proteome proﬁling of cancer-associated fibroblasts...

Imaging, Diagnosis, Prognosis

Proteome Profiling of Cancer-Associated FibroblastsIdentifies Novel Proinflammatory Signatures and PrognosticMarkers for Colorectal Cancer

Sofia Torres1, Rub�en A. Bartolom�e1, Marta Mendes1, Rodrigo Barderas1, M. Jes�us Fernandez-Ace~nero3,Alberto Pel�aez-García1, Cristina Pe~na4, María Lopez-Lucendo2, Roi Villar-V�azquez1,Antonio García de Herreros5, Felix Bonilla4, and J. Ignacio Casal1

AbstractPurpose: Cancer-associated fibroblasts (CAF) are essential components of the stroma that play a critical

role in cancer progression. This study aimed to identify novel CAFs markers that might contribute to the

invasion and the prognosis of colorectal cancer.

Experimental Design: The azoxymethane/dextran sodium sulfate mouse model of sporadic colon

cancer represents an adequate source for the isolation of CAFs and normal fibroblasts. By using the explants

technique,wepurifiedCAFs andnormal fibroblasts fromcolon tissues.Whole-cell extracts and supernatants

were subjected to in-depth quantitative proteomic analysis by tandem mass spectrometry. Further valida-

tions of upregulated proteins inCAFswere carried out by chemokinemicroarray and immunohistochemical

analyses of mouse and human tissues.

Results: Using a fold-change of 1.4 or more, we found 132 and 125 differentially expressed proteins in

whole-cell extracts and supernatants, respectively. We found CAFs-associated proinflammatory and des-

moplastic signatures. The proinflammatory signature was composed of several cytokines. Among them,

CCL2 and CCL8 caused an increase in migration and invasion of colorectal cancer KM12 cells. The

desmoplastic signature was composed of 30 secreted proteins. Inmouse and human samples, expression of

LTBP2, CDH11, OLFML3, and, particularly, FSTL1 was significantly increased in the tumoral stroma,

without significant expression in the cancer epithelial cells. The combination of CALU and CDH11 stromal

expression showed a significant association with disease-free survival and poor prognosis.

Conclusion: We have identified LTBP2, CDH11, OLFML3, and FSTL1 as selective biomarkers of cancer

stroma, and CALU and CDH11 as candidate stromal biomarkers of prognostic significance in colon cancer.

Clin Cancer Res; 19(21); 6006–19. �2013 AACR.

IntroductionThe tumor stroma comprisesmost of the cancermass and

is mainly composed of fibroblasts and endothelial cells,

although it also contains infiltrating immune cells andpericytes (1). Stroma nurtures cancer cells and facilitatestumor development and invasion. Clinical and experimen-tal data support the hypothesis that tumor stromapromotesinvasion and cancer metastasis (2). Within stroma, fibro-blasts are key components for cancer progression. Stromalfibroblasts are called activated fibroblasts, myofibroblasts,or cancer-associated fibroblasts (CAF), and they acquire aparticular phenotype similar to fibroblasts present in skinwounds. Carcinoma progression is associated with anincrease in the production of fibrosis, known as desmopla-sia, similar to that present in wound healing (3). CAFsrespond to profibrotic and promigratory factors, such asTGF-b, platelet-derived growth factor (PDGF), hepatocytegrowth factor (HGF), or fibroblast growth factor 2 (FGF2),thereby promoting cancer progression. They are character-ized by increased expression of myofibroblastic markerssuch as a-smooth muscle actin (a-SMA), FSP1, or prolyl-4-hydroxilase (4, 5). Cancer fibroblasts proliferate more thantheir normal counterparts and secrete more constituentsof the extracellular matrix (ECM) and ECM-degrading

Authors' Affiliations: 1Department of Cellular and Molecular Medicine;2Proteomics Core Facility, Centro de Investigaciones Biol�ogicas (CIB-CSIC); 3Department of Pathology, Fundaci�on Jim�enez Díaz; 4Departmentof Oncology, Hospital Puerta de Hierro Majadahonda, Madrid; and 5IMIM-Hospital del Mar, Barcelona, Spain

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

S. Torres and R.A. Bartolom�e contributed equally to this work.

Current address for R. Barderas: Department of Biochemistry and Molec-ular Biology, I. Universidad Complutense, Madrid, Spain.

Corresponding Author: J. Ignacio Casal, Department of Cellular andMolecular Medicine, Centro de Investigaciones Biol�ogicas (CIB-CSIC),Ramiro de Maeztu, 9, 28040 Madrid, Spain. Phone: 34-918373112; Fax:34-915360432; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-13-1130

�2013 American Association for Cancer Research.

ClinicalCancer

Research

Clin Cancer Res; 19(21) November 1, 20136006

on November 20, 2018. © 2013 American Association for Cancer Research.clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst September 11, 2013; DOI: 10.1158/1078-0432.CCR-13-1130

http://clincancerres.aacrjournals.org/

proteases (6). In addition, CAFs can mediate inflammationand angiogenesis, as reported in skin cancer (7). Despite thesignificant number ofmarkers and secreted proteins alreadyassociated with activated fibroblasts, the study of theircontribution to tumor growth and invasion would benefitof additional markers for cell selection, prognosis, andinvasion prediction (8).With regard to human colon adenocarcinomas, CAFs

synthesize ECM components such as fibronectin, tenascin,collagens types I, III, IV, V, and XII, and proteoglycans(biglycan, fibromodulin, perlecan, and versican; refs. 9,10). In addition, they contribute to the formation of base-mentmembranes by secreting collagen type IV and laminin.For a more comprehensive molecular analysis, we proposethe use of sensitive proteomic techniques combined with amouse model of colon cancer, cell isolation, and massspectrometry (MS). We have selected the well-establishedmurine model of colitis-associated cancer (CAC) based onthe use of azoxymethane (AOM) and dextrane sodiumsulfate (DSS). Although the model does not progress tometastasis (11), it mimics quite well many steps in cancerprogressionof sporadic colorectal cancer. In fact, thismodelhas been very useful for the elucidation of the role of TNF-a,interleukin (IL)-6, NF-kB, and other molecules in the ini-tiation and progression of inflammation-associated cancer(see ref. 12 for a review).Fibroblast characterization has been largely complicated

by the necessity to isolate primary fibroblasts from thecolonic tissue. In other publications, the characterizationof CAFs was carried out by coculturing normal fibroblastsand colon cancer cell lines (10), cell sorting (7), or cell

immortalization (13). We propose the use of direct intes-tinal explants in culture for fibroblast isolation andproteomic characterization (14). This strategy stronglyreduces sample complexity and heterogeneity. The useof an inbred mouse model reduces biologic variation dueto genetic and environmental heterogeneity. Becauseprimary fibroblasts duplicate only a few times beforeentering senescence, we discarded metabolic labeling forquantification. Therefore, iTRAQ (isobaric tag for relativeand absolute quantification) was preferred because of itsreproducibility and reliability (15).

Here, we compared the protein component of thewhole-cell extracts and conditioned medium of primaryCAFs and normal fibroblasts (NF) isolated from AOM/DSS–induced sporadic colorectal cancer mice or controls,respectively. We identified 132 and 125 proteins deregu-lated in whole-cell extracts and conditioned medium,respectively, of CAFs versus normal fibroblasts. In silicostudies demonstrated a predominant association of upre-gulated proteins to deposition of ECM, wound healing,and inflammation. Proinflammatory and desmoplasticsignatures, specific for colon cancer, were defined forCAFs-associated proteins. A number of proteins wereidentified as promising tumor-associated stromal prog-nostic markers. Among these proteins, high expression ofCALU and CDH11 was associated to poor survival inhuman cancer samples.

Materials and MethodsAOM/DSS–induced colon cancer model

All mouse studies were performed under the approval ofthe Animal Ethics Committee of the Spanish ResearchCouncil (CSIC, Madrid, Spain). Animals were housedunder pathogen-free conditions, andwere given autoclavedfood and water ad libitum. For the AOM/DSSmodel, FVB/Nmice (4- to 6-week-old) were weighed and given a singleintraperitoneal injection of azoxymethane (10 mg/kg) orvehicle (PBS). Five days later, animals received either 2.5%DSS or normal drinking water. Chronic colitis-derivedcolon cancer was induced after three cycles of DSS treat-ment, which consisted of 5 days with 2.5%DSS followed by16 days with normal water. At the end of the protocol,animals were sacrificed and distal colons were longitudi-nally cut, rinsed twice with ice-cold PBS, and cut into smallpieces. Pieces of intestine were either cultured to isolatefibroblasts or fixed in 10% buffered formalin overnight forimmunohistochemical analysis.

Cell line cultureKM12C and KM12SM human colon cancer cells were

obtained from the laboratory of I. Fidler (MD AndersonCancer Center, Houston, TX). A large batch of workingaliquots of KM12 cells was stored in liquid nitrogen. Foreach experiment, cells were thawed and kept in culture fora maximum of 10 passages in Dulbecco’s modified Eaglemedium (DMEM; Gibco-Life Technologies) containing10% fetal calf serum (FCS) and antibiotics at 37�C. Cellswere tested forMycoplasma but not authenticated, because

Translational RelevanceOur knowledge of cancer-associated stroma is rather

limited in comparisonwith that of cancer epithelial cells.Stroma constitutes the cancer microenvironment thatnurtures cancer cells and facilitates invasion and metas-tasis. Moreover, stromal myofibroblasts have beenassociated with recurrence and survival. Novel biomar-kers are required for the isolation and characterizationof colorectal cancer–associated fibroblasts (CAF) toimprove overall survival and recurrence prediction incolorectal cancer. We, here, describe a strategy based onan in-depth proteome profiling of CAFs obtained from amouse model of human sporadic colorectal cancer. Theuse of this model facilitates the isolation of pure popula-tions of CAFs, avoiding other cell contaminants. Thevalidity of this strategy is supported by the identificationof multiple stromal biomarkers previously described inhumans. In addition, we report a collection of novelstromal biomarkers aswell as comprehensive proinflam-matory and desmoplastic CAFs signatures for colorectalcancer. The value and relevance of these novel stromalbiomarkers for colon cancer prognosis and survival werevalidated in human samples.

Molecular Signatures of Colon Cancer–Associated Fibroblasts

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 6007




these lines were originally isolated in Dr Fidler’s labora-tory. CT26 murine colon cancer cell line was obtaineddirectly from the American Type Culture Collection(which authenticated the cell line by short-tandem repeatprofiling), and passaged for fewer than 6 months afterreceipt or resuscitation according to the manufacturer’sinstructions.

Fibroblast isolation and cultureCAFs and normal fibroblasts cultures were established

from colonic tissues from AOM/DSS–treated mice (n¼ 16)and vehicle-treated (n ¼ 16) by the explant technique(14, 16). Tissues were cut into 2 to 3 mm fragments andplanted in fibroblast growthmedium-2 (Lonza) containing2% FCS and 1% penicillin/streptomycin/amphotericin B(Invitrogen) at 37�C in a 5% CO2-humidified atmosphere.Normal fibroblasts and CAFs grew around the explants andwere cultured for approximately 3 weeks. Then, tissue frag-ments were removed, and cells were trypsinized and seededat 70% to 90% confluence and cultured in complete medi-um. When cells reached 95% to 100% confluence (�1.7–2.0� 104 cells/cm2), theywere washedwith PBS, incubatedwith serum-free medium for 1 hour, washed again, andincubated for 24 hours in serum-free medium. Then, cellsand conditioned medium were collected, pooled in twodifferent batches for normal fibroblasts and CAFs, andmixed for performing the different experiments. We recov-ered 7.0 and 13.5 � 106 cells from normal fibroblasts andCAFs cultures, respectively.

Sample preparation and iTRAQ labelingFor a detailed description of sample preparation, see

Supplementary Methods. iTRAQ labeling was carried outusing iTRAQ Reagent 4-Plex kit (AB SCIEX), according tothe manufacturer’s instructions but with minor modifica-tions. Briefly, 100 mg of protein fromwhole-cell lysate or 50mg of protein from conditioned medium from normalfibroblasts and CAFs were digested with 5 mg trypsin (Pro-mega) overnight at 37�C. Peptide mixtures were labeledwith iTRAQ reagents: 114 and 116 for normal fibroblastsand 115 and 117 for CAFs as illustrated in SupplementaryFig. S1. To remove interfering substances, strong cationexchange (SCX) chromatography was carried out using aResource S column (GE Healthcare). To adjust the pHbetween 2.5 and 3.3, the sample mixture was diluted with15 volume of 10 mmol/L KH2PO4 pH 3.0 containing 25%acetonitrile (ACN). Peptides were eluted in 1 mL of elutionbuffer (25% ACN, 350 mmol/L KCl, 10 mmol/L KH2PO4,pH 3.0). Then, peptides were desalted with Sep-Pak C18cartridges (Waters), dried, and reconstituted in OFFGELsolution (Agilent Technologies). All chemicals were obt-ained from Sigma-Aldrich.

Peptide fractionation and MS analysisPeptides were recovered in 12 fractions by isoelectric

focusing using a low-resolution strip (pH 3–10) in a 3100OFF-GEL fractionator (Agilent Technologies). Five micro-liters of 25% trifluoroacetic acid (TFA) were added to each

fraction, and desalting was performed with Zip-Tip. Pep-tide mixtures were vacuum-dried and reconstituted with 6mL 0.1% formic acid. Then, peptides were trapped onto aprecolumn C18-A1 ASY-Column (2 cm, ID100 mm, 5mm;Thermo Scientific) and run with a linear gradient of 2%to 35% ACN in 0.1% aqueous solution of formic acid.The gradient was performed over 180 minutes using anEasy-nLC (Proxeon) at a flow-rate of 300 nL/min onto aBiosphere C18 column (75 mm, 16 cm, and 3 mm;NanoSeparations). Then, peptides were scanned and frag-mented with a linear ion trap-Orbitrap Velos (ThermoScientific). The Orbitrap Velos was operated in a data-dependent mode to automatically switch between MS andtandemmass spectrometry (MS–MS). Survey full-scan MSspectra were acquired from m/z 400 to 1,600 after accu-mulation to a target value of 106 in the Orbitrap at aresolution of 60,000 at m/z 400. For internal mass cali-bration, we used the 445.120025 ion for lock mass.Charge-state screening was enabled, and precursors withcharge state unknown or 1 were excluded. After the surveyscan, the 10 most intense precursor ions were selectedfor collision-induced dissociation—higher energy colli-sion dissociation (CID-HCD) MS–MS fragmentation.Peptide identification was performed in both CID andHCD spectra and quantification of iTRAQ reporter ionswas performed in HCD. HCD was carried out with excessof collision energy for effectively maximizing abundanceof the reporter ions. For CID fragmentation, the targetvalue was set to 10,000 and normalized collision energyto 35%. For HCD, target value was set to 50,000 andcollision energy was set to 55%. Dynamic exclusion wasapplied during 30 seconds. All MS data were analyzed andquantified with Proteome Discoverer (version 1.3.0.339;Thermo Scientific) using standardized workflows (seeSupplementary Materials and Methods for a moredetailed description).

Mouse cytokine arrayConditioned medium from CAFs and normal fibroblasts

were collected after 24 hours in serum-free medium andincubated with Mouse Cytokine Antibody Array C6 forsemiquantitative analysis of 97 mouse cytokines (RayBio-tech) according to the manufacturer’s instructions. Then,membraneswere scanned and analyzedusing Redfin, a two-dimensional gel image analysis software (Ludesi). Relativesemiquantitative cytokine intensities were normalized incomparison with control spots on the same membrane.Expression ratios were calculated comparing the signalintensities for each spot of the different cytokines. The limitof detection, sensitivity, and the dynamic range of themeasured cytokines were more than 1 pg/mL, where mostof cytokines can be detected.

Cell proliferation, adhesion, migration, and invasionassays

KM12C and KM12SM human colon cancer cells (17)were cultured in DMEM (Life Technologies) containing10% FCS and antibiotics at 37�C in a 5% CO2-humidified

Torres et al.

Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research6008




atmosphere. Invasion, migration, adhesion, and cell pro-liferation assays were carried out by following establishedprocedures (18), with the addition of the indicated cyto-kines: CCL2 (10 ng/mL), CCL8 (50 ng/mL), and IL-9(2 ng/mL). See Supplementary Materials and Methods fora detailed description.

Immunohistochemical analysisPieces of intestine from AOM/DSS–treated or control

mice were fixed in 10% buffered formalin overnight toperform hematoxylin and eosin (H&E) and immunohis-tochemical staining. For the prognostic studies in humans,a total of 80 patients, diagnosed and treated of colorectaladenocarcinoma between 2001 and 2006 in Fundaci�onJim�enez D��az (Madrid, Spain) and followed in the longterm, were included for the study. Human samples wereprepared as described previously (18). All human biopsieswere obtained with the patient’s consent and approval ofthe Ethical Committee of Hospital Fundaci�on Jim�enezD��az in accordance with the official Spanish regulations.We reviewed the clinical records of the patients to deter-mine tumor stage at the time of diagnosis and outcome(18). Each sample was deparaffinized for antigen retrievalusing the PT Link Module (Dako) at high pH for 20minutes, rehydrated, and then incubated with the primaryantibody against a-SMA, Snail1, S100A4, CALU, CDH11,LTBP2, FSTL1, or OLFML3 (Supplementary Table S1). Thereaction was revealed using 3,30-diaminobenzidine (DAB)as the chromogen and hematoxylin for counterstaining,and observed in an Olympus microscope. ImageJ softwarewas used to quantify the DAB staining of immunohisto-chemical images.

BioinformaticsIngenuity Pathway Analysis (IPA; Ingenuity systems,

Inc.) was used to analyze the predicted biologic functionsof the proteins deregulated in CAFs and to determineprotein interactions as well as for network analysis. Bio-GPS software was used to determine the top-5 tissues withhighest expression of each altered protein compartmentto determine the desmoplastic signature (10). FatiGowas used to get further insight in deregulated functionsusing a cutoff value of less than 0.1 for the adjusted Pvalue (19).

ResultsIsolation and characterization of fibroblasts fromAOM/DSS colon cancer tumorsAfter AOM/DSS treatment, development of adenocar-

cinomas in the distal colon of mice was visible by visualinspection and histologic staining (Fig. 1A). Mouse ade-nocarcinomas were equivalent to human colorectal ade-nocarcinomas, corresponding to a well-differentiated infil-trating enteroid adenocarcinoma, with the presence oflarge numbers of flat and polypoid malignant tumors. Themucosa from control mice was normal, without dysplasticchanges (Fig. 1A). On immunohistochemical analysis,AOM/DSS tumors showed high expression level of a-SMA,

S100A4, and Snail, considered markers of activated fibro-blasts in desmoplastic tumors (Fig. 1B). To establish CAFsand normal fibroblasts in cultures, distal colons from 16AOM/DSS–treated and vehicle-treated mice were cut intosmall fragments and placed in culture. Fibroblasts migratedoutside the explants and expanded (Fig. 1C). Cells at 95%to 100% confluence (�2 � 104 cells/cm2) were collected,divided into two fractions, and mixed for iTRAQ experi-ments. On Western blot analysis, it was confirmed thatAOM/DSS–isolated fibroblasts had higher expression ofmarkers corresponding to activated fibroblasts such asa-SMA, S100A4, and Snail1 (Fig. 1D). The expression wasequivalent to the immunohistochemical staining in theoriginal tumor (Fig. 1B). As expected, purified fibroblastsexpressed vimentin, a mesenchymal marker, but not theepithelial marker Epcam.

Protein identification, iTRAQ quantification, GeneOntology, and functional networks alterations

Briefly, 100 (cell lysates) and 50 mg (conditionedmedia) of proteins from cultures of CAFs and normalfibroblasts were trypsin-digested. Peptides were labeledwith iTRAQ for relative quantification and fractionatedwith OFFGEL, pH 3–10. To avoid biases in peptidelabeling that could affect the final quantification, weperformed two replicate analyses using the 114 and116 iTRAQ tags for normal fibroblasts peptides and115 and 117 for CAFs peptides (Supplementary Fig.S1). In total, 3,846 and 734 peptides corresponding to1,353 and 295 proteins were identified in the whole-cellextract and conditioned medium, respectively. For quan-tification, peptide ratios were calculated by comparingthe intensity of the reporter ions in the MS–MS spectra,and normalized either using the ratios corresponding toseven different housekeeping proteins in the whole-cellextract or the median in the secretome (SupplementaryFig. S2). A total of 1,102 and 250 proteins were quan-tified in the whole-cell extract and conditioned medium,respectively (Supplementary Tables S2 and S3). We usedseveral criteria for protein selection, including: (i) a fold-change of 1.4 or more relative to control samples; (ii)proteins present in both replicates; (iii) expression ratiosfollowing a similar trend in both replicates; and (iv)proteins identified with a single unique peptide or var-iability higher than 50% were manually inspected toverify that the peptide only corresponded to a singleprotein. We selected fold-change of 1.4 as a minimumvalue to include all relevant proteins that showed sys-tematic change and to compensate for the compression ofiTRAQ ratios, which leads to underestimation of fold-changes (20). Using a fold-change of 1.4 or more, wefound 132 proteins deregulated in complete cell lysates,109 upregulated, and 23 downregulated (SupplementaryTable S4), as well as 125 proteins in the secretome, with72 upregulated and 53 downregulated (SupplementaryTable S5).

Gene Ontology analysis of whole-cell extracts resultsfrom CAFs showed a clear upregulation of ECM






constituents (collagen a1 types I, V, III, and XII, collagena2 types I and V, and fibulin 2), proteins for matrixassembly (decorin, prolyl 4-hydroxylase a2, and leprecan1), proteins related to TGF-b signaling, such as LTBP2(latent TGF-b–binding protein 2), cadherin 11, fibrillin 1,and insulin-like growth factor (IGF) 2 receptor (IGF2R;Supplementary Table S6). In the secretome, 26 of 125proteins were related to ECM. In addition, we identifieddifferentially expressed chemokines, such as CCL11,CCL8, CXCL5, or CCL2, IGF-related proteins such asIGFBP7, or IGF1, cell adhesion, and wound response(Supplementary Table S7). Collectively, these data sug-gest that activated fibroblasts participate in ECM remo-deling, wound healing, and epithelial differentiation, andare important regulators of inflammation.

Altered biologic functions, networks, and pathways wereanalyzed using IPA (Supplementary Table S8) and FatiGO(Supplementary Table S9). Using IPA in whole-cell extract,"Connective Tissue disorders," related with ECM remodel-ing, was the top altered function (n ¼ 21 proteins andP values ranging from 1.59E�11 to 1.34E�02). In the

secretome, we found "Cancer" to be the most representedbiologic function altered (n ¼ 97 proteins and P valuesranging from 2.82E�14 to 1.35E�04), and "OrganismalInjuries and Abnormalities" as the top biologic functionaltered (n ¼ 54 proteins and P values ranging from2.55E�14 to 1.54E�04). With regard to altered networkfunction "Dermatological Diseases and Conditions, Cellu-lar Assembly and Organization, Cellular Function andMaintenance" with 31 proteins emerged as identified incell extracts (Fig. 2A), and "Cell morphology, CellularAssembly and Organization, Cellular Function and Main-tenance" was significantly altered in the secretome, with 28proteins identified (Fig. 2B).

With FatiGo, we found that oxidative phosphorylationand molecular functions related to redox processes werealtered in whole-cell extract. Transport of proteins andtheir cellular localization emerged as the top alteredbiologic processes. In the secretome, the ECM and focaladhesion were the most altered processes. In addition, weobserved a focus toward modulation of molecular func-tions through receptor binding, whereas the most altered

AC

on

tro

lA

OM

/DS

S

DAY 1 DAY 7 DAY 15 DAY 25C

NF

s

CA

Fs

S100-A4

αα--Tub

αα--SMA

Vim

Epcam

Snail1

1.0 5.0

1.0 2.0

1.0 26.0

D

1.0 1.0

0.0 0.0

BC

on

tro

lA

OM

/DS

SH&E

NF

sC

AF

s

αα--SMA Snail1 S100A4

Co

ntr

ol

AO

M/D

SS

Ratio* 4.8 2.66.2

Ratio*

2.0

5.0

26.0

1.0

–

Figure 1. Isolation and culture of fibroblast from colorectal tissue fromAOM/DSS–treatedmice. A, mice treatedwith AOM/DSS showed engrossed intestinewith polypoid tumors, whereas mice treated with vehicle displayed a normal colon (left). H&E staining of tumoral and normal colon (right). B,immunohistochemical analysis of cancer and normal tissue using antibodies against S100A4, Snail1, and a-SMA. C, fibroblast isolation from colonictissues by the explants technique between days 1 and 25. Tissues were cut into 2- to 3-mm fragments and placed in culture medium for 3 weeks toallow fibroblast migration. Top, control explants; bottom, AOM/DSS explants. Explants are indicated with white arrows. D, Western blot analysis ofisolated fibroblasts using antibodies against S100A4, Snail1, and a-SMA. Vimentin was used as positive fibroblast marker and EpCAM as negative control.B and D, ratio of AOM-DSS/control was calculated for Snail1, a-SMA, and S100A4 using ImageJ software or QuantityOne 1-D Analysis Software,respectively.

Torres et al.





A

B

Cell morphology, Cellular Assembly and Organization,

Cellular Function and Maintenance

Dermatological Diseases and Conditions, Cellular

Assembly and Organization, Cellular Function and

Maintenance

Figure 2. Network functions affected in CAFs.Protein networks were identified by IPA usingthe 132 and 125 differentially expressedproteins identified in whole-cell extract andsecretome, respectively. A, "DermatologicalDiseases and Conditions, Cellular Assemblyand Organization, Cellular Function andMaintenance" network was identified in whole-cell extracts with a score of 47. The networkconsisted of 26 upregulated and fivedownregulated proteins from 34 direct-interacting proteins. B, "Cell Morphology,Hematological System Development andFunction, Inflammatory Response" wassignificantly altered in the secretome with ascore of 44. The network consisted of 27upregulated and one downregulated proteinsfrom 35 direct-interacting proteins.






biologic processes were the response to wounding, exter-nal stimulus, and inflammatory response. Collectively,these protein networks indicate the association of CAFproteins with desmoplastic and proinflammatory signa-tures, respectively.

Validation and fibroblast specificity of selectedbiomarkers

After previous network analyses and data-mining usingthe GeneCards database, we selected top upregulatedproteins that were not previously associated to coloncancer and pertained to major networks (Fig. 2 andSupplementary Tables S8 and S9). Therefore, 11 differ-entially expressed proteins (five from whole-cell extractand six from secretome) were validated usingWestern blotanalysis. We selected a relatively large number of proteinsfrom both locations to make a more comprehensiveanalysis, as they would represent different family proteinsand functionalities. Results of Western blot analysis wereconsistent with the iTRAQ quantification data (Fig. 3Aand B). Molecules such as CRABP1, LTBP2, FDPS, andCDH11 were upregulated in whole-cell extracts (Fig. 3A),whereas OLFML3, CALU, AEBP1, LMOD1, SPON2, andFSTL1 were upregulated in CAFs supernatants (Fig. 3B).To confirm that purified fibroblasts were equivalent tothose present in the tumor, we validated a subset of thesemarkers in the original tissues of the AOM/DSS model byquantitative real-time PCR (RT-PCR) and Western blotanalysis. Protein and mRNA expression values were sim-ilar between purified fibroblasts and original tissues (Sup-plementary Fig. S3).

To confirm fibroblast specificity of stromal biomarkers,we compared the presence of these proteins in CAFs withregard to CT26—a colon adenocarcinoma murine cellline—by Western blot analysis. Expression of LTBP2,FDPS, CDH11, and OLFML3 was specific or clearly upre-gulated in fibroblasts with regard to the epithelial CT26cells (Fig. 3C). When we tested the supernatants, SPON2and FSTL1 were also specific for fibroblasts. Finally, weinvestigated the capacity of CDH11 for cell sorting offibroblasts. PDGF receptor a (PDGFRa) was used as apositive control (7). CDH11 antibodies were able toisolate fibroblast cell lines, such as mouse NIH-3T3 orhuman BJ-hTERT, but did not work with epithelial celllines such as CT26 or RKO (Fig. 3D). These results sug-gested that all these six proteins were stromal biomarkersspecific for colon cancer.

Cytokines and growth factors deregulated in thesecretome of activated fibroblasts areproinflammatory and protumorigenic on coloncancer cells

CAFs secrete chemokines that control critical steps of theadhesion–invasion–metastasis cascade and affect recruit-ment of inflammatory cells and enhancement of angiogen-esis (21, 22). ByMS,weobserved a significant increase in theexpression of cytokines such as CCL11 (Eotaxin-1), CCL8(MCP-2), CCL2 (MCP-1), SAA3, and CXCL5 (ENA-78),

whereas CSF1 was downregulated in CAFs (SupplementaryTable S7). To investigate other additional changes in cyto-kines, we used a mouse chemokine microarray for CAF/NFsupernatants (Fig. 4A). CCL11, CCL8, SAA3, and CXCL5antibodies were not present in the arrays. Microarray semi-quantitative data confirmed the increase in CCL2 expres-sion, although not as relevant as in iTRAQ, and showed anincrease in other proinflammatory cytokines such as IL-6,CXCL2, CCL20, and osteopontin (OPN; Fig. 4B). Overall,our chemokine profile was similar to the one described forcancer skin fibroblasts that showed NF-kB–dependent can-cer-promoting activity (7).

To be activated by cytokines, cancer cells should expressthe corresponding receptors. To detect CCL2 and CCL8receptors (CCR1, 2, 3, and 5) as well as the receptor for IL-9 (IL-9R), we performed RT-PCR assays with KM12C andSM cells. These assays showed that both cell lines expressIL-9R as well as CCR1 and CCR3, whereas CCR2 andCCR5 were detected only in KM12C and KM12SM,respectively (Fig. 4C). Then, we investigated the effect ofCCL2, CCL8, and IL-9 on cell proliferation, adhesion,migration, and invasion of colon cancer cell lines (Fig.4D). A more than 2-fold increase was observed at theadhesion, migration, and invasion level in KM12 cells,after addition of CCL2 and CCL8, but there was no effectfor IL-9. In contrast, IL-9 induced a significant increase inthe proliferation of KM12 cell lines. Collectively, thesedata suggest that CCL2 and CCL8 play an important rolein tumorigenesis and the acquisition of invasive capacityof cancer cells.

Desmoplastic signature in AOM/DSS colon fibroblastsand prognostic value

Because activated fibroblasts are characterized by theacquisition of a myofibroblast phenotype consisting ofdifferent smooth–muscle-like proteins, such as a-SMA,one aim of this work was to define a desmoplasticsignature for colon cancer. We searched for desmoplasticmarkers on the differentially expressed proteins using theBio-GPS database. In the secretome, we found 30 secretedproteins (24% of the total) preferentially expressed inmyofibroblasts (Table 1), including seven collagens, col-lagen metabolism-related proteins (LOX and PCOLCE),fibronectin, proteoglycans (perlecan and byglican), cyto-kines (CCL11, CCL8, and CCL2), and calcium metabo-lism proteins [calumenin (CALU)], IGF-binding proteins(IGFBP6 and -7), and FSTL1. The cell lysate showed asimilar profile.

Then, we studied the value of the proteins in this desmo-plastic signature as prognostic biomarkers in humanpatients. Given the large number of candidates, we carriedout a meta-analysis using databases from the "cBio CancerGenomics Portal" (23), which contains mRNA expressiondatasets from 274 samples of colon cancer. Using Kaplan–Meier for overall survival analysis, we found some combi-nations of proteins with prognostic value from whole-cellextract (Supplementary Fig. S4A) and the secretome (Sup-plementary Fig. S4B). The most significant combination

Torres et al.





NFsA

C D

BWB

Olfml3

Olfml3

Calu

Calu

Cell Iysates Supernatants

Aebp1

Aebp1

Lmod1

Lmod1

Spon2

Spon2

Fstl1

CT26kDa kDa

10

180

48

48

130

100

75

75

63

3535

35

0 0100 100101 101102 102103 103

35

63

CAFs CT26

Cell

num

ber

1.34 1.44

0.63 0.59

0.790.88

0.32

2.010.31

0.31

CT26

CCD18

RKO

Fiuorescence intensity

PDGFRαCDH11

NIH-3T3

BJ-hTERT

CAFs

RhoGDI

Fstl1

3.6

2.2

2.1

8.515.06

11.88

11.75

11.03

10.33

3.08

8.1

7.9

3.7

3.8

2.4

3.93

2.42

2.07

4.9

0.1

1.2

2.03

1.19

0.29

iTRAQ WBiTRAQ

RatioCAFs/NFs

RatioCAFs/NFsCAFs NFs CAFs

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Inte

nsity

[co

un

ts](

10

^3)

Crabp1

Crabp1

Ltbp2

Ltbp2

Fdps

Fdps

Cdh11

Cdh11

Thy1

α-Tub

α-Tub

300

200

100

114114

m/z

m/z

m/z

m/z

m/z

115115

116116

117

114m/z

115 116 117

114m/z

115 116 117

114m/z

115 116 117

114m/z

115 116 117

114 115 116 117

117

114 115 116 117

114 115 116 117

m/z114 115 116 117

m/z114 115 116 117

m/z114 115 116 117

200

100

0

250

150

50

140

80

20

15

10

5

120

60

20

600

400

200

600

400

200

2.5

1.5

0.5

20

10

35

25

15

5

100

50

10

Figure 3. Western blotting validation of iTRAQ data and biomarker expression on stromal cells. Protein samples from whole-cell extract (A) or concentratedconditioned media (B) of CAFs and normal fibroblasts were separated on SDS–PAGE gels, transferred to nitrocellulose membranes, and probed with theindicated antibodies. Tubulin was used as a control. Protein abundance was quantified by densitometry, and CAF/NF ratios were compared with theiTRAQ ratios.C,Western blot analysis of the expressionof selected stromal biomarkers in themurine adenocarcinomacell lineCT26andCAFsusing completecell extracts and concentrated supernatants, respectively. D, flow cytometric analysis for testing CDH11 and PDGFRa specificity (gray areas) onfibroblasts and cancer cell lines. An irrelevant antibody was used as a control (white areas).






A

B

0

5,000

10,000

15,000

20,000

25,000

30,000

35,000

40,000

45,000

50,000

Arb

itra

ry u

nit

s

NFs

CAFs

KM12C KM12SM

Ad

here

d c

ells/m

m2

0

50

100

150

200

250***

***

0

50

100

150

200

250

300

*** ***

Inv

as

ive

ce

lls

0

20

40

60

80

100

120

0

50

100

150

200

250

300

350***

*** ***

***

C

ell

mig

rati

on

sp

ee

d (

μμm/h

)

0

2

4

6

8

10

1

0

2

3

4*** ***

****** ***

***

Ab

so

rban

ce (5

60 n

m)

***

*** *

0

0.1

0.2

0.3

0.4

0.5

*

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8* ***

Proliferation

Migration Invasion

Adhesion

KM12C KM12SM

NFs CAFs

CXCL-2 CXCL-2

IL-6 IL-6

CCL-2 CCL-2

CCL-20 CCL-20Galectin-1 Galectin-1

CCL-5

EotaxinEotaxin

CCL-5

D

CCR1 CCR2 CCR3 IL−9R GAPDHCCR5

C

Torres et al.





included RPN2, IVL, RP1, and CALU (Supplementary Fig.S4C), which displayed a highly significant association tooverall survival (log-rank test P ¼ 0.000095). In contrast,the association with survival in ovarian or lung adenocar-cinoma did not deliver significant P values (data notshown).

Verification of stromal markers and prognosticassociation in human cancer samplesImmunohistochemical staining was performed to vali-

date the expression of the five selected candidate stromalbiomarkers CALU, CDH11, FSTL1, OLFML3, and LTBP2 inAOM/DSS–induced tumor samples. SPON2 was not testedbecause of the lack of suitable antibodies. The selectedmarkers were strongly expressed in the stroma of AOM/DSS–induced murine adenocarcinoma, with little or noexpression in normal colon mucosa (Fig. 5A). The expres-sion of these markers was more intense in the leading edgeof tumors, which could be associated with the desmoplasticinvasion front.Finally, to investigate the relevance of these markers in

human colon cancer samples, we used a tissue micro-array. Clinical information, together with the stromalexpression of the selected CAF markers, is summarizedin Supplementary Table S10. We observed an increasedexpression for CALU, CDH11, FSTL1, OLFML3, andLTBP2 in the tumoral respect to normal stroma (Fig.5B). All of them showed stronger expression in theleading edge of the tumors, where the tumor infiltratesthe surrounding areas. Tumoral epithelial cells showed anincreased expression in more than 70% of the tumors forCALU, CDH11, and OLFML3, but no significant expres-sion was detected for FSTL1 (Fig. 5B). In contrast, in thestroma of normal tissue, the staining of these markers wasmainly observed to be negative or weak in contrast totumoral stroma where the markers were highly overex-pressed (Fig. 5B).Then,we determined if the expression of these proteins in

the stroma was associated with survival in human cancerusing the tissuemicroarray data for patients followed on thelong term (>5 years). The series was retrospectively selected(Supplementary Table S10). High stromal expression ofCALU correlated significantly with lymph node involve-ment at the moment of diagnosis (P ¼ 0.034; Fig. 5C) andwith poor prognosis (P ¼ 0.010; Fig. 5D). In addition,stromal expression of CDH11 correlated with disease-freesurvival (P ¼ 0.051; Fig. 5E), but not with prognosis (P ¼0.105). In addition, the combination of CALUwithCDH11

expression displayed a significant association with disease-free survival (log-rank test P ¼ 0.015) and overall survival(log-rank test P ¼ 0.009) in human cancer samples (Fig.5D). The rest of the biomarkers did not reach significantcorrelation with metastasis or survival (Supplementary Fig.S5A–K).

DiscussionThe use of the AOM/DSS murine model of sporadic

colon cancer for the isolation of purified and homoge-neous populations of CAFs, together with in-depth prote-omic analysis, allowed the identification of an elevatednumber of proteins deregulated after fibroblast activationin cancer. The identification of many of the currentlyknown stromal biomarkers confirmed the validity ofour experimental approach. As an example, we found anincrease in collagens type III and XII, which are mainlyimplicated in desmoplasia and colorectal cancer metas-tasis (10). At least six identified proteins: LTBP2, CDH11,FDPS, OLFML3, SPON2, and FSTL1 were novel candidatestromal biomarkers, as no expression was observed in themurine adenocarcinoma cell line CT26. In addition,CDH11 was useful to isolate colon cancer fibroblasts.The relevance of these biomarkers for colorectal cancerprognosis was confirmed by studies with human tissuemicroarrays. The expression of CDH11, FSTL1, OLFML3,and LTBP2, together with CALU, was increased in thetumoral stroma of patients with colorectal cancer. More-over, we observed an association of CALU and CDH11with poor survival and prognosis in human patients withcancer. Together, these data support the value of theAOM/DSS murine model for the discovery of stromalbiomarkers applicable in human patients.

Despite this, the azoxymethane model presents somelimitations. The most relevant is its inability to developmetastasis. Recently, an adenomatous polyposis coli (APC)mutant mouse model of sporadic colon cancer was used toanalyze the proximal fluid proteome of whole tumor tis-sues, without isolation of pure cell populations (24). Look-ing at the candidate protein biomarkers identified in thatmodel, we have observed several coincidences in ECMfibroblast-related proteins: biglycan, cingulin, collagens,decorin, or lamin A/C, confirming the relative similarity ofboth models.

CAFs exhibited specific proinflammatory and desmo-plastic protein signatures for colon cancer. Several reportssupport the role of the desmoplastic microenvironmentin tumor promotion (2). For instance, Tsujino and

Figure 4. Chemokine expression in conditioned medium from CAFs and normal fibroblasts. Functional studies with selected chemokines. A, arepresentative image of a chemokinemicroarray after screening of conditionedmedium from CAFs and normal fibroblasts. White arrows indicate themostderegulated cytokines in CAFs and normal fibroblasts. B, most abundant chemokines and cytokines differentially expressed in CAFs versus normalfibroblasts. Bar graphs were calculated in arbitrary units with the mean value of the duplicates present in the microarrays. C, RT-PCR analysis of theindicated cytokine receptors in KM12C and KM12SM cell lines. D, proliferation, adhesion, migration, and invasion of colon cancer KM12 cells inpresence or absence of the indicated chemokines. Proliferation was determined by MTT assays after 24 hours of culture. Cell adhesion to Matrigel wasperformed after 2 minutes of incubation with the indicated chemokines and 4 minutes of adhesion. Migration speed was calculated as the distancecovered by the cells in 24 hours. Cell invasion was done across Matrigel of the KM12 cell lines treated as indicated. Experiments were performed threetimes. Error bars, SD (�, P < 0.05; ��, P < 0.005).






colleagues (25) reported a strong association between thepresence of abundant stromal myofibroblasts, livermetastasis, and shorter disease-free survival. In addition,it has been reported that tumor-derived ECM is stifferthan normal ECM. Moreover, increased matrix stiffnessand ECM remodeling was observed in premalignant tis-sue. This increase was shown to contribute to malignanttransformation in the breast (26). In colorectal cancer,previous studies showed a correlation between the pres-ence of myofibroblasts (25), vimentin (27), or expressionof fibroblast activation protein (FAP; ref. 28) with prog-nosis and recurrence.

Here, we identified several new prognostic markers asso-ciated with fibroblasts. Calumenin is a calcium-bindingprotein that regulates the activity of coagulation factors(29), and was reported as being overexpressed in coloncancer (30, 31). Although calumenin expression is notrestricted to the stroma, we have found a good associationwith stromal expression, lymph node invasion, and lower

survival, alone or in combination with other proteins suchas CDH11. Cadherin-11 is a mesenchymal cadherin, upre-gulated byTGF-b, whichhas been identified as anovel targetfor pulmonary fibrosis and associated with invasion insquamous cell carcinoma (32, 33). The combination ofCALU with CDH11 significantly improved the prognosticvalue of disease-free survival.

We found several other TGF-b–related deregulated pro-teins as stromal biomarkers. OLFML3 belongs to the olfac-tomedin domain-containing proteins. They are knownBMP antagonists (34), and also showed proangiogenicactivity in cancer (35). FSTL1 expression was the mostspecific for the human stromal compartment. FSTL1 isa TGF-b–inducible gene, SPARC-related, that enhancesinflammatory cytokine/chemokine expression (36) andseems to be implicated in angiogenesis and revasculariza-tion (37). Another TGF-b signaling-related protein wasLTBP2, which colocalizes with fibronectin and collagentype I for ECM regulation and remodeling (38, 39). LTBP2

Table 1. Myofibroblastic signature associated with proteins in the CAF secretomea

Accessionnumber Description Name

Averagefold-change SD

P48298 Eotaxin Ccl11 25.37 25.83Q61581 IGFBP7 Igfbp7 11.06 5.21Q9Z121 C–C motif chemokine 8 Ccl8 11.00 10.87Q61554 Fibrillin-1 Fbn1 7.81 5.60Q60847 Collagen a-1(XII) chain Col12a1 6.12 9.61O35887 Calumenin Calu 6.11 2.59P11087 Collagen a-1(I) chain Col1a1 5.51 2.09P08121 Collagen a-1(III) chain Col3a1 5.48 0.38D3YYT0 Cadherin-2 Cdh2 4.46 1.41P28301 Protein-lysine 6-oxidase Lox 4.31 1.98G5E8M2 Fibronectin 1, isoform CRA_d Fn1 4.27 2.17P10148 C–C motif chemokine 2 Ccl2 4.11 1.42Q01149 Collagen a-2(I) chain Col1a2 3.73 1.02Q61398 Procollagen C-endopeptidase enhancer 1 Pcolce 3.35 0.63P98063 Bone morphogenetic protein 1 Bmp1 3.28 1.06P28653 Biglycan Bgn 3.21 0.56P08122 Collagen a-2(IV) chain Col4a2 3.17 1.55E9QL02 Heparan sulfate proteoglycan 2 Hspg2 3.13 0.49Q91X24 Igfbp6 protein Igfbp6 3.12 0.95Q9WVJ9 EGF-containing fibulin-like ECM protein 2 Efemp2 3.04 1.41P10493 Nidogen-1 Nid1 2.14 0.26P35441 Thrombospondin-1 Thbs1 2.08 0.36Q3U962 Collagen a-2(V) chain Col5a2 2.05 0.30P07214 SPARC Sparc 1.81 0.21Q8BND5 Sulfhydryl oxidase 1 Qsox1 1.80 0.38P28862 Stromelysin-1 Mmp3 1.80 0.67O88207 Collagen a-1(V) chain Col5a1 1.71 0.38P12032 Metalloproteinase inhibitor 1 Timp1 1.69 0.45Q62356 Follistatin-related protein 1 Fstl1 1.59 0.16P10639 Thioredoxin Txn 0.19 0.06

aDifferentially expressed proteins were identified as preferentially expressed in smooth muscle if they were among the top five tissueswith highest expression using the Bio-GPS database.

Torres et al.





Controlmouse

AOM/DSSmouse

CA

LU

CD

H11

FS

TL1

OL

FM

L3

Humannormal mucosa

Humancancer mucosa

LT

BP

2BA

D 1.0

0.8

0.6

0.4

0.2

0 20 100 120 14040 60 80

NegativePositive

Time (mo) Time (mo)

Cu

m s

urv

iva

l

CALU

0.0 P = 0.010

1.0

0.8

0.6

0.4

0.2

0 20 100 120 14040 60 80

NegativePositive

Time (mo)

Cu

m s

urv

iva

l

CALU

CDH11

0.0 P = 0.009

1.0

0.8

0.6

0.4

0.2

0 20 100 120 14040 60 80

NegativePositive

Time (mo)

Dis

ea

se

-fre

e s

urv

iva

l

CALU

CDH11

0.0 P = 0.015

1.0

0.8

0.6

0.4

0.2

0 20 100 120 14040 60 80

NegativePositive

Dis

ea

se

-fre

e s

urv

iva

l

CDH11

0.0P = 0.051

C

NegativeModerate

0

20

40

60

80

100

NO YES

Lymph node

involvement

% p

atients

P = 0.034

Intense

CALU staining

Figure 5. Validation of stromal biomarkers and prognostic value of CALU andCDH11 in patients with colorectal cancer. A, immunohistochemical staining ofAOM/DSS cancer and normal mouse tissues using antibodies against CALU, CDH11, FSTL1, OLFML3, and LTBP2. B, immunohistochemical analysis ofCALU, CDH11, FSTL1, OLFML3, and LTBP2 expression in human tissue microarrays. White arrows indicate the leading edge of the tumor. Pictureswere taken at �200 magnification. C, percentage of patients with invasion of regional lymph nodes at time of diagnosis in function of CALU stromalexpression; P values were calculated with x

2 test. D, Kaplan–Meier analyses of overall and disease-free survival of patients according to the stromalexpression of CALU or CDH11 and their combination, respectively. For statistical analysis of significance, we used the log-rank test.






is overexpressed in pancreatic cancer, being mainly locatedin the cancer stroma (40). TGF-b activates CAFs by changingthe secretion profile of chemokines to assist colorectalcancer cells in the progression of the disease. An inflam-matorymicroenvironment canpromotemalignant progres-sion and participate in the recruitment and retention ofinfiltrating leukocytes to inhibit the resolution of inflam-mation (41, 42).

Evidence was provided of a significant increase inproinflammatory chemokines (CCL2, CCL11, CCL8,CCL20, CXCL5, and IL-9) in the conditioned media ofmouse CAFs, either by MS or chemokine microarrays. Inaddition to function as leukocyte chemoattractants andproangiogenic properties, human CCL2, CCL8, and IL-9promoted tumorigenesis in two human colorectal cancercell lines. CCL8—also called MCP2—a pluripotent che-mokine, attracts monocytes (43) and lymphocytes (44).From our results, CCL2 and CCL8 regulated the prolif-eration, adhesion, migration, and invasion of two colo-rectal cancer cell lines, whereas IL-9 was only efficient intumor growth proliferation. Because CCL2 and CCL8share at least three receptors (CCR2, CCR3, and CCR5;ref. 45), this could explain the similar effect induced byboth chemokines in colorectal cancer. In fact, severalsolid cancers, including colorectal cancer, showed expres-sion of functional chemokine receptors on tumor cells,able to promote proliferation and metastasis (46). There-fore, CCL8 and IL-9 are new colon cancer–associatedchemokines that deserve further investigation. Stromalchemokines might work through either paracrine orautocrine pathways. CAFs and myofibroblasts could dif-ferentiate from resident tissue fibroblasts under variousparacrine stimuli (47). In this regard, resident monocytesor other immune cells might secrete chemokines thatactivate the proinflammatory signature in CAFs by para-crine stimulation. In addition, carcinoma cells may pro-duce chemokines that affect CAFs. In turn, CAFs secretecytokines that can modulate tumor cells or other fibro-blasts in different ways.

Our data provide a comprehensive view of the manymolecules secreted by CAFs that favor the progression andinvasion of tumors, from angiogenesis to desmoplasiaand profibrotic mechanisms. A panel of new stromal

markers in colon cancer was identified. Among them, thecombination of CDH11 with CALU gave a significantprognostic value. In addition, verification of the resultswith human samples confirmed the value of the modeland the strategy for identification of novel candidateprognostic biomarkers with clinical potential. In summa-ry, this study defines a new set of stromal biomarkers withgood prognostic value that might constitute an alternativeto current markers such as a-SMA or the presence ofmyofibroblasts in colon cancer.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: M.J. Fernandez-Ace~nero, A. Garc��a de Herreros,I. CasalDevelopment of methodology: S. Torres, M. Mendes, M.J. Fernandez-Ace~nero, C. Pe~naAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): S. Torres, R.A. Bartolom�e, R. Barderas, M.J.Fernandez-Ace~nero, A. Pel�aez-Garc��a, M. Lopez-Lucendo, R. Villar-V�azquez,A. Garc��a de Herreros, F. BonillaAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): S. Torres, R.A. Bartolom�e, M. Mendes, R.Barderas, M.J. Fernandez-Ace~nero, R. Villar-V�azquez, I. CasalWriting, review, and/or revisionof themanuscript: S. Torres,M.Mendes,M.J. Fernandez-Ace~nero, F. Bonilla, I. CasalAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): S. Torres, C. Pe~na, M. Lopez-Lucendo, R. Villar-V�azquezStudy supervision: I. Casal

Grant SupportThis work was financially supported by a grant to established research

groups of the Asociaci�on Espa~nola Contra el C�ancer (AECC); to R.A.Bartolom�e, and by the Spanish Ministry of Science and Innovation (grantno. BIO2012-31023 and the Comunidad deMadrid (grant no. S2010/BMD-2344/Colomics2). S. Torres is the recipient of a Juan de la Cierva programmefellowship. M.Mendes was the recipient of a ProteoRed contract. R. Barderaswas a recipient of a JAE-DOC/FSE Contract (CSIC), and currently has afellowship of the Ram�on y Cajal programme. A. Pel�aez has a Formaci�on delPersonal Investigador (FPI) fellowship from theMINECO. R. Villar-Vazquezhas a Formaci�on del Personal Universitario (FPU) fellowship from theMinistry of Education and Culture (MEC).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received April 30, 2013; revised August 9, 2013; accepted August 24, 2013;published OnlineFirst September 11, 2013.

References1. Hanahan D, Coussens LM. Accessories to the crime: functions of

cells recruited to the tumor microenvironment. Cancer Cell 2012;21:309–22.

2. DeWever O, Demetter P, Mareel M, BrackeM. Stromal myofibroblastsare drivers of invasive cancer growth. Int J Cancer 2008;123:2229–38.

3. Dvorak HF. Tumors: wounds that do not heal. Similarities betweentumor stroma generation and wound healing. N Engl J Med 1986;315:1650–9.

4. Kalluri R, Zeisberg M. Fibroblasts in cancer. Nat Rev Cancer 2006;6:392–401.

5. Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, Carey L,et al. Reconstruction of functionally normal and malignant humanbreast tissues in mice. Proc Natl Acad Sci U S A 2004;101:4966–71.

6. Liao D, Luo Y, Markowitz D, Xiang R, Reisfeld RA. Cancer associatedfibroblasts promote tumor growth and metastasis by modulating thetumor immunemicroenvironment in a 4T1murinebreast cancermodel.PLoS One 2009;4:e7965.

7. Erez N, Truitt M, Olson P, Arron ST, Hanahan D. Cancer-associatedfibroblasts are activated in incipient neoplasia to orchestrate tumor-promoting inflammation in an NF-kappaB-dependent manner. CancerCell 2010;17:135–47.

8. Swartz MA, Iida N, Roberts EW, Sangaletti S, Wong MH, Yull FE, et al.Tumor microenvironment complexity: emerging roles in cancer ther-apy. Cancer Res 2012;72:2473–80.

9. Coulson-Thomas VJ, Coulson-Thomas YM, Gesteira TF, de Paula CA,Mader AM, Waisberg J, et al. Colorectal cancer desmoplastic reaction

Torres et al.





up-regulates collagen synthesis and restricts cancer cell invasion. CellTissue Res 2011;346:223–36.

10. Karagiannis GS, Petraki C, Prassas I, Saraon P, Musrap N, Dimitroma-nolakis A, et al. Proteomic signatures of the desmoplastic invasion frontreveal collagen type XII as a marker of myofibroblastic differentiationduring colorectal cancer metastasis. Oncotarget 2012;3:267–85.

11. Nambiar PR, Girnun G, Lillo NA, Guda K,Whiteley HE, Rosenberg DW.Preliminary analysis of azoxymethane induced colon tumors in inbredmice commonly used as transgenic/knockout progenitors. Int J Oncol2003;22:145–50.

12. Terzic J, Grivennikov S, Karin E, Karin M. Inflammation and coloncancer. Gastroenterology 2010;138:2101–14.

13. Goetz JG, Minguet S, Navarro-Lerida I, Lazcano JJ, Samaniego R,Calvo E, et al. Biomechanical remodeling of the microenvironment bystromal caveolin-1 favors tumor invasion and metastasis. Cell 2011;146:148–63.

14. Dolznig H, Rupp C, Puri C, Haslinger C, Schweifer N, Wieser E, et al.Modeling colon adenocarcinomas in vitro a 3D co-culture systeminduces cancer-relevant pathways upon tumor cell and stromal fibro-blast interaction. Am J Pathol 2011;179:487–501.

15. Li Z, Adams RM, Chourey K, Hurst GB, Hettich RL, Pan C. Systematiccomparison of label-free, metabolic labeling, and isobaric chemicallabeling for quantitative proteomics on LTQorbitrap velos. J ProteomeRes 2012;11:1582–90.

16. Beddy D, Mulsow J, Watson RW, Fitzpatrick JM, O'Connell PR.Expression and regulation of connective tissue growth factor bytransforming growth factor beta and tumour necrosis factor alpha infibroblasts isolated from strictures in patients with Crohn's disease. BrJ Surg 2006;93:1290–6.

17. MorikawaK,Walker SM, Jessup JM, Fidler IJ. In vivo selection of highlymetastatic cells from surgical specimens of different primary humancolon carcinomas implanted into nude mice. Cancer Res 1988;48:1943–8.

18. Barderas R, Bartolom�eRA, Fernandez-Ace~neroMJ, Torres S, Casal JI.High expression of IL-13 receptor alpha2 in colorectal cancer isassociatedwith invasion, livermetastasis, andpoor prognosis. CancerRes 2012;72:2780–90.

19. Al-Shahrour F, Diaz-Uriarte R, Dopazo J. FatiGO: aweb tool for findingsignificant associations of GeneOntology terms with groups of genes.Bioinformatics 2004;20:578–80.

20. Jankova L, Chan C, Fung CL, Song X, Kwun SY, Cowley MJ, et al.Proteomic comparison of colorectal tumours and non-neoplasticmucosa frompaired patient samples using iTRAQmass spectrometry.Mol Biosyst 2011;7:2997–3005.

21. Karnoub AE, Weinberg RA. Chemokine networks and breast cancermetastasis. Breast Dis 2006;26:75–85.

22. Cook AC, Tuck AB, McCarthy S, Turner JG, Irby RB, Bloom GC, et al.Osteopontin induces multiple changes in gene expression that reflectthe six "hallmarks of cancer" in a model of breast cancer progression.Mol Carcinog 2005;43:225–36.

23. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al.The cBio cancer genomics portal: an open platform for exploringmultidimensional cancer genomics data. Cancer Discov 2012;2:401–4.

24. Fijneman RJ, de Wit M, Pourghiasian M, Piersma SR, Pham TV,Warmoes MO, et al. Proximal fluid proteome profiling of mouse colontumors reveals biomarkers for early diagnosis of human colorectalcancer. Clin Cancer Res 2012;18:2613–24.

25. Tsujino T, Seshimo I, Yamamoto H, Ngan CY, Ezumi K, Takemasa I,et al. Stromal myofibroblasts predict disease recurrence for colorectalcancer. Clin Cancer Res 2007;13:2082–90.

26. Levental KR, Yu H, Kass L, Lakins JN, EgebladM, Erler JT, et al. Matrixcrosslinking forces tumor progression by enhancing integrin signaling.Cell 2009;139:891–906.

27. Ngan CY, Yamamoto H, Seshimo I, Tsujino T, Man-i M, Ikeda JI, et al.Quantitative evaluation of vimentin expression in tumour stroma ofcolorectal cancer. Br J Cancer 2007;96:986–92.

28. Henry LR, Lee HO, Lee JS, Klein-Szanto A, Watts P, Ross EA, et al.Clinical implications of fibroblast activation protein in patients withcolon cancer. Clin Cancer Res 2007;13:1736–41.

29. Wajih N, Owen J,Wallin R. Enhanced functional recombinant factor VIIproduction by HEK 293 cells stably transfected with VKORC1 wherethe gamma-carboxylase inhibitor calumenin is stably suppressed byshRNA transfection. Thromb Res 2008;122:405–10.

30. Galamb O, Gyorffy B, Sipos F, Spisak S, Nemeth AM, Miheller P, et al.Inflammation, adenoma and cancer: objective classification of colonbiopsy specimens with gene expression signature. Dis Markers2008;25:1–16.

31. Mikula M, Rubel T, Karczmarski J, Goryca K, Dadlez M, Ostrowski J.Integrating proteomic and transcriptomic high-throughput surveys forsearch of new biomarkers of colon tumors. Funct Integr Genomics2011;11:215–24.

32. Schneider DJ,WuM, Le TT, Cho SH, BrennerMB, BlackburnMR, et al.Cadherin-11 contributes to pulmonary fibrosis: potential role in TGF-beta production and epithelial to mesenchymal transition. FASEB J2012;26:503–12.

33. Vered M, Dayan D, Yahalom R, Dobriyan A, Barshack I, Bello IO, et al.Cancer-associated fibroblasts and epithelial-mesenchymal transitionin metastatic oral tongue squamous cell carcinoma. Int J Cancer2010;127:1356–62.

34. Inomata H, Haraguchi T, Sasai Y. Robust stability of the embryonicaxial pattern requires a secreted scaffold for chordin degradation. Cell2008;134:854–65.

35. Miljkovic-Licina M, Hammel P, Garrido-Urbani S, Lee BP, MeguenaniM, Chaabane C, et al. Targeting olfactomedin-like 3 inhibits tumorgrowth by impairing angiogenesis and pericyte coverage. Mol CancerTher 2012;11:2588–99.

36. Chaly Y, Marinov AD, Oxburgh L, Bushnell DS, Hirsch R. FSTL1promotes arthritis in mice by enhancing inflammatory cytokine/che-mokine expression. Arthritis Rheum 2012;64:1082–8.

37. Ouchi N, Oshima Y, Ohashi K, Higuchi A, Ikegami C, Izumiya Y, et al.Follistatin-like 1, a secreted muscle protein, promotes endothelial cellfunction and revascularization in ischemic tissue through a nitric-oxidesynthase-dependent mechanism. J Biol Chem 2008;283:32802–11.

38. Mangasser-Stephan K, Gartung C, Lahme B, Gressner AM. Expres-sion of isoforms and splice variants of the latent transforming growthfactor beta binding protein (LTBP) in cultured human liver myofibro-blasts. Liver 2001;21:105–13.

39. Gabrielsen A, Lawler PR, Yongzhong W, Steinbruchel D, Blagoja D,Paulsson-Berne G, et al. Gene expression signals involved in ischemicinjury, extracellular matrix composition and fibrosis defined by globalmRNA profiling of the human left ventricular myocardium. J Mol CellCardiol 2007;42:870–83.

40. Turtoi A, Musmeci D,Wang Y, Dumont B, Somja J, Bevilacqua G, et al.Identification of novel accessible proteins bearing diagnostic andtherapeutic potential in human pancreatic ductal adenocarcinoma. JProteome Res 2011;10:4302–13.

41. Buckley CD, Pilling D, Lord JM, Akbar AN, Scheel-Toellner D, SalmonM. Fibroblasts regulate the switch from acute resolving to chronicpersistent inflammation. Trends Immunol 2001;22:199–204.

42. Calon A, Espinet E, Palomo-Ponce S, Tauriello DV, Iglesias M,Cespedes MV, et al. Dependency of colorectal cancer on a TGF-beta-driven program in stromal cells for metastasis initiation. CancerCell 2012;22:571–84.

43. Uguccioni M, D'Apuzzo M, Loetscher M, Dewald B, Baggiolini M.Actions of the chemotactic cytokines MCP-1, MCP-2, MCP-3,RANTES, MIP-1 alpha and MIP-1 beta on human monocytes. Eur JImmunol 1995;25:64–8.

44. Taub DD, Proost P, Murphy WJ, Anver M, Longo DL, van Damme J,et al. Monocyte chemotactic protein-1 (MCP-1), -2, and -3 are che-motactic for human T lymphocytes. J Clin Invest 1995;95:1370–6.

45. Erreni M, Bianchi P, Laghi L, Mirolo M, Fabbri M, Locati M, et al.Expression of chemokines and chemokine receptors in human coloncancer. Methods Enzymol 2009;460:105–21.

46. Wang D, Dubois RN, Richmond A. The role of chemokines inintestinal inflammation and cancer. Curr Opin Pharmacol 2009;9:688–96.

47. Kunz-Schughart LA,KnuechelR. Tumor-associated fibroblasts (part I):Active stromal participants in tumor development and progression?Histol Histopathol 2002;17:599–621.






2013;19:6006-6019. Published OnlineFirst September 11, 2013.Clin Cancer Res Sofia Torres, Rubén A. Bartolomé, Marta Mendes, et al. Colorectal CancerNovel Proinflammatory Signatures and Prognostic Markers for Proteome Profiling of Cancer-Associated Fibroblasts Identifies

Updated version

10.1158/1078-0432.CCR-13-1130doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2013/09/10/1078-0432.CCR-13-1130.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/19/21/6006.full#ref-list-1

This article cites 47 articles, 10 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/19/21/6006.full#related-urls

This article has been cited by 8 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/19/21/6006To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-13-1130

http://clincancerres.aacrjournals.org/content/suppl/2013/09/10/1078-0432.CCR-13-1130.DC1

http://clincancerres.aacrjournals.org/content/19/21/6006.full#ref-list-1

http://clincancerres.aacrjournals.org/content/19/21/6006.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/19/21/6006


proteome proﬁling of cancer-associated fibroblasts...

Documents