Metabolism and cell shape in cancer: A fractal analysis

The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 Contents lists available at ScienceDirect The International Journal of Biochemistry & Cell Biology journal homepage: www.elsevier.com/locate/biocel Metabolism and cell shape in cancer: A fractal analysis Fabrizio D’Anselmi a,c , Mariacristina Valerio b,g , Alessandra Cucina c , Luca Galli d , Sara Proietti e , Simona Dinicola e , Alessia Pasqualato f , Cesare Manetti g , Giulia Ricci h , Alessandro Giuliani i , Mariano Bizzarri e,∗ a ASI, Italian Space Agency, Roma, Italy Area of Molecular Medicine Scientific Director’s Office, Italian National Cancer Institute “Regina Elena”, Roma, Italy Dept. of Surgery “Pietro Valdoni” - University La Sapienza, Roma, Italy d Advanced Computer Systems A.C.S. S.p.A., Roma, Italy e Dept. of Experimental Medicine - University La Sapienza, Roma, Italy f Dept. of Basic and Applied Medical Science. University “G. D’Annunzio”, Chieti-Pescara, Italy g Dept. of Chemistry - University La Sapienza, Roma, Italy h Dept. of Experimental Medicine - Second University, Naples, Italy i Environment and Health Department, Istituto Superiore di Sanita’, Roma, Italy b c a r t i c l e i n f o Article history: Available online 9 May 2010 Keywords: Fractal morphology Phenotype reversion Bending energy Tumour metabolome Morphogenetic field a b s t r a c t Fractal analysis in cancer cell investigation provided meaningful insights into the relationship between morphology and phenotype. Some reports demonstrated that changes in cell shape precede and trigger dramatic modifications in both gene expression and enzymatic function. Nonetheless, metabolomic pattern in cells undergoing shape changes have been not still reported. Our study was aimed to investigate if modifications in cancer cell morphology are associated to relevant transition in tumour metabolome, analyzed by nuclear magnetic resonance spectroscopy and principal component analysis. MCF-7 and MDA-MB-231 breast cancer cells, exposed to an experimental morphogenetic field, undergo a dramatic change in their membrane profiles. Both cell lines recover a more rounded shape, loosing spindle and invasive protrusions, acquiring a quite “normal” morphology. This result, quantified by fractal analysis, shows that normalized bending energy (a global shape characterization expressing the amount of energy needed to transform a specific shape into its lowest energy state) decreases after 48 h. Later on, a significant shift from a high to a low glycolytic phenotype was observed on both cell lines: glucose flux begins to drop off at 48 h, leading to reduced lactate accumulation, and fatty acids and citrate synthesis slow-down after 72 h. Moreover, de novo lipidogenesis is inhibited and nucleotide synthesis is reduced, as indicated by the positive correlation between glucose and formate. In conclusion, these data indicate that the reorganization of cell membrane architecture, induced by environmental cues, is followed by a relevant transition of the tumour metabolome, suggesting cells undergo a dramatic phenotypic reversion. © 2010 Elsevier Ltd. All rights reserved. 1. Introduction The tumour metabolome – mainly characterized by the glycolytic phenotype – confers to the evolving cancer cell population an advantage and contributes to tissue invasion and metastasis spreading (De Berardinis et al., 2008). However the glycolytic phenotype is not confined to cancer cells: embryonic tissues, as well as highly proliferating cells, like lymphocytes, share a similar pattern Abbreviations: EMF, experimental morphogenetic field; NBE, normalized bending energy; PCA, principal component analysis; NMR, nuclear magnetic resonance. ∗ Corresponding author at: Dept. of Experimental Medicine, University La Sapienza, viale Regina Elena 324, 00161 Roma, Italy. Tel.: +39 06 49766606; fax: +39 06 49766603. E-mail address: [email protected] (M. Bizzarri). 1357-2725/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocel.2010.05.002 (Wang et al., 1976). Moreover, cancer cell metabolism is significantly affected by cell cycle phase and confluence or sub-confluence culture conditions, displaying high plasticity to adapt in presence of adverse microenvironmental conditions (Tomassini et al., 2006). These data suggest that tumour metabolome might be considered a dynamic reversible phenotypic trait, likely governed by the non-linear interplays of several both genomic and non-genomic factors: epigenome, nutrient availability, oxygen and blood supply, stiffness and diffusion gradients shaping the microenvironmental constraints. On the other hand, it is reasonable to infer that the modification of microenvironmental cues, could influence tumour metabolism in order to force cancer cells loose (partly or entirely) their malignant features. Physical – extracellular matrix stiffness, microgravity – as well as chemical microenvironmental factors, have been reported to induce relevant morphological changes on cell architecture (Chen F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 et al., 1997). In turn, cell shape transition has proven to precede and to influence significantly both gene expression and enzymatic reactions (Carmeliet and Bouillon, 1999; Boonstra, 1999). Moreover, by controlling the cellular environment with microfabricated patterning, studies on mammary epithelial cell tissue morphogenesis demonstrated to modify nuclear organization and subsequently modulate cellular and tissue phenotype (Lelièvre et al., 1998). Furthermore, microenvironmental-induced shape changes in chondrocyte nuclei correlate with collagen synthesis (Thomas et al., 2002) or changes in cartilage composition and density (Guilak, 1995). Therefore, as clearly stated by Ingber (1999), cell shape should be considered as “[. . .] the most critical determinant of cell function [. . .] cell shape per se appears to govern how individual cells will respond to chemical signals (soluble mitogens and insoluble ECM molecules) in their local microenvironment.” Cell shape modifications are thought to ‘integrate’ a wide set of stimuli and to drive cell switching between different cell fates, i.e. different phenotypes. Broadly speaking, cell shape could be viewed as the “structural” result of the interacting influences between both internal and external constraints. Moreover, quantitative shape descriptors (i.e. fractal dimension), possess thermodynamics meaning and they could provide insights into the complexity score of the observed system. In this context, cell shape is thought to be the morphological expression of an integrated system, and a specific morphology could be assigned to every functional cell state: in other words, every phenotype or differentiated cell possesses a well-defined shape, described by both classic morphological as well thermodynamics parameters. Yet, an understandable link between shape and metabolic or genomic function has been never proposed. This is partly due to the limited knowledge about how biochemical reactions are associated to the cytoskeleton (i.e. the internal topology of structures-linked reactions), and, on the other hand, to a lack of a standardized and wide-accepted measure of cell shape complexity. A quantitative method that lends itself particularly useful for characterizing complex irregular structures is fractal analysis. NonEuclidean objects are better described by fractal geometry, which has the ability to quantify the irregularity and complexity of objects with a measurable value called the “fractal dimension”. Fractal dimension differs from our intuitive notion of dimension (i.e. topological dimension) in what can be considered as a non-integer value, and more irregular and complex an object is, more higher is its fractal dimension in relation to its topological dimension (Mandelbrott, 1982). Fractal geometry is well suited to quantify those morphological features which pathologists have long used (and are still using today!) in a qualitative way to describe malignancies. So far, during the last decade, several reviews of fractal measures application in pathology have outlined that fractal analysis could provide reliable and helpful information (Losa et al., 2002; Baish and Jain, 2000; Cross, 1997). Our study was intended to investigate if modifications in cancer cell morphology are associated to relevant transition in tumour metabolomic profile, analyzed by means of nuclear magnetic resonance (NMR) spectroscopy and flux principal component analysis (PCA) (exometabolome study). 2. Materials and methods 2.1. Cell culture and EMF MCF-7 and MDA-MB-231 human breast carcinoma cell lines (ECACC, Sigma–Aldrich, St. Louis, MO, USA) were cultured in Dulbecco modified Eagle’s medium (DMEM, Catalog no. EC B7501L; Euroclone Ltd., Cramlington, UK), supplemented with 10% fetal 1053 calf serum (FCS, Euroclone), non-essential aminoacids and antibiotics (penicillin 100 IU/ml, streptomycin 100 ␮g/ml, gentamycin 200 ␮g/ml, all from Euroclone). MCF-7 cells used for experiments were from passages 30 to 40, and MDA-MB-231 cells were from passages 5 to 15. The different number of passages for the two cell lines used in our experiments simply represents the passages performed in our laboratory from the purchase of each cell line. EMF (experimental morphogenetic field) is a set of proteins extracted from chicken egg’s albumen, easily soluble in culture media. EMF is now patented. We dissolved EMF in DMEM to reach a final concentration of 10% of the volume. Cells were cultured for 168 h in DMEM + 10% FCS, as a control condition, and DMEM + 10% FCS + 10% EMF as experimental arm. 2.2. Cell proliferation assay MCF-7 and MDA-MB-231 cells were seeded in 12-well culture plates (Falcon, Becton Dickinson Labware) at a concentration of 2 × 104 cells/well in DMEM + 10% FCS. The following day, the cells were refed with DMEM + 10% FCS and DMEM + 10% FCS + EMF 10%. The plates were incubated for 168 h at 37 ◦ C in an atmosphere of 5% CO2 . Every day the cells were trypsinized and centrifuged, and cell pellets were resuspended in PBS. Cell count was performed by a particle count and size analyzer (Beckman Coulter, Inc. Fullerton, CA, USA) and by a Thoma hemocytometer. Three replicate wells were used for each data point, and the experiments were performed six times. 2.3. Optical microscopy Both MCF-7 and MDA-MB-231 cells were grown on six well tissue culture plate (Becton Dickinson Labware, Franklin Lake, NJ, USA) and were photographed every 24 h with Nikon Coolpix 995 digital camera coupled with Nikon Eclipse TS100 optical microscope (Nikon Corporation, Japan). Photomicrographs at different magnifications, were saved as TIFF files and used for Normalized Bending Energy (NBE) analysis. 2.4. Electron microscopy MCF-7 and MDA cells, cultured with or without EMF, were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4), postfixed in 1% OsO4 in Zetterquist buffer, de-hydrated in ethanol, and embedded in epoxy resin. Ultrathin sections were contrasted in aqueous uranyl-acetate and lead-hydroxide, studied and photographed by a Hitachi 7000 Transmission Electron Microscope (Hitachi, Tokyo, Japan). 2.5. Shape analysis Shape analysis via NBE computation of nuclear and cell membrane of breast cancer cells has been performed in a semi-automatic way. First, cell membranes have been manually segmented delineating the initial contour; secondly, accurate membrane segmentation is automatically performed by using an Active Contour Model (“Snake”) (Sebri et al., 2007). Snake models basically consist of a curve, which can dynamically conform to object shapes in response to internal and external forces. These forces could be seen as the result of a functional global minimization processor based on local information. The conventional Snake model has some limitation, in that the initial contour must be placed close to the object to prevent it from converging to a local minimum. In order to overcome this problem, the Gradient Vector Flow GVF-Snake method (Xu and Prince, 1998) has been used, which, besides having a large capturing range, is currently also the most efficient one in terms of precision, for the tracking of breast cancer cells (Fig. 1). The 1054 F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 Fig. 1. Examples of breast cancer cell membrane segmentation using the GVF-Snake method. (A) MCF-7 cells in control and (B) in EMF condition after 96 h. (C) MDA-MB-231 cells in control and (D) in EMF condition after 96 h. NBE shape descriptor has been computed considering 36 cell membranes, randomly chosen from both the MCF-7 and MDA-MB-231 breast cancer cell lines for each experimental time. The computations were performed by in-house software developed at ACS (Advanced Computer Systems, S.p.A., Rome, Italy). 2.6. Metabolomic analysis For metabolomic analysis both MCF-7 and MDA-MB-231 cells were grown on INTEGRIDTM (150 mm × 25 mm) tissue culture dish (Becton Dickinson Labware, Franklin Lake, NJ, USA). Cells were stopped at 48, 72 and 96 h; the medium for each time point was collected and stored at −80 ◦ C. Medium dried samples were dissolved in 600 ␮l of 1 mM TSP [sodium salt of 3-(trimethylsilyl) propionic-2,2,3,3-d4 acid] solution in D2 O PBS buffer (pH 7.4) to avoid chemical-shift changes due to pH variation. 2D 1 H J-resolved (JRES) NMR spectra were acquired on a 500 MHz DRX Bruker Avance spectrometer (Bruker Biospin, Rheinstetten, Germany), using a double spin echo sequence with 8 transients per increment for 32 increments. After 2D JRES processing, the 1D skyline projections exported were aligned and then reduced into spectral bins with widths ranging from 0.01 to 0.03 ppm. The total spectral area (excluding water signal) was normalized to unity. NMR data was processed using Bruker’s XWINNMR software and custom-written MATLAB code (Version 7.0; The MathWorks, Natick, MA). All data were expressed in terms of net balances, i.e. as difference between the single time points (48, 72 and 96 h) and time 0, allowing the analysis of fluxes, representing the actual utilization of the substrate. Each 142 bins spectrum can be considered as a vector of 142 dimensions, being each dimension the Y value at each specific X (ppm) location. In the case of N samples, the matrix having the spectra as rows (statistical units) and the 142 Y values as variables (columns) constituted the starting data matrix collecting all the relevant information. This matrix undergoes a principal component analysis (PCA) procedure in order to extract the main features of the data set in terms of between variables correlation structure: each extracted component points to a direction of variation maximizing the between variables correlation and thus correspondent to a given metabolic pathway. Between groups differences on principal component scores were considered as statistically significant at p < 0.05 in the t-test. PCA was conducted using PLS Toolbox (Version 5.2; Eigenvector Research, Manson, WA) within MATLAB. Inferential analysis was performed by using GraphPad Instat software (GraphPad Software, Inc., San Diego, CA, USA). 3. Results 3.1. Cell proliferation Both MCF-7 and MDA-MB-231 control cells display high rate of proliferation. A slightly, even if significant decrease in proliferative trend was observed in EMF-treated cells, after the first 48 h (data not shown). 3.2. Shape analysis MCF-7 and MDA-MB-231 cells growing in an experimental morphogenetic field progressively undergo dramatic changes of cell membrane shape. After 48 h, membrane profiles change, evolving into a more rounded shape, loosing spindle and invasive protrusions. NBE values while not changing in MCF-7 control cells, are slightly decreasing in MDA-MB-231 control samples. On the contrary, a dramatic reduction was observed for NBE levels in treated samples. Differences were statistically significant starting from 72 h and 120 h for MDA-MB-231 and MCF-7, respectively (Table 1). Even if the differences, above mentioned, are not statistically significant, they were already emerging at 48 h. Indeed, NBE profiles of cell F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 1055 Fig. 2. Mean normalized bending energy values (calculated for cell membrane) in MCF-7 cell line computed at different experimental time in controls and treated (EMF) conditions. The error bars refer to the standard error of the mean NBE values. Fig. 3. Mean normalized bending energy values (calculated for cell membrane) in MDA-MB-231 cell line computed at different experimental time in controls and treated (EMF) conditions. The error bars refer to the standard error of the mean NBE values. membrane MCF-7 and MDA-MB-231 (Figs. 2 and 3) show that cell shape in treated samples undergoes a phase-transition between 48 and 72 h, as indicated by the high variance values. Moreover, starting from 72 h, MCF-7 cells treated with EMF were able to form both tight (absent in control cells) and gap-junctions, as confirmed by electron microscopy (Fig. 4). 3.3. Metabolomic analysis of MCF-7 cells Fig. 4. Electron microscopy pictures of MCF-7 breast cells treated with EMF. Gap (grey arrow, left panel) and tight junctions (white arrows, right panel) are shown in photomicrographs. PCA extracted seven relevant components, together explaining about 75% of the total variability of the system. To compare all the specific pairs of control and treated groups a t-test was applied to the component scores. The discrimination between control and treated groups was achieved in the first two PCs at each experimental time and in the PC3 and PC4 at 48 and 96 h as well as in PC6 at 72 and 96 h (Table 2). The analysis of the major order parameters present in the data, namely PC1 and PC2 (21% and 19% of variation explained, respectively), allowed us to determine the metabolite pattern discriminating the two groups (Fig. 5, Table 3). PC1 most correlated variables were net balances of fatty acids, 3-hydroxybutyrate and acetate (these entering with negative loading). This profile points, in EMF-treated MCF-7 cells, to a reduced de novo lipidogenesis and a greater utilization of these metabolites for energetic purposes, compared to control ones, particularly at 48 h (differences in PC1 scores: 3.1 at 48 h, 0.5 at 72 h, 0.9 at 96 h). Starting from 72 h, the discrimination between control and treated samples is carried out by PC2, pointing to a shift from a high glycolytic phenotype to a metabolism in which energy requirements are mainly fulfilled by glutamine and fatty acids consumption. Indeed, PC2 points to a lower consumption of glucose – with reduced lactate release – and a greater utilization of glutamine for biosynthetic purposes in EMF-treated cells, particularly after 72 h (differences in PC2 scores: 0.6 at 48 h, 2.1 at 72 h, 2.2 at Table 1 Normalized bending energy shape descriptor t-test probability, comparing control versus treated cells values for MCF-7 and MDA-MB-231 breast cancer lines for each experimental time (threshold p < 0.05). Experimental time (h) MCF-7 MDA-MB-231 0 24 48 72 96 120 144 168 0.8766 0.92874 0.5432 0.1973 0.7345 0.4773 0.9789 0.0164 0.0890 0.0000 0.0322 0.0000 0.0383 0.0000 0.0021 0.0000 1056 F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 Table 2 t-Test comparing control versus treated cells (MCF7). In parentheses the percent of variance explained by each principal component is reported (threshold p < 0.05). Experimental time (h) PC1 (21%) PC2 (19%) PC3 (11%) PC4 (9%) PC5 (6%) PC6 (5%) PC7 (4%) 48 72 96 <0.00001 0.0007 <0.00001 0.0003 <0.00001 <0.00001 0.021 0.614 0.0002 0.0002 0.338 0.0001 0.169 0.420 0.720 0.424 0.002 <0.00001 0.878 0.973 0.697 Table 3 Most correlated regions of 1 H NMR spectra to PC1 and PC2. PC1 PC2 a ppm Factor loading Metabolite ppma Factor loadingb Metabolitec 0.89, 1.29, 2.75 1.20 1.92 0.81 −0.81 −0.73 Lipid (c) 3-Hydroxybutyrate (c) Acetate (c) 3.23, 3.41, 3.71, 3.75, 3.78, 3.83, 3.95, 4.66, 5.23 1.47 2.14, 2.43 8.45 0.75 −0.67 −0.80 0.80 Glucose (c) Alanine (p) Glutamine (c) Formate (p) a b c b c Mid-spectral integral region, i.e. 3.26 represents ppm 3.23–3.27; only spectral regions containing only one metabolite are reported. Values are given as mean of loadings obtained for all spectral regions of each metabolite. Consumption and production are indicated as (c) and (p), respectively. Fig. 5. Overview of the PCA model built on the NMR dataset of medium samples collected from control and EMF-treated MCF-7 cells at 48, 72 and 96 h. The score plot of the first two components (PC1 versus PC2) showing differences among groups is shown. 96 h). Moreover, the positive correlation as for PC2 between glucose and formate, suggests an inhibition of de novo nucleotide synthesis in EMF-treated cells. The absence of correlation between glutamine and lactate shows the former is preferentially driven along anabolic pathways in EMF-treated samples. 3.4. Metabolomic analysis of MDA-MB-231 cells For MDA-MB-231 cells, five principal components (PCs) were extracted, together explaining the 80% of the total variance. A t-test, applied to the component scores to compare control and treated cells, highlighted significant differences between the two groups on the first four PCs at each experimental time and on the PC5 at 48 and 96 h (Table 4). Analysis of the PC1/PC2 plane (Fig. 6) showed that PC1 is by far the major order parameter present in the data (42% of variance explained) and corresponds to the core energy metabolism as evident from its positive loading (correlation coefficient between original variable and component) with glucose utilization and its negative loading with lactate (see Table 5). This correlation structure implies the samples having higher PC1 scores correspond to those samples with a lower use of glucose; on the contrary, those ones with lower scores are the statistical units endowed with the higher glucose utilization and lactate produc- Fig. 6. Overview of the PCA model built on the NMR dataset of medium samples collected from control and EMF-treated MDA-MB-231 cells at 48, 72 and 96 h. The score plot of the first two components (PC1 versus PC2) showing differences among groups is shown. The major metabolic difference between control and treated groups at 96 h is highlighted by the black line. tion. As evident in Fig. 6, the by far maximal difference between control and treated groups is at the 96 h point, where control samples display much higher glucose consumption (high glycolytic phenotype). In the other time points too, control samples show consistently lower values of PC1 with respect to treated samples, but differences are noticeably lower. This emerges by the average differences in PC1 scores between control and treated groups at different times (0.6 at 48 h, 1.0 at 72 h, 2.6 at 96 h). Moreover, after 72 h, PC2 scores obtained from EMF-treated cells, displayed a clear metabolomic reversion, mainly characterized by reduced glycolytic fluxes that, concomitantly with a reduced citrate bioconversion, lead to increased ␤-oxidation rates and reduced fatty acids synthesis. 4. Discussion Microenvironmental factors and proteins extracted from both embryonic and maternal morphogenetic fields, have been shown to inhibit cancer progression, inducing programmed cell death or morphologic and phenotype reversion (Cucina et al., 2006; Hendrix et al., 2007). Moreover, environmental cues and physical forces have been shown to deeply influence both cell shape and architecture (Ingber, 2005). Significant changes in cell morphology have F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 1057 Table 4 t-Test comparing control versus treated cells (MDA-MB-231). In parentheses the percent of variance explained by each principal component is reported (threshold p < 0.05). Experimental time (h) PC1 (42%) PC2 (15%) PC3 (12%) PC4 (7%) PC5 (4%) 48 72 96 <0.00001 <0.00001 <0.00001 0.007 <0.00001 <0.00001 <0.00001 <0.00001 0.006 <0.00001 <0.00001 0.001 0.003 0.326 0.044 Table 5 Most correlated regions of 1 H NMR spectra to PC1 and PC2. PC1 PC2 a ppm Factor loading Metabolite ppma Factor loadingb Metabolitec 3.23, 3.41, 3.71, 3.75, 3.78, 3.83, 3.95, 4.66, 5.23 1.33, 4.12 2.14, 2.43 0.97 −0.85 −0.87 Glucose (c) Lactate (p) Glutamine (c) 0.89, 1.29, 1.59, 2.05, 2.25 2.14 2.65 0.81 0.73 0.76 Lipid (c) Acetoacetate (c) Citrate (c) a b c b c Mid-spectral integral region, i.e. 3.26 represents ppm 3.23–3.27; only spectral regions containing only one metabolite are reported. Values are given as mean of loadings obtained for all spectral regions of each metabolite. Consumption and production are indicated as (c) and (p), respectively. been obtained in 3D culture of cancer cells (Kenny and Bissell, 2003); however, this is the first report in which on a 2D culture, a morphogenetic field, constituted by an experimental set of proteins extracted from egg albumen, induces both phenotype and morphology reversion in breast cancer cells. Cell morphology in EMF-treated samples undergoes significant changes, evolving from a spindle to a round profile. Fractal measures demonstrated cell shape acquires a “less-dissipative” architecture in terms of reduced levels of NBE. The “curvegram” obtained by using digital signal processing gives a multiscale representation of the curvature, providing an useful resource for translation and rotation-invariant shape classification, as well as a mean of deriving quantitative information about the complexity of the shapes being investigated (Cesar and Costa, 1997). Bending Energy is a global shape characterization correspondent to the amount of energy needed to transform the specific shape under analysis into its lowest energy state (i.e. a circle) (Bowie and Young, 1977). Bending energy is thus proportional to the fractal dimension of the studied object that in turn has to do with the amount of complexity. We can consider fractal dimension in much the same way that thermodynamics might view intensive measures as temperature. Therefore, NBE provides an understandable link between cell morphology and thermodynamics features of the system (Smith et al., 1996). In our study, membrane profiles of control cancer cells exhibit high basal NBE values. In MCF-7 control samples, after 96 h, no significant changes were recorded in NBE levels, whereas in MDAMB-231 control cells, a slight reduction is observed after 72 h. In the latter case, it could be argued that the high population density reached by confluent cells exerts a physical constraint, limiting the “protrusive” phenotype and determining a more “compact” cell morphology (crowding effect). On the contrary, EMF treatment induces in both cancer cell lines an impressive morphological modification and a significant reduction of NBE, starting from 48 h. These effects emerged earlier in MDA-MB-321 than in MCF7 breast cancer cells. Indeed MCF7 are widely recognized to present only some malignant features, showing a low aggressive behaviour. These data suggest that more evident are the morphological abnormalities, more sensitive are the cells to the effects triggered by environmental cues. The experimental morphogenetic field induces an overall reorganization of cell architecture leading to a shape “normalization”, characterized by rounded, smooth profiles, and reduced levels of dissipative energy. It is tempting to speculate modifications like that produce several consequences on the overall system, leading to changes in cell to cell relationships. Bending Energy is known to be inversely correlated with surface tensions, and surface tensions are reflective of intercellular adhesive intensities (Foty et al., 1994). Therefore, it can be assumed that, while reducing bending energy, EMF treatment increases intercellular adhesion forces. This aspect is noteworthy, considering that surface tensions influence both embryonic and tumour cell spreading (Foty et al., 1996). Modification in bending energy highlights the transition from a “protrusive” towards a compact shape, the former associated to an invasive pattern (Rohrschneider et al., 2007). Hence, we hypothesize that reversion of tumour shape towards more “physiologic” fractal dimension, implies a reduced morphologic instability and increased cell connectivity. This statement is further supported by the re-establishment of both tight and gap-junctions (absent in control cells), observed in MCF-7 EMF-treated cells. It is well known that gap junctional intercellular structures integrate and modulate both cellular and microenvironmental signals in order to inhibit proliferation and enhancing differentiating processes (Trosko et al., 2004). It could be therefore argued that some of the modifications observed in tumour metabolism, could be partly attributed to the de novo reconstituted communication between cells. It is noteworthy that current cytotoxic anticancer treatments induce a significant increase in cell shape fractal dimension and “may unwittingly contribute to tumour morphologic instability and consequent tissue invasion” (Cristini et al., 2005). Indeed, whereas mild chemotherapeutic regimens did not modify tumour fractal dimension, intensive cytotoxic chemotherapy increases fractal and NBE values, and enhances tissue disorder and chaotic tumour behaviour, eventually promoting selection of more malignant phenotypes (Ferreira et al., 2003). Shape “normalization” was accompanied by a meaningful metabolomic shift in cancer cells exposed to EMF. Glycolytic fluxes were reduced, concomitantly with a decrease in lactate, glutathione, glutamine and other compounds. Namely for EMFtreated MDA-MB-231 cells, when cell proliferation slows-down and cell shape reaches a new stable configuration with low values of Bending Energy, cancer cells undergo a metabolomic reversion, characterized by the inhibition of both lipidogenesis and de novo nucleotide synthesis. It is well known that cancer cells share an increased glutamine use (McKeehan, 1982). However, in EMF-treated cells, glutaminolysis increase does not correlate with a simultaneous increase in lactate, nor to an increase in fatty acid synthesis. Keeping in mind that proliferation is slightly inhibited in EMF-treated cells, these results outline that glutaminolysis increased rates can not be explained by proliferative needs: this implies the treated cells devote a higher portion of chemical energy 1058 F. D’Anselmi et al. / The International Journal of Biochemistry & Cell Biology 43 (2011) 1052–1058 to a different anabolic work. Indeed, excess of glutamine is preferentially transformed into proteins and does not appear as lactate. This interpretation is given a proof of concept by the observation of the development of differentiating pathways (the increase of E-cadherin and ␤-casein protein biosynthesis) and differentiated structures (ducts and hollow acini, mainly in MCF-7 cells) in treated cells at later times (96–168 h) (data not shown). Our results should be related to a previously published report (Meadows et al., 2008), in which glucose uptake was measured in human normal mammary epithelial 48R cells and MCF-7 cells. Glycolytic fluxes have been correlated with biomass, cell morphology and medium exposed surface, demonstrating that both morphological traits and medium exposed surface were the main driving force of glucose uptake in cells. It is worth noting that shape modification leads to a lessdissipative architecture, as it is showed by the significant reduction in NBE values. So far, fractal measures enable to emphasize the link between cell morphology and thermodynamics. According to the Prigogine–Wiame theory of development (Prigogine and Wiame, 1946), during carcinogenesis, a living system constitutively deviates from a steady state trajectory. This deviation is accompanied by an increase in the system dissipation function (« ) at the expense of coupled processes in other parts of the organism, where « = q0 + qgl (meaning, respectively, q0 oxygen consumption and qgl glycolysis intensity). The metabolomic data collected in this study displayed a significant reduction in glycolysis activity, with unchanged values of oxygen consumption. Therefore in EMF condition « decreased significantly, until a stable state was attained, characterized by a minimum in the rate of energy dissipation (principle of minimum energy dissipation) (Zotin and Zotin, 1997). Thereby, the reduction in both dissipative function and NBE levels, suggests cell system achieves, after shape transition, a new thermodynamic stable state. This behaviour is exactly the opposite of what happens in growing cancer cells and experimentally observed in tumour control samples. Pursuing this perspective, the “Warburg effect” (i.e. high glycolytic fluxes and high glucose anaerobic metabolism) should not be longer considered as a “linear” consequence of gene deregulation or an adaptation to hypoxia, but a “system property” of cancer cells, influenced by both internal and microenvironmental constraints (Cascante et al., 2002). Cell energy metabolism changes along different cell cycle phases, being more “dissipative” during wound healing, fast growth (specifically during embryonic development), and cancer progression. Keeping in mind that thermodynamic dissipative function is correlated with both glucose metabolism and cell shape, we suggest that the latter could interfere with metabolic pathways. Cell shape is known to influence several generegulatory pathways through architectural rearrangement, thereby representing a relevant independent factor controlling tissue fate and cell commitment to quiescence, apoptosis or proliferation. Our data demonstrated that the morphogenetic field is able in inducing dramatic changes in breast cancer cell shape expressed by fractal measures. Consequently, meaningful changes in “tumour metabolome” were observed by NMR-spectroscopy and PCA flux analysis. These data indicate cell shape “normalization” is associated with a reversion in tumour metabolic phenotype. Further metabolomic studies are clearly warranted to better correlate metabolism and shape morphology, and to handle these two set of parameters into an integrated dynamical description of tumour cell biology. References Baish JW, Jain RK. Fractals and cancer. Cancer Res 2000;60:3683–8. Boonstra J. Growth factor-induced signal transduction in adherent mammalian cells is sensitive to gravity. FASEB 1999;13:S35–42. Bowie JE, Young IT. An analysis technique for biological shape. Acta Cytol 1977;21:739–46. Carmeliet G, Bouillon R. The effect of microgravity on morphology and gene expression of osteoblasts in vitro. FASEB 1999;13:S129–34. Cascante M, Boros LG, Comin-Anduix B, de Atauri P, Centelles JJ, Lee PW. Metabolic control analysis in drug discovery and disease. Nat Biotechnol 2002;20:243–9. Cesar Jr MR, da Fontoura Costa L. The application and assessment of multiscale bending energy for morphometric characterization of neural cells. Rev Sci Instrum 1997;68:2177–86. Chen CS, Mrksich M, Huang S, Withesides GM, Ingber DE. Geometric control of cell life and death. Science 1997;276:1425–8. Cristini V, Frieboes HB, Gatenby R, Caserta S, Ferrari M, Sinek J. Morphologic instability and cancer invasion. Clin Cancer Res 2005;11(19):6772–9. Cross SS. Fractals in pathology. J Pathol 1997;182:1–8. Cucina A, Biava PM, D’Anselmi F, Coluccia P, Conti F, Di Clemente R, et al. Zebrafish embryo proteins induce apoptosis in human colon cancer cells (Caco2). Apoptosis 2006;11(9):1617–28. De Berardinis R, Lum JJ, Hatzivassiliou G, Thompson CB. The biology of cancer: metabolic reprogramming fuels cell growth and proliferation. Cell Metab 2008;7:11–20. Ferreira SC, Martins ML, Villa MJ. Morphology transitions induced by chemotherapy in carcinomas in situ. Phys Rev E 2003;67:1–9. Foty R, Forgacs G, Pfleger C, Steimberg M. Liquid properties of embryonic tissues: measurement of interfacial tensions. Phys Rev Lett 1994;72:2298–301. Foty R, Pfleger C, Forcas G, Steimberg M. Surface tensions of embryonic tissues predict their mutual envelopment behaviour. Development 1996;122:1611–20. Guilak F. Compression-induced changes in the shape and volume of the chondrocyte nucleus. J Biomech 1995;28:1529–41. Hendrix MJ, Seftor EA, Seftor RE, Kasemeier-Kulesa J, Kulesa PM, Postovit LM. Reprogramming metastatic tumour cells with embryonic microenvironments. Nat Rev Cancer 2007;7(4):246–55. Ingber DE. How cells (might) sense microgravity. FASEB 1999;13:S3–15. Ingber DE. Mechanical control of tissue growth: function follows form. Proc Natl Acad Sci USA 2005;102(33):11571–2. Kenny PA, Bissell MJ. Tumor reversion: correction of malignant behaviour by microenvironmental cues. Int J Cancer 2003;107:688–95. Lelièvre SA, Weaver VM, Nickerson JA, Larabell CA, Bhaumik A, Petersen OW, et al. Tissue phenotype depends on reciprocal interactions between the extracellular matrix and the structural organization of the nucleus. Proc Natl Acad Sci USA 1998;95:14711–6. Losa GA, Merloini D, Nonnenmacher TF, Weibel ER. Fractals in biology and medicine. 1st ed. Basel: Birkhauser Verlag AG; 2002. Mandelbrott BB. The fractal geometry of the nature. New York: WH Freeman; 1982. McKeehan WL. Glycolysis, glutaminolysis and cell proliferation. Cell Biol Int Rep 1982;18:3275–82. Meadows AL, Kong B, Berdichevsky M, Roy S, Rosiva R, Blanch HW, et al. Metabolic and morphological differences between rapidly proliferative cancerous and normal breast epithelial cells. Biotechnol Prog 2008;24:334–41. Prigogine I, Wiame JM. Biologie et Thermodynamique des phenomenes irreversibles. Experientia 1946;2:451–3. Rohrschneider M, Scheuermann G, Hoehme S, Drasdo D. Shape characterization of extracted and simulated tumor samples using topological and geometric measures. In: Proceedings of the 29th annual international conference of the IEEE EMBS. Engineering in Medicine and Biology Society; 2007. p. 6271–7. Sebri A, Malek J, Tourki R. Automated breast cancer diagnosis based on GVF-snake segmentation, wavelet features extraction and neural network classification. J Comput Sci 2007;3(8):600–7. Smith TG, Lange GD, Marks WB. Fractal methods and results in cellular morphology – dimensions, lacunarity and multifractals. J Neurosci Methods 1996;69:123–36. Thomas CH, Collier JH, Sfeir CS, Healy KE. Engineering gene expression and protein synthesis by modulation of nuclear shape. Proc Natl Acad Sci USA 2002;99:1972–7. Tomassini A, Miccheli A, Di Clemente R, Valerio M, Coluccia P, Bizzarri M, et al. NMRbased metabolic profiling of human hepatoma cells in relation to cell growth. Biochim Biophys Acta 2006;1760(11):1723–31. Trosko JE, Chang CC, Upham BL, Tai MH. Ignored hallmarks of carcinogenesis: stem cells and cell–cell communication. Ann NY Acad Sci 2004;1028:192–201. Wang T, Marquardt C, Foker J. Aerobic glycolysis during lymphocite proliferation. Nature 1976;261:702–5. Xu C, Prince JL. Generalized gradient vector flow external forces for active contours. Signal Process 1998;71(2):131–9. Zotin AA, Zotin AI. Phenomenological theory of ontogenesis. Int J Dev Biol 1997;41:917–21.