have been assessed by a confocal laser microscope (DMRE, Leica, Germany) equipped with a 63X oil immersion objective (numerical aperture: 1.40) or a 40X oil immersion objective (numerical aperture: 1.25). In order to cover the entire cell height, a range of about 3-4 μ m, for each sample, was captured by a z-stacking interval of 250 nm. The confocal microscopy technique has been chosen in order to associate the presence of MNPs (as fluorescence emitted by the 2-NBDG functionalizing agent) at the same focus plan of the actin filaments, allowing to confirm the MNPs internalization by the cells. The specific 2-NBDG excitation wavelength is in the blue region (λ = 480 nm), while the emission is in the green one (λ = 510-550 nm) 49 . However, in the Figures MNPs are shown in the pseudocolor red to better appreciate their presence. The analysis of the total fluorescence content of each cell was performed by ImageJ "ROI Manager" after single-cell contour selection by the "freehand selection" tool, for each sample. The quantification was performed on the image reporting the average value of fluorescence in the 2-NBDG channel, relative to the total confocal images recorded. The fluorescence coefficient, C t,n , has been calculated for treated (t) and non-treated (n) samples respectively: where, FluoCell represents the fluorescence intensity of a single selected cell for the channel related to the CoFe 2 O 4 -2-NBDG NPs emission, FluoBk is the background intensity corresponding to the same cell-contour area and N is the total number of analyzed cells (N = 200 cells from almost 20 regions, per sample). The uptake ratio, R u , is then calculated as: To evaluate the internalization of non-fluorescent MNPs, the Perls' iron staining, that produces a Prussian-blue deposition in presence of reduced iron, was chosen 50 . The Perls' iron staining has been carried out by a semi-automated instrument (Artysan ™ Link Pro, Dako), on PFA-fixed samples either permeabilized or not by a membrane permeabilization procedure (10 minutes with a 0.1% TritonX-100 solution), in order to assess intracellular -or cell surface bound-MNPs, respectively (see Supplementary Fig. 5).
To assess the glucose function in MNPs uptake and so avoid possible 2-NBDG dye contribution, D-glucose and citrate coated CoFe 2 O 4 NPs were used. The uptake process, for each MNPs functionalization type, has been tracked after 5, 15, 30, 60, 120, 180 and 240 minutes, in both glucose medium conditions. The samples were prepared on duplicate and the experiments were repeated three times. The quantification of cells positive for internalized MNPs has been performed on pictures acquired via optical microscope (Leica DMD 108 equipped with a 10×, 20× and 40× objectives, numerical aperture 0.40, 0.70 and 0.95, respectively). Cells were counted as positive in case of Prussian blue deposition (compatible as to be MNPs) regardless of the number of events per cell. The quantifications are reported as total counted Prussian-blue-cells respect to the total amount of cells in the area. At least 200 cells per sample were analyzed. ImageJ software was employed to overlay single-channel images and adjust the brightness/contrast settings, while to compose the final images Adobe Photoshop was utilized.

Uptake of MNPs investigated by SEM-FIB technique.
Breast cancer cell lines have been further analyzed at the nanometer level by SEM, to evaluate the presence on intracellular MNPs. Cell samples were fixed with 4% PFA for 15 minutes and dehydrated by increasing concentration of Et-OH (from 30% to 99.9%, purchased from Sigma-Aldrich Co LLC). Samples were dried by gently nitrogen blow and coated with a 20 nm thick Chromium film by sputtering. Samples were transferred in a Cross Beam Microscope (Zeiss International, Germany) for Gallium Focused Ion Beam (FIB) precise milling and Scanning Electron Microscopy (SEM) inspection. The chromium coating, via sputtering, allowed delineating the cell morphology and afterward the FIB cuts were performed on random positions, onto cells, and on visible aggregates as well, in order to examine whether and where the MNPs were internalized.
Pharmacological inhibition of the Glucose Transporter Channel 1. To inhibit GLUT1, breast cancer cell lines have been treated with the specific inhibitor STF-31 (Sigma-Aldrich Co. LLC). The evaluation of toxicity has been performed via the MTT assay, whereas the effect of STF-31 on NPs uptake was evaluated both by Perls' iron staining and immuno-fluorescence microscopy as described above. The protocol followed for the STF-31 toxicity evaluation is described in Supplementary Methods. The uptake inhibition effect has been observed at 5 μ M and 10 μ M concentrated inhibitor in MCF7 and MDA-MB-231, respectively. The time points analyzed were 2, 4, 6 and 8 hours for MCF7 and 8, 16, 24 and 32 hours for MDA-MB-231 and MNPs were added 2 hours before the PFA fixation for each time step. Each treatment was carried in 2 wells (in a 24 well-plate) and repeated 3 times. transfection reagent with a control siRNA (Invitrogen). The samples were prepared on triplicate and the experiments were repeated four times. Cellular levels of GLUT1 were quantified by immunofluorescence staining and FACS analysis (details are described in the previous paragraphs).

Cell Local Heating Through NPs Laser Beam Absorption. A continuous-wave infrared (IR) laser
beam (IPG Photonics) was focused through the microscope lens (Nikon, 60X water immersion objective, numerical aperture: 1.25) onto the selected MNPs-bearing cell. The setup was built on a Nikon microscope (Eclipse TE2000-E, Nikon, Japan) and images recorded on CMOS camera (Hamamatsu ORCA-Flash 4.0). The wavelength of the laser was 1064 nm, for which cell adsorption is minimum. In each recorded movie an IR filter was inserted in order to better appreciate the effects caused by the laser. It was, moreover, used to avoid the IR-laser entrance into recording device and thus allowing good video quality. The power on the sample has been regulated from 0 to 15 mW to evaluate the increasing outcomes. Cobalt ferrite nanoparticles absorbance in IR is lower than visible or UV wavelength 51 but in the same time much higher than that of the cells 52 .