Tumor mechanics and metabolic dysfunction

Jason C. Tung a, J. Matthew Barnes a, Shraddha R. Desai b,Christopher Sistrunk b, Matthew W. Conklin c, Pepper Schedin d, Kevin W. Eliceiri e,Patricia J. Keely c, Victoria L. Seewaldt b,1, Valerie M. Weaver a,f,g,h,i,n,1

a Center for Bioengineering and Tissue Regeneration, Department of Surgery, University of California at San Francisco, San Francisco, CA 94143, USAb Department of Medicine, Duke University, Durham, NC 27710, USAc Department of Biomedical Engineering, University of Wisconsin Carbone Comprehensive Cancer Center, Wisconsin Institute for Medical Research,University of Wisconsin at Madison, Madison, WI 53706, USAd Department of Cell, Developmental, and Cancer Biology, Oregon Health & Science University, Portland, OR 97239, USAe Laboratory for Optical and Computational Instrumentation, Laboratory for Cell and Molecular Biology, University of Wisconsin at Madison, Madison, WI 53706, USAf Department of Anatomy, University of California at San Francisco, San Francisco, CA 94143, USAg Department of Bioengineering and Therapeutic Sciences, University of California at San Francisco, San Francisco, CA 94143, USAh Eli and Edythe Broad Center of Regeneration Medicine and Stem Cell Research, University of California at San Francisco, San Francisco, CA 94143, USAi Helen Diller Comprehensive Cancer Center, University of California at San Francisco, San Francisco, CA 94143, USA

a r t i c l e i n f o

Article history:Received 21 May 2014Received in revised form1 November 2014Accepted 25 November 2014Available online 19 December 2014

Keywords:Tumor microenvironmentTumor metabolismMechanosignalingCancerECM stiffnessFree radicals

a b s t r a c t

Desmosplasia is a characteristic of most solid tumors and leads to fibrosis through abnormal extracellularmatrix (ECM) deposition, remodeling, and posttranslational modifications. The resulting stiff tumor stromanot only compromises vascular integrity to induce hypoxia and impede drug delivery, but also promotesaggressiveness by potentiating the activity of key growth, invasion, and survival pathways. Intriguingly, manyof the protumorigenic signaling pathways that are mechanically activated by ECM stiffness also promoteglucose uptake and aerobic glycolysis, and an altered metabolism is a recognized hallmark of cancer. Indeed,emerging evidence suggests that metabolic alterations and an abnormal ECMmay cooperatively drive cancercell aggression and treatment resistance. Accordingly, improved methods to monitor tissue mechanics andmetabolism promise to improve diagnostics and treatments to ameliorate ECM stiffening and elevatedmechanosignaling may improve patient outcome. Here we discuss the interplay between ECM mechanicsand metabolism in tumor biology and suggest that monitoring these processes and targeting their regulatorypathways may improve diagnostics, therapy, and the prevention of malignant transformation.

& 2014 Elsevier Inc. All rights reserved.

Contents

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270Altered ECM dynamics, increased collagen cross-linking, and mechanical signaling in the tumor microenvironment . . . . . . . . . . . . . . . . . . . . . . . . 270

ECM stiffening as a hallmark of cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270Mechanosignaling and cellular tension drive cancer aggressiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271Experimental methods to evaluate ECM stiffness and tissue mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

Metabolism of cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272Oncogenesis and glycolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272Tumor redox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273Therapeutic targeting of tumor metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273Clinical metabolic imaging techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274

A proposed role for tissue mechanics-mediated regulation of tumor cell metabolism and cancer progression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275Physical forces influence cellular metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275Tissue stiffness regulates metabolism and drives cancer progression: a developing direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/freeradbiomed

Free Radical Biology and Medicine

http://dx.doi.org/10.1016/j.freeradbiomed.2014.11.0200891-5849/& 2014 Elsevier Inc. All rights reserved.

n Corresponding author at: Center for Bioengineering and Tissue Regeneration, Department of Surgery, University of California at San Francisco, San Francisco, CA 94143,USA. Fax: þ(415) 476 3985.

E-mail address: Valerie.Weaver@ucsfmedctr.org (V.M. Weaver).1 Shared senior authors.

Free Radical Biology and Medicine 79 (2015) 269–280

Conclusions and future perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

Introduction

Over the past several decades, an appreciation for the importanceof cell and tissue mechanics in cancer initiation, progression, andmetastasis has emerged [14,94,118,123,132,134]. The protumorigeniccell and tissue mechanophenotype manifests as both an intrinsicalteration of cell and tissue structure and mechanics and changes inthe biophysical properties of the tumor microenvironment (i.e., mech-anics, geometry, and topology of the extracellular matrix) [82,97,105].In particular, the noncellular extracellular matrix (ECM) component ofthe tumormicroenvironment plays a critical role in promoting invasionand metastasis [11]. What is now emerging is that interactionsbetween the cell and its associated ECM create a dynamic mechanicalrelationship mediated by a balance of the cell's contractility and thephysical state (i.e., elastic vs rigid) of the ECM microenvironment. Thisbiophysical equilibrium serves to regulate a variety of critical cellularprocesses, including cell differentiation, proliferation, and motility. Notsurprisingly, perturbation in the biophysical dynamics between theepithelium and the ECM potentiates the activity of key signalingpathways that regulate tumor growth, invasion, and survival [88].Interestingly, many of the signaling pathways that promote the aggr-essive behavior of cancer cells also regulate glucose uptake andglycolysis. Deregulated cellular energetics is an emerging hallmark ofaggressive cancers and reflects the metabolic reprogramming eventthat has been closely linked to tumor cell proliferation, under thehostile conditions of the tumor microenvironment [13,50]. In thisreview, we provide a brief overview of the independent roles played byan altered tissue metabolism and an aberrant, stiffened ECM in cancerinitiation and progression and highlight emerging data suggesting aregulatory connection between these two critical cancer regulators. Weargue that novel tractable biomarkers and efficacious prevention andtherapy programs may be developed by targeting the reciprocalfeedback loop between cancer metabolism and the aberrant, mechani-cally modified ECM.

Altered ECM dynamics, increased collagen cross-linking, andmechanical signaling in the tumor microenvironment

ECM stiffening as a hallmark of cancer

Tumors are typically fibrotic and are often characterized by incr-eased and abnormal deposition, turnover, and posttranslational mod-ifications of the ECM that progressively stiffen the stroma [25,84,87,96]. Accordingly, the architecture and physical properties of tumor-associated ECM are fundamentally different from those of the normaltissue stroma. In particular, collagens are often deposited in greaterabundance early during cancer development [19,25,68,122,155] andare major contributors to the tissue stiffness that stimulate mech-anosignaling in stromal fibroblasts and the cancerous epithelium[84,113,121]. The extracellular processing of collagen is mediated thro-ugh cleavage of collagen propeptides by specific proteinases. There-after, enzymes such as lysyl oxidases and lysyl hydroxylases catalyzecovalent intermolecular cross-links between collagens and elastin,modifying their elasticity and tensile strength, profoundly changingvarious cellular behaviors. The collagen fibers surrounding the normalepithelial structures of soft tissues, such as the mammary gland andlung, are typically curly and anisotropic [25]. For example, in the breast,collagen I levels are tightly regulated with oriented type I collagen

fibers hugging in parallel the epithelial tubules and relaxed nonorien-ted fibrils surrounding the terminal ductal lobular regions. By contrast,transformed epithelial tissue is characterized by an abundance ofhighly linearized type I collagen bundles that are often oriented perpe-ndicular to the bulk of the noninvading tumor epithelial mass, andalong which tumor cells can be seen migrating on thick fibersprojecting into the parenchyma [58,84,113]. Consistently, malignancyis linked to the expression of ECM remodeling enzymes, includingheparanases, 6-O-sulfatases, cysteine cathepsins, and matrix metallo-proteases [56,69]. Breast tumors, as well as head and neck cancer, alsoshow elevated levels of collagen-processing enzymes, including lysyloxidases and lysyl hydroxylases, and chronic activation of lysyl oxidase,which is a poor prognostic marker [4,78]. Elevated lysyl oxidase-mediated cross-linking also stiffens the tissue ECM through collagenfiber stabilization and enhancement of the mechanical properties,leading to invasion, metastasis, and poor patient survival [4,27,84]. As aresult of this altered architecture and deregulated remodeling, tumorstroma is typically stiffer than normal stroma. Indeed, the ECM asso-ciated with some breast cancer tissue is robustly stiffer than normal[84,87,112]. Interestingly, collagen architecture has also been linkedwith the reduction in breast cancer risk associated with early-agepregnancy [90]. Despite an increased abundance of collagen I in parousstroma, the matrix is actually less linearized and associated with adecrease in tissue stiffness. These aspects of extracellular collagenprocessing serve to inhibit the invasive phenotype, despite elevateddeposition.

Tumor progression is linked to a dynamic and reciprocal remodelingof the local ECM by the resident cancer cells, stromal fibroblasts, andinfiltrating endothelium and immune cells [88]. For instance, a localstiffening of the ECM stimulates actomyosin cell contractility, which,through elevated localized traction forces, simultaneously modifies theadjacent ECM and promotes cell growth, invasion, and survival byfostering focal adhesion maturation and enhancing growth factorreceptor and G-protein-coupled receptor signaling [21,99]. Indeed, theincreased tumor cell and fibroblast contractility initiates tension-dependent matrix remodeling and promotes the linearization ofcollagen fibrils surrounding the invasive front of the tumor. In additionto influencing tissue mechanics, these linearized collagen fibrils alsofacilitate tumor cell invasion and metastasis. For example, migratingtransformed mammary epithelial cells have been observed travelingalong these linear bundles of collagen [58,114,152]. Consistently, stiffen-ing the ECM in vitro and in vivo promotes the malignant progression ofoncogenically modified, premalignant mammary epithelial cells [84].Furthermore, inhibiting tissue fibrosis and stiffening by preventing coll-agen cross-linking enhanced the latency of tumor formation in Her2/Neu mice and significantly reduced tumor incidence and grade [84].Inhibition of lysyl oxidase activity has also been shown to suppresstumor cell metastasis and can modulate the phenotypic characteristicsof basal cells [110]. Clearly, altered ECM dynamics are not merely afeature of the tumor microenvironment, but actively contribute totumor cell aggressiveness.

Because altered ECM dynamics are a fundamental characteristic ofcancer, new approaches that exploit these aberrations are of greatclinical value. Tumor-associated collagen signatures (TACSs) aim todefine the procession of changes with respect to collagen that havebeen observed and classified as markers of mammary carcinomaprogression. TACS-1, when mammary tumors exhibit a localizedincrease in the deposition of collagen near the tumor lesion, occurs

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280270

very early in tumor formation. As tumors increase in size, the colla-gen fibers align parallel to the tumor boundary straight and aredenoted TACS-2. The highest risk signature, TACS-3, occurs whencollagen fibers are bundled and aligned perpendicular to the tumorborder. This signature has the highest stiffness and predicts poorsurvival in women with invasive breast cancer. In preclinical modelsand human breast cancer, when TACS-3 is present, breast cancer cellspreferentially invade along the stiff collagen fibers into the adjacentstroma, with TACS-3 collagen fibers identifying focal sites of breastcancer cell microinvasion [90,125,126]. These studies highlight theclinical importance of altered ECM dynamics in defining and under-standing cancer aggressiveness.

Mechanosignaling and cellular tension drive cancer aggressiveness

Cells sense and convert exogenous forces into signaling pathwaysthrough a mechanism termed mechanotransduction [48]. Throughmechanotransduction, cancer cells respond to mechanical changes intheir microenvironment, including increased compression, hydrostaticpressure, shear stress, and elevated ECM stiffness or cell–cell tension.Whereas tissue compression and shear stress are tissue-level forces,the increased ECM stiffness that cancer cells and their associated

stromal cells experience begins at the nanoscale and progresses to thetissue level to gradually promote malignant transformation [84,112].A shift in the amount or composition of the ECM, an increase in thelevel of cross-linked matrix, or the reorientation of matrix fibrilscooperate to stiffen the tumor tissue, thereby enhancing the activity ofkey signaling pathways that stimulate cell growth and survival andpromote invasion and migration [96,106,115].

Cells sense, process, and respond to mechanical and other biop-hysical cues from the ECM using a complex web of systems thatinclude specialized mechanosensitive ion channels and adhesionreceptors (e.g., integrins) and their associated adhesion plaque pro-teins, which propagate signals deep into the cell through an inter-connected cytoskeletal–molecular motor network. This orchestratedsystem allows cells to modulate their shape and nuclear architectureand tune their actomyosin-generated cellular tension that ultimatelyremodels their local ECM [7,41,57,61,67,81]. The best understoodmatrix-directed mechanotransduction mechanism is initiated throu-gh ECM-dependent integrin activation and clustering, resulting intalin recruitment, and through vinculin activation promotes focaladhesion assembly, focal adhesion kinase (FAK) and paxillin phos-phorylation, subsequent Rho–ROCK-dependent actin remodeling, andreciprocal actomyosin-mediated cell contractility (Fig. 1a) [37,106,114,

Fig. 1. Mechanosignaling and cellular tension drive cancer aggressiveness. (a) Both normal and malignant cells sense and respond to mechanical cues from the ECM viaadhesion receptors (e.g., integrins), intracellular focal adhesions, cytoskeletal networks, and molecular motors. Cell adhesion and focal adhesion in turn collaborate withG-protein-coupled receptors and transmembrane receptor tyrosine kinases to enhance cell growth and survival and invasion by altering phosphoinositide 3-kinase (PI3K)signaling and downstream p-AKT. (b) In compliant 3D gels with material properties similar to those measured in the normal mammary gland, nonmalignant humanmammary epithelial cells form growth-arrested, polarized acini analogous to the terminal ductal lobular units observed at the end buds of the differentiated breast.Incremental stiffening of the basement membrane gel progressively compromises tissue morphogenesis and alters epidermal growth factor-dependent growth of these cells.Thus, colony size progressively increases, lumen formation is compromised, cell–cell junctions are disrupted, and tissue polarity is inhibited. Arrows indicate loss ofendogenous basement membrane and disruption of basal polarity. (c) Nonmalignant human epithelial cells grown on stiff substrates show greater cell contractility,measured via traction force microscopy, compared to the same cells grown on softer substrates. Reprinted from [106] by permission of the publisher.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280 271

121]. Focal adhesions thereafter collaborate with G-protein-coupledreceptors and transmembrane receptor tyrosine kinases to enhancecell growth, survival, migration, and invasion through ERK, JNK, andphosphoinositide 3-kinase (PI3K) signaling (Fig. 1b) [84,106,151]. Inparticular, PI3K regulates cell growth, invasion, and survival and hasbeen implicated in tumor aggression and treatment resistance[39,40]. Not surprisingly, ECM stiffness and elevated epithelialmechanosignaling enhance growth factor receptor-dependent PI3Ksignaling to foster malignant cell behavior [96].

The phosphatase and tensin homolog (PTEN) protein, a tumorsuppressor frequently altered in human cancers, is the masternegative regulator of the PI3K signaling pathway, and thereforeplays a critical role in cell growth, proliferation, and survival [51,80,89]. Not surprisingly, even a modest reduction in PTEN is sufficientto permit the malignant transformation of a tissue and can accel-erate tumor progression and treatment resistance [2]. Indeed, theexquisite cellular sensitivity to PTEN levels and function is reflectedby the diverse regulatory mechanisms controlling its expression.For instance, although normal human mammary cells express highlevels of PTEN protein, nonmalignant mammary epithelial cells(MECs) propagated in vitro on conventional tissue culture plasticexpress low levels of PTEN. However, PTEN protein expression isrestored and presumably stabilized when MECs are grown within athree-dimensional reconstituted basement membrane (3D rBM) toassemble the “differentiated” acini observed in normal breast tissuein vivo [32,86]. Intriguingly, the elastic modulus of tissue culture su-bstrates is superphysiologic, whereas the material properties of rBMhydrogels are remarkably similar to those of the normal humanbreast [63,106]. Indeed, the ECM within a human breast tumor isstiffer than normal tissue, implying that matrix compliance maystabilize PTEN protein. Consistently, we recently showed that a stiffECM increases levels of miR-18a, which directly targets PTENmRNA, both directly and indirectly, by decreasing levels of HOXA9,to regulate malignant transformation and tumor metastasis inculture and in vivo by increasing PI3K/AKT signaling [96]. Thesefindings highlight a direct relationship between PTEN, the biophy-sical properties of the ECM, and PI3K-dependent malignancy.

Compared to their nonmalignant counterparts, transformed epith-elial cells show striking differences in their mechanoresponsivenessthat we and others showed can facilitate their growth, invasion, andsurvival in response to the increased stiffness found in primary tumors[106,138,147]. Many tumor cells exhibit differential intermediatefilament profiles and cytoskeletal architectures that modify theirmigratory behavior. The replacement of a keratin-based cytoskeletonwith a vimentin-based cytoskeleton is a hallmark of epithelial-to-mesenchymal transition (EMT) in mammary tissue and contributessignificantly to invasive potential [74,103,150]. Cancer cells are alsofrequently highly contractile and this elevated cytoskeletal tensioncontributes to their ability to grow, invade, and survive in the hostiletumor microenvironment. For instance, transformed cells plated oncompliant substrates spread rapidly, grow, and survive, whereas theirnormal counterparts do not [147]. Tumor cells also generate high tra-ction forces that disrupt cell–cell junctions, compromise tissue polar-ity, promote anchorage-independent survival, and enhance invasion(Fig. 1c) [106]. High cellular tension also increases Rho/ROCK activity,which, via engagement of its downstream effectors, promotes tumorcell proliferation, migration, and survival [121]. These findings demon-strate the reciprocal relationship between ECMmicromechanical sign-als and cell structural control, how these signals propagate throughsignal transduction networks, and the subsequent modulation of keybehaviors connected with tumorigenesis, invasion, and metastasis.

Experimental methods to evaluate ECM stiffness and tissue mechanics

The degree of tumor stiffness, which is typically determined bymeasuring its elastic modulus, has been exploited as a means to detect

malignant lesions in an otherwise normal tissue. The elastic modulusis a measure of how easily a material deforms and is calculated bydividing stress by strain; the greater the elastic modulus, the stiffer thematerial. Thus, the elastic modulus offers a quantitative method fordetermining mechanical differences between tissues.

Multiple techniques have been utilized to characterize tissue levelchanges in elastic modulus in tumors with varying degrees of success.Unconfined compression and shear rheology testing have both beenused to quantify incremental stiffening of the mammary gland as ittransitions from normal to premalignant to invasive cancer, and havedemonstrated that the stromal tissue adjacent to the invadingepithelium is substantially stiffer than normal [84,106]. Shear rheol-ogy testing has also been used to monitor changes in tissue stiffnessduring the progression of liver fibrosis [38,108]. These techniqueslend themselves to gross tissue level assays in which resolution andsensitivity are less demanding. However, recent findings that stressthe functional implications of ECM stiffness and an altered celltension on transformation potential and treatment response demandmore precise measurements. Thus, although unconfined compressionand shear rheology provide valuable quantitative measurements oftissue elastic modulus, they are limited by their millimeter resolutioncapabilities.

Atomic force microscopy (AFM) is one imaging modality thatpermits higher resolution imaging of tissue elastic modulus. AFMmeasures the interaction force between a sample surface, such as aliving cell, and a microscale to nanoscale spring-like cantilever. Tip–sample interactions generate a force that deflects the cantilever,which can then be optically tracked via a photodiode and convertedto an interaction force using the spring constant of the cantilever. Theforce measurement capability of AFM has been used to measure theregional tissue elastic modulus of cultured brain slices [26], pancrea-tic ductal adenocarcinoma tissue [102], and excised mammary tissue[87,90,96,110]. These measurements yielded high-resolution forceheat maps that demonstrate the existence of stiffness tracts thatregister with regions of collagen fiber enlargement and linearization.Although AFM has effectively characterized changes in tissue elasticmodulus at the microscale, the clinical applicability of the techniqueis severely limited by the requirement of unprocessed samples.

Over the past decade, elastography, which can employ eitherultrasound or magnetic resonance imaging (MRI) techniques, hasbeen utilized clinically to map tissue elastic modulus. This techniquehas found utility in tumor diagnosis [70], as well as intraoperativelocalization of tumor tissue during resection of gliomas [127,141].Despite its widespread usage, elastography is limited by its lowresolution compared with the length scales relevant to many diseases(hundreds of micrometers for ultrasound elastography and milli-meters for MRI elastography), the long acquisition times associatedwith MRI elastography, and issues with tissue penetrance and uservariability with ultrasound elastography. Because of the limitationsdescribed above, new imaging techniques must be developed tomeasure tissue elastic modulus as a means of precisely and accuratelydiagnosing and monitoring cancer.

Metabolism of cancer

Oncogenesis and glycolysis

Transformed cells exhibit aberrant growth rates that demandelevated levels of ATP to support the biogenesis of the cells' fourmajor macromolecules: carbohydrates, proteins, lipids, and nucleicacids. The physician scientist Otto Warburg hypothesized early on thattumor cells must adjust their metabolic machinery to accommodatethese energy requirements. Warburg found that tumor cells transitionfrom the conventional ATP-producing system, oxidative phosphoryla-tion (OXPHOS), to glycolysis in the presence of adequate oxygen. The

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280272

oddity of this observation, known as the Warburg Effect or aerobicglycolysis, is that glycolysis is usually reserved for anaerobic ATPproduction and, per unit of glucose, is a much less efficient producer ofATP than OXPHOS [148,149]. Although apparently counterintuitive,glycolysis allows much faster, on-demand, ATP production thanOXPHOS, but at the demand of more glucose uptake. Despite havingmajor implications in basic tumor cell biology as well as clinicaldiagnostics and therapy, tumor metabolism was largely ignored fordecades until recently, when transforming oncogenes were found tohave both direct and indirect effects on the cell's central metabolism.

At present a growing number of oncogenes and tumor suppressorgenes have been functionally linked to cell metabolism (for anexcellent comprehensive review, see [13]). Ras and Src were the firstoncogenes associated with the Warburg Effect owing to their induc-tion of glucose transporter (GLUT genes) expression [30]. Sim-ilarly PI3K/AKT upregulation leads to increased GLUT1 and GLUT4expression, thus allowing more uptake of glucose to fuel glycolysis butalso activating (phosphorylates) a number of metabolic enzymesincluding hexokinase and phosphofructokinase 2, the former of whichcatalyzes glucose to glucose 6-phosphate, the first step of glycolysis.Additionally, by activating the mammalian target of rapamycin(mTOR), AKT activity leads to increased mRNA translation and proteinbiogenesis [77]. Another common oncoprotein, Myc, drives aerobicglycolysis directly through the upregulation of glucose transportersand phosphofructokinase-2 [101] and cooperatively with hypoxia-inducible factor 1 (HIF-1) to induce hexokinase 2 and pyruvatedehydrogenase [71]. Not surprisingly loss of function of tumorsuppressors can mimic the metabolic phenotype driven by oncogeneexpression. Through the TP53-inducible glycolysis and apoptosisregulator, p53 reduces glycolysis and enhances OXPHOS throughSCO2 (cytochrome c oxidase 2) [6,93]. P53 also maintains expressionof PTEN, which is the major regulatory brake on the PI3K/AKT/mTORpathway [133]. Further, PTEN loss or AKT activation leads to increased

glycolysis through depletion of ATP stores [29]. These findings high-light the pleiotropic regulation of metabolism by classical oncogenicand tumor-suppressive signaling networks (Fig. 2).

Tumor redox

A major component of the tumor microenvironment and a poorpredictor of outcome is hypoxia, which occurs from the combinationof the enhanced metabolic and replicative states of tumor cells with acompromised vasculature [142,143]. Through activation of HIF-1,hypoxia acts to reinforce glycolysis, and sustained hypoxia leads tomitochondrial dysfunction [104]. Collectively, this results in an abnor-mal cellular redox status. Redox, which is the balance of the reducedand oxidized states of a cell, is governed by the levels of reactiveoxygen (ROS) and nitrogen species and antioxidant molecules. As aconsequence of increased ATP production and macromolecule biogen-esis, tumor cells accumulate abnormally high levels of ROS [3,135,139].Through posttranslational modifications, these oxidative moleculescontribute to tumor progression by augmenting the activity ofoncogenic proteins such as PI3K, mitogen-activated protein kinases,and Src while further inhibiting tumor suppressors such as PTEN andPTP [137]. Further, ROS support the transcription of HIF-1α and GLUT1,thus creating a feed-forward mechanism that enhances the WarburgEffect in tumor cells [5]. Although increased ROS production is usuallythought of as protumorigenic, certain thresholds of oxidative damageto proteins and nucleic acids can trigger cell death through apoptosis[13,136], thus revealing another potential therapeutic opportunity inthe treatment of tumors.

Therapeutic targeting of tumor metabolism

An exciting development in tumor therapy is the exploitation ofabnormal tumor metabolism [35]. Metformin is an agonist of AMPK(AMP-activated protein kinase), a finely tuned sensor of the cellularAMP/ATP ratio. At high ratios, AMPK is activated and negativelyregulates mTOR, thus reducing protein production and proliferation.Secondarily AMPK activity reduces insulin production and gluconeo-genesis [128]. Interestingly, before the mechanism of action ofmetformin was studied, it was already found to improve insulinsensitivity and reduce blood glucose in diabetics. Today this is themost common drug in the treatment of type 2 diabetes [24]. Fromdecades of diabetic patient data, large meta-analyses were performedshowing that those patients on metformin have significantly reducedrisk of cancer incidence and cancer-associated death [10,28], whichhas led to clinical trials in several solid tumors, including colorectal,endometrial, prostate, breast, and brain (recently reviewed by Kasz-nicki et al. [66]). Whereas mouse tumor models and cell cultureexperiments show that metformin induces mTOR inhibition, growtharrest, and apoptosis of tumor cells [59,85], patient data are quitelimited. In breast cancer patients, metformin reduces mTOR tran-scription and the Ki67 staining proliferative index [47] and alsoimproves treatment response rates [62]. Metformin treatment alsoreduces the rate of colorectal cancer formation and levels of prostatecancer death [52,107]. Many ongoing clinical trials will elucidate theefficacy of metformin as a first line of treatment, in combinationtherapy, and in tumor prevention studies. In addition to metformin,multiple drugs that target various aspects of metabolism are inpreclinical and clinical trials [35].

Because metformin negatively regulates pathways downstreamof AKT [46], which are important in tumor ECM remodeling andmechanosignaling, it is possible that use of this drug may help toreduce tissue fibrosis and thereby temper the malignant effects ofa stiffened ECM. Exciting recent data in support of this hypothesisinclude findings that metformin treatment in heart failure-pronerats reduced perivascular fibrosis and prevented heart failure.Additionally, in a mouse model of cardiac fibrosis, metformin

Fig. 2. Regulation of metabolism in tumor cells. Normal cells use the TCA cycle(OXPHOS) for ATP production, whereas oncogenic signaling drives a glycolyticphenotype in tumor cells. PI3K activates AKT, which stimulates glycolysis bydirectly regulating glycolytic enzymes and by activating mTOR. mTOR has pleio-tropic effects on metabolism, but facilitates the glycolytic phenotype by enhancingHIF-1 activity, which engages a hypoxia-adaptive transcriptional program. HIF-1induces the expression of glucose transporters (GLUTs), glycolytic enzymes, andPDK1, which blocks the entry of pyruvate into the TCA cycle. The LKB1 tumorsuppressor activates AMPK, which opposes the glycolytic phenotype by inhibitingmTOR. Myc cooperates with HIF to induce several genes that encode glycolyticproteins, but also increases mitochondrial metabolism. The tumor suppressor p53opposes the glycolytic phenotype by suppressing glycolysis through TIGAR,increasing mitochondrial metabolism via SCO2 and supporting expression of PTEN.OCT1 acts in an opposing manner to activate the transcription of genes that driveglycolysis and suppress oxidative phosphorylation. The switch to the PKM2 isoformaffects glycolysis by slowing the pyruvate kinase reaction and diverting substratesinto alternative biosynthetic and reduced NADPH-generating pathways. TCA,tricarboxylic acid; HIF-1, hypoxia-inducible factor 1; LKB1, liver kinase B1; PDK1,pyruvate dehydrogenase kinase, isozyme 1; AMPK, AMP-activated protein kinase;PKM2, pyruvate kinase M2; TIGAR, TP53-induced glycolysis and apoptosis reg-ulator; MCT, monocarboxylate transporter; PDH, pyruvate dehydrogenase. Thedashed lines indicate the shift from TCA to glycolysis under the loss of p53function. Reprinted from [13] by permission of the publisher.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280 273

treatment led to reduced levels of TGFβ and activated Smad3, thusdecreasing collagen deposition and fibrosis [18,154]. These studieshint at connections between altered tumor mechanics and meta-bolism, which are discussed below.

Clinical metabolic imaging techniques

Links between dysfunctional metabolism and malignancy havefostered the development of a number of noninvasive clinicallytractable approaches to image tissue metabolism in situ. By deliveringradioactively tagged glucose to patients, tumors can be detected bytheir abnormally high uptake of the metabolite. [18F]Fluorodeoxyglu-cose (FDG) is taken up through GLUT transporters and phosphorylatedby hexokinase. Owing to the lack of a 20-hydroxyl group, thisradiolabeled analog of glucose is not metabolized further, allowingfor the identification of metabolically active tissue with positronemission tomography (PET) imaging. Clinically, FDG–PET has beenused successfully in diagnosis, staging, monitoring response to therapy,and predicting outcome in tumors of the breast, colon, and lung[36,60]. Clinical histopathology and experimental xenograft studieshave shown that FDG uptake is correlated with increased HIF-1α andGLUT expression [1,64] as well as Ki67 proliferative status [49].

MRI and magnetic resonance spectroscopy (MRS) allow quantita-tive analysis of metabolites in a tissue. This technique has beenparticularly useful in the diagnosis and prognosis of brain tumors

[95]. An adaptation of MRS is hyperpolarized 13C MRS, which allowstracing of the fate of 13C-radiolabeled metabolites. Although not yetFDA approved, this technique is safe in patients [98] and has beenused in cell culture and mouse models to study glutaminase activityin hepatocellular carcinoma cells [33] and the abnormal metabolismof α-ketoglutarate in IDH1 mutant gliomas [16]. Current advances inthis technology and reduction in 13C reagent costs will probably pushthis technique into the clinic [75]. Metabolic changes occur earlyduring tumorigenesis; thus advancements in FDG–PET and MRSmetabolic imaging have the potential to help diagnose various tumortypes at much earlier stages.

Another approach that is useful in the metabolic analysis ofarchived biopsy tissue is the fluorescence imaging of NADH/NADPHand FAD. These ubiquitous metabolic cofactors, which play key rolesin glycolysis, OXPHOS, and ROS scavenging, are autofluorescent andreadily imaged because of their relatively high concentrations withinthe cell. Biochemical means have uncovered changes in the redoxstate of these molecules in tumor cells during early stages ofcarcinoma and when tumors progress [130,144,156]. In addition tointensity, the fluorescence lifetime imaging (FLIM) of NADPH and FADserves as a readout of metabolic state, as each of these molecules hasa unique FLIM value depending on whether it is free in the cytosol orbound to protein (Fig. 3). Metabolic imaging of NADPH and FAD byFLIMmatches metabolic perturbations in culture, and this technologyis able to distinguish basal cell skin sarcomas from surrounding

Fig. 3. Fluorescence lifetime imaging of a high-risk breast biopsy. (a) Standard hematoxylin and eosin-stained section with a region of atypia. (b) FLIM of the unstainedadjacent section in which the weighted average lifetime of NADH (τm) and (c) the fraction of free NADH (a1) is shown as a color map. (d) Quantitative histograms of the pixel-by-pixel data from the regions of interest outlined in (b). An increase in a1, which corresponds to a decrease in τm, in epithelium displaying atypia is an indication these cellshave either increased glycolysis or decreased oxidative phosphorylation.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280274

normal skin, as well as glioma from normal brain [8,22,34,83,146].FLIM of NADPH and FAD discerns early carcinoma in situ of themammary gland in mouse [20], suggesting these approaches may beuseful for in situ determinations of early changes associated withaerobic glycolysis. Consistent with this finding, our recent datademonstrate that changes in NADPH fluorescence lifetime valuesmay serve as an early biomarker for biochemical changes in tissuethat is deemed morphologically normal by standard pathologicalassessment. The clinical relevance of these approaches is furtherhighlighted by the finding that metabolic imaging by FLIM corre-sponds to tamoxifen responsiveness in both breast xenographs andhuman head and neck squamous cell carcinoma [129,146].

A proposed role for tissue mechanics-mediated regulationof tumor cell metabolism and cancer progression

Physical forces influence cellular metabolism

A stiffened ECM is becoming appreciated as a critical component ofthe tumor microenvironment and data are accumulating to show that

mechanics and cell adhesion regulate metabolism. Upon integrin-mediated ECM adhesion, cells exhibit a spike of ROS that is necessaryfor subsequent integrin signaling and Src activation [17], and quench-ing of ROS reduces v-Src-driven tumorigenicity and invasion [42]. Inaddition to adhesion, the compliance of ECM has marked effects oncellular growth and proliferation [106,138] processes, which demandincreased metabolic flux. Interestingly, the EMT, a process heavilyimplicated in invasion, is favored on stiffer substrates [79,138], andmesenchymal cell types exhibit increased aerobic glycolysis throughincreased Zeb1-induced expression of glucose transporters [92] andSnail-induced methylation of fructose 1,6-bisphosphate [23]. More-over, stiffness-induced Rac1b facilitates formation of the NADPHoxidase complex, which drives ROS production and enhances theEMT phenotype in breast epithelial cells [79].

The interplay between physical forces and metabolism is perhapsbest understood in endothelial cells. Analogous to the regulation ofhomeostasis by ECM compliance, physiologic (regular pulsatile orlaminar flow) shear stress helps regulate gene expression and redoxof endothelial cells; however, under conditions of elevated orirregular (turbulent or oscillatory) shear stress, these cells upregulateRac1 activity, which fuels ROS production, driving NF-κB activation

Fig. 4. Breast malignancy associates with miR-18a and reduced PTEN expression. (a) HOXA9 (top, red), PTEN (middle, red), p-AKT substrate (bottom, red), and DAPI (blue)staining of human nonmalignant breast samples and breast tumor samples. The epithelium of normal breast tissue coexpressed appreciable quantities of nuclear HOXA9 andcytoplasmic PTEN proteins, as did the more differentiated luminal A breast tumors. In contrast, the less differentiated luminal B, basal-like, and HER2þ tumors showedreduced expression of HOXA9 and PTEN and higher levels of activated AKT. Scale bar, 100 μm. (b) As measured by atomic force microscopy, miR-18a expression is correlatedwith breast ECM elastic modulus in human patient samples from both normal and transformed breast tissue. Reprinted from [96] by permission of the publisher.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280 275

and facilitating atherosclerosis [140]. Further, irregular shear stress inischemia leads to PI3K/AKT activation and ROS production in a feed-forward mechanism [15]. Clear connections between cell adhesion,mechanical signaling, and metabolism exist; however, more studiesare needed to understand the molecular mechanism by which theseprocesses influence one another.

Tissue stiffness regulates metabolism and drives cancer progression: adeveloping direction

Both altered metabolism and deregulated ECM dynamics clearlyact as fundamental drivers of cancer progression. However, emergingevidence suggests that these two characteristic features of cancer mayalso be interconnected, driving the progression of cancer via asynergistic, cooperative network. The PI3K signaling pathway iscentral for regulation of glucose metabolism. In fact, activation ofthe PI3K/AKT pathway renders cells dependent on high levels ofglucose flux, a classic biochemical phenomenon of cancer cells [12]. Inaddition to activation of mTOR-mediated signaling, the PI3K pathwayalso regulates glucose uptake and utilization. In insulin-independenttissues, PI3K signaling through AKT regulates the expression ofglucose transporters, GLUT1 and GLUT4 [73], via hexokinase, therebystimulating phosphofructokinase activity and increasing glucoseuptake [9,119], an action critical for enhanced aerobic glycolysis. How-ever, PI3K activation also occurs in the context of increased matrixstiffness via integrin-mediated activation of FAK and downstreampathways [45,65,116,117,153]. Thus, regulation of PIP3 by PI3K andsubsequent activation of AKT andmTOR represent mechanistic meansby which matrix stiffness and increased ECM deposition may influ-ence cellular metabolism, thereby promoting both cell proliferationand survival [106,151], as well as facilitating oncogenic transformationand tumor metastasis [84]. Matrix stiffness may also regulate meta-bolism through effects on p53, as AKT drives p53 degradation viaMDM2 [53]. Moreover, p53 is further regulated by reduction ofcysteine residues and may become inactivated as the redox balanceof the cell shifts [31].

Matrix stiffness-mediated PI3K activation of AKT is also reinforcedthrough β-catenin and Myc-dependent matrix stiffness-inducedexpression of miR18a, which inhibits PTEN expression and stimulatescancer progression [96]. Consistent with these findings, knockout ofFAK in mammary tumors downregulates the activity of PI3K and AKTwhile enhancing PTEN expression and disrupting tumor growth andmetastasis [114]. Activation of PI3K by growth factors alone has alsobeen shown to be insufficient to induce accumulation of PIP3, andoxidation-mediated inactivation of PTEN is required to increase theconcentration of PIP3 sufficiently to trigger downstream signalingevents [76]. Thus, oxidation- and stiffness-mediated inactivation ofPTEN may work cooperatively to trigger the PIP3 downstreamsignaling events that drive cancer progression.

Evidence also suggests a dynamic relationship between cell attach-ment to the ECM and metabolic activity regulation. Detachment ofmammary epithelial cells from ECM causes an ATP deficiency owing tothe loss of glucose transport. However, this deficiency can be rescuedvia overexpression of ErbB2, which restores glucose uptake throughstabilization of the epidermal growth factor receptor and PI3K activa-tion via glucose-stimulated flux through the antioxidant-generatingpentose phosphate pathway. Interestingly, ATP deficiency can also berescued by antioxidant treatment without rescue of glucose uptake,which is dependent on stimulation of fatty acid oxidation, a processinhibited by detachment-induced reactive oxygen species [124]. Thesefindings demonstrate that cell attachment to the ECM directly regu-lates metabolic activity, providing a means for improving cell survivalin the disrupted ECM tumor microenvironment through antioxidantrestoration of ATP generation.

Intriguing microarray data generated from culturing normalmouse mammary gland cells in dense versus compliant 3D collagen

matrices suggests an enhanced oxidative phosphorylation signaturein stiffer environments [115]. Although this may reflect the differencebetween normal and transformed cells, there is increasing evidencethat oxidative phosphorylation is also increased in tumors [131].Increased oxidative phosphorylation in the context of increasedaerobic glycolysis allows cells a robust metabolism that may be nece-ssary for the substantial energy demands of proliferation, migra-tion, and metastasis. Indeed, AKT/mTOR activation of 4EBP1 leads toincreased mitochondrial biogenesis [44] and may represent anadditional means by which matrix stiffness regulates metabolismthrough a FAK/PI3K/AKT pathway.

Enhanced ECM deposition, aligned collagen, collagen cross-link-ing, and fibrosis are all key aspects of breast cancer progression. Asdescribed here, not only are there solid data suggesting that ECMstiffness stimulates aberrant cell growth, survival, invasion, andmetastasis through modulation of the PI3K/AKT network, but ECMstiffness may also contribute to altered cellular metabolism.

Conclusions and future perspectives

Increased deposition, remodeling, and cross-linking of ECM wit-hin the breast, and the resulting increased tensile forces, are charac-teristic of breast cancer progression and also serve to activatesignaling pathways critical for proliferation, survival, invasion, andmetastasis. Emerging evidence also shows that some of these sameECM-activated signaling pathways promote aerobic glycolysis andglucose uptake. Here we provide evidence that tissue mechano-signaling activates signaling networks that simultaneously promotebreast carcinogenesis and metabolic reprogramming. Both alteredECM dynamics and metabolic reprogramming are critical drivers ofcancer aggressiveness. Thus, research focusing on their intercon-nected feedback loop may be a promising direction for biomarkerdiscovery and drug development.

Emerging clinical data have begun to link stiffer matrices withincreased cancer aggressiveness in human patients, emphasizinghowmatrix stiffness and deregulated mechanosignaling componentscould provide useful strategies for clinical intervention and to iden-tify novel biomarkers [96]. In basal-like or luminal B samples, breastECM stiffness and miR-18a expression were found to be significantly

Fig. 5. Metabolically active triple-negative breast cancer from women at high-riskfor breast cancer. (a) High glucose uptake as measured by 18FDG–PET. (b) Timed3.0-T MRI demonstrates the vascularity of the cancer.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280276

higher compared to levels measured in the normal and luminal Atumor biopsies. Owing to this mechanically mediated miR-18aexpression profile, PTEN expression was found to be expressed inappreciable quantities in the more differentiated luminal B cancersand normal breast tissue, whereas less differentiated luminal B,basal-like, and HER2þ tumors showed reduced expression of PTEN,along with higher levels of activated AKT (Fig. 4). MiR-18a levels alsodistinguished luminal A from luminal B tumors and predicted clinicaloutcome in patients with luminal type breast cancer. These findingssuggest that macro-level tumor stiffness may parallel future diseaseaggression, with tumor size, grade, and subtype as independent fac-tors influencing tissue mechanics. Furthermore, ECM stiffness acts asa biophysical parameter that can distinguish luminal A from luminalB tumors, with miR-18a and PTEN presenting as tractable biomar-kers. Unfortunately, imaging techniques that characterize tissuestiffness in the clinical setting are severely limited. However, becauseof the interconnected relationship of metabolism and tissue stiffness,the identification of surrogate metabolic markers for tissue stiffnessmay serve to accurately monitor cancer progression.

Altered metabolism and tissue elastic modulus have typically beenidentified as independent features of cancer. In fact, activation of PI3K/AKT, excess metabolic energy, and glycolysis are all associated withpoor prognosis in triple-negative breast cancer [72,109,120]. Interest-ingly, these factors seem to influence clinical outcome through similarand overlapping pathways. In response to increased mechanicalstiffness, integrin clustering, ERK activation, cytoskeletal remodeling,and Rho GTPase-dependent contractility are engaged and foster theassembly of focal adhesions, which in turn collaborate with G-protein-coupled receptors and transmembrane receptor tyrosine kinases toalter PI3K signaling and downstream AKT signaling. Mutations in thekey component PI3K lead to subsequent downstream activation ofAKT and predict poor survival and risk of recurrence [43,100]. Geneticdefects in mitochondrial respiration and increase in glycolysis alsoactivate AKT signaling via increases in metabolic potential mediated byNADH, which promotes recruitment of GLUTs to the cell surface andincreased glucose uptake and glycolysis [91]. These high glucose levels

have also been shown to promote the aggressive behavior of triple-negative breast cancer, including proliferation, activation of oncogenicsignaling, apoptosis resistance, and reduced drug efficacy [145].Interestingly, women at high risk for developing triple-negative breastcancer show altered metabolic state [54]. Random periareolar fine-needle aspiration samples taken from high-risk patients were shownto express high levels of activated proteins representing RTK/AKT/mTOR, RTK/AKT/ERK, and mitochondrial apoptosis, suggesting thatsignificant metabolic dysfunction exists long before tumors can bedetected in these patients. The strong correlation between ECMstiffness and AKT suggests that these high-risk women may also havealtered ECM, making them prime candidates for elastography imagingstudies. PET imaging is also now being used to determine glucoseuptake in these high-risk patients (Fig. 5). Many of these high-riskpatients are also involved in chemotherapy prevention trials thatinclude metformin. It will be interesting to see whether this early stateof metabolic dysfunction is reversible and whether early targeting ofaltered metabolism is an effective preventative strategy. Takentogether, these studies highlight the central role of AKT and redoxactivation in defining the aggressive behavior of triple-negative breastcancer. Given the key role that matrix stiffness plays in regulatingPI3K/AKT signaling, these studies also provide evidence for the conve-rgence of mechanosignaling and redox activation in aggressive can-cers. Because AKT-network signaling is active in premalignant tissuefrom women at high risk for breast cancer before the development ofan invasive phenotype [55,111], this opens the possibility that combi-nation therapies that target both cellular metabolism and tissuestiffness may serve to more effectively halt disease progression.

Altered metabolism and ECM stiffening are two chief componentsof the tumor microenvironment that have a profound impact oncancer progression (Fig. 6). Although new evidence suggests that bothare intricately linked through their downstream signaling pathways,additional studies will need to be conducted to fully investigate theircomplex relationship. Nevertheless, this newly developing area ofcancer research may provide fresh avenues to pursue for the devel-opment of tractable biomarkers and novel cancer therapeutics.

Fig. 6. Proposed model by which a stiffened extracellular matrix (ECM) regulates metabolism. Through mechanical activation of pathways highlighted in Figs. 1 and 2, ECMstiffness (and other perturbed forces within the tumor microenvironment) drives a glycolytic phenotype.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280 277

Acknowledgments

This work was supported by National Institutes of Health GrantsNHLBI T32 HL007544 to J.C.T. and F32 CA174319 to J.M.B. as well as R01CA155664, R01 CA158668, and R01 CA170851 to V.L.S. and R01CA136590, R01 CA114462, R01 CA142833, and NIBIB R21 EB008811to P.J.K. and U54CA163155-01, U54CA143836, CA138818-01A1, R01CA174929, R33 CA183685-01, and U01 ES019458 to V.M.W. Support forthe review was also provided by the Susan G. Komen Breast CancerAwards KG091020 to V.L.S. and KG110560PP to V.M.W. and a V-Foundation Award (http://www.jimmyv.org/) to V.L.S. and by CDMRPBreast Cancer Research Program Grant BC074970 to P.J.K. and Era ofHope Scholar Expansion W81XWH-13-1-0216 to V.M.W. The fundershad no role in the study design, data collection and analysis, decision topublish, or preparation of the manuscript.

References

[1] Ai, B.; Pan, T.; et al. [The relationship of expression of GLUT1, HIF-1alpha andthe uptake of FDG in non-small cell lung cancer]. Zhongguo Fei Ai Za Zhi10:508–512; 2007.

[2] Alimonti, A.; Carracedo, A.; et al. Subtle variations in Pten dose determinecancer susceptibility. Nat. Genet. 42:454–458; 2010.

[3] Bae, Y. S.; Kang, S. W.; et al. Epidermal growth factor (EGF)-inducedgeneration of hydrogen peroxide: role in EGF receptor-mediated tyrosinephosphorylation. J. Biol. Chem. 272:217–221; 1997.

[4] Barker, H. E.; Chang, J.; et al. LOXL2-mediated matrix remodeling inmetastasis and mammary gland involution. Cancer Res. 71:1561–1572; 2011.

[5] Bell, E. L.; Klimova, T. A.; et al. The Qo site of the mitochondrial complex III isrequired for the transduction of hypoxic signaling via reactive oxygen speciesproduction. J. Cell Biol. 177:1029–1036; 2007.

[6] Bensaad, K.; Tsuruta, A.; et al. TIGAR, a p53-inducible regulator of glycolysisand apoptosis. Cell 126:107–120; 2006.

[7] Bershadsky, A. D.; Balaban, N. Q.; et al. Adhesion-dependent cell mechan-osensitivity. Annu. Rev. Cell Dev. Biol. 19:677–695; 2003.

[8] Bird, D. K.; Yan, L.; et al. Metabolic mapping of MCF10A human breast cellsvia multiphoton fluorescence lifetime imaging of the coenzyme NADH.Cancer Res. 65:8766–8773; 2005.

[9] Biswas, S.; Lunec, J.; et al. Non-glucose metabolism in cancer cells—is it all inthe fat? Cancer Metastasis Rev. 31:689–698; 2012.

[10] Bowker, S. L.; Majumdar, S. R.; et al. Increased cancer-related mortality forpatients with type 2 diabetes who use sulfonylureas or insulin: response toFarooki and Schneider. Diabetes Care 29:1990–1991; 2006.

[11] Butcher, D. T.; Alliston, T.; et al. A tense situation: forcing tumourprogression. Nat. Rev. Cancer 9:108–122; 2009.

[12] Buzzai, M.; Bauer, D. E.; et al. The glucose dependence of Akt-transformedcells can be reversed by pharmacologic activation of fatty acid beta-oxidation. Oncogene 24:4165–4173; 2005.

[13] Cairns, R. A.; Harris, I. S.; et al. Regulation of cancer cell metabolism. Nat. Rev.Cancer 11:85–95; 2011.

[14] Castano, Z.; Tracy, K.; et al. The tumor macroenvironment and systemicregulation of breast cancer progression. Int. J. Dev. Biol. 55:889–897; 2011.

[15] Chatterjee, S.; Browning, E. A.; et al. Membrane depolarization is the triggerfor PI3K/Akt activation and leads to the generation of ROS. Am. J. Physiol.Heart Circ. Physiol 302:H105–114; 2012.

[16] Chaumeil, M. M.; Larson, P. E.; et al. Hyperpolarized [1-13C]glutamate: ametabolic imaging biomarker of IDH1 mutational status in glioma. CancerRes. 74:4247–4257; 2014.

[17] Chiarugi, P.; Pani, G.; et al. Reactive oxygen species as essential mediators ofcell adhesion: the oxidative inhibition of a FAK tyrosine phosphatase isrequired for cell adhesion. J. Cell Biol. 161:933–944; 2003.

[18] Cittadini, A.; Napoli, R.; et al. Metformin prevents the development ofchronic heart failure in the SHHF rat model. Diabetes 61:944–953; 2012.

[19] Conklin, M. W.; Eickhoff, J. C.; et al. Aligned collagen is a prognosticsignature for survival in human breast carcinoma. Am. J. Pathol. 178:1221–1232; 2011.

[20] Conklin, M. W.; Provenzano, P. P.; et al. Fluorescence lifetime imaging ofendogenous fluorophores in histopathology sections reveals differencesbetween normal and tumor epithelium in carcinoma in situ of the breast.Cell. Biochem. Biophys. 53:145–157; 2009.

[21] Croft, D. R.; Sahai, E.; et al. Conditional ROCK activation in vivo inducestumor cell dissemination and angiogenesis. Cancer Res. 64:8994–9001; 2004.

[22] De Beule, P. A.; Dunsby, C.; et al. A hyperspectral fluorescence lifetime probefor skin cancer diagnosis. Rev. Sci. Instrum 78:123101; 2007.

[23] Dong, C.; Yuan, T.; et al. Loss of FBP1 by Snail-mediated repression providesmetabolic advantages in basal-like breast cancer. Cancer Cell 23:316–331;2013.

[24] Dowling, R. J.; Goodwin, P. J.; et al. Understanding the benefit of metforminuse in cancer treatment. BMC Med. 9:33; 2011.

[25] Egeblad, M.; Rasch, M. G.; et al. Dynamic interplay between the collagenscaffold and tumor evolution. Curr. Opin. Cell Biol. 22:697–706; 2010.

[26] Elkin, B. S.; Azeloglu, E. U.; et al. Mechanical heterogeneity of the rathippocampus measured by atomic force microscope indentation. J. Neuro-trauma 24:812–822; 2007.

[27] Erler, J. T.; Giaccia, A. J. Lysyl oxidase mediates hypoxic control of metastasis.Cancer Res. 66:10238–10241; 2006.

[28] Evans, J. M.; Donnelly, L. A.; et al. Metformin and reduced risk of cancer indiabetic patients. BMJ 330:1304–1305; 2005.

[29] Fang, M.; Shen, Z.; et al. The ER UDPase ENTPD5 promotes proteinN-glycosylation, the Warburg effect, and proliferation in the PTEN pathway.Cell 143:711–724; 2010.

[30] Flier, J. S.; Mueckler, M. M.; et al. Elevated levels of glucose transport andtransporter messenger RNA are induced by ras or src oncogenes. Science235:1492–1495; 1987.

[31] Forsberg, K.; Di Giovanni, S. Cross talk between cellular redox status,metabolism, and p53 in neural stem cell biology. Neuroscientist 20:326–342;2014.

[32] Fournier, M. V.; Fata, J. E.; et al. Interaction of E-cadherin and PTEN regulatesmorphogenesis and growth arrest in human mammary epithelial cells.Cancer Res. 69:4545–4552; 2009.

[33] Gallagher, F. A.; Kettunen, M. I.; et al. 13C MR spectroscopy measurementsof glutaminase activity in human hepatocellular carcinoma cells using hyperpo-larized 13C-labeled glutamine. Magn. Reson. Med. 60:253–257; 2008.

[34] Galletly, N. P.; McGinty, J.; et al. Fluorescence lifetime imaging distinguishesbasal cell carcinoma from surrounding uninvolved skin. Br. J. Dermatol.159:152–161; 2008.

[35] Galluzzi, L.; Kepp, O.; et al. Metabolic targets for cancer therapy. Nat. Rev.Drug Discovery 12:829–846; 2013.

[36] Gambhir, S. S.; Czernin, J.; et al. A tabulated summary of the FDG PETliterature. J. Nucl. Med. 42(5 Suppl.):1S–93S; 2001.

[37] Geiger, B.; Spatz, J. P.; et al. Environmental sensing through focal adhesions.Nat. Rev. Mol. Cell Biol. 10:21–33; 2009.

[38] Georges, P. C.; Hui, J. J.; et al. Increased stiffness of the rat liver precedesmatrix deposition: implications for fibrosis. Am. J. Physiol. Gastrointest. LiverPhysiol. 293:G1147–1154; 2007.

[39] Gershtein, E. S.; Scherbakov, A. M.; et al. Phosphatidylinositol 3-kinase/AKTsignalling pathway components in human breast cancer: clinicopathologicalcorrelations. Anticancer Res. 27:1777–1782; 2007.

[40] Gershtein, E. S.; Shatskaya, V. A.; et al. Expression of phosphatidylinositol-3kinase in lung cancer. Bull. Exp. Biol. Med. 130:1166–1168; 2000.

[41] Giancotti, F. G.; Ruoslahti, E. Integrin signaling. Science 285:1028–1032; 1999.[42] Giannoni, E.; Buricchi, F.; et al. Intracellular reactive oxygen species activate

Src tyrosine kinase during cell adhesion and anchorage-dependent cellgrowth. Mol. Cell. Biol. 25:6391–6403; 2005.

[43] Gonzalez-Angulo, A. M.; Ferrer-Lozano, J.; et al. PI3K pathway mutations andPTEN levels in primary and metastatic breast cancer. Mol. Cancer Ther10:1093–1101; 2011.

[44] Goo, C. K.; Lim, H. Y.; et al. PTEN/Akt signaling controls mitochondrialrespiratory capacity through 4E-BP1. PLoS One 7:e45806; 2012.

[45] Guan, J. L. Integrin signaling through FAK in the regulation of mammary stemcells and breast cancer. IUBMB Life 62:268–276; 2010.

[46] Gwinn, D. M.; Shackelford, D. B.; et al. AMPK phosphorylation of raptormediates a metabolic checkpoint. Mol. Cell 30:214–226; 2008.

[47] Hadad, S. M.; Dewar, J. A.; et al. Gene signature of metformin actions onprimary breast cancer within a window of opportunity randomized clinicaltrial. J. Clin. Oncol. 28(15 Suppl.):560; 2010.

[48] Hamill, O. P.; Martinac, B. Molecular basis of mechanotransduction in livingcells. Physiol. Rev. 81:685–740; 2001.

[49] Han, B.; Lin, S.; et al. Correlation of 18F-FDG PET activity with expressions ofsurvivin, Ki67, and CD34 in non-small-cell lung cancer. Nucl. Med. Commun.30:831–837; 2009.

[50] Hanahan, D.; Weinberg, R. A. Hallmarks of cancer: the next generation. Cell144:646–674; 2011.

[51] Hollander, M. C.; Blumenthal, G. M.; et al. PTEN loss in the continuum ofcommon cancers, rare syndromes and mouse models. Nat. Rev. Cancer11:289–301; 2011.

[52] Hosono, K.; Endo, H.; et al. Metformin suppresses colorectal aberrant crypt fociin a short-term clinical trial. Cancer Prev. Res. (Philadelphia) 3:1077–1083; 2010.

[53] Hu, W.; Feng, Z.; et al. The regulation of multiple p53 stress responses ismediated through MDM2. Genes Cancer 3:199–208; 2012.

[54] Ibarra-Drendall, C.; Dietze, E. C.; et al. Metabolic syndrome and breastcancer risk: is there a role for metformin? Curr. Breast Cancer Rep 3:142–150;2011.

[55] Ibarra-Drendall, C.; Troch, M. M.; et al. Pilot and feasibility study:prospective proteomic profiling of mammary epithelial cells from high-riskwomen provides evidence of activation of pro-survival pathways. BreastCancer Res. Treat. 132:487–498; 2012.

[56] Ilan, N.; Elkin, M.; et al. Regulation, function and clinical significance ofheparanase in cancer metastasis and angiogenesis. Int. J. Biochem. Cell Biol.38:2018–2039; 2006.

[57] Ingber, D. E. Tensegrity-based mechanosensing from macro to micro. Prog.Biophys. Mol. Biol. 97:163–179; 2008.

[58] Ingman, W. V.; Wyckoff, J.; et al. Macrophages promote collagen fibrillogen-esis around terminal end buds of the developing mammary gland. Dev. Dyn.235:3222–3229; 2006.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280278

[59] Isakovic, A.; Harhaji, L.; et al. Dual antiglioma action of metformin: cell cyclearrest and mitochondria-dependent apoptosis. Cell. Mol. Life Sci. 64:1290–1302;2007.

[60] Jadvar, H.; Alavi, A.; et al. 18F-FDG uptake in lung, breast, and colon cancers:molecular biology correlates and disease characterization. J. Nucl. Med.50:1820–1827; 2009.

[61] Janmey, P. A. The cytoskeleton and cell signaling: component localizationand mechanical coupling. Physiol. Rev. 78:763–781; 1998.

[62] Jiralerspong, S.; Giordano, S. H.; et al. The effects of metformin on pathologiccomplete response (pCR) rates in diabetic breast cancer (BC) patients receivingneoadjuvant systemic therapy (NST). J. Clin. Oncol. 26(15 Suppl.):528; 2008.

[63] Johnson, K. R.; Leight, J. L.; et al. Demystifying the effects of a three-dimensional microenvironment in tissue morphogenesis. Methods Cell Biol.83:547–583; 2007.

[64] Kaira, K.; Murakami, H.; et al. Biological correlation of 18F-FDG uptake onPET in pulmonary neuroendocrine tumors. Anticancer Res. 33:4219–4228;2013.

[65] Kallergi, G.; Agelaki, S.; et al. Activation of FAK/PI3K/Rac1 signaling controlsactin reorganization and inhibits cell motility in human cancer cells. Cell.Physiol. Biochem. 20:977–986; 2007.

[66] Kasznicki, J.; Sliwinska, A.; et al. Metformin in cancer prevention andtherapy. Ann. Transl. Med 2:57; 2014.

[67] Katsumi, A.; Orr, A. W.; et al. Integrins in mechanotransduction. J. Biol. Chem.279:12001–12004; 2004.

[68] Kauppila, S.; Stenback, F.; et al. Aberrant type I and type III collagen geneexpression in human breast cancer in vivo. J. Pathol. 186:262–268; 1998.

[69] Kessenbrock, K.; Plaks, V.; et al. Matrix metalloproteinases: regulators of thetumor microenvironment. Cell 141:52–67; 2010.

[70] Khaled, W.; Reichling, S.; et al. Ultrasonic strain imaging and reconstructiveelastography for biological tissue. Ultrasonics 44(Suppl. 1):e199–202; 2006.

[71] Kim, J. W.; Gao, P.; et al. Hypoxia-inducible factor 1 and dysregulated c-Myccooperatively induce vascular endothelial growth factor and metabolicswitches hexokinase 2 and pyruvate dehydrogenase kinase 1. Mol. Cell. Biol.27:7381–7393; 2007.

[72] Kim, M. J.; Kim, D. H.; et al. Expression of metabolism-related proteins intriple-negative breast cancer. Int. J. Clin. Exp. Pathol. 7:301–312; 2014.

[73] Kohn, A. D.; Summers, S. A.; et al. Expression of a constitutively active AktSer/Thr kinase in 3T3-L1 adipocytes stimulates glucose uptake and glucosetransporter 4 translocation. J. Biol. Chem. 271:31372–31378; 1996.

[74] Kokkinos, M. I.; Wafai, R.; et al. Vimentin and epithelial–mesenchymaltransition in human breast cancer—observations in vitro and in vivo. CellsTissues Organs 185:191–203; 2007.

[75] Kurhanewicz, J.; Vigneron, D. B.; et al. Analysis of cancer metabolism byimaging hyperpolarized nuclei: prospects for translation to clinical research.Neoplasia 13:81–97; 2011.

[76] Kwon, J.; Lee, S. R.; et al. Reversible oxidation and inactivation of the tumorsuppressor PTEN in cells stimulated with peptide growth factors. Proc. Natl.Acad. Sci. USA 101:16419–16424; 2004.

[77] Laplante, M.; Sabatini, D. M. mTOR signaling in growth control and disease.Cell 149:274–293; 2012.

[78] Le, Q. T.; Harris, J.; et al. Validation of lysyl oxidase as a prognostic markerfor metastasis and survival in head and neck squamous cell carcinoma:Radiation Therapy Oncology Group trial 90-03. J. Clin. Oncol. 27:4281–4286;2009.

[79] Lee, K.; Chen, Q. K.; et al. Matrix compliance regulates Rac1b localization,NADPH oxidase assembly, and epithelial–mesenchymal transition. Mol. Biol.Cell 23:4097–4108; 2012.

[80] Leevers, S. J.; Vanhaesebroeck, B.; et al. Signalling through phosphoinositide3-kinases: the lipids take centre stage. Curr. Opin. Cell Biol. 11:219–225; 1999.

[81] Lele, T. P.; Sero, J. E.; et al. Tools to study cell mechanics and mechan-otransduction. Methods Cell Biol. 83:443; 2007.

[82] Lelievre, S. A.; Weaver, V. M.; et al. Tissue phenotype depends on reciprocalinteractions between the extracellular matrix and the structural organiza-tion of the nucleus. Proc. Natl. Acad. Sci. USA 95:14711–14716; 1998.

[83] Leppert, J.; Krajewski, J.; et al. Multiphoton excitation of autofluorescencefor microscopy of glioma tissue. Neurosurgery 58:759–767; 2006. (discussion759-767).

[84] Levental, K. R.; Yu, H.; et al. Matrix crosslinking forces tumor progression byenhancing integrin signaling. Cell 139:891–906; 2009.

[85] Liu, B.; Fan, Z.; et al. Metformin induces unique biological and molecularresponses in triple negative breast cancer cells. Cell Cycle 8:2031–2040;2009.

[86] Liu, H.; Radisky, D. C.; et al. Polarity and proliferation are controlled bydistinct signaling pathways downstream of PI3-kinase in breast epithelialtumor cells. J. Cell Biol. 164:603–612; 2004.

[87] Lopez, J. I.; Kang, I.; et al. In situ force mapping of mammary glandtransformation. Integr. Biol. (Cambridge) 3:910–921; 2011.

[88] Lu, P.; Weaver, V. M.; et al. The extracellular matrix: a dynamic niche incancer progression. J. Cell Biol. 196:395–406; 2012.

[89] Maehama, T.; Dixon, J. E. The tumor suppressor, PTEN/MMAC1, depho-sphorylates the lipid second messenger, phosphatidylinositol 3,4,5-trispho-sphate. J. Biol. Chem. 273:13375–13378; 1998.

[90] Maller, O.; Hansen, K. C.; et al. Collagen architecture in pregnancy-inducedprotection from breast cancer. J. Cell Sci. 126:4108–4110; 2013.

[91] Manning, B. D.; Cantley, L. C. AKT/PKB signaling: navigating downstream. Cell129:1261–1274; 2007.

[92] Masin, M.; Vazquez, J.; et al. GLUT3 is induced during epithelial–mesench-ymal transition and promotes tumor cell proliferation in non-small cell lungcancer. Cancer Metab 2:11; 2014.

[93] Matoba, S.; Kang, J. G.; et al. p53 regulates mitochondrial respiration. Science312:1650–1653; 2006.

[94] Mbeunkui, F.; Johann Jr. D. J. Cancer and the tumor microenvironment: areview of an essential relationship. Cancer Chemother. Pharmacol. 63:571–582; 2009.

[95] McKnight, T. R.; Lamborn, K. R.; et al. Correlation of magnetic resonancespectroscopic and growth characteristics within Grades II and III gliomas.J. Neurosurg. 106:660–666; 2007.

[96] Mouw, J. K.; Yui, Y.; et al. Tissue mechanics modulate microRNA-dependentPTEN expression to regulate malignant progression. Nat. Med 20:360–367;2014.

[97] Nelson, C. M.; Bissell, M. J. Modeling dynamic reciprocity: engineering three-dimensional culture models of breast architecture, function, and neoplastictransformation. Semin. Cancer Biol. 15:342–352; 2005.

[98] Nelson, S. J.; Kurhanewicz, J.; et al. Metabolic imaging of patients withprostate cancer using hyperpolarized [1-13C]pyruvate. Sci. Transl. Med. 5;2013. (198ra108).

[99] O’Brien, L. E.; Jou, T. S.; et al. Rac1 orientates epithelial apical polaritythrough effects on basolateral laminin assembly. Nat. Cell Biol. 3:831–838;2001.

[100] Ocana, A.; Vera-Badillo, F.; et al. Activation of the PI3K/mTOR/AKT pathwayand survival in solid tumors: systematic review and meta-analysis. PLoS One9:e95219; 2014.

[101] Osthus, R. C.; Shim, H.; et al. Deregulation of glucose transporter 1 andglycolytic gene expression by c-Myc. J. Biol. Chem. 275:21797–21800; 2000.

[102] Ozdemir, B. C.; Pentcheva-Hoang, T.; et al. Depletion of carcinoma-associatedfibroblasts and fibrosis induces immunosuppression and accelerates pan-creas cancer with reduced survival. Cancer Cell 25:719–734; 2014.

[103] Pagan, R.; Martin, I.; et al. Vimentin filaments follow the preexistingcytokeratin network during epithelial–mesenchymal transition of culturedneonatal rat hepatocytes. Exp. Cell Res. 222:333–344; 1996.

[104] Papandreou, I.; Cairns, R. A.; et al. HIF-1 mediates adaptation to hypoxia byactively downregulating mitochondrial oxygen consumption. Cell Metab.3:187–197; 2006.

[105] Paszek, M. J.; Boettiger, D.; et al. Integrin clustering is driven by mechanicalresistance from the glycocalyx and the substrate. PLoS Comput. Biol. 5:e1000604; 2009.

[106] Paszek, M. J.; Zahir, N.; et al. Tensional homeostasis and the malignantphenotype. Cancer Cell 8:241–254; 2005.

[107] Patel, T.; Hruby, G.; et al. Clinical outcomes after radical prostatectomy indiabetic patients treated with metformin. Urology 76:1240–1244; 2010.

[108] Perepelyuk, M.; Terajima, M.; et al. Hepatic stellate cells and portalfibroblasts are the major cellular sources of collagens and lysyl oxidases innormal liver and early after injury. Am. J. Physiol. Gastrointest. Liver Physiol.304:G605–614; 2013.

[109] Phipps, A. I.; Chlebowski, R. T.; et al. Body size, physical activity, and risk oftriple-negative and estrogen receptor-positive breast cancer. Cancer Epide-miol. Biomarkers Prev 20:454–463; 2011.

[110] Pickup, M. W.; Laklai, H.; et al. Stromally derived lysyl oxidase promotesmetastasis of transforming growth factor-beta-deficient mouse mammarycarcinomas. Cancer Res. 73:5336–5346; 2013.

[111] Pilie, P. G.; Ibarra-Drendall, C.; et al. Protein microarray analysis ofmammary epithelial cells from obese and nonobese women at high risk forbreast cancer: feasibility data. Cancer Epidemiol. Biomarkers Prev 20:476–482;2011.

[112] Plodinec, M.; Loparic, M.; et al. The nanomechanical signature of breastcancer. Nat. Nanotechnol 7:757–765; 2012.

[113] Provenzano, P. P.; Eliceiri, K. W.; et al. Collagen reorganization at the tumor–stromal interface facilitates local invasion. BMC Med. 4:38; 2006.

[114] Provenzano, P. P.; Inman, D. R.; et al. Mammary epithelial-specific disruption offocal adhesion kinase retards tumor formation and metastasis in a transgenicmouse model of human breast cancer. Am. J. Pathol. 173:1551–1565; 2008.

[115] Provenzano, P. P.; Inman, D. R.; et al. Matrix density-induced mechanor-egulation of breast cell phenotype, signaling and gene expression through aFAK–ERK linkage. Oncogene 28:4326–4343; 2009.

[116] Provenzano, P. P.; Inman, D. R.; et al. Collagen density promotes mammarytumor initiation and progression. BMC Med. 6:11; 2008.

[117] Provenzano, P. P.; Keely, P. J. The role of focal adhesion kinase in tumorinitiation and progression. Cell Adhes. Migr 3:347–350; 2009.

[118] Quail, D. F.; Joyce, J. A. Microenvironmental regulation of tumor progressionand metastasis. Nat. Med 19:1423–1437; 2013.

[119] Rathmell, J. C.; Fox, C. J.; et al. Akt-directed glucose metabolism can preventBax conformation change and promote growth factor-independent survival.Mol. Cell. Biol. 23:7315–7328; 2003.

[120] Rose, D. P.; Haffner, S. M.; et al. Adiposity, the metabolic syndrome, andbreast cancer in African-American and white American women. Endocr. Rev.28:763–777; 2007.

[121] Samuel, M. S.; Lopez, J. I.; et al. Actomyosin-mediated cellular tension drivesincreased tissue stiffness and beta-catenin activation to induce epidermalhyperplasia and tumor growth. Cancer Cell 19:776–791; 2011.

[122] Santala, M.; Simojoki, M.; et al. Type I and III collagen metabolites aspredictors of clinical outcome in epithelial ovarian cancer. Clin. Cancer Res.5:4091–4096; 1999.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280 279

[123] Sautes-Fridman, C.; Cherfils-Vicini, J.; et al. Tumor microenvironment ismultifaceted. Cancer Metastasis Rev. 30:13–25; 2011.

[124] Schafer, Z. T.; Grassian, A. R.; et al. Antioxidant and oncogene rescue ofmetabolic defects caused by loss of matrix attachment. Nature 461:109–113;2009.

[125] Schedin, P.; Borges, V. Breaking down barriers: the importance of the stromalmicroenvironment in acquiring invasiveness in young women's breastcancer. Breast Cancer Res. 11:102; 2009.

[126] Schedin, P.; Keely, P. J. Mammary gland ECM remodeling, stiffness, andmechanosignaling in normal development and tumor progression. ColdSpring Harbor Perspect Biol 3:a003228; 2011.

[127] Selbekk, T.; Bang, J.; et al. Strain processing of intraoperative ultrasoundimages of brain tumours: initial results. Ultrasound Med. Biol. 31:45–51; 2005.

[128] Shackelford, D. B.; Shaw, R. J. The LKB1–AMPK pathway: metabolism andgrowth control in tumour suppression. Nat. Rev. Cancer 9:563–575; 2009.

[129] Shah, A. T.; Beckler, M. Demory; et al. Optical metabolic imaging oftreatment response in human head and neck squamous cell carcinoma. PLoSOne 9:e90746; 2014.

[130] Skala, M. C.; Riching, K. M.; et al. In vivo multiphoton microscopy of NADHand FAD redox states, fluorescence lifetimes, and cellular morphology inprecancerous epithelia. Proc. Natl. Acad. Sci. USA 104:19494–19499; 2007.

[131] Solaini, G.; Sgarbi, G.; et al. Oxidative phosphorylation in cancer cells.Biochim. Biophys. Acta 1807:534–542; 2011.

[132] Spano, D.; Zollo, M. Tumor microenvironment: a main actor in the metastasisprocess. Clin. Exp. Metastasis 29:381–395; 2012.

[133] Stambolic, V.; MacPherson, D.; et al. Regulation of PTEN transcription by p53.Mol. Cell 8:317–325; 2001.

[134] Sung, S. Y.; Hsieh, C. L.; et al. Tumor microenvironment promotes cancerprogression, metastasis, and therapeutic resistance. Curr. Probl. Cancer31:36–100; 2007.

[135] Szatrowski, T. P.; Nathan, C. F. Production of large amounts of hydrogenperoxide by human tumor cells. Cancer Res. 51:794–798; 1991.

[136] Takahashi, A.; Ohtani, N.; et al. Mitogenic signalling and the p16INK4a–Rbpathway cooperate to enforce irreversible cellular senescence. Nat. Cell Biol.8:1291–1297; 2006.

[137] Thomas, S. R.; Witting, P. K.; et al. Redox control of endothelial function anddysfunction: molecular mechanisms and therapeutic opportunities. Antioxid.Redox Signaling 10:1713–1765; 2008.

[138] Tilghman, R. W.; Cowan, C. R.; et al. Matrix rigidity regulates cancer cellgrowth and cellular phenotype. PLoS One 5:e12905; 2010.

[139] Toyokuni, S.; Okamoto, K.; et al. Persistent oxidative stress in cancer. FEBSLett. 358:1–3; 1995.

[140] Tzima, E.; Del Pozo, M. A.; et al. Activation of Rac1 by shear stress inendothelial cells mediates both cytoskeletal reorganization and effects ongene expression. EMBO J. 21:6791–6800; 2002.

[141] Unsgaard, G.; Rygh, O. M.; et al. Intra-operative 3D ultrasound in neurosur-gery. Acta Neurochir. (Wien) 148:235–253; 2006. (discussion 253).

[142] Vaupel, P. Prognostic potential of the pre-therapeutic tumor oxygenationstatus. Adv. Exp. Med. Biol. 645:241–246; 2009.

[143] Vaupel, P.; Mayer, A.; et al. Tumor hypoxia and malignant progression.Methods Enzymol. 381:335–354; 2004.

[144] Villette, S.; Pigaglio-Deshayes, S.; et al. Ultraviolet-induced autofluorescencecharacterization of normal and tumoral esophageal epithelium cells withquantitation of NAD(P)H. Photochem. Photobiol. Sci. 5:483–492; 2006.

[145] Wahdan-Alaswad, R.; Fan, Z.; et al. Glucose promotes breast canceraggression and reduces metformin efficacy. Cell Cycle 12:3759–3769; 2013.

[146] Walsh, A. J.; Cook, R. S.; et al. Optical metabolic imaging identifies glycolyticlevels, subtypes, and early-treatment response in breast cancer. Cancer Res.73:6164–6174; 2013.

[147] Wang, H. B.; Dembo, M.; et al. Substrate flexibility regulates growth andapoptosis of normal but not transformed cells. Am. J. Physiol. Cell Physiol 279:C1345–C1350; 2000.

[148] Warburg, O. On the origin of cancer cells. Science 123:309–314; 1956.[149] Warburg, O.; Wind, F.; et al. The metabolism of tumors in the body. J. Gen.

Physiol. 8:519–530; 1927.[150] Willipinski-Stapelfeldt, B.; Riethdorf, S.; et al. Changes in cytoskeletal

protein composition indicative of an epithelial–mesenchymal transition inhuman micrometastatic and primary breast carcinoma cells. Clin. Cancer Res.11:8006–8014; 2005.

[151] Wozniak, M. A.; Desai, R.; et al. ROCK-generated contractility regulates breastepithelial cell differentiation in response to the physical properties of a three-dimensional collagen matrix. J. Cell Biol. 163:583–595; 2003.

[152] Wyckoff, J. B.; Wang, Y.; et al. Direct visualization of macrophage-assistedtumor cell intravasation in mammary tumors. Cancer Res. 67:2649–2656;2007.

[153] Xia, H.; Nho, R. S.; et al. Focal adhesion kinase is upstream of phosphatidy-linositol 3-kinase/Akt in regulating fibroblast survival in response to con-traction of type I collagen matrices via a beta 1 integrin viability signalingpathway. J. Biol. Chem. 279:33024–33034; 2004.

[154] Xiao, H.; Ma, X.; et al. Metformin attenuates cardiac fibrosis by inhibiting theTGFbeta1–Smad3 signalling pathway. Cardiovasc. Res. 87:504–513; 2010.

[155] Zhu, G. G.; Risteli, L.; et al. Immunohistochemical study of type I collagenand type I pN-collagen in benign and malignant ovarian neoplasms. Cancer75:1010–1017; 1995.

[156] Zipfel, W. R.; Williams, R. M.; et al. Live tissue intrinsic emission microscopyusing multiphoton-excited native fluorescence and second harmonic gen-eration. Proc. Natl. Acad. Sci. USA 100:7075–7080; 2003.

J.C. Tung et al. / Free Radical Biology and Medicine 79 (2015) 269–280280