1

TargetingtheMicroenvironmentinAdvancedColorectalCancer1

2

DanieleV.F.Tauriello1andEduardBatlle1,23

4

1.InstituteforResearchinBiomedicine(IRBBarcelona).TheBarcelonaInstituteofScience5

andTechnology.BaldiriiReixac10,08028Barcelona,Spain6

2.InstitucióCatalanadeRecercaiEstudisAvançats(ICREA),Barcelona,Spain.7

8

9

10

11

12

13

14

15

16

Correspondence:Daniele.Tauriello@irbbarcelona.org&Eduard.Batlle@irbbarcelona.org17

18

19

20

Keywords:stroma,cancerecology,molecularsubtypes,pre-clinicalmousemodels,21

immuneoncology,TGF-β22

23

2

Abstract1

Colorectal cancer (CRC) diagnosis often occurs at late stages when tumour cells have2

alreadydisseminated.Current therapiesarepoorlyeffective formetastaticdisease, themain3

causeofdeathinCRC.Despitemountingevidenceimplicatingthetumourmicroenvironmentin4

CRCprogressionandmetastasis,clinicalpracticeremainspredominantlyfocussedontargeting5

the epithelial compartment. As CRCs remain largely refractory to current therapies, we are6

compelled to devise alternative strategies. TGF-β has emerged as a key architect of the7

microenvironment in poor prognosis cancers. Disseminated tumour cells show a strong8

dependencyonaTGF-β-activatedstromaduringtheestablishmentandsubsequentexpansion9

of metastasis. Here, we will review and discuss the development of integrated approaches10

focusedontargetingtheecosystemofpoorprognosisCRCs.11

12

Trends13

• In the search for new paradigms on which to base novel therapeutic strategies for14

advancedCRC,focushasincreasinglybeencentredonthetumourmicroenvironment.15

• Thereisanemergingnotionthatinteractionsbetweenepithelialcancercellsandtheir16

environmentcanbeunderstoodapplyingaconceptualframeworksimilartothatused17

tostudyecosystems.18

• In CRC patients, a stromal-expressed gene programme enriched for TGF-β and19

downstreamtargetshasbeenlinkedtopoorprognosisandmetastasisformation.20

• TGF-β signalling plays key roles in instructing the tumour microenvironment of late21

stageCRCs,yetinhibitionofTGF-βsignallingasatherapeuticstrategyremainsscarcely22

exploredintheclinic.23

3

Thetumourepitheliumasprimarytargetofstandardtherapies1

Colorectal cancer (CRC) originates from benign lesions known as adenomas: localised2

glandularovergrowths in theepithelial liningof the largebowel.Over time, someadenomas3

accumulate mutations in signalling pathways critical to stem cell maintenance, cellular4

proliferation,andtumoursuppression [1,2]andtheycanevolve into invasiveCRCs thatmay5

ultimatelyspreadtodistantorgans.Thisprocesscalledmetastasis(seeBox1)occursinabout6

40-50%ofpatientsandconfersalowprobabilityofsurvival.7

In the clinic, patients are classified in 4 stages (see Glossary). Treatment involves8

aggressive surgical resection of the primary tumour,which cures a large proportion of early9

stagepatients.However, somestage II andmanystage IIIpatientswill relapse, frequently in10

theformofmetastasis,asaconsequenceoftumourcellsthatdisseminatedbeforeresection.11

Unliketheprimarytumour,metastasesarelessfrequentlyremovedbysurgeryunlesslimitedin12

number and extent [3]. Therefore patients that present with metastases at the time of13

diagnosis (stage IV)andthoseatperceivedriskof relapsereceivecytotoxicchemotherapy: in14

most cases a combination of folinic acid, 5-fluorouracil, and oxaliplatin or ironotecan. This15

strategyaimstokillhighlyproliferativecancercellsandhasbeenastapleinthetreatmentof16

solidcancersfordecades[4].AlthoughadjuvantchemotherapyisbeneficialtostageIIIpatients17

and has modest potential for improved survival in stage II [5, 6], it performs poorly in the18

metastaticsetting,almostinvariablygivingrisetodrugresistanceanddiseaseprogression.19

In second-line treatment, chemotherapy is increasingly combinedwith targeted therapy,20

designed to intervene with specific signalling pathways or cellular mechanisms [4]. A key21

exampleinCRCistheuseofantibodiestargetingtheepidermalgrowthfactor(EGF)receptor,22

4

oftenexploitedbycancercells to stimulateproliferation.Apart fromeventual resistance, the1

mainproblemwithtargetedtherapies isthevariableresponses inpatients,requiringabetter2

griponpredictivebiomarkers[7].Forinstance,mutationsthatactivatetheMAPK/ERKpathway3

downstream of EGFR, such as those frequently occurring in the KRAS andNRAS GTPases or4

BRAF,canrenderanti-EGFRtherapyineffective[8].5

A better understanding of advanced cancer may lead to more effective therapies for6

metastaticCRC,butisalsohighlyrelevanttoimprovethemanagementofearlierstagesofthe7

disease.ArguablythemostimportantquestionforstageI-IIIpatientsiswhetherornottotreat8

at all andwhich therapeutic strategywill bebeneficial toprevent recurrence.Assessmentof9

probability of relapse after therapy on the individual level remains a major challenge [9].10

Anatomicalandhistopathologicalfeaturesofthetumoursuchasextentofinvasion(T),number11

of lymph node metastases (N), bowel perforation or obstruction, the presence of12

lymphovascularinvasion,andapoorlydifferentiatedhistologyareusedtoidentifyCRCpatients13

atriskofrecurrence.However, theseparametersonlyholdamoderatepredictivepowerand14

donothelpselectoptimaltreatmentoptions.Furthermore,withtheexceptionofchromosomal15

or microsatellite instability (MSI) [10, 11] and mutations in the BRAF oncogene [12], no16

molecularfeatureisrobustlyassociatedwithprognosisandthereforeusedroutinelyinclinical17

practice[9,13].18

19

20

5

TranscriptomicsredefiningCRCclassification1

Toimprovepatientstratificationandidentifypotentialmoleculartargetsfortherapy,the2

field has generated large collections of transcriptomic datasets from tumour samples,which3

have enabled the identification of CRC subtypes based on distinctive global gene expression4

profiles[14-19]. Inanattempttoconsolidatethesedata,arecentmeta-analysisdividedCRCs5

into4definedconsensusmolecular subtypes (CMS), representinganMSI-like class (CMS1), a6

canonical WNT/MYC class (CMS2), a metabolically dysregulated class (CMS3), and a7

mesenchymalclass(CMS4)[20].Thelattersubtype,towhichabout1inevery4CRCsbelongs,is8

ofparticularinterestasitrelapseswithhigherfrequencythantheothers.9

Theelevatedexpressionofmesenchymalgenesinanepithelialcancerledtotheproposal10

that epithelial-to-mesenchymal transition (EMT) characterizes poor prognosis in CRC [21].11

However,thetranscriptomeofatumoursamplenotonlyreflectstheexpressionprogrammeof12

epithelial cancer cells but also the profile of mesenchymal cells present in the tumour13

microenvironment(TME).Indeed,ourgroupandthegroupofEnzoMedicorecentlydiscovered14

thatexpressionofmesenchymalgenesintranscriptomicCRCdataislargelycontributedbycells15

oftheTME,mainlybycancer-associatedfibroblasts(CAFs),ratherthanbycancercells[22,23].16

In agreement with these observations, CMS4 cancers show a higher degree of stromal17

infiltration than the other subtypes [20]. Furthermore, our analyses indicate that a large18

proportionof theseCAF-expressed genes are strongly associated to cancer relapse andpoor19

prognosisinCRCcohorts[22,23].Ofnote,thefactthattheexpressionofmesenchymalgenes,20

includingmanyEMTmasterregulatorssuchSNAI1,TWISTandZEB1,iscontributedbystromal21

6

cellsintranscriptomicprofilesdoesnotruleoutthepossibilitythatsubsetsoftumourcellsmay1

undergoEMT,particularlyatinvasionfronts.2

3

TreatingCancerasanEcosystem4

Focussing on the tumour stroma is not novel. At the end of the 19th century, English5

physician Stephen Paget – observing preferential patterns of metastatic dissemination to6

particularorgans–proposedthe‘seedandsoil’hypothesistoexplainthisphenomenonasthe7

combinationof the colonizing cancer cells’ adaptability anda congenialmicroenvironment in8

thedistantorgan[24].HesuggesteditwouldbeusefultostudythepropertiesoftheTME,as9

wellasitscrosstalkwithcancercells.Indeed,theroleoftheTMEbothattheoriginalsiteandin10

thedistantorganhasbecomeatopicofintenseinvestigationinrecentdecades[25].Similarto11

howintestinalcryptstemcellsareregulatedbyaspecializedstemcellmicroenvironmentcalled12

aniche,theTMEcanfostercancerstemcellsandthusdriverecurrenceandmetastasis(Box1)13

[26,27],aswellasprovidemechanismsfordrugresistance[28,29].Itisnowacceptedthatthe14

TMEisanactiveplayerintumourmalignizationandthatitprovidesafertilegroundtodelvefor15

therapeutictargets[30,31].16

The study of complex interactions between cancer cells and their environment inspires17

parallelswiththe interrelationshipsrecognized inanecosystem.Theapplicationofecological18

conceptstooncologymayhelpexplainwhyarigidfocusonepithelialcancercellsalonehasled19

toahighnumberoffailednewdrugsandclinicalstrategies.Justasecologyistheframeworkof20

choice for understanding themechanisms of organismal evolution through natural selection,21

ecologyworking on the level of cellular populations serves to clarify the dynamics of cancer22

evolution. This includes natural selection due to limited resources as well as the artificial23

7

selectionofsubclonesresistanttotherapy.Hence,thefitnessofacellularpopulation(orofa1

setof cancermutations) shouldbeconsidered in thecontextof its localenvironment,which2

consistsofcellular interactions,physicalparameters(oxygenlevels,pH,mechanicalpressure),3

andmoleculeslikeextracellularmatrixcomponents,growthfactorsandcytokines.4

Moreover,ecologicaltypesofinteractioncandescribevariouslevelsofcrosstalkintumour5

dynamics. Examples of competitive and cooperative interactions have been reported both6

betweendifferentepithelialsubclonesandbetweenepithelialcancercellsandthestroma[32-7

34].Additionally,theimmunesystemcanbeconvertedfromapredatortoanaccomplice(see8

Box2).Indeed,thestromalenvironmentisnotstaticandco-evolutionisacompellingconcept:9

parallel evolution of neoplastic epithelial cancer cells and cells in the microenvironment,10

subject to epigenetic and gene expression changes [35]. Furthermore, the stromal11

compartment contributes significantly to intratumoral heterogeneity, with important12

implications for disease progression and therapeutic responses [36]. Thus, cancer ecology13

integrates complex interactions between epithelial cancer cell populations and their14

environmenttopromiseadeeperunderstandingofthebiologyofcancer[37,38].15

16

ModulatingtheCRCecosystemfortherapeuticpurposes17

The compositionof the TMEduring colorectal tumorigenesis and in advanced cancers is18

subject to intensifying inquiry (for a recent overview see refs: [39, 40]) and is found to be19

different from normal intestinal stroma (see Figure 1, Key Figure). As recently reported, the20

majorityofgenesthatpredictcancerrecurrenceinCRCpatientsareexpressedinCAFsandnot21

bythecancercells[22,41].ThestrikingprognosticpoweroftheTMEraisestheopportunityto22

moreaccuratelyestimate the riskof relapseofanygiven stage I-IIIpatient. It alsopaves the23

8

wayforthedevelopmentofadditionallyrefinedmolecularclassificationsbasedonthecontents1

of the ecosystem. A relevant early example of such an ecological classification of CRC was2

establishedbyGalonandothers[42-44],whoobservedthatinfiltrationofparticularsubsetsof3

Tcellsinthetumourmasspredictslongerdiseasefreesurvivalintervalsaftertherapy.4

Moreover,novelCRCclassificationsbasedonthetypeandfeaturesofstromalcellswillbe5

key to stratifypatients for treatments that target theTME.Basedon the idea thatCRC cells6

dependonparticularstromalfactorstoeffectivelydisseminate,therapeuticsthatmodulatethe7

CRCecosystemmaybeeffectiveintreatingorpreventingmetastasis.Anadditionaladvantage8

of targeting the TME is its genetic stability, making drug resistance less likely to occur.9

Furthermore,thecommonalityofmostelementsinthestromaacrosscancertypessuggestsa10

broadapplicabilityofeffective therapies.Ahandfulof therapiesbasedon theaboveconcept11

have already been developed and are currently being applied. For instance, blood vessel12

formation and endothelial cells have been successfully targeted by anti-angiogenic therapy13

[45]. Indeed, a monoclonal antibody against vascular endothelial growth factor (VEGF, i.e.14

bevacizumab) inhibits angiogenesis and has been shown to improve survival for stage IV15

patients[46,47].Additionally,metronomicchemotherapy–aimingtoforestalldrugresistance16

and suppress neo-angiogenesis – has shown promise for prolonging survival and limiting17

metastasisinpreclinicalmodels,andisbeingtestedinclinicaltrials[48,49].18

AnotherprimeexampleoftherapydirectedtotheTMEis immunotherapy(Box2). Ithas19

beenproposedthatseveral(chemo-)therapiesthatdonotdirectlytargetimmunity,suchas5-20

fluorouracil and platinum compounds, in fact do rely on immune components. These21

9

conventional treatments are effective at least in part by reinforcing tumour1

immunosurveillance,especiallywhentheyinduceimmunogeniccancercelldeath[50].2

3

TGF-βasamainorganizerofthemetastaticCRCecosystem4

TGF-β has a dual role in CRC, with tumour suppressive functions in early stages and5

promotingdiseaseprogressioninadvanceddisease(seeBox3).Ourgroupdemonstratedthat6

elevatedTGF-βsignallingintheTMElinkstopoorprognosisinCRC[22,41].ThestromalTGF-β7

responseencodesageneprogrammethatincludesaplethoraofcytokines,growthfactorsand8

extracellularmatrixproteins,manyofwhichhavebeenshowntoplaykeyrolesduringdisease9

progressionandmetastasisinseveralcancertypes.Amongthesepotentialtherapeutictargets10

is interleukin-11(IL-11),acytokinesecretedbyTGF-β-stimulatedfibroblasts that inducespro-11

survival response inCRCcellsduringcancerprogressionandmetastatic colonization [41,51].12

The picture that emerges is that key functions required to complete the formation of13

metastasisareprovidedbytheTGF-β-activatedmicroenvironment(Figure1).14

Because disseminated tumour cells thrive on a TGF-β-activated environment, this15

dependencycouldbeexploitedintheclinicalsettingtoimprovepatienttreatment;ourgroup16

showedforthefirsttimethatpharmacologicalinhibitionofstromalTGF-βsignallingiseffective17

in blocking metastatic colonization in mice [22, 41]. Besides activating a pro-metastatic18

programme in CAFs [52], stromal TGF-β signalling induces a pro-tumorigenic phenotype in19

neutrophils and macrophages [53-55]. In addition, TGF-β directly suppresses the anticancer20

response of the adaptive immune system, including the deregulation of cytotoxic T21

lymphocytes [56-58].Also,neutralizingtheTGF-βpathway inTcells ledto immune-mediated22

10

eradicationoftumoursinamodelofmelanoma[59].ThatthedualroleofTGF-βsignallingin1

instructing immune cells and CAFs might be more related than is immediately obvious, is2

suggestedbyreportinwhichCAFswereshowntohaveimmunesuppressiveeffects[60,61].3

Thus, it will be interesting to assess the efficacy of pharmacological intervention in this4

pathwayon theCRC ecosystem. Several therapeutic strategies targeting TGF-β are in clinical5

development (see Box 4) [62-64]. This spurs the hope that if we can block TGF-β6

pharmacologically, we may prevent a vicious cycle of stroma activation, expression of pro-7

tumorigenic secreted factors, and the generation of an immunosuppressive environment8

(Figure 1). Indeed, TGF-β inhibition may work as (or synergize with other types of)9

immunotherapy(Box2).10

11

CaveatsaboutthedualroleofTGF-βsignallinginCRC12

DuetothetumoursuppressivefunctionofTGF-βsignallinginepithelialcomponentofCRC13

(see Box 3), elevated TGF-β levels necessary to instruct the poor prognosis-associated14

microenvironmentmightnotbecompatiblewithtumourgrowth.Wespeculatethatthelossof15

TGF-βresponseintheepithelialcomponentofthecancermayfacilitatetheelevationofTGF-β16

levels during tumour progression. A functional link between these two events has been17

observedinexperimentalmodels[65].InabiobankofCRCorganoids,independenceofTGF-β-18

mediatedgrowthinhibitionwasassociatedwithadvancedstagesofthedisease[66],whereas19

lossofSMAD4inepithelialCRCcellscorrelatedwithpoorprognosisinseveralstudies[67,68].20

Yet,abrogationofthecytostaticTGF-βresponseinCRCcellsisnotexclusivelyexplainedbyloss-21

of-function mutation in pathway components suggesting additional mechanisms of TGF-β22

11

resistance [66]. Also, a proportion of TGF-β-reactive CRCs may respond to the cytokine by1

undergoing EMT while bypassing the cytostatic effect. Apparently, this effect is relevant to2

trigger invasionofaparticular subsetofearly lesionsnamedsessile serratedadenomas [69],3

whichmightthereforebenefitfromanti-TGF-βtherapy.Conversely,itremainsunclearwhether4

anti-TGF-β therapy would be safe for patients with cancer cells that have an intact TGF-β5

pathway–eveniftheyrepresentasubclone–andarekeptincheckbyitscytostaticeffects.6

Although the TGF-βpathwayhas emerged as a powerful architect of thepro-metastatic7

poorprognosisTME, itseffecton individualstromalcell types ispleiotropicand incompletely8

charted. Accordingly, we propose that, to treat the cancer ecosystemwithmore integrated9

approaches, we need to study and understand its complexity in greater detail. As such,10

manipulatingamasterregulatorlikeTGF-βinsufficientlycomplexmodelsystemsofmetastatic11

CRCshouldnotonlyprovidetheneededrationaleforclinicaltranslation,butwouldadditionally12

giveusrelevantparameterstohelpmapthecrosstalkinthecancerecosystem.13

14

Needforbetterpreclinicalmodelsmovingforward15

To study the efficacy of above-mentioned and other stroma-directed therapies in the16

preclinicalsetting,thereisademandformorecomprehensive,predictivemodels.Althoughthe17

fieldhasmovedfromsubcutaneously injecting2D-culturedhumanCRCcell linestoexploiting18

3Dorganoidcultureandpatient-derivedxenografts(PDX)–approachesthatpromisetoretain19

tumour heterogeneity in genetics, architecture and drug responses [70-72] – these in vivo20

models lack an intact TME. This is due to a mismatch between some murine and human21

signallingmoleculesandbecauseoftheabsentorabrogatedimmunesystemrequiredforthe22

12

engraftment of human tumour tissue in murine hosts. To address these problems, mouse1

modelsarebeinggeneratedthatcorrectknownsignallingdeficitsaswellasfeaturehumanized2

immune system components, potentially leading to patient-specific immune environments.3

Theseadvancespromisetostronglyenhancecurrentmodelsforcancerbiologyandpreclinical,4

personalizedoncology.5

Analternativeapproach to study theCRCTMEandcancer immunity involvesgenetically6

modified mouse models (GEMMs), which feature an intact, immunocompetent TME yet7

typicallyfailtocaptureadvancedCRC.Thisleavesagap:totestcurrentideasanddiscoverand8

develop novel therapeutic strategies we are in need of CRC models that are humanlike in9

molecular characteristics, genetics and histopathology, and that are metastatic in a fully10

immunocompetentenvironment.Theidealscenariowouldincludeatransplantable(organoid)11

system in the commonlyusedC57BL/6 strain, so that it canbe leveragedacross awealthof12

existinggeneticmodelstodissectspecificfunctionsofstromalpathways.Furthermore,sucha13

systemwouldbecriticaltotestingnoveltherapiesthattargetthetumourecosystem.14

15

Concludingremarks16

The CRC oncology field has evolved in the past decades. From histological analyses to17

molecularsubtyping,fromcytotoxicchemotherapytopersonalizedcombinationswithtargeted18

therapy.Whathasnotchanged,unfortunately,isthatbesidessurgerythereisverylittlewecan19

dotoimproveoverallsurvivalforadvanceddisease.20

However, the paradigm shift that points to the TME as a critical factor in determining21

metastatic success opens the door to a more integrated understanding of the biology of22

13

metastasis. One that invites us to consider cancer as a complex ecosystem rather than as a1

tenacious weed. In this context, therapeutic targeting of stromal TGF-signalling might be2

expectedtoreprogrammethepoorprognosiscancerecosystem(Figure1),withthepotential3

to make metastatic lesions unsustainable or to prevent metastatic initiation altogether.We4

hopethatthisemergingconceptwillfindatimelytranslationtotheclinic.Acknowledgingthe5

many challenges ahead (see also Outstanding Questions), we need more sophisticated and6

predictive CRC models to show proof-of-principle results that may catalyse the effort of7

targetingtheTME.8

9

OutstandingQuestions10

• WhattriggerstheupregulationofTGF-βinsometumoursbutnotothers?11

• Whatcell type is (orwhichcell typesare) the sourceof secretedandactivatedTGF-β12

thatreprogramstheCRCecosystem?13

• AsnotallCRCshaveknownmutations in theTGF-βpathway,andsometumoursmay14

haveco-existingmutantandsensitivesubclones,towhatextentistheresponsivenessof15

epithelial cells to TGF-β (mainly associated to early disease stages) still relevant in16

advancedCRC?17

• Andwithinthiscontext,isTGF-βtherapysafeforallpatients?18

• Given thedependencyofmetastatic initiationonaTGF-β-activatedTME, is theTGF-β19

therapeuticwindowlimitedtoadjuvanttherapy–toinhibitmetastaticcolonization–or20

couldTGF-βtherapyalsotreatestablishedmetastases,e.g.asimmunotherapy?21

22

14

Box1.CRCmetastasis:biologyandtherapeuticscope1

Initial transformation of the colon epithelium and subsequent transitions through the2

adenoma-carcinoma sequence are thought to require successive genetic alterations in 43

signalling pathways:WNT, EGFR, TGF-β and P53 [1, 2]. In the healthy colonicmucosa, these4

signallingpathwaysregulatethebehaviourofintestinalstemcells(ISCs).Therefore,duringCRC5

progression,tumourcellsbecomeindependentfromthesecryptnichesignals[66].About40%6

of all CRCdisseminate to other organs.Metastasis can either bepresent duringdiagnosis or7

reveal itselfmonthsoryearsaftersurgeryasa recurrence. Inmanycases,distant recurrence8

targets the liver, at least in part because intestinal blood vessels drain into the portal vein,9

whichconnectstotheliver.Metastasisisamultistepprocessthatrequirescellstodetachfrom10

their epithelial neighbours, invade the submucosa, intravasateand survive in thevasculature11

(bloodorlymphaticvessels),extravasateandsurviveinanalienorgan,toeventuallyreinitiatea12

thrivingtumour[73,74].Thereismuchofthebiologyofmetastasisthatwedon’tunderstand,13

butfromwhatweknow,itisaveryharshandinefficientprocess[75,76].Arguably,theprocess14

selectsforthefiercestandmostresilientcancer(stem)cellsanddiscoveringwhatmakesthem15

successfulmightuncovertherapeutictargets.Importantly,themetastaticprocesshasnotbeen16

linked to specific genetic alterationswithin epithelial CRC cells [77]. Therefore, studies have17

shiftedfocusontotheroleoftheTMEandthemechanismbywhichthemetastaticcellexploits18

the TME for survival, immune evasion, and growth stimuli [78, 79]. Perhaps, success in19

metastasisisdefinedaswhocanbestremodeltheirmicroenvironment.20

21

15

Box2.Cancerinflammation,immunityandimmunotherapy1

Inflammation is a well-described risk factor in gastrointestinal tumorigenesis [80], and2

encompassesdisparatesignallingpathwaysbyavarietyofimmunecellsthatimpacteverystep3

of cancer progression andmetastasis [81]. Although the immune system can be a powerful4

tumour suppressor by the coordinated elimination of aberrant cells, tumours find a way to5

surviveinspiteofimmunosurveillance.Thereisevidenceforimmunoselectioninmanycancers,6

includingCRC–wherethenumberofobservedneoantigens(inremainingsubclones) is lower7

thanexpectedbasedonmutationrates–andcancercellscanacquireresistancetoimmunity8

bye.g.abrogatingantigenpresentation[82,83].9

Moreover,ithasbecomeincreasinglyclearthatinflammatorymechanismsintumourscan10

alsohelpdrivecancerprogression[81].Bystudyinghowprogressingcancerssuppresstheanti-11

tumour response and subvert immune regulation to foster immune suppressor cells and12

produce pro-tumorigenic factors that support cancer cells, the expansive field of cancer13

immunology has introduced a multitude of novel therapeutic strategies [84, 85]. On the14

conceptuallevel,manysuchtherapiesaimtoenhanceendogenousanticancerimmunitythatis15

somehowinsufficientordysfunctional[86].Toachievethiseffect,atherapycanreducecellular16

or molecular crosstalk mechanisms of immune suppression and/or increase T cell survival,17

proliferation,infiltration,andactivationintheTME[87].18

The promise of immuno-oncology has recently been demonstrated by long-lasting19

responsestoimmunecheckpointinhibitors.Theseprovedbeneficialinvarioustypesofcancer20

byunleashingTcells thatwerecrippledbysignallingeither from immunesuppressorcellsor21

cancercellshijackinganti-inflammatorymechanisms [88,89]. It is stillunclearwhether these22

16

therapieswillbroadlybenefitCRCpatients.However,onestudyonmismatchrepair-deficient1

(MSI) cancers, includingof the colon,did show therapeutic responses [90]. This supports the2

concept that tumour cells with higher frequency of neoantigens, such as in microsatellite3

instable cancers, are generally more vulnerable to T cell-mediated anti-tumour immunity,4

possiblyexplainingtheirrelativelygoodprognosis[91,92].Ifcurrentimmunotherapiesturnout5

nottohaveabroadapplicabilityassingleagentsinCRC,thereisstillhopethatacombination6

of treatments that both enhances intratumoral immune infiltration aswell as activationwill7

farebetter[93,94].8

9

Box3.Thedual,biphasicroleofTGF-βinCRC10

Thepro-metastaticeffectsofTGF-βsignallingonthetumourmicroenvironmentcanoccur11

independently of signalling in epithelial cancer cells, as many cancers silence the epithelial12

pathwaysoastoprogress.TGF-βsignallingisconsideredatumoursuppressorpathwayincolon13

carcinogenesisbecauseittriggersapotentcytostaticresponseinepithelialcells[22,65,95-97].14

Indeed,about40%ofCRCsdisplay loss-of-functionmutations inTGF-βpathwaycomponents,15

which are acquired around the adenoma carcinoma transition [2, 98]. An early clue for a16

positive role for TGF-β signalling indiseaseprogressionwas foundbyKawataand colleagues17

whoreportedthatserumlevelsofTGF-β1predictrelapseinCRCpatients[99].Later,ourgroup18

showedthatTGFB1,TGFB2andTGFB3mRNAlevelsassociatewithshorterdiseasefreesurvival19

intervalsaftertherapy instageI-IIIpatients[41].ElevatedTGFB1-3expressioncorrelateswith20

upregulationofTGF-βresponsegenesignatures(TBRS)incellsoftheTME,specificallyinTcells,21

macrophages, endothelial cells and most prominently CAFs. These TBRSs hold a striking22

17

predictivepowerforcancerrecurrenceinCRC[41]anddetectionbyimmunohistochemistryof1

severalproteinsupregulatedinTGF-β-activatedCAFsidentifiespatientsathighriskofrelapse2

[22].Furthermore,comparisonofthemolecularsubtypesrevealedthatthemainsignallingaxis3

deregulated in the stromal/mesenchymal subtype (CMS4) is the TGF-β pathway [20]. A4

significant proportion of the genes differentially expressed in CMS4 cancers that predict a5

higherprobabilityofrecurrenceareincludedintheTBRSsandcorrespondtogenesinducedby6

TGF-β in CAFs. The role of a TGF-β-activated ecosystem in disease progression has been7

confirmed in experimentalmodels of advanced CRC, where enforced TGF-β signalling in the8

TMEstronglyincreasesmetastaticburden[41].9

10

Box4.TGF-βtherapeuticsinclinicaldevelopment11

Several methods of targeting the TGF-β pathway have been described. These include12

antisense oligonucleotides to TGF-β ligand mRNA, small molecule inhibitors of the receptor13

kinasedomains,andmonoclonalantibodiesagainstligandsorreceptors.Inaddition,thereare14

vaccines and adoptive cell transfer strategies that target TGF-β signalling as part of their15

mechanism. For anoverviewof all (pre-) clinical strategies, see references [62-64].Here,we16

selectedTGF-βtargetingtherapiesthatarecurrentlyinclinicaltrialsforCRC(TableI).17

18

19

20

18

TableI.TGF-βtargetedtherapiesinclinicaltrialsforCRC1Type Drug MoA Setting(Phase) TrialIdentifierAntisenseoligos Trebedersen

(AP12009 AntisensePharma)

ASO to TGF-β2mRNA

Melanoma,pancreas,CRC(I)

NCT00844064

ReceptorinhibitorsKinaseinhibitors

Galunisertib(LY2157299EliLilly)

TβRI(ALK5)kinaseinhibitor

Rectalcancer(II)Study cancerimmunology in solidtumours(I)Checkpoint inhibitorcombination in solidtumours(I/II)

NCT02688712NTT02304419NCT02423343

TEW-7197,MedPactor

TβRI(ALK5)kinaseinhibitor

Solidtumours(I) NCT02160106

MAbs IMC-TR1(LY3022859EliLilly)

TβRII neutralizingMAb

Solidtumours(I) NCT01646203

PF-03446962(Pfizer)

ALK-1 neutralizingMAb

Metastatic CRC (I)Solidtumours(I)

NCT02116894NCT00557856

ImmunotherapyVaccines FANG vaccine

(Gradalis)shRNA againstfurin*

MetastaticCRC(II)

NCT01505166

TAGvaccine TGF-β2-antisense/GMCSF modifiedautologoustumourcells

Metastatic solidtumours(I)

NCT00684294

MoA:Modeofaction.ASO:antisenseoligo.ALK1:Activinreceptor-likekinase1(non-canonical2TGF-βreceptortype I inendothelialcells).*Aspartof itsMoA,autologoustumourcellscarry3bifunctionalshRNAblockingfurin,leadingtolowerTGF-β1andTGF-β2levels.45

6

7

19

Glossary1

Adjuvantchemotherapy:chemotherapygivenafterprimarytreatment(surgeryinCRC)to2

lowertheriskofthecancercomingback.3

Cancer evolution: the theory that within a tumour, accumulating mutations and/or4

localizedselectivepressureleadstothegenerationofsubclonesthatcompetewitheachother5

forlimitedresources.Furthermore,atthetimeofdiagnosisa(minority)subclonemayalready6

carryamutationthatrendersit(partially)resistanttochemotherapy.Boththisconceptandthe7

realization that this theory helps explain the unique nature of each individual tumourmake8

cancerevolutionahighlyrelevantconceptthatformsthebasisformoreintegrated(ecological)9

therapeuticstrategies.10

CRC staging: The American Joint Committee on Cancer (AJCC) has established a general11

cancerstagingprotocolcalledtheTNMStagingSystem.Tstandsfortumourandevaluatesthe12

original (primary) tumour in gradesof aggressiveness;N stands for lymphNode involvement13

andqualifiesthepresenceofcancercellsspreadingtonearby lymphnodes.Maddressesthe14

presence or absence of distantmetastasis (spreading of cancer to distant organs). Based on15

these3categories,patientsaregroupedintofourstagesindicatedbyRomannumerals(I,II,III16

and IV).While statistically representing significantly different risk groups for recurrence and17

cancer-relateddeath,thestagingsystemdoesnotaccuratelypredictonthelevelofindividual18

patients.19

Cytokines: large class of small secreted signalling proteins involved in cellular crosstalk,20

especiallyrelevantfortheorchestrationofimmuneresponsesandcancer.21

Ecology: the comprehensive study of the distribution, abundance and dynamics of22

organisms, their interactions with others and with their physical environment. It is most23

frequently applied on the organism level,where populations and their environment forman24

ecosystemandthelargestsuchsystemstudiedistheecosphereearth,yetitcanalsobeapplied25

to cells in a tissue or in a cancer. Central concepts in ecology include cooperation and26

competition,parasitismandpredation,andevolution.27

Immunosurveillance:theprocessbywhichmalignantcellsarekeptincheckbyanticancer28

immunity. Cancer immunity is understood to have an initial elimination phase before the29

20

balance is tipped and tumour cells start escaping from immunosurveillance. Between1

eliminationandescape,theremaybeaprolongedstateofequilibriumbetweencancergrowth2

andimmunity-relatedkilling.3

Relapse/Recurrence:thereturnofatumouraftersurgicalremovaloftheoriginaltumour4

(whichoftenentailedremovalofpartsor thewholeof thecolon).Thetumourcanreappear,5

oftenasadistantmetastasis.6

Metronomicchemotherapy:frequentorcontinuousadministrationoflownon-toxicdoses7

withno interruptions for longperiods toprevent resistanceandtargetbothepithelialcancer8

cellsandthegrowingtumourvasculature.9

MicroenvironmentorStroma:supportiveconnectivetissuewithstructuralandfunctional10

roles both during homeostasis and in disruption such as wounding or disease. The stroma11

includesfibroblasts,bloodandlymphaticvessels,immunecells,andtheextracellularmatrix.In12

thecontextofcancer,thestromaisincreasinglyunderstoodasacomplex,dynamicentitythat13

istransformedbyandco-evolveswithcancercellsandcandrivemalignantprogression.14

Microsatellite instability (MSI): due to a deficiency in DNAmismatch-repair genes,MSI15

patients accumulate spontaneous errors in regions of their genome that are repetitive and16

therebychallengingtoreplicate.17

18

21

Figurelegend1

Figure 1 (Key Figure). The CRC ecosystem and the progression-driving role of TGF-β.2

Besides fibroblasts (CAF), endothelial cells, pericytes and mesenchymal stem cells (MSC),3

colorectal cancers are infiltrated by a variety of innate and adaptive immune cells, including4

lymphocytes (T cells, B cells, natural killer cells), monocytes, macrophages, dendritic cells,5

granulocytes (neutrophils, basophils, eosinophils, and mast cells), and myeloid-derived6

suppressorcells(MDSC).7

TGF-β has emerged as a central architect that drivesmalignization of the cancer ecosystem,8

activating stromal cells towards pro-tumorigenic differentiation, inducing the expression of9

secretedfactors,andfacilitatingtheaccumulationofimmunesuppressiveanddrugresistance10

functionalities.Effectiveecological treatmentmight includeacombinationofTGF-β inhibition11

withotherstroma-directedtherapies,suchaschemotherapyandtargeted(immuno-)therapy.12

13

Acknowledgements14

Workintheauthors’labhasbeensupportedbygrantSAF2014-53784_RandaJuandela15

Cierva fellowship (D.V.F.T) from the Spanish Ministry of Economy and Competitiveness16

(MINECO),andbytheDr.JosefSteinerFoundation. IRBBarcelona istherecipientofaSevero17

OchoaAwardofExcellencefromtheMINECO.WewouldliketothankElenaSanchoforcritical18

readingofthismanuscript,aswellastheBatllelabforfruitfuldiscussions.19

20

22

References1

21Fearon,E.R.andVogelstein,B.(1990)Ageneticmodelforcolorectaltumorigenesis.Cell61,3759-76742Fearon,E.R.(2011)Moleculargeneticsofcolorectalcancer.AnnuRevPathol6,479-50753Kopetz,S.,etal. (2009) Improvedsurvival inmetastaticcolorectalcancer isassociatedwith6adoptionofhepaticresectionandimprovedchemotherapy.JClinOncol27,3677-368374VanCutsem,E.,etal.(2014)Metastaticcolorectalcancer:ESMOClinicalPracticeGuidelines8fordiagnosis,treatmentandfollow-up.AnnOncol25Suppl3,iii1-995Gray,R.,etal.(2007)Adjuvantchemotherapyversusobservationinpatientswithcolorectal10cancer:arandomisedstudy.Lancet370,2020-2029116Andre,T.,etal.(2009)Improvedoverallsurvivalwithoxaliplatin,fluorouracil,andleucovorin12asadjuvanttreatmentinstageIIorIIIcoloncancerintheMOSAICtrial.JClinOncol27,3109-133116147Sinicrope,F.A.,etal.(2016)MolecularBiomarkersinthePersonalizedTreatmentofColorectal15Cancer.ClinGastroenterolHepatol14,651-658168Bertotti,A.,etal.(2015)ThegenomiclandscapeofresponsetoEGFRblockadeincolorectal17cancer.Nature526,263-267189Labianca,R.,etal.(2013)Earlycoloncancer:ESMOClinicalPracticeGuidelinesfordiagnosis,19treatmentandfollow-up.AnnOncol24Suppl6,vi64-722010Carethers,J.M.andJung,B.H.(2015)GeneticsandGeneticBiomarkersinSporadicColorectal21Cancer.Gastroenterology149,1177-1190e11732211 Mouradov, D., et al. (2013) Survival in stage II/III colorectal cancer is independently23predictedbychromosomalandmicrosatellite instability,butnotbyspecificdrivermutations.24AmJGastroenterol108,1785-17932512 Samowitz,W.S., et al. (2005) Poor survival associated with the BRAF V600Emutation in26microsatellite-stablecoloncancers.CancerRes65,6063-60692713 Jass, J.R. (2007) Classification of colorectal cancer based on correlation of clinical,28morphologicalandmolecularfeatures.Histopathology50,113-1302914DeSousaEMelo,F.,etal. (2013)Poor-prognosiscoloncancer isdefinedbyamolecularly30distinctsubtypeanddevelopsfromserratedprecursorlesions.NatMed19,614-6183115 Marisa, L., et al. (2013) Gene expression classification of colon cancer into molecular32subtypes:characterization,validation,andprognosticvalue.PLoSMed10,e10014533316 Sadanandam, A., et al. (2013) A colorectal cancer classification system that associates34cellularphenotypeandresponsestotherapy.NatMed19,619-6253517 Roepman, P., et al. (2014) Colorectal cancer intrinsic subtypes predict chemotherapy36benefit,deficientmismatchrepairandepithelial-to-mesenchymal transition. Int JCancer134,37552-5623818 Budinska, E., et al. (2013) Gene expression patterns unveil a new level of molecular39heterogeneityincolorectalcancer.JPathol231,63-764019Schlicker,A.,etal.(2012)Subtypesofprimarycolorectaltumorscorrelatewithresponseto41targetedtreatmentincolorectalcelllines.BMCMedGenomics5,6642

23

20Guinney, J.,etal. (2015)Theconsensusmolecularsubtypesofcolorectalcancer.NatMed121,1350-1356221Kalluri,R.andWeinberg,R.A.(2009)Thebasicsofepithelial-mesenchymaltransition.JClin3Invest119,1420-1428422 Calon, A., et al. (2015) Stromal gene expression defines poor-prognosis subtypes in5colorectalcancer.NatGenet47,320-329623 Isella, C., et al. (2015) Stromal contribution to the colorectal cancer transcriptome. Nat7Genet47,312-319824Paget,S.(1889)TheDistributionofSecondaryGrotwhinCanceroftheBreast.Lancet133,9571-5731025 Fidler, I.J. (2003) The pathogenesis of cancer metastasis: the 'seed and soil' hypothesis11revisited.NatRevCancer3,453-4581226 Biswas, S., et al. (2015) Microenvironmental control of stem cell fate in intestinal13homeostasisanddisease.JPathol237,135-1451427 Medema, J.P. and Vermeulen, L. (2011) Microenvironmental regulation of stem cells in15intestinalhomeostasisandcancer.Nature474,318-3261628 Tredan, O., et al. (2007) Drug resistance and the solid tumor microenvironment. J Natl17CancerInst99,1441-14541829Klemm,F.andJoyce,J.A.(2015)Microenvironmentalregulationoftherapeuticresponsein19cancer.TrendsCellBiol25,198-2132030 Fang, H. and Declerck, Y.A. (2013) Targeting the tumor microenvironment: from21understandingpathwaystoeffectiveclinicaltrials.CancerRes73,4965-49772231Quail,D.F.andJoyce,J.A. (2013)Microenvironmentalregulationoftumorprogressionand23metastasis.NatMed19,1423-14372432Hanahan,D.andCoussens,L.M.(2012)Accessoriestothecrime:functionsofcellsrecruited25tothetumormicroenvironment.CancerCell21,309-3222633 Marusyk, A., et al. (2014) Non-cell-autonomous driving of tumour growth supports sub-27clonalheterogeneity.Nature514,54-582834Cleary,A.S.,etal.(2014)Tumourcellheterogeneitymaintainedbycooperatingsubclonesin29Wnt-drivenmammarycancers.Nature508,113-1173035 Polyak, K., et al. (2009) Co-evolution of tumor cells and their microenvironment. Trends31Genet25,30-383236 Junttila, M.R. and de Sauvage, F.J. (2013) Influence of tumour micro-environment33heterogeneityontherapeuticresponse.Nature501,346-3543437Merlo,L.M.,etal.(2006)Cancerasanevolutionaryandecologicalprocess.NatRevCancer356,924-9353638Korolev,K.S.,etal.(2014)Turningecologyandevolutionagainstcancer.NatRevCancer14,37371-3803839Peddareddigari,V.G.,etal.(2010)Thetumormicroenvironmentincolorectalcarcinogenesis.39CancerMicroenviron3,149-1664040Quante,M., et al. (2013) The gastrointestinal tumormicroenvironment.Gastroenterology41145,63-784241Calon,A., et al. (2012)Dependencyof colorectal cancerona TGF-beta-drivenprogram in43stromalcellsformetastasisinitiation.CancerCell22,571-58444

24

42Naito,Y.,etal.(1998)CD8+Tcellsinfiltratedwithincancercellnestsasaprognosticfactor1inhumancolorectalcancer.CancerRes58,3491-3494243Galon,J.,etal.(2006)Type,density,andlocationofimmunecellswithinhumancolorectal3tumorspredictclinicaloutcome.Science313,1960-1964444Galon,J.,etal.(2014)Towardstheintroductionofthe'Immunoscore'intheclassificationof5malignanttumours.JPathol232,199-209645 Carmeliet, P. and Jain, R.K. (2011) Molecular mechanisms and clinical applications of7angiogenesis.Nature473,298-307846Mathonnet,M.,etal.(2014)Hallmarksincolorectalcancer:angiogenesisandcancerstem-9likecells.WorldJGastroenterol20,4189-41961047 Hurwitz, H., et al. (2004) Bevacizumab plus irinotecan, fluorouracil, and leucovorin for11metastaticcolorectalcancer.NEnglJMed350,2335-23421248 Hackl, C., et al. (2013) Metronomic oral topotecan prolongs survival and reduces liver13metastasis in improvedpreclinicalorthotopicandadjuvant therapycoloncancermodels.Gut1462,259-2711549Pasquier,E.,etal.(2010)Metronomicchemotherapy:newrationalefornewdirections.Nat16RevClinOncol7,455-4651750 Zitvogel, L., et al. (2013) Mechanism of action of conventional and targeted anticancer18therapies:reinstatingimmunosurveillance.Immunity39,74-881951 Putoczki, T.L., et al. (2013) Interleukin-11 is the dominant IL-6 family cytokine during20gastrointestinaltumorigenesisandcanbetargetedtherapeutically.CancerCell24,257-2712152 Calon, A., et al. (2014) TGF-beta in CAF-mediated tumor growth and metastasis. Semin22CancerBiol25,15-222353Fridlender,Z.G.,etal.(2009)Polarizationoftumor-associatedneutrophilphenotypebyTGF-24beta:"N1"versus"N2"TAN.CancerCell16,183-1942554 Gong, D., et al. (2012) TGFbeta signaling plays a critical role in promoting alternative26macrophageactivation.BMCImmunol13,312755 Pickup,M., et al. (2013) The roles of TGFbeta in the tumourmicroenvironment.Nat Rev28Cancer13,788-7992956 Thomas, D.A. andMassague, J. (2005) TGF-beta directly targets cytotoxic T cell functions30duringtumorevasionofimmunesurveillance.CancerCell8,369-3803157Yang,L.,etal.(2010)TGF-betaandimmunecells:animportantregulatoryaxisinthetumor32microenvironmentandprogression.TrendsImmunol31,220-2273358Chen,M.L., etal. (2005)RegulatoryT cells suppress tumor-specificCD8T cell cytotoxicity34throughTGF-betasignalsinvivo.ProcNatlAcadSciUSA102,419-4243559 Gorelik, L. and Flavell, R.A. (2001) Immune-mediated eradication of tumors through the36blockadeoftransforminggrowthfactor-betasignalinginTcells.NatMed7,1118-11223760 Kraman,M., et al. (2010) Suppression of antitumor immunity by stromal cells expressing38fibroblastactivationprotein-alpha.Science330,827-8303961 Feig, C., et al. (2013) Targeting CXCL12 from FAP-expressing carcinoma-associated40fibroblastssynergizeswithanti-PD-L1immunotherapyinpancreaticcancer.ProcNatlAcadSci41USA110,20212-202174262Akhurst,R.J.andHata,A. (2012)Targeting theTGFbetasignallingpathway indisease.Nat43RevDrugDiscov11,790-81144

25

63 Neuzillet, C., et al. (2015) Targeting the TGFbeta pathway for cancer therapy.Pharmacol1Ther147,22-31264 Smith, A.L., et al. (2012)Molecular pathways: targeting the TGF-beta pathway for cancer3therapy.ClinCancerRes18,4514-4521465Munoz, N.M., et al. (2006) Transforming growth factor beta receptor type II inactivation5inducesthemalignanttransformationofintestinalneoplasmsinitiatedbyApcmutation.Cancer6Res66,9837-9844766Fujii,M.,etal.(2016)AColorectalTumorOrganoidLibraryDemonstratesProgressiveLossof8NicheFactorRequirementsduringTumorigenesis.CellStemCell18,827-838967Yan,P.,etal.(2016)ReducedExpressionofSMAD4IsAssociatedwithPoorSurvivalinColon10Cancer.ClinCancerRes22,3037-30471168Alazzouzi,H.,etal. (2005)SMAD4asaprognosticmarker incolorectalcancer.ClinCancer12Res11,2606-26111369Fessler,E.,etal. (2016)TGFbetasignalingdirectsserratedadenomas to themesenchymal14colorectalcancersubtype.EMBOMolMed8,745-7601570Jung,P.,etal.(2011)Isolationandinvitroexpansionofhumancolonicstemcells.NatMed1617,1225-12271771 Sato, T., et al. (2009) Single Lgr5 stemcells build crypt-villus structures in vitrowithout a18mesenchymalniche.Nature459,262-2651972 Siolas, D. andHannon,G.J. (2013) Patient-derived tumor xenografts: transforming clinical20samplesintomousemodels.CancerRes73,5315-53192173Chambers,A.F., etal. (2002)Disseminationandgrowthof cancer cells inmetastatic sites.22NatRevCancer2,563-5722374Massague, J.andObenauf,A.C. (2016)Metastaticcolonizationbycirculating tumourcells.24Nature529,298-3062575Luzzi,K.J.,etal.(1998)Multistepnatureofmetastaticinefficiency:dormancyofsolitarycells26aftersuccessfulextravasationand limitedsurvivalofearlymicrometastases.AmJPathol153,27865-8732876Chaffer,C.L.andWeinberg,R.A.(2011)Aperspectiveoncancercellmetastasis.Science331,291559-15643077Jones,S.,etal.(2008)Comparativelesionsequencingprovidesinsightsintotumorevolution.31ProcNatlAcadSciUSA105,4283-42883278 Mlecnik, B., et al. (2016) The tumor microenvironment and Immunoscore are critical33determinantsofdisseminationtodistantmetastasis.SciTranslMed8,327ra3263479Steeg,P.S.(2016)Targetingmetastasis.NatRevCancer16,201-2183580 Terzic, J., et al. (2010) Inflammation and colon cancer.Gastroenterology 138, 2101-211436e21053781Grivennikov,S.I.,etal.(2010)Immunity,inflammation,andcancer.Cell140,883-8993882Rooney,M.S.,etal.(2015)Molecularandgeneticpropertiesoftumorsassociatedwithlocal39immunecytolyticactivity.Cell160,48-614083Schumacher,T.N.andSchreiber,R.D.(2015)Neoantigensincancerimmunotherapy.Science41348,69-744284Galluzzi,L.,etal.(2014)Classificationofcurrentanticancerimmunotherapies.Oncotarget5,4312472-1250844

26

85 Finn, O.J. (2012) Immuno-oncology: understanding the function and dysfunction of the1immunesystemincancer.AnnOncol23Suppl8,viii6-9286Palucka,A.K.andCoussens,L.M.(2016)TheBasisofOncoimmunology.Cell164,1233-1247387 Joyce, J.A. and Fearon, D.T. (2015) T cell exclusion, immune privilege, and the tumor4microenvironment.Science348,74-80588 Pardoll, D.M. (2012) Theblockadeof immune checkpoints in cancer immunotherapy.Nat6RevCancer12,252-264789Topalian,S.L.,etal.(2015)Immunecheckpointblockade:acommondenominatorapproach8tocancertherapy.CancerCell27,450-461990 Le,D.T., etal. (2015)PD-1Blockade inTumorswithMismatch-RepairDeficiency.NEngl J10Med372,2509-25201191Llosa,N.J.,etal. (2015)Thevigorous immunemicroenvironmentofmicrosatellite instable12coloncancerisbalancedbymultiplecounter-inhibitorycheckpoints.CancerDiscov5,43-511392Kloor,M.,etal. (2010) Immuneevasionofmicrosatelliteunstablecolorectalcancers. Int J14Cancer127,1001-10101593Sharma,P.andAllison,J.P.(2015) Immunecheckpointtargeting incancertherapy:toward16combinationstrategieswithcurativepotential.Cell161,205-2141794Vanneman,M.andDranoff,G.(2012)Combiningimmunotherapyandtargetedtherapiesin18cancertreatment.NatRevCancer12,237-2511995Massague,J.(2008)TGFbetainCancer.Cell134,215-2302096Markowitz,S.,etal.(1995)InactivationofthetypeIITGF-betareceptorincoloncancercells21withmicrosatelliteinstability.Science268,1336-13382297 Takaku, K., et al. (1998) Intestinal tumorigenesis in compoundmutantmiceof bothDpc423(Smad4)andApcgenes.Cell92,645-6562498 The Cancer Genome Atlas Network (2012) Comprehensive molecular characterization of25humancolonandrectalcancer.Nature487,330-3372699Tsushima,H., etal. (2001)Circulating transforminggrowth factorbeta1asapredictorof27livermetastasisafterresectionincolorectalcancer.ClinCancerRes7,1258-1262282930

TGF-β inhibitionImmunotherapy

Stromal therapy

Cancer activates Stroma

TME drives Immune Evasion & Drug Resistance

Stroma secretesProtumorigenic factors

EpithelialCancer cell

Lymphoctes

MonocyteMDSC Macrophage

Granulocyte

Dendritic Cell

MSC Extracellular Matrix

Endothelial cell Pericyte

CAF