R

Cfif

PMD

a

ARRAA

KCBPLIG

1

atammtr

crs

cbt

t0

h1

Seminars in Cancer Biology 35 (2015) 96–106

Contents lists available at ScienceDirect

Seminars in Cancer Biology

jo ur nal ho me pag e: www.elsev ier .com/ locate /semcancer

eview

omplexity in the tumour microenvironment: Cancer associatedbroblast gene expression patterns identify both common and unique

eatures of tumour-stroma crosstalk across cancer types

aolo Gandellini1, Francesca Andriani1, Giuseppe Merlino, Francesca D’Aiuto, Luca Roz,aurizio Callari ∗

epartment of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy

r t i c l e i n f o

rticle history:eceived 28 June 2015eceived in revised form 17 August 2015ccepted 21 August 2015vailable online 28 August 2015

eywords:

a b s t r a c t

Cancer is a complex disease, driven by the accumulation of several somatic aberrations but fostered bya two-way interaction between tumour cells and the surrounding microenvironment. Cancer associ-ated fibroblasts (CAFs) represent one of the major players in tumour-stroma crosstalk. Recent in vitroand in vivo studies, often conducted by employing high throughput approaches, have started unravel-ling the key pathways involved in their functional effects. This review focus on open challenges in thestudy of CAF properties and function, highlighting at the same time the existence of common mecha-

ancer associated fibroblasts (CAF)reast cancerrostate cancerung cancernter-diseaseene expression

nisms as well as peculiarities in different cancer types (breast, prostate and lung cancer). Although stilllimited by current experimental models, which are unable to deal with the full level of complexity ofthe tumour microenvironment, a better understanding of these mechanisms may enable the identifica-tion of new biomarkers and therapeutic targets, to improve current strategies for cancer diagnosis andtreatment.

© 2015 Elsevier Ltd. All rights reserved.

. Introduction

Tumorigenesis has been classically considered as a cellutonomous mechanism triggered by the accumulation of muta-ions able to confer a growth advantage to transformed cellsnd the capability to invade surrounding tissues and eventuallyetastasize. However, cancer cells are integrated in a complexicroenvironment and the importance of stromal cells in con-

ributing to tumour initiation and progression is increasinglyecognized [1–5].

Tumour microenvironment (TME) is an ensemble of different

ellular and structural factors including vasculature and immune-elated cells, fibroblasts and the extracellular matrix (ECM),howing either tumour promoting or anti-tumour activity in an

Abbreviations: CAF, cancer associated fibroblast; NF, normal fibroblasts; AF, adja-ent fibroblasts; ECM, extracellular matrix; TME, tumour microenvironment; BC,reast cancer; PC, prostate cancer; LC, lung cancer; EMT, epithelial-mesenchymalransition.∗ Corresponding author. Present address: Cancer Research UK Cambridge Insti-

ute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge CB2RE, United Kingdom. Tel.: +44 1223 769531; fax: +44 1223 769510.

E-mail address: maurizio.callari@cruk.cam.ac.uk (M. Callari).1 Equal contribution.

ttp://dx.doi.org/10.1016/j.semcancer.2015.08.008044-579X/© 2015 Elsevier Ltd. All rights reserved.

intricate network of signals hard to dissect and to characterizeexperimentally. However, its relevance in several aspects of tumourprogression, including response to pharmacological treatments andits potential role as a direct therapeutic target [6,7] has moti-vated growing research efforts to investigate the key molecularmechanisms involved in tumour-stroma crosstalk. Such crosstalkis mediated either by cell–cell interactions or soluble factors, butmore recently other key players have been identified, as secretedmiRNAs, metabolites and exosomes [8–12].

In this review, we mainly focus on one of the major players inthe TME, usually referred to as cancer associated fibroblasts (CAFs).Fibroblasts, beside their involvement in development, tissue repairand inflammatory response, have been shown to participate inhuman tumorigenesis by providing a permissive environment forproliferation and survival of epithelial cells, and by remodellingECM to promote tumour growth and invasiveness [13–16]. Ris-ing evidence however also informs on phenotypic and functionalheterogeneity within the stromal compartment of the TME [17],indicating that, even for a single cell type such as fibroblast, differ-ent subpopulations might exist and exert distinct functions [18].

Intriguingly, recent studies support the notion that cells other thannormal fibroblasts are a possible source of CAFs, redefining thismicroenvironment component more like a cell state rather than acell type. Among them are mesenchymal stem cells, smooth muscle

n Cancer Biology 35 (2015) 96–106 97

ce

ecctdck(rmtw

rcreca

2

obouhwgtniSdn(oit

2

ciocCtnpnr

d[

m[mF

Box 1: Challenges in defining and studying cancer asso-ciated fibroblasts and their interplay with tumour cells

• Multiple players in the TME: An increasing bulk of liter-ature demonstrates the relevance of tumour-CAF crosstalk,however many other cell types (e.g. adaptive and innateimmune cells or endothelial cells) populate primary ormetastatic sites contributing to a complex network of hetero-typic interactions resulting in a global tumour-promoting orinhibiting effect. Such a complexity is still hard to reproduceor dissect with current available models and techniques.

• Limits of available models: Although studies on co-culture models are generating significant advances in theunderstanding of the biology of cancer, thus providinga controllable technique to study tumour-stroma interac-tions, models with increasing complexity will be needed toimprove our knowledge on heterotypic interactions and clar-ify the role of other important factors such as ECM stiffnessand mechanical forces and the relevance of the 3D architec-ture of the tissue [127].

• Identification of CAFs: Detecting fibroblasts having aso-called ‘activated’ phenotype able to promote cancer pro-gression is still a challenge. Although several markers havebeen proposed and some of them largely used across differ-ent cancer types (e.g. �-SMA and FAP), none of them seemsto be specific enough and further research is needed to iden-tify new markers or the best combination of them [19].

• Quantity vs quality: Gene expression profiling has beenextensively used to identify or validate CAF-derived signa-tures. However it is challenging to distinguish variationsin expression levels caused by the presence of specificcell phenotypes from variations ascribable to variability oftumour sampling and therefore mainly reflecting a higher orlower amount of stroma. Unfortunately, although potentiallyuseful datasets are nowadays available, only a minority hasinformation on tumour cellularity, which might help duringdata analysis and interpretation.

P. Gandellini et al. / Seminars i

ells, endothelial cells and cancer cells themselves after undergoingpithelial-mesenchymal transition (EMT) [19,20].

Complexity of interactions within the microenvironment is alsoxemplified by recent data centred on selective ablation of spe-ific stromal elements in pancreatic cancer. Selective targeting ofells expressing the fibroblast activation protein (FAP) with FAP-argeted chimeric antigen receptor T cells was able to reduceesmoplasia and impair growth of murine models of pancreaticancer [21]. However, strategies targeting cells expressing anotherey protein involved in CAF function, �-smooth muscle actin�-SMA) [22], or reducing sonic hedgehog signalling [23] (bothesulting in reduction of stromal elements and fibrosis) resulted inore aggressive tumours. These novel findings further strengthen

he notion that a number of different aspects should be evaluatedhen investigating tumour-stroma crosstalk.

The role of CAFs in tumour progression, the regulation ofesponse to therapies and the prognostic relevance of markers asso-iated with their activity in different tumours have been recentlyeviewed [17,24–26]. This review will largely concentrate on thevaluation of the complexity of the interaction between CAFs andancer cells across different cancer types to highlight similaritiesnd specificities of this crucial tumour promoting crosstalk.

. The CAF-tumour interplay in different cancers types

In several different solid cancers, CAFs have been shown toften display properties associated with the activated myofibro-lasts found at sites of wound healing [14,27] Accurate definitionf markers of CAF activation and functional dissection of the molec-lar bases of their pro-tumorigenic properties are however stillampered by several experimental challenges (Box 1). Nonetheless,hat we have learned so far is detailed in the following para-

raphs, principally for breast, prostate and lung cancer. For theseumour types, an extensive characterization of the CAF compo-ent exists in the literature and they were also the most studied

n our laboratories, with several in-house datasets available (seeection 3). Furthermore, they could be seen as representative ofifferent major etiological factors (or drivers) in their predomi-ant histologies: i.e. heavy exposure to environmental carcinogenslung), hormone dependency (breast and prostate) and possiblyrgan senescence (prostate). Investigation of stromal componentsn these three entities may therefore offer the opportunity to inves-igate common versus unique features in different tumour types.

.1. Breast cancer

In the mammary gland, fibroblasts are one of the most importantomponents of the connective tissue that contribute to structuralntegrity. Similarly to the wound healing process, with the onsetf breast cancer (BC), fibroblasts adjacent to tumour cells (the soalled CAFs) become activated and acquire myofibroblast features.AFs have indeed a drastically different phenotype with respecto their normal counterpart, in terms of morphology, immunophe-otype, proliferation rate, released cytokines, deposition of ECMroteins and gene expression profiles [28–30]. Such acquired phe-otype makes them able to sustain tumour progression and affectesponse to treatment [14,31].

BC CAFs represent a very heterogeneous cell population, andifferent hypotheses have been formulated to explain their origin32–34], with no evidence on a main dominant derivation.

One of the open challenges is the identification of reliable

arkers able to identify the CAF population. In addition to �-SMA

35], the characteristic fingerprint of myofibroblasts, several CAFarkers have been proposed, such as Platelet-Derived Growth

actor Receptor-� (PDGFR-�) and FAP, a type II integral membrane

protein with peptidase, gelatinase and collagenase activity, highlyexpressed in reactive stroma of solid tumour and wounded tissue[36,37]. However, the combination of different markers seemsto be advantageous compared to any single marker for theiridentification. High expression of PDGFR-� was associated withshorter recurrence and survival in a cohort of BC patients [38].On the contrary, FAP was significantly associated with longerdisease-free and overall survival [39].

In keeping with the prominent changes affecting the ECM dur-ing carcinogenesis, other proposed CAF markers are ECM-relatedmolecules as Tenascin-C (TNC) and different matrix metallopro-teinases (MMPs), the latter responsible for the degradation of thebasement membrane during tumour invasion. Several other CAFmarkers have been proposed, such as neural/glial antigen 2 (NG2),podoplanin (PDPN) and fibroblast specific protein (FSP), but theirspecificity for CAFs is questionable. Such a lack of consensus mightbe a consequence of intrinsic phenotype heterogeneity of CAFs.

The mechanisms of fibroblasts activation in BC are still elusive.The major tumour cell-derived factors that have been reported toactivate stromal fibroblasts are TGF� and CXCL12/SDF1. Kojimaet al. showed an elegant mechanism by which TGF�, released bycarcinoma cells [40], enhances endogenous TGF� and SDF-1 pro-duction via T�R-Smad signalling and induces CXCR4 expressionin stromal fibroblasts, generating two autocrine signalling loopswhich cross-communicate to maintain the myofibroblastic phe-

notype [32]. Other pro-fibrotic factors released by breast tumourcells, such as PDGF-�/� [41], or IL6 [42] can also act on residentfibroblasts and induce their activation.

9 in Can

atvpfts

Ctamcbwefigdafecd

mfTcrc

ieTcttstis

lmettmmuhm

tpmmficgcl

c

8 P. Gandellini et al. / Seminars

Beside paracrine mechanisms, also epigenetic alterations, suchs the down-regulation of TP53, CDKN1A (p21), PTEN and CAV1umour suppressor genes reviewed in [43], are involved in acti-ation of CAFs. Especially, in BC, CAFs have a lower level of CAV1rotein compared to normal fibroblasts (NFs) [44,45]. The trans-ormation of NIH 3T3 fibroblast cells by various oncogenes leads tohe reduction of CAV1, which correlates very well with the biggerize of colonies formed by these transformed cells [46].

In a well-annotated cohort of 160 BC patients, a loss of stromalAV1 in CAFs was associated with a dramatically increased risk ofumour recurrence, lymph node metastasis, tamoxifen-resistance,nd poor clinical outcome [47]. Bonuccelli et al. tried to explainechanistically how the loss of CAV1 in fibroblasts, mediates these

linical effects, introducing the new concept named “Reverse War-urg effect”. They showed that CAV1-deficient stromal fibroblastsere sufficient to promote both tumour growth and angiogen-

sis. Furthermore, proteomic analysis of CAV1-deficient stromalbroblasts indicated that these cells up-regulate the expression oflycolytic enzymes, shifting towards aerobic glycolysis. Thus, CAV1-eficient stromal fibroblasts may contribute to tumour growth andngiogenesis by providing energy-rich metabolites in a paracrineashion, feeding tumour cells with pyruvate and lactate, which cannter into TCA cycle (oxidative phosphorylation) in the epithelialancer cells [48]. Interestingly a similar mechanism has also beenescribed in prostate cancer (see below).

CAFs regulate breast cancer stem cells with a paracrineechanism, promoting their self-renewal or reprogramming dif-

erentiated tumour cells to a stem cell-like phenotype [49]. Indeed,suyada et al. showed that co-culturing CAF with breast cancer cellsauses an increase of sphere formation respect to NF, due to theelease of CCL2, which induces NOTCH1 activation in breast cancerells [50].

Since BC includes different molecular subtypes, Park et al.nvestigated their relationship with CAF phenotypes [51]. They cat-gorized 642 tumour samples into luminal A, luminal B, HER2 orriple Negative Breast Cancer (TNBC), whereas tumour stroma waslassified as desmoplastic, sclerotic, normal-like, or inflammatoryype. They showed that the expression of CAF-related proteins inhe stroma depends on BC molecular subtypes and marker expres-ion also varied across the four stroma classes, overall supportinghe existence of heterogeneous CAF subpopulations and suggest-ng that a crosstalk between stromal component and BC cells isubtype-specific.

Accordingly, we found that the paracrine stimulation of BC cellsines belonging to different molecular subtypes with conditioned

edium derived from CAFs and NFs produced subtype-specificffects. Indeed, luminal and HER2+ subtypes, but not TNBC sub-ypes, sensed the CAF interaction by drastically changing theirranscriptome and secretome and by increasing their proliferation,

igration, invasion, compared with NF stimulation (Merlino et al.,anuscript in preparation). This also indicates that secreted sol-

ble factors are important in driving at least some aspects of theeterotypic interaction between epithelial and stromal compart-ent.Allinen et al. tried to delineate epithelial-stroma interaction at

he molecular level by comparing gene expression and genomicrofiles of epithelial, myofibroblasts and stromal cells from nor-al tissue, in situ and invasive carcinoma. They showed that theost dramatic changes occurred in myoepithelial cells and myo-

broblasts during the transition from normal breast tissue to in situarcinoma stage and that the majority of differentially expressedenes encoded for secreted and cell surface proteins, such as the

hemokines CXCL12 and CXCL14, stimulate in a paracrine way pro-iferation, invasion and migration of tumour cells [29].

Bauer et al. isolated CAFs and NFs from six primary human breastarcinoma specimens and matched normal. Using gene expression

cer Biology 35 (2015) 96–106

profiling, an up-regulation of WISP1, collagen type-X (COL10A) andTGF� isoforms in CAFs was found [52]. Based on the work of Pen-nica et al. [53] and Kim et al. [54], they speculated that WISP1expression in CAF could activate paracrine Wnt-1 signalling inhuman BC.

Another study profiling primary cultured CAFs and NFs derivedfrom early stages of BC tissues, showed that CAFs induce an increaseof proliferation and invasion of MDA-MB-231 with respect to NF aswell as displayed abnormal signal pathways involved in cell cycle,cell adhesion and secreted factors [30]. Overall, several evidencessupport that BC-associated stroma is characterized by increasedexpression of cytokines (EGF, HGF, PDGF, TLL-12, SSP-1, CXCL12,CXCL14), extracellular matrix molecules (FBN1, FB2M, SPARC, ADM,POSTN) and proteases (MMP1, MMP2, MMP13) [29,55–58]. Most ofchanges in these genes seem have been observed in early stages (i.e.stroma from in situ BC compared with normal stroma) and not inthe transition between in situ and invasive carcinoma, supportinga major role of CAFs during BC carcinogenesis [29,58].

2.2. Prostate cancer

Prostate stroma consists of smooth muscle cells (SMC), fibro-blasts, endothelial cells, nerves and immune cells [59]. SMCs(which express �-SMA, calponin, desmin, myosin) and fibroblasts(which express vimentin and laminin) share the primary func-tion of synthesizing structural and regulatory components of theECM. Both cell types can reversibly regulate their proliferation rateand phenotype to maintain tissue homeostasis [59]. For example,during wound repair, fibroblasts and SMCs convert to a commonmyofibroblast phenotype, characterized by �-SMA and vimentinexpression, formation of contractile filaments, well developedrough endoplasmic reticulum and Golgi apparatus, cell attach-ments to the, gap junctions and up-regulated synthesis of ECMand ECM remodelling proteases [59]. A reactive stroma exhibit-ing alterations typical of wound healing has been also observedin precancerous lesions affecting prostate gland (e.g. prostaticintraepithelial neoplasia, PIN) and definitely in prostate cancer(PC). While normal prostate stroma is mainly composed of SMCs,tumour-associated stroma is enriched with myofibroblasts andfibroblasts [60]. Prostate CAFs are indeed characterized by co-expression of vimentin (mesenchymal cell intermediate filament)and �-SMA, without expression of Calponin, a later stage markerof smooth muscle differentiation. Interestingly, the percentage ofmyofibroblasts is further increased in the stroma of poorly differen-tiated Gleason 4 cancers compared to Gleason 3 tumours [60]. Thepresence of myofibroblasts in reactive stroma indicates that theremay be increased production of ECM components and overall ECMremodelling during PC progression. Consistently, CAFs express highlevels of MMP2 and MMP9 and reduced levels of their inhibitors,such as TIMP1 and TIMP2. Prostate CAFs are also characterized byenhanced expression of TNC, fibronectin and FAP.

By secretion of soluble signals, including cytokines and growthfactors, and by remodelling of ECM, CAFs has been shown to influ-ence each step of PC development, growth and metastasis [61].Indirect evidence of this is the finding by Ayala et al. [62], who,by analysing tissue microarrays from 847 PC patients, showed thatboth volume of reactive stroma in the tumour as well as �-SMAexpression were significant predictors of poor disease-free sur-vival. Similarly, Jia et al. [63] found that numerous changes in geneexpression occurring in tumour adjacent stroma at the time of diag-nosis correlate with the chance of relapse following prostatectomy.

A direct evidence of the active role of CAFs in prostate tumori-

genesis was elegantly provided by Olumi et al. in their seminalarticle dating 1999. The authors showed that CAFs, but not nor-mal fibroblasts distal to the tumour, stimulated cancer growthof initiated/immortalized non-tumorigenic epithelial cells [64].

n Can

Rccgfi

luaeetatbsC[

tEncttnoatora

cgoasf

uoolpcbdfnp

otttgpccrcdivb

P. Gandellini et al. / Seminars i

esponsible for the proliferative spur exerted by CAFs on tumourells is certainly the secretion of growth factors such as hepato-yte growth factors (HGF), insulin-like growth factors (IGF), nerverowth factors (NGF), Wnt1, epidermal growth factor (EGF) andbroblasts growth factors 2 (FGF2) [61].

In addition, CAFs have been shown to support cancer cell pro-iferation through a mutual metabolic reprogramming. In fact, aspon the contact with PC cells, CAFs undergo Warburg metabolismnd mitochondrial oxidative stress, thus resulting in increasedxpression of glucose transporter GLUT1, lactate production, andxtrusion of lactate by de novo expressed monocarboxylateransporter-4 [65]. In turn, tumour cells reprogram toward anerobic metabolism and increase lactate upload via the lactateransporter monocarboxylate transporter-1. Gradually, PC cellsecome independent of glucose consumption, becoming able tourvive in low glucose environments, while developing addiction toAF-secreted lactate as fuel for anabolic pathways and cell growth65].

Besides stimulating cancer cell growth, CAFs have been showno substantially contribute to PC progression by both remodellingCM and by influencing tumour cell plasticity. In this regard, Gian-oni et al. [66] have shown that CAFs, through secretion of MMPs,an elicit EMT in PC cells. Indeed, co-injection with CAFs was showno enhance the local growth of subcutaneous PC cell xenografts andhe development of spontaneous metastases in mice [66]. Mecha-istically, CAFs induce a Rac1b/cycloxygenase-2-mediated releasef reactive oxygen species in carcinoma cells, which activates NF-�Bnd HIF1 ̨ [67]. We recently showed that HIF1 ̨ directly represseshe transcription of miR-205 [68], thus relieving the brake imposedn the miRNA target genes, including E-cadherin transcriptionalepressors [69], and ultimately leading to establishment of EMTnd enhancement of metastasis.

An intimate link between the EMT and the gain of epithelial stemell properties has been reported [70]. Accordingly, PC cells under-oing CAF-induced EMT are characterized by increased expressionf stem cell markers and enhanced ability to form prostaspheresnd to self-renew [66]. Prostate CAFs have been shown to boostpheroid formation also from bona fide cancer stem cells isolatedrom prostate tumours based on stemness markers [71].

It is noteworthy that the action of stroma on tumour cells is notnidirectional, rather malignancy is driven by the establishmentf a vicious tumour-stroma interplay. Though the possible cells ofrigin of CAFs may be diverse (including resident SMCs, endothe-ial cells, pericytes or circulating bone marrow-derived fibroblastrogenitors or mesenchymal stem cells) there is evidence that PCells may themselves activate surrounding fibroblasts to myofibro-lasts. Due to its pivotal role in wound repair and fibroproliferativeiseases, TGF� has been long considered the major cancer-derivedactor able to sustain tumour reactive stroma [60]. However, Gian-oni et al. [66] recently proposed a role for IL6 in the activation ofrostate CAFs.

Analysis of the transcriptome of microdissected tumour stromar isolated fibroblasts has been pursued in the last years to clarifyhe phenotype of CAFs, the pathways relevant for their reac-ive behaviour and eventually identify the factor responsible forheir activation [72–74]. In this regard, by comparatively analysingene expression profiles of CAFs and NFs obtained from radicalrostatectomies, we found that tumour reactive fibroblasts areharacterized by positive enrichment of genes that regulate actinytoskeleton remodelling and muscle contractility and gene setselated to glycolysis/gluconeogenesis pathway, in accord with theirharacteristic functional features [75]. To clarify which tumour-

erived stimuli are able to induce the reactive phenotype seen

n vivo, gene expression analysis was extended to fibroblasts acti-ated in vitro with TGF� or IL6. Globally, a major overlap was foundetween transcriptomes of IL6-activated NFs and CAFs compared to

cer Biology 35 (2015) 96–106 99

TGF�-NFs, though both in vitro activated fibroblasts showed phe-notypic similarities with CAFs.

Accumulating evidence points towards a role for changes occur-ring with increasing age in prostatic stroma as promoters ofprostate epithelial cell growth [76]. Notably, advancing age is asso-ciated with substantial increases in the incidence rates of commondiseases affecting the prostate gland including benign prostatichyperplasia and PC [77]. Bavik et al. demonstrated that senes-cent prostate fibroblasts stimulate the growth of pre-neoplasticand neoplastic prostate epithelial cells mainly through the secre-tion of paracrine factors [78]. In keeping with these data, in vivostudies demonstrated that regions adjacent to cancer epitheliumare enriched in senescent stromal cells [79]. Moreover, the exist-ence of a “trait d’union” between cell senescence and fibroblastactivation has been reported. Indeed, various senescence-inducingstimuli, even unrelated to replicative senescence (i.e. oxidativestress), have been described to induce myofibroblast differentia-tion [80,81]. miR-210, up-regulated along replicative senescence,is able per se to induce senescence features (p16, p21, �-H2AX foci)in pre-senescent prostatic fibroblasts, and concomitantly convertthem into CAF-like cells [82]. Actually, miR-210 overexpressingfibroblasts are characterized by increased �-SMA and collagen typeI expression and the capability to promote cancer cell EMT, to sup-port angiogenesis and to recruit endothelial precursor cells andmonocytes/macrophages [82]. The other way around, even patient-derived CAFs show features of senescence, such as �-H2AX foci and�-galactosidase staining, thus confirming the intimate relationshipbetween senescence and stromal activation [82]. In this regard, IL6is one of the cytokines secreted by senescent cells in the contextof the so-called senescence associated secretory phenotype (SASP)and a role in the same context has been recently reported for IL1�[83]. Potentially, either senescent epithelial cells or fibroblasts mayrelease IL6 in the TME, thus contributing to establish or maintain asenescent and reactive stroma able to support PC progression.

2.3. Lung cancer

The TME of human non-small cell lung cancer (NSCLC) is com-posed of different subpopulations of stromal cells and abundantextracellular matrix (ECM), mainly produced and remodelled byfibroblasts [84]. As a model to study contribution of fibroblasts tolung tumorigenesis we and other groups [85,86] have establishedshort-term cultures of fibroblasts from primary NSCLC speci-mens, obtaining cancer-associated fibroblasts (CAF, from cancertissue), adjacent fibroblasts (AF, derived from lung tissue adjacentto the tumour with challenging histological characteristics) andfibroblasts from histologically normal lung tissue (NFs) [85–87].Immunophenotypical characterization of these cells revealed het-erogeneous populations expressing different markers, such as FSP,�-SMA, vimentin or FAP, identifying a phenotype consistent withthat of activated myofibroblasts. Interestingly, the expression ofsuch markers was also evident in fibroblasts derived from normaltissue, indicating the presence of an early stroma activation in LC[87]. Using in vitro co-cultures of fibroblasts and NSCLC cells andselective profiling the purified fibroblastic fraction, Fromigue et al.identified a critical role of fibroblasts in the expression of genesrelated to growth and survival of cancer cells (i.e. IGF-BP, regula-tors of TGF� family), in proteins that can affect changes in the statusof cell proliferation, differentiation and migration (i.e. kinases PLK,PRKR, MLKC) or factors that regulate angiogenesis (ˇFGF, IL8, VEGF)[88]. The secretion by lung fibroblasts of soluble factors such ascytokines and chemokines that directly stimulate tumour growth

has been also reported. A recent cross-species functional analysisof CAFs described a novel CAF gene signature with significant rele-vance in human NSCLCs, showing that CAF-derived CLCF1 and IL6exert specific paracrine effects to promote tumour growth in vivo

1 in Can

atpou[

kesscbherlt�tpcfiassbeotmbup

etftaLtaatinr[top

iuarwstcspnaf

00 P. Gandellini et al. / Seminars

nd may be particularly important components of the CAF secre-ome [89]. Moreover, CXCL-12/CXCR4 axis [90,91] and IGF-II/IGFR1aracrine signals [92] have been shown to mediate crucial aspectsf the interaction between CAFs and NSCLC, including the mod-lation of the stemness phenotype and dissemination properties91].

Heterogeneity within the fibroblast population appears to be aey issue from a phenotypic and functional perspective [24]. Sev-ral studies based on genome-wide expression profiling assessedignificant differences among fibroblasts originated from differentites as well as within fibroblasts separated from the same anatomi-al site [93], which are not composed of a homogeneous population,ut rather consist of subsets of different fibroblasts defining theireterogeneity [94]. In particular, Pechkovsky et al. described genexpression profiles of fibroblasts isolated from different anatomicalegions of human lung, normal human proximal bronchi and distalung parenchyma. Fibroblasts originated at lung parenchyma spon-aneously developed a myofibroblasts phenotype with increased-SMA expression and responsible for ECM proteins deposition

hat could be critical for development and progression of idiopathiculmonary fibrosis and other fibro-proliferative diseases as well asancer [95]. These studies point toward the existence of site speci-city of signals and extracellular proteins expressed by fibroblastsnd boost the interesting possibility that these signals could repre-ent “home address” identified by neoplastic cells to support theirurvival, proliferation and migration. Additional pattern of fibro-lasts heterogeneity have been identified through the study of genexpression profiling of fibroblasts population in an in vivo modelf bleomycin induced lung fibrosis [96]. In this study, osteopon-in (OPN) was highly overexpressed and represented an activation

arker of lung fibroblasts during fibrosis. Interestingly, OPN hadeen previously described as prominently increased in fibroblastsndergoing senescence and as a crucial mediator of the tumourromoting effects of senescent stroma [97].

Beside their well-established effect as functional essential play-rs in the carcinogenesis process, an increasing interest to evaluatehe prognostic role of stromal cells within the tumour has sur-aced in the last few years. Accumulating evidence also showedhat proteins expressed by CAFs, such as MMP2 [98], �-SMA [99]nd carbonic anydrase IX [100] may correlate with prognosis ofC. The potential prognostic role of fibroblasts has been also bet-er clarified in an interesting study by Ito et al., which considered

large series of patients with stage I lung adenocarcinoma. Thenalysis of 329 tumours by immunohistochemistry revealed thathe presence of PDPN-positive CAFs was the most powerful andndependent recurrence predictor in patients with stage I lung ade-ocarcinoma and may be useful for identifying patients with a highisk of recurrence who might benefit from adjuvant chemotherapy101]. A more recent study suggested that PDPN-positive CAFs athe invasive front of the tumour might promote the local invasionf cancer cells through a RhoA-mediated mechanism that could berevented by ROCK inhibition [102].

Gene-expression based strategies have been also implementedn the last few years to define a “stroma signature” and betternderstand the mechanism underlying the tumour-stroma inter-ctions to identify novel stroma-related biomarkers of prognosticelevance, suitable as therapeutic targets. Interestingly, an ‘activeound’ related expression program obtained comparing serum-

timulated and serum-starved fibroblasts was found to be relatedo the intrinsic features of the tumour cells and to predict can-er progression [103]. The prognostic power of the ‘wound healingignature’ was tested in different types of human cancer and, in

articular, data from 62 patients with stage I and II lung ade-ocarcinoma demonstrated that signature-positive tumours weressociated with a worse prognosis. Another interesting study per-ormed by Navab et al. identified 46 genes differentially expressed

cer Biology 35 (2015) 96–106

between paired CAF and NF cultures established from 15 LC patients[86]. Identified genes could be associated to pathways related tosignal transduction, response to stress, cell adhesion as well asECM–protein interaction (IGTA11, THBS2, and COL11A1) or TGF�signalling pathway (THBS2 and BMP4). Interestingly, a CAF progno-stic signature based on 11 genes (ICAM1, THBS2, MME, OXTR, PDE3B,B3GALT2, EVI2B, COL14A1, GAL, MVTP2) was able to discriminatepatients with different clinical outcome.

Edlund et al. used published gene expression data sets including203 stroma-related genes from different cancer types to exam-ine prognostic value of protein expression in the stroma of 190NSCLC patients and subsequently in an independent cohort of 240patients. Among 12 proteins differentially expressed in the stroma(BGN, CD99, DCN, EMILIN1, FBN1, PDGFRB, PDLIM5 POSTN, SPARC,TAGLN, TNC and VCAM) only CD99 was found to be associated witha negative prognostic impact in LC patients [104]. To identify fac-tors responsible for a proficient cross-talk between fibroblasts andcancer cells in lung tumours we also performed expression profil-ing of 60 cultures of fibroblasts isolated from surgical specimensderived from different areas of the lung of cancer patients (tumoraltissue, peri-tumoral normal tissue and normal tissue distant fromthe surgical margin) and endowed with different pro-tumorigenicpotential as assessed by both in vivo and in vitro assays. From classcomparison analyses, we identified factors potentially involvedin different aspects of fibroblast phenotype such as differencesbetween CAF and NF, or between activated and non-activatedfibroblasts (using high expression of �-SMA as indication of activa-tion), or in vivo functional activity such the ability to elicit growthof LC cells (pro-tumorigenic property) or to stimulate cancercells dissemination to distant organs (pro-dissemination prop-erty). Functional annotation of the differentially regulated genesidentified pathways relevant for fibroblast-cancer cells commu-nication including regulation of cell–cell contacts (Andriani et al.,manuscript in preparation).

2.4. Other cancer types

The relevance of the interaction between fibroblasts and cancercells for tumour progression has been investigated also in othertumour types where specific gene expression patterns associatedwith CAFs have been described.

In particular, pancreatic carcinoma is characterized by a promi-nent stromal content rich in myofibroblast-like pancreatic stellatecells that have been shown to increase tumour progression [105].Among the pathways identified as mediators of the tumour-promoting activity of stromal cells, a central role has been ascribedto IL1� mediated signalling [106] and to activation of the sonichedgehog pathway [107]. In gene-expression studies, enrichmentin stroma-related genes has also been described to correlate togemcitabine resistance and poor survival [108]. CAFs from pancre-atic cancers, similarly to those from skin and breast carcinomas, arealso enriched for a pro-inflammatory gene signature (characterizedby high levels of CXCL2, IL6 and OPN) and could contribute to sustaintumour-promoting inflammation through an NF-�B dependentmechanism [109].

In colon cancer, ‘CAF signatures’ have been obtained by func-tional evaluation of properties of cultures of CAFs revealingpotential heterogeneity within CAFs and prognostic value in clin-ical samples [110,111]. Interestingly, the previously describedpoor prognosis of specific colorectal cancer transcriptional sub-types, including the stem/serrated/mesenchymal subtype, has alsorecently been associated to abundant stromal content suggesting

contribution of CAFs as a major determinant of clinical outcome[112,113].

In fibroblasts derived from oral squamous cell carcinomas differ-ences have been detected according to the stage of the disease and

P. Gandellini et al. / Seminars in Cancer Biology 35 (2015) 96–106 101

Table 1Datasets included in the inter-disease analysis.

Cancer tissue Dataset Reference Number of samples Description Comparisons

Breast

MERLINO Merlino et al.(manuscript, inpreparation)

9 Human normal dermal fibroblasts(HNDF) ± breast cancer cell linesconditioned medium

Conditioned vs unconditionedHNDF

MA [58] 66Microdissected stroma from normaltissue, DCIS or IDC clinical specimens

IDC-associated vs Normal stromaDCIS-associated vs Normal stroma

BAUER [52] 12 CAF and normal fibroblasts (NF) fromprimary tumors

CAF vs NF

PLANCHE [57] 12 Tumor-associated stroma and matchednormal stroma

Tumor vs normal stroma

Prostate

DOLDI [75] 9CAF and human prostate fibroblasts(HPF) from clinical specimens

CAF vs HPFIL6 treated HPF vs untreated HPFTGF� treated HPF vs untreated HPF

PLANCHE [57] 12 Tumor-associated stroma and matchednormal stroma

Tumor vs normal stroma

PASCAL [73] 7 Tumor-associated and normal stromalcells

Tumor-associated vs normalstromal cells

ASHIDA [72] 20 Cultured stromal cells fromtumortissue or normal peripheral zone

Tumor-associated vs normalstromal cells

Lung ANDRIANI [87] 48

CAF, adjacent fibroblasts (AF) and NF;also classified according to alpha-SMApositivity, in vitro growth anddissemination potential

CAF vs NFCAF vs AFAF vs NFAlpha-SMA: high vs lowIn vivo growth: high vs lowDissemination: high vs low

gclalawm

fmrdst

3e

tTotspietldt

[md

NAVAB [86] 30

ene signatures associated with genetically unstable oral squamousell carcinoma (including ˛-SMA and ITGA6) were found to corre-ate to poor patient prognosis in an independent dataset of headnd neck cancers [114]. Finally, the genes most frequently deregu-ated in fibroblasts derived from oesophageal cancer specimens arelso related to ECM remodelling and immune response regulationith expression of FGFR2 apparently related to the creation of aicroenvironment conductive for cancer cells proliferation [115].The interplay with tumour microenvironment is also crucial

or the progression of hematologic malignancies and the involvedolecules largely overlap those described for solid tumours, as

eviewed in [116]. More recently, Paggetti et al. showed that CLL-erived exosomes can promote disease progression by modulatingeveral functions of surrounding stromal cells which acquire fea-ures of CAFs [117].

. Inter-disease meta-analysis to highlight CAF genexpression similarities

Typically, cancer research projects focus on a specific cancerype and this has been also valid for studies characterizing theME and in particular CAFs. Although fibroblasts from differentrgans can have very divergent transcriptional programs, part ofhe tumour-induced signalling might be shared by this major con-tituent of TME independently of the tumour site. This might beartially evinced from the previous paragraphs, but we run an

nter-pathology bioinformatic analysis to directly compare genexpression profiles of CAFs/microdissected stroma from differentumour types generated by our and other groups, possibly high-ighting common transcriptional programs. Datasets included areescribed in Table 1 and have been discussed in the previous sec-ions.

For publicly available datasets we started from raw data. RSN118] and RMA [119] normalizations have been applied to Illu-

ina and Affymetrix microarray platforms, respectively. Internalatasets were processed as described elsewhere [75] (Andriani

Primary fibroblast cell lines fromNSCLC or normal lung

Tumor-associated vs normalstromal cells

et al., manuscript in preparation; Merlino et al., manuscript inpreparation).

To prevent technical biases, we preferred to cross-evaluatethe output of differential expression analysis, instead of mergingexpression data derived from different platforms. Class compari-son analysis was performed using the limma Bioconductor package[120]. Overall, stroma interacting with tumour cells was comparedwith normal stroma, although analysed specimens ranged frommicro-dissected clinical specimens to primary or in vitro stimulatedcells of the stromal compartment (Table 1). For each comparison,all genes were ranked according to the modified t-statistics val-ues obtained. These ranked gene lists were subjected to a GeneSet Enrichment Analysis (GSEA, v.4.0) [121] to identify Gene Ontol-ogy terms or Canonical Pathways (BIOCARTA, KEGG, REACTOME)significantly enriched. A total of 18 comparisons were analysed(Table 1) and 41 gene sets were positively enriched in at least 50%of the comparisons (Fig. 1A).

Most of commonly enriched gene sets are related to ECMmolecules and ECM organization (e.g. collagens), cell–cell interac-tion (integrins and focal adhesions), cytoskeleton (e.g. actin fila-ments) and muscle contraction, identifying these biological func-tions as the most up-regulated in CAFs across different cancer types.

Remarkable observations can be made examining how thedifferent comparisons clustered. The strongest enrichments werefound for several comparisons from BC and LC stroma. Notably,also the comparison AF versus NF in Andriani dataset showed asimilar pattern, suggesting that fibroblasts adjacent to tumours,even if in histologically non-neoplastic areas are characterized bytranscriptional changes typical of reactive fibroblasts. However,only some of the comparisons in PC suggest major similaritywith CAFs from other tumours, i.e. Ashida dataset or the TGF�-stimulated prostate NFs, indirectly supporting TGF� as a masterregulator of CAF activation in multiple cancer types. On the con-

trary, several comparisons in PC showed a mild activation of theselected gene sets, suggesting the existence of CAFs with differenttranscriptomes. Probably, different stimuli may be predominantin inducing activation of a reactive stroma depending on several

102 P. Gandellini et al. / Seminars in Cancer Biology 35 (2015) 96–106

Fig. 1. Inter-disease evaluation of tumour-normal stroma transcriptomic changes. (A) Gene set enrichment analysis was performed for 18 class comparisons detailed inTable 1. Gene sets significantly enriched in at least 50% of evaluated comparisons are reported in the heatmap. (B) ECM3 signature was tested for enrichment in the same setof class comparisons and obtained p-values are reported as barplot. Significance threshold: p < 0.05 (dashed green line).

n Can

fogbcfipa

cwtipfisa

wdgGwrnEaewf

4

wecci

ntpitpichrtfaacsc

ocdt

C

P. Gandellini et al. / Seminars i

actors like tumour stage, patient age, cancer aetiology (e.g. historyf inflammatory diseases, such as prostatitis) or even tumourenetic background. For example, TGF�-dependent activation maye prevalent in conditions that mimic wound healing, such ashronic inflammation. In contrast, the evidence that IL6-activatedbroblasts showed a senescent-like phenotype (induction of p16,21, �-H2AX foci and H3 lysine 9 trimethylation [75]) may suggestn aging-related activation process.

Lung fibroblasts which highly support in vivo growth of cancerells show similarities with IL6-activated prostatic fibroblasts, asell as lung CAFs compared with AFs. A possible interpretation is

hat IL6 is not involved in initial activation of lung fibroblasts butn later stages, coupled with the acquisition of tumour-promotingroperties. Finally, outlier behaviour can be noticed for lungbroblasts showing pro-dissemination ability, with most of geneets not enriched or even negatively enriched possibly indicatingctivation of specific mechanisms relating to metastasis.

In light of the fact that several gene sets contained ECM genes,e specifically tested the so-called ECM3 gene signature originallyerived by studying expression changes of extracellular matrixenes in a cohort of BC clinical specimens [122,123]. By applyingSEA as described above to the same set of comparisons in Table 1,e found that the ECM3 signature is significantly enriched (i.e. up-

egulated) in most of tumour stroma/fibroblasts compared with theormal counterpart (Fig. 1B). More in detail, up-regulation of theCM signature mainly occurs in CAFs probably driven by TGF�, withn exception represented by CAF versus AF comparison in LC. Gen-rally, no enrichment was found in IL6-driven CAFs and, in keepingith the previous analysis, even a negative enrichment was found

or LC fibroblasts showing pro-dissemination ability.

. Conclusions

It has become clear that cancer is an ‘in miniature ecosystem’ith several ‘species’ (i.e. cell types) actively interacting and influ-

ncing each other. Although cancer is basically a genomic diseaseaused by the accumulation of somatic aberrations, a two-wayrosstalk between cancer cells and microenvironment componentss crucial during both carcinogenesis and tumour progression.

We just started to unravel such a complex network of sig-als, however, several challenges still need to be overcome due tohe presence of heterogeneous population of CAFs, which are stilloorly defined at molecular levels (Box 1). Nevertheless, the game

s worth the candle, since remarkable biomarkers and therapeu-ic targets might be veiled in this intricate network. A pivotal andromising example of effective targeting of the microenvironment

s the inhibition of immune checkpoints in melanoma and otherancer types [124–126]. Remarkably, high throughput strategiesave been encouraged in the last few years to identify novel stroma-elated biomarkers of prognostic relevance, suitable as therapeuticargets [26]. It is worth noticing that studies published so far haveocused on specific cancer types but some of the molecular mech-nisms underlying fibroblasts activation or their pro-tumorigenicctivity might be shared by several cancer types as shown and dis-ussed in this review, enabling the opportunity of identifying broadpectrum new therapeutic approaches, an opposite and probablyomplementary trend compared to precision medicine.

In the future it is envisaged that integrated studies and technol-gy advancements may allow a more accurate examination of theomposition of the stromal compartment and ultimately permit theevelopment of therapeutic strategies based on interference with

umour-stroma crosstalk with greater potential for cancer control.

onflict of interest

The authors declare that there are no conflicts of interest.

cer Biology 35 (2015) 96–106 103

Acknowledgements

This work was supported by the AIRC 5X1000 (Associ-azione Italiana Ricerca sul Cancro) project “EDERA” Tumourmicroenvironment-related changes as new tools for early detec-tion and assessment of high-risk diseases [ED12162] and by AIRCgrant IG13403 (to L.R.).

References

[1] Egeblad M, Nakasone ES, Werb Z. Tumors as organs: complex tissues thatinterface with the entire organism. Dev Cell 2010;18:884–901,http://dx.doi.org/10.1016/j.devcel.2010.05.012.

[2] Pietras K, Ostman A. Hallmarks of cancer: interactions with the tumorstroma. Exp Cell Res 2010;316:1324–31, http://dx.doi.org/10.1016/j.yexcr.2010.02.045.

[3] Polyak K, Haviv I, Campbell IG. Co-evolution of tumor cells and theirmicroenvironment. Trends Genet 2009;25:30–8.

[4] Hanahan D, Coussens LM. Accessories to the crime: functions of cellsrecruited to the tumor microenvironment. Cancer Cell 2012;21:309–22,http://dx.doi.org/10.1016/j.ccr.2012.02.022.

[5] Quail DF, Joyce JA. Microenvironmental regulation of tumor progression andmetastasis. Nat Med 2013;19:1423–37, http://dx.doi.org/10.1038/nm.3394.

[6] Casey SC, Amedei A, Aquilano K, Benencia F, Bhakta D, Boosani CS, et al.Cancer prevention and therapy through the modulation of the tumormicroenvironment. Semin Cancer Biol 2015, http://dx.doi.org/10.1016/j.semcancer.2015.02.007.

[7] Östman A. The tumor microenvironment controls drug sensitivity. Nat Med2012;18:1332–4, http://dx.doi.org/10.1038/nm.2938.

[8] Fong MY, Zhou W, Liu L, Alontaga AY, Chandra M, Ashby J, et al.Breast-cancer-secreted miR-122 reprograms glucose metabolism inpremetastatic niche to promote metastasis. Nat Cell Biol 2015;17:183–94,http://dx.doi.org/10.1038/ncb3094.

[9] Webber JP, Spary LK, Sanders AJ, Chowdhury R, Jiang WG, Steadman R, et al.Differentiation of tumour-promoting stromal myofibroblasts by cancerexosomes. Oncogene 2015;34:290–302, http://dx.doi.org/10.1038/onc.2013.560.

[10] Webber J, Yeung V, Clayton A. Extracellular vesicles as modulators of thecancer microenvironment. Semin Cell Dev Biol 2015;40:27–34,http://dx.doi.org/10.1016/j.semcdb.2015.01.013.

[11] Kohlhapp FJ, Mitra AK, Lengyel E, Peter ME. MicroRNAs as mediators andcommunicators between cancer cells and the tumor microenvironment.Oncogene 2015, http://dx.doi.org/10.1038/onc.2015.89.

[12] Wang T, Liu G, Wang R. The intercellular metabolic interplay between tumorand immune cells. Front Immunol 2014;5:358, http://dx.doi.org/10.3389/fimmu.2014.00358.

[13] Elenbaas B, Weinberg RA. Heterotypic signaling between epithelial tumorcells and fibroblasts in carcinoma formation. Exp Cell Res 2001;264:169–84,http://dx.doi.org/10.1006/excr.2000.5133.

[14] Kalluri R, Zeisberg M. Fibroblasts in cancer. Nat Rev Cancer 2006;6:392–401,http://dx.doi.org/10.1038/nrc1877.

[15] Bhowmick NA, Neilson EG, Moses HL. Stromal fibroblasts in cancer initiationand progression. Nature 2004;432:332–7, http://dx.doi.org/10.1038/nature03096.

[16] Orimo A, Weinberg RA. Stromal fibroblasts in cancer: a noveltumor-promoting cell type. Cell Cycle 2006;5:1597–601.

[17] Junttila MR, de Sauvage FJ. Influence of tumour micro-environmentheterogeneity on therapeutic response. Nature 2013;501:346–54,http://dx.doi.org/10.1038/nature12626.

[18] Sugimoto H, Mundel TM, Kieran MW, Kalluri R. Identification of fibroblastheterogeneity in the tumor microenvironment. Cancer Biol Ther2006;5:1640–6.

[19] Madar S, Goldstein I, Rotter V. Cancer associated fibroblasts—more thanmeets the eye. Trends Mol Med 2013;19:447–53, http://dx.doi.org/10.1016/j.molmed.2013.05.004.

[20] Augsten M. Cancer-associated fibroblasts as another polarized cell type ofthe tumor microenvironment. Front Oncol 2014;4:62, http://dx.doi.org/10.3389/fonc.2014.00062.

[21] Lo A, Wang L-CS, Scholler J, Monslow J, Avery D, Newick K, et al.Tumor-promoting desmoplasia is disrupted by depleting FAP-expressingstromal cells. Cancer Res 2015, http://dx.doi.org/10.1158/0008-5472.CAN-14-3041.

[22] Ozdemir BC, Pentcheva-Hoang T, Carstens JL, Zheng X, Wu C-C, Simpson TR,

et al. Depletion of carcinoma-associated fibroblasts and fibrosis inducesimmunosuppression and accelerates pancreas cancer with reduced survival.Cancer Cell 2014;25:719–34, http://dx.doi.org/10.1016/j.ccr.2014.04.005.

[23] Rhim AD, Oberstein PE, Thomas DH, Mirek ET, Palermo CF, Sastra SA, et al.Stromal elements act to restrain, rather than support, pancreatic ductal

1 in Can

04 P. Gandellini et al. / Seminars

adenocarcinoma. Cancer Cell 2014;25:735–47, http://dx.doi.org/10.1016/j.ccr.2014.04.021.

[24] Öhlund D, Elyada E, Tuveson D. Fibroblast heterogeneity in the cancerwound. J Exp Med 2014;211:1503–23, http://dx.doi.org/10.1084/jem.20140692.

[25] Valcz G, Sipos F, Tulassay Z, Molnar B, Yagi Y. Importance ofcarcinoma-associated fibroblast-derived proteins in clinical oncology. J ClinPathol 2014;67:1026–31, http://dx.doi.org/10.1136/jclinpath-2014-202561.

[26] Paulsson J, Micke P. Prognostic relevance of cancer-associated fibroblasts inhuman cancer. Semin Cancer Biol 2014;25:61–8,http://dx.doi.org/10.1016/j.semcancer.2014.02.006.

[27] Hinz B, Phan SH, Thannickal VJ, Galli A, Bochaton-Piallat M-L, Gabbiani G.The myofibroblast: one function, multiple origins. Am J Pathol2007;170:1807–16, http://dx.doi.org/10.2353/ajpath.2007.070112.

[28] Polyak K, Kalluri R. The role of the microenvironment in mammary glanddevelopment and cancer. Cold Spring Harb Perspect Biol 2010;2:a003244,http://dx.doi.org/10.1101/cshperspect.a003244.

[29] Allinen M, Beroukhim R, Cai L, Brennan C, Lahti-Domenici J, Huang H, et al.Molecular characterization of the tumor microenvironment in breast cancer.Cancer Cell 2004;6:17–32, http://dx.doi.org/10.1016/j.ccr.2004.06.010.

[30] Peng Q, Zhao L, Hou Y, Sun Y, Wang L, Luo H, et al. Biological characteristicsand genetic heterogeneity between carcinoma-associated fibroblasts andtheir paired normal fibroblasts in human breast cancer. PLoS One2013;8:e60321, http://dx.doi.org/10.1371/journal.pone.0060321.

[31] Bissell MJ, Radisky D. Putting tumours in context. Nat Rev Cancer2001;1:46–54, http://dx.doi.org/10.1038/35094059.

[32] Kojima Y, Acar A, Eaton EN, Mellody KT, Scheel C, Ben-Porath I, et al.Autocrine TGF-beta and stromal cell-derived factor-1 (SDF-1) signalingdrives the evolution of tumor-promoting mammary stromal myofibroblasts.Proc Natl Acad Sci USA 2010;107:20009–14, http://dx.doi.org/10.1073/pnas.1013805107.

[33] Spaeth EL, Dembinski JL, Sasser AK, Watson K, Klopp A, Hall B, et al.Mesenchymal stem cell transition to tumor-associated fibroblastscontributes to fibrovascular network expansion and tumor progression.PLoS One 2009;4:e4992, http://dx.doi.org/10.1371/journal.pone.0004992.

[34] Zeisberg EM, Potenta S, Xie L, Zeisberg M, Kalluri R. Discovery of endothelialto mesenchymal transition as a source for carcinoma-associated fibroblasts.Cancer Res 2007;67:10123–8, http://dx.doi.org/10.1158/0008-5472.CAN-07-3127.

[35] Surowiak P, Murawa D, Materna V, Maciejczyk A, Pudelko M, Ciesla S, et al.Occurence of stromal myofibroblasts in the invasive ductal breast cancertissue is an unfavourable prognostic factor. Anticancer Res2007;27:2917–24.

[36] Lee J, Fassnacht M, Nair S, Boczkowski D, Gilboa E. Tumor immunotherapytargeting fibroblast activation protein, a product expressed intumor-associated fibroblasts. Cancer Res 2005;65:11156–63,http://dx.doi.org/10.1158/0008-5472.CAN-05-2805.

[37] Park JE, Lenter MC, Zimmermann RN, Garin-Chesa P, Old LJ, Rettig WJ.Fibroblast activation protein, a dual specificity serine protease expressed inreactive human tumor stromal fibroblasts. J Biol Chem 1999;274:36505–12.

[38] Paulsson J, Sjöblom T, Micke P, Pontén F, Landberg G, Heldin C-H, et al.Prognostic significance of stromal platelet-derived growth factorbeta-receptor expression in human breast cancer. Am J Pathol2009;175:334–41, http://dx.doi.org/10.2353/ajpath.2009.081030.

[39] Ariga N, Sato E, Ohuchi N, Nagura H, Ohtani H. Stromal expression offibroblast activation protein/seprase, a cell membrane serine proteinase andgelatinase, is associated with longer survival in patients with invasive ductalcarcinoma of breast. Int J Cancer 2001;95:67–72.

[40] Jakowlew SB. Transforming growth factor-beta in cancer and metastasis.Cancer Metastasis Rev 2006;25:435–57,http://dx.doi.org/10.1007/s10555-006-9006-2.

[41] Bronzert DA, Pantazis P, Antoniades HN, Kasid A, Davidson N, Dickson RB,et al. Synthesis and secretion of platelet-derived growth factor by humanbreast cancer cell lines. Proc Natl Acad Sci USA 1987;84:5763–7.

[42] Hugo HJ, Lebret S, Tomaskovic-Crook E, Ahmed N, Blick T, Newgreen DF,et al. Contribution of fibroblast and mast cell (afferent) and tumor (efferent)IL-6 effects within the tumor microenvironment. Cancer Microenviron 2012,http://dx.doi.org/10.1007/s12307-012-0098-7.

[43] Mao Y, Keller ET, Garfield DH, Shen K, Wang J. Stromal cells in tumormicroenvironment and breast cancer. Cancer Metastasis Rev2013;32:303–15, http://dx.doi.org/10.1007/s10555-012-9415-3.

[44] Mercier I, Casimiro MC, Wang C, Rosenberg AL, Quong J, Minkeu A, et al.Human breast cancer-associated fibroblasts (CAFs) show caveolin-1downregulation and RB tumor suppressor functional inactivation:Implications for the response to hormonal therapy. Cancer Biol Ther2008;7:1212–25.

[45] Martinez-Outschoorn UE, Pavlides S, Whitaker-Menezes D, Daumer KM,Milliman JN, Chiavarina B, et al. Tumor cells induce the cancer associatedfibroblast phenotype via caveolin-1 degradation: implications for breastcancer and DCIS therapy with autophagy inhibitors. Cell Cycle2010;9:2423–33.

cer Biology 35 (2015) 96–106

[46] Koleske AJ, Baltimore D, Lisanti MP. Reduction of caveolin and caveolae inoncogenically transformed cells. Proc Natl Acad Sci USA 1995;92:1381–5.

[47] Witkiewicz AK, Dasgupta A, Sotgia F, Mercier I, Pestell RG, Sabel M, et al. Anabsence of stromal caveolin-1 expression predicts early tumor recurrenceand poor clinical outcome in human breast cancers. Am J Pathol2009;174:2023–34, http://dx.doi.org/10.2353/ajpath.2009.080873.

[48] Bonuccelli G, Whitaker-Menezes D, Castello-Cros R, Pavlides S, Pestell RG,Fatatis A, et al. The reverse Warburg effect: glycolysis inhibitors prevent thetumor promoting effects of caveolin-1 deficient cancer associatedfibroblasts. Cell Cycle 2010;9:1960–71.

[49] Bibber B, Sinha G, Lobba ARM, Greco SJ, Rameshwar P. A review of stem celltranslation and potential confounds by cancer stem cells. Stem Cells Int2013;2013:241048, http://dx.doi.org/10.1155/2013/241048.

[50] Tsuyada A, Chow A, Wu J, Somlo G, Chu P, Loera S, et al. CCL2 mediatescross-talk between cancer cells and stromal fibroblasts that regulates breastcancer stem cells. Cancer Res 2012;72:2768–79,http://dx.doi.org/10.1158/0008-5472.CAN-11-3567.

[51] Park SY, Kim HM, Koo JS. Differential expression of cancer-associatedfibroblast-related proteins according to molecular subtype and stromalhistology in breast cancer. Breast Cancer Res Treat 2015;149:727–41,http://dx.doi.org/10.1007/s10549-015-3291-9.

[52] Bauer M, Su G, Casper C, He R, Rehrauer W, Friedl A. Heterogeneity of geneexpression in stromal fibroblasts of human breast carcinomas and normalbreast. Oncogene 2010;29:1732–40,http://dx.doi.org/10.1038/onc.2009.463.

[53] Pennica D, Swanson TA, Welsh JW, Roy MA, Lawrence DA, Lee J, et al. WISPgenes are members of the connective tissue growth factor family that areup-regulated in wnt-1-transformed cells and aberrantly expressed inhuman colon tumors. Proc Natl Acad Sci USA 1998;95:14717–22.

[54] Kim YC, Clark RJ, Ranheim EA, Alexander CM. Wnt1 expression inducesshort-range and long-range cell recruitments that modify mammary tumordevelopment and are not induced by a cell-autonomous beta-catenineffector. Cancer Res 2008;68:10145–53,http://dx.doi.org/10.1158/0008-5472.CAN-08-2992.

[55] Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H, et al.Stromal gene expression predicts clinical outcome in breast cancer. Nat Med2008;14:518–27, http://dx.doi.org/10.1038/nm1764.

[56] Farmer P, Bonnefoi H, Anderle P, Cameron D, Wirapati P, Wirapati P, et al. Astroma-related gene signature predicts resistance to neoadjuvantchemotherapy in breast cancer. Nat Med 2009;15:68–74,http://dx.doi.org/10.1038/nm.1908.

[57] Singer CF, Gschwantler-Kaulich D, Fink-Retter A, Haas C, Hudelist G,Czerwenka K, et al. Differential gene expression profile in breastcancer-derived stromal fibroblasts. Breast Cancer Res Treat2008;110:273–81, http://dx.doi.org/10.1007/s10549-007-9725-2.

[58] Ma X-J, Dahiya S, Richardson E, Erlander M, Sgroi DC. Gene expressionprofiling of the tumor microenvironment during breast cancer progression.Breast Cancer Res 2009;11:R7, http://dx.doi.org/10.1186/bcr2222.

[59] Tuxhorn JA, Ayala GE, Rowley DR. Reactive stroma in prostate cancerprogression. J Urol 2001;166:2472–83.

[60] Tuxhorn JA, Ayala GE, Smith MJ, Smith VC, Dang TD, Rowley DR. Reactivestroma in human prostate cancer: induction of myofibroblast phenotypeand extracellular matrix remodeling. Clin Cancer Res 2002;8:2912–23.

[61] Taddei ML, Giannoni E, Comito G, Chiarugi P. Microenvironment and tumorcell plasticity: an easy way out. Cancer Lett 2013;341:80–96,http://dx.doi.org/10.1016/j.canlet.2013.01.042.

[62] Ayala G, Tuxhorn JA, Wheeler TM, Frolov A, Scardino PT, Ohori M, et al.Reactive stroma as a predictor of biochemical-free recurrence in prostatecancer. Clin Cancer Res 2003;9:4792–801.

[63] Jia Z, Rahmatpanah FB, Chen X, Lernhardt W, Wang Y, Xia X-Q, et al.Expression changes in the stroma of prostate cancer predict subsequentrelapse. PLoS One 2012;7:e41371,http://dx.doi.org/10.1371/journal.pone.0041371.

[64] Olumi AF, Grossfeld GD, Hayward SW, Carroll PR, Tlsty TD, Cunha GR.Carcinoma-associated fibroblasts direct tumor progression of initiatedhuman prostatic epithelium. Cancer Res 1999;59:5002–11.

[65] Fiaschi T, Marini A, Giannoni E, Taddei ML, Gandellini P, De Donatis A, et al.Reciprocal metabolic reprogramming through lactate shuttle coordinatelyinfluences tumor-stroma interplay. Cancer Res 2012;72:5130–40,http://dx.doi.org/10.1158/0008-5472.CAN-12-1949.

[66] Giannoni E, Bianchini F, Masieri L, Serni S, Torre E, Calorini L, et al. Reciprocalactivation of prostate cancer cells and cancer-associated fibroblastsstimulates epithelial-mesenchymal transition and cancer stemness. CancerRes 2010;70:6945–56, http://dx.doi.org/10.1158/0008-5472.CAN-10-0785.

[67] Giannoni E, Bianchini F, Calorini L, Chiarugi P. Cancer associated fibroblastsexploit reactive oxygen species through a proinflammatory signatureleading to epithelial mesenchymal transition and stemness. Antioxid RedoxSignal 2011;14:2361–71, http://dx.doi.org/10.1089/ars.2010.3727.

[68] Gandellini P, Giannoni E, Casamichele A, Taddei ML, Callari M, Piovan C,et al. miR-205 hinders the malignant interplay between prostate cancercells and associated fibroblasts. Antioxid Redox Signal 2014;20:1045–59,http://dx.doi.org/10.1089/ars.2013.5292.

n Can

P. Gandellini et al. / Seminars i

[69] Gandellini P, Folini M, Longoni N, Pennati M, Binda M, Colecchia M, et al.miR-205 exerts tumor-suppressive functions in human prostate throughdown-regulation of protein kinase Cepsilon. Cancer Res 2009;69:2287–95,http://dx.doi.org/10.1158/0008-5472.CAN-08-2894.

[70] Mani SA, Guo W, Liao M-J, Eaton EN, Ayyanan A, Zhou AY, et al. Theepithelial-mesenchymal transition generates cells with properties of stemcells. Cell 2008;133:704–15, http://dx.doi.org/10.1016/j.cell.2008.03.027.

[71] Liao C-P, Adisetiyo H, Liang M, Roy-Burman P. Cancer-associated fibroblastsenhance the gland-forming capability of prostate cancer stem cells. CancerRes 2010;70:7294–303, http://dx.doi.org/10.1158/0008-5472.CAN-09-3982.

[72] Ashida S, Orloff MS, Bebek G, Zhang L, Zheng P, Peehl DM, et al. Integratedanalysis reveals critical genomic regions in prostate tumormicroenvironment associated with clinicopathologic phenotypes. ClinCancer Res 2012;18:1578–87,http://dx.doi.org/10.1158/1078-0432.CCR-11-2535.

[73] Pascal LE, Ai J, Vêncio RZN, Vêncio EF, Zhou Y, Page LS, et al. Differentialinductive signaling of CD90 prostate cancer-associated fibroblastscompared to normal tissue stromal mesenchyme cells. Cancer Microenviron2011;4:51–9, http://dx.doi.org/10.1007/s12307-010-0061-4.

[74] Planche A, Bacac M, Provero P, Fusco C, Delorenzi M, Stehle J-C, et al.Identification of prognostic molecular features in the reactive stroma ofhuman breast and prostate cancer. PLoS One 2011;6:e18640,http://dx.doi.org/10.1371/journal.pone.0018640.

[75] Doldi V, Callari M, Giannoni E, D’Aiuto F, Maffezzini M, Valdagni R, et al.Integrated gene and miRNA expression analysis of prostate cancerassociated fibroblasts supports a prominent role for interleukin-6 infibroblast activation. Oncotarget 2015, in press.

[76] Dean JP, Nelson PS. Profiling influences of senescent and aged fibroblasts onprostate carcinogenesis. Br J Cancer 2008;98:245–9,http://dx.doi.org/10.1038/sj.bjc.6604087.

[77] Bianchi-Frias D, Vakar-Lopez F, Coleman IM, Plymate SR, Reed MJ, Nelson PS.The effects of aging on the molecular and cellular composition of theprostate microenvironment. PLoS One 2010;5,http://dx.doi.org/10.1371/journal.pone.0012501.

[78] Bavik C, Coleman I, Dean JP, Knudsen B, Plymate S, Nelson PS. The geneexpression program of prostate fibroblast senescence modulates neoplasticepithelial cell proliferation through paracrine mechanisms. Cancer Res2006;66:794–802, http://dx.doi.org/10.1158/0008-5472.CAN-05-1716.

[79] Yang G, Rosen DG, Zhang Z, Bast RC, Mills GB, Colacino JA, et al. Thechemokine growth-regulated oncogene 1 (Gro-1) links RAS signaling to thesenescence of stromal fibroblasts and ovarian tumorigenesis. Proc Natl AcadSci USA 2006;103:16472–7, http://dx.doi.org/10.1073/pnas.0605752103.

[80] Liu D-H, Chen Y-M, Liu Y, Hao B-S, Zhou B, Wu L, et al. Rb1 protectsendothelial cells from hydrogen peroxide-induced cell senescence bymodulating redox status. Biol Pharm Bull 2011;34:1072–7.

[81] Waghray M, Cui Z, Horowitz JC, Subramanian IM, Martinez FJ, Toews GB,et al. Hydrogen peroxide is a diffusible paracrine signal for the induction ofepithelial cell death by activated myofibroblasts. FASEB J 2005;19:854–6,http://dx.doi.org/10.1096/fj.04-2882fje.

[82] Taddei ML, Cavallini L, Comito G, Giannoni E, Folini M, Marini A, et al.Senescent stroma promotes prostate cancer progression: the role ofmiR-210. Mol Oncol 2014;8:1729–46,http://dx.doi.org/10.1016/j.molonc.2014.07.009.

[83] Laberge R-M, Sun Y, Orjalo AV, Patil CK, Freund A, Zhou L, et al. MTORregulates the pro-tumorigenic senescence-associated secretory phenotypeby promoting IL1A translation. Nat Cell Biol 2015;17:1049–61,http://dx.doi.org/10.1038/ncb3195.

[84] Bremnes RM, Dønnem T, Al-Saad S, Al-Shibli K, Andersen S, Sirera R, et al.The role of tumor stroma in cancer progression and prognosis: emphasis oncarcinoma-associated fibroblasts and non-small cell lung cancer. J ThoracOncol 2011;6:209–17, http://dx.doi.org/10.1097/JTO.0b013e3181f8a1bd.

[85] Kim S-H, Choe C, Shin Y-S, Jeon M-J, Choi S-J, Lee J, et al. Human lungcancer-associated fibroblasts enhance motility of non-small cell lung cancercells in co-culture. Anticancer Res 2013;33:2001–9.

[86] Navab R, Strumpf D, Bandarchi B, Zhu C-Q, Pintilie M, Ramnarine VR, et al.Prognostic gene-expression signature of carcinoma-associated fibroblasts innon-small cell lung cancer. Proc Natl Acad Sci USA 2011;108:7160–5,http://dx.doi.org/10.1073/pnas.1014506108.

[87] Andriani F, Bertolini G, Caserini R, Bursomanno S, Gasparini P, Pastorino U,et al. Contribution of fibroblasts isolated from primary lung cancers andnormal lung tissue to tumor growth. In: Proc. 102nd AACR Annu. Meet.2011. p. 530.

[88] Fromigué O, Louis K, Dayem M, Milanini J, Pages G, Tartare-Deckert S, et al.Gene expression profiling of normal human pulmonary fibroblasts followingcoculture with non-small-cell lung cancer cells reveals alterations related tomatrix degradation, angiogenesis, cell growth and survival. Oncogene2003;22:8487–97, http://dx.doi.org/10.1038/sj.onc.1206918.

[89] Vicent S, Sayles LC, Vaka D, Khatri P, Gevaert O, Chen R, et al. Cross-species

functional analysis of cancer-associated fibroblasts identifies a critical rolefor CLCF1 and IL-6 in non-small cell lung cancer in vivo. Cancer Res2012;72:5744–56, http://dx.doi.org/10.1158/0008-5472.CAN-12-1097.

[90] Wald O, Izhar U, Amir G, Kirshberg S, Shlomai Z, Zamir G, et al. Interactionbetween neoplastic cells and cancer-associated fibroblasts through the

cer Biology 35 (2015) 96–106 105

CXCL12/CXCR4 axis: role in non-small cell lung cancer tumor proliferation. JThorac Cardiovasc Surg 2011;141:1503–12,http://dx.doi.org/10.1016/j.jtcvs.2010.11.056.

[91] Bertolini G, D’Amico L, Moro M, Landoni E, Perego P, Miceli R, et al.Microenvironment modulated metastatic CD133+/CXCR4+/EpCAM- lungcancer initiating cells sustain tumor dissemination and correlate with poorprognosis. Cancer Res 2015, in review.

[92] Chen W-J, Ho C-C, Chang Y-L, Chen H-Y, Lin C-A, Ling T-Y, et al.Cancer-associated fibroblasts regulate the plasticity of lung cancer stemnessvia paracrine signalling. Nat Commun 2014;5:3472,http://dx.doi.org/10.1038/ncomms4472.

[93] Chang HY, Chi J-T, Dudoit S, Bondre C, van de Rijn M, Botstein D, et al.Diversity, topographic differentiation, and positional memory in humanfibroblasts. Proc Natl Acad Sci USA 2002;99:12877–82,http://dx.doi.org/10.1073/pnas.162488599.

[94] Kotaru C, Schoonover KJ, Trudeau JB, Huynh M-L, Zhou X, Hu H, et al.Regional fibroblast heterogeneity in the lung: implications for remodeling.Am J Respir Crit Care Med 2006;173:1208–15,http://dx.doi.org/10.1164/rccm.200508-1218OC.

[95] Pechkovsky DV, Hackett TL, An SS, Shaheen F, Murray LA, Knight DA. Humanlung parenchyma but not proximal bronchi produces fibroblasts withenhanced TGF-beta signaling and alpha-SMA expression. Am J Respir CellMol Biol 2010;43:641–51, http://dx.doi.org/10.1165/rcmb.2009-0318OC.

[96] Tsukui T, Ueha S, Abe J, Hashimoto S, Shichino S, Shimaoka T, et al.Qualitative rather than quantitative changes are hallmarks of fibroblasts inbleomycin-induced pulmonary fibrosis. Am J Pathol 2013;183:758–73,http://dx.doi.org/10.1016/j.ajpath.2013.06.005.

[97] Pazolli E, Luo X, Brehm S, Carbery K, Chung J-J, Prior JL, et al. Senescentstromal-derived osteopontin promotes preneoplastic cell growth. CancerRes 2009;69:1230–9, http://dx.doi.org/10.1158/0008-5472.CAN-08-2970.

[98] Ishikawa S, Takenaka K, Yanagihara K, Miyahara R, Kawano Y, Otake Y, et al.Matrix metalloproteinase-2 status in stromal fibroblasts, not in tumor cells,is a significant prognostic factor in non-small-cell lung cancer. Clin CancerRes 2004;10:6579–85, http://dx.doi.org/10.1158/1078-0432.CCR-04-0272.

[99] Chen Y, Zou L, Zhang Y, Chen Y, Xing P, Yang W, et al. Transforming growthfactor-�1 and �-smooth muscle actin in stromal fibroblasts are associatedwith a poor prognosis in patients with clinical stage I-IIIA nonsmall cell lungcancer after curative resection. Tumour Biol 2014;35:6707–13,http://dx.doi.org/10.1007/s13277-014-1908-y.

[100] Nakao M, Ishii G, Nagai K, Kawase A, Kenmotsu H, Kon-No H, et al.Prognostic significance of carbonic anhydrase IX expression bycancer-associated fibroblasts in lung adenocarcinoma. Cancer2009;115:2732–43, http://dx.doi.org/10.1002/cncr.24303.

[101] Ito M, Ishii G, Nagai K, Maeda R, Nakano Y, Ochiai A. Prognostic impact ofcancer-associated stromal cells in patients with stage I lungadenocarcinoma. Chest 2012;142:151–8,http://dx.doi.org/10.1378/chest. 11-2458.

[102] Neri S, Ishii G, Hashimoto H, Kuwata T, Nagai K, Date H, et al.Podoplanin-expressing cancer-associated fibroblasts lead and enhance thelocal invasion of cancer cells in lung adenocarcinoma. Int J Cancer 2015,http://dx.doi.org/10.1002/ijc.29464.

[103] Chang HY, Sneddon JB, Alizadeh AA, Sood R, West RB, Montgomery K, et al.Gene expression signature of fibroblast serum response predicts humancancer progression: similarities between tumors and wounds. PLoS Biol2004;2:E7, http://dx.doi.org/10.1371/journal.pbio.0020007.

[104] Edlund K, Lindskog C, Saito A, Berglund A, Pontén F, Göransson-Kultima H,et al. CD99 is a novel prognostic stromal marker in non-small cell lungcancer. Int J Cancer 2012;131:2264–73, http://dx.doi.org/10.1002/ijc.27518.

[105] Hwang RF, Moore T, Arumugam T, Ramachandran V, Amos KD, Rivera A,et al. Cancer-associated stromal fibroblasts promote pancreatic tumorprogression. Cancer Res 2008;68:918–26,http://dx.doi.org/10.1158/0008-5472.CAN-07-5714.

[106] Tjomsland V, Spångeus A, Välilä J, Sandström P, Borch K, Druid H, et al.Interleukin 1� sustains the expression of inflammatory factors in humanpancreatic cancer microenvironment by targeting cancer-associatedfibroblasts. Neoplasia 2011;13:664–75.

[107] Walter K, Omura N, Hong S-M, Griffith M, Vincent A, Borges M, et al.Overexpression of smoothened activates the sonic hedgehog signalingpathway in pancreatic cancer-associated fibroblasts. Clin Cancer Res2010;16:1781–9, http://dx.doi.org/10.1158/1078-0432.CCR-09-1913.

[108] Garrido-Laguna I, Uson M, Rajeshkumar NV, Tan AC, de Oliveira E, Karikari C,et al. Tumor engraftment in nude mice and enrichment in stroma-relatedgene pathways predict poor survival and resistance to gemcitabine inpatients with pancreatic cancer. Clin Cancer Res 2011;17:5793–800,http://dx.doi.org/10.1158/1078-0432.CCR-11-0341.

[109] Erez N, Truitt M, Olson P, Arron ST, Hanahan D. Cancer-associated fibroblastsare activated in incipient neoplasia to orchestrate tumor-promoting

inflammation in an NF-kappaB-dependent manner. Cancer Cell2010;17:135–47, http://dx.doi.org/10.1016/j.ccr.2009.12.041.

[110] Berdiel-Acer M, Sanz-Pamplona R, Calon A, Cuadras D, Berenguer A, SanjuanX, et al. Differences between CAFs and their paired NCF from adjacent

1 in Can

06 P. Gandellini et al. / Seminars

colonic mucosa reveal functional heterogeneity of CAFs, providingprognostic information. Mol Oncol 2014;8:1290–305,http://dx.doi.org/10.1016/j.molonc.2014.04.006.

[111] Herrera M, Islam ABMMK, Herrera A, Martín P, García V, Silva J, et al.Functional heterogeneity of cancer-associated fibroblasts from human colontumors shows specific prognostic gene expression signature. Clin CancerRes 2013;19:5914–26, http://dx.doi.org/10.1158/1078-0432.CCR-13-0694.

[112] Calon A, Lonardo E, Berenguer-Llergo A, Espinet E, Hernando-Momblona X,Iglesias M, et al. Stromal gene expression defines poor-prognosis subtypesin colorectal cancer. Nat Genet 2015;47:320–9, http://dx.doi.org/10.1038/ng.3225.

[113] Isella C, Terrasi A, Bellomo SE, Petti C, Galatola G, Muratore A, et al. Stromalcontribution to the colorectal cancer transcriptome. Nat Genet2015;47:312–9, http://dx.doi.org/10.1038/ng.3224.

[114] Lim KP, Cirillo N, Hassona Y, Wei W, Thurlow JK, Cheong SC, et al. Fibroblastgene expression profile reflects the stage of tumour progression in oralsquamous cell carcinoma. J Pathol 2011;223:459–69,http://dx.doi.org/10.1002/path.2841.

[115] Zhang C, Fu L, Fu J, Hu L, Yang H, Rong T-H, et al. Fibroblast growth factorreceptor 2-positive fibroblasts provide a suitable microenvironment fortumor development and progression in esophageal carcinoma. Clin CancerRes 2009;15:4017–27, http://dx.doi.org/10.1158/1078-0432.CCR-08-2824.

[116] Raffaghello L, Vacca A, Pistoia V, Ribatti D. Cancer associated fibroblasts inhematological malignancies. Oncotarget 2015;6:2589–603.

[117] Paggetti J, Haderk F, Seiffert M, Janji B, Distler U, Ammerlaan W, et al.

Exosomes released by chronic lymphocytic leukemia cells induce thetransition of stromal cells into cancer-associated fibroblasts. Blood 2015,http://dx.doi.org/10.1182/blood-2014-12-618025.

[118] Du P, Kibbe WA, Lin SM. Lumi: a pipeline for processing Illumina microarray.Bioinforma 2008;24:1547–8.

cer Biology 35 (2015) 96–106

[119] Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U,et al. Exploration, normalization, and summaries of high densityoligonucleotide array probe level data. Biostat 2003;4:249–64.

[120] Smyth GK. Linear models and empirical bayes methods for assessingdifferential expression in microarray experiments. Stat Appl Genet Mol Biol2004;3:Article3.

[121] Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA,et al. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc Natl Acad Sci USA2005;102:15545–50.

[122] Bergamaschi A, Tagliabue E, Sørlie T, Naume B, Triulzi T, Orlandi R, et al.Extracellular matrix signature identifies breast cancer subgroups withdifferent clinical outcome. J Pathol 2008;214:357–67, http://dx.doi.org/10.1002/path.2278.

[123] Triulzi T, Casalini P, Sandri M, Ratti M, Carcangiu ML, Colombo MP, et al.Neoplastic and stromal cells contribute to an extracellular matrix geneexpression profile defining a breast cancer subtype likely to progress. PLoSOne 2013;8:e56761, http://dx.doi.org/10.1371/journal.pone.0056761.

[124] Cohen J, Sznol M. Therapeutic combinations of immune-modulatingantibodies in melanoma and beyond. Semin Oncol 2015;42:488–94,http://dx.doi.org/10.1053/j.seminoncol.2015.02.014.

[125] Eggermont AMM, Maio M, Robert C. Immune checkpoint inhibitors inmelanoma provide the cornerstones for curative therapies. Semin Oncol2015;42:429–35, http://dx.doi.org/10.1053/j.seminoncol.2015.02.010.

[126] Reck M, Paz-Ares L. Immunologic checkpoint blockade in lung cancer. Semin

Oncol 2015;42:402–17, http://dx.doi.org/10.1053/j.seminoncol.2015.02.013.

[127] Sung KE, Su X, Berthier E, Pehlke C, Friedl A, Beebe DJ. Understanding theimpact of 2D and 3D fibroblast cultures on in vitro breast cancer models.PLoS One 2013;8:e76373, http://dx.doi.org/10.1371/journal.pone.0076373.