Microenvironment-Modulated Metastatic LungCancer ... · 3636–49. 2015 AACR. Introduction Despite novel therapeutic strategies, survival rates for lung cancer have not greatly improved

Tumor and Stem Cell Biology

Microenvironment-Modulated MetastaticCD133þ/CXCR4þ/EpCAM� Lung Cancer–InitiatingCells Sustain Tumor Dissemination and Correlatewith Poor PrognosisGiulia Bertolini1, Lucia D'Amico2, Massimo Moro1, Elena Landoni3, Paola Perego4,Rosalba Miceli3, Laura Gatti4, Francesca Andriani1, Donald Wong5, Roberto Caserini1,Monica Tortoreto4, Massimo Milione6, Riccardo Ferracini2, Luigi Mariani3, Ugo Pastorino7,Ilaria Roato2, Gabriella Sozzi1, and Luca Roz1

Abstract

Metastasis is the main reason for lung cancer–related mor-tality, but little is known about specific determinants of suc-cessful dissemination from primary tumors and metastasisinitiation. Here, we show that CD133þ/CXCR4þ cancer-initi-ating cells (CIC) directly isolated from patient-derived xeno-grafts (PDX) of non–small cell lung cancer are endowed withsuperior ability to seed and initiate metastasis at distant organs.We additionally report that CXCR4 inhibition successfullyprevents the increase of cisplatin-resistant CD133þ/CXCR4þ

cells in residual tumors and their metastatization. Immuno-phenotypic analysis of lung tumor cells intravenously injectedor spontaneously disseminated to murine lungs demonstratedthe survival advantage and increased colonization ability of aspecific subset of CD133þ/CXCR4þ with reduced expression ofepithelial cell adhesion molecule (EpCAM�), which also showsthe greatest in vitro invasive potential. We next prove thatrecovered disseminated cells from lungs of PDX-bearing mice

enriched for CD133þ/CXCR4þ/EpCAM� CICs are highlytumorigenic and metastatic. Importantly, microenvironmentstimuli eliciting epithelial-to-mesenchymal transition, includ-ing signals from cancer-associated fibroblasts, are able toincrease the dissemination potential of lung cancer cellsthrough the generation of the CD133þ/CXCR4þ/EpCAM� sub-set. These findings also have correlates in patient sampleswhere disseminating CICs are enriched in metastatic lymphnodes (20-fold, P ¼ 0.006) and their detection in primarytumors is correlated with poor clinical outcome (disease-free survival: P ¼ 0.03; overall survival: P ¼ 0.05). Overall,these results highlight the importance of specific cellularsubsets in the metastatic process, the need for in-depth char-acterization of disseminating tumor cells, and the potentialof therapeutic strategies targeting both primary tumor andtumor–microenvironment interactions. Cancer Res; 75(17);3636–49. �2015 AACR.

IntroductionDespite novel therapeutic strategies, survival rates for lung

cancer have not greatly improved in recent years, mostly due tolate diagnosis and limited efficacy of available pharmacologictreatments (1). Cancer-initiating cells (CIC), functionally definedas the subset of cancer cells responsible for tumor generation andmaintenance (2, 3), could also account for tumor recurrence andmetastatic seeding due to their inherent stem-like and chemore-sistant properties (4).

In this context, specific subsets of tumor cells endowed withboth stemness and motility properties have been defined asmigrating CICs (5, 6), and activation of the epithelial-to-mes-enchymal transition (EMT) process, which is pivotal in meta-static progression, has been shown to modulate stemnessproperties of cancer cells, indicating a possible connectionbetween external cues and the generation of potential seedsof metastasis (7–9). The subset of CICs deputed to tumordissemination could therefore be specifically modulated bytumor microenvironment stimuli, as recently observed forcolon metastatic CICs (10). Studies in transgenic mouse mod-els also indicate that dissemination is an early event and is

1Tumor Genomics Unit, Department of Experimental Oncology andMolecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori,Milan, Italy. 2CeRMS (Center for Research and Medical Studies) A.O.U.San Giovanni Battista, Turin, Italy. 3Medical Statistics, Biometry andBioinformatics, Unit of Clinical Epidemiology and Trial Organization,Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 4Molec-ular Pharmacology Unit, Department of Experimental Oncology andMolecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori,Milan, Italy. 5British Canadian BioSciences Corp, Vancouver, BritishColumbia,Canada. 6PathologyUnit, Fondazione IRCCS IstitutoNazio-naledei Tumori,Milan, Italy. 7Thoracic SurgeryUnit, Fondazione IRCCSIstituto Nazionale dei Tumori, Milan, Italy.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

L. Roz and G. Sozzi contributed equally to this article.

Corresponding Authors: Luca Roz, Fondazione IRCCS Istituto Nazionale deiTumori, Via Venezian 1, Milan 20133, Italy. Phone: 39-02-23902875; Fax: 39-02-23902928; E-mail: [email protected]; and Ilaria Roato, Center forResearch and Medical Studies A.O.U. San Giovanni Battista, Via Santena 5, Turin10126, Italy. Phone: 39-01-16334672; Fax: 39-01-16336887; E-mail:[email protected]

doi: 10.1158/0008-5472.CAN-14-3781

�2015 American Association for Cancer Research.

CancerResearch

Cancer Res; 75(17) September 1, 20153636

on October 4, 2020. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst July 3, 2015; DOI: 10.1158/0008-5472.CAN-14-3781

http://cancerres.aacrjournals.org/

supported by tumor cells that have undergone EMT and possessCIC features (11, 12).

Increasing evidence also shows that chemotherapy mightselect for intrinsically chemoresistant CICs that could beresponsible for primary tumor recurrence (13, 14). According-ly, we previously demonstrated that a distinct subset ofCD133þ lung CIC coexpressing the chemokine receptor CXCR4is spared by cisplatin treatment (15). The CXCL12/CXCR4 axisplays a crucial role in mediating tumor cell survival and themetastatic process (16); more recently, it has been proven tosustain CICs, through activation of self-renewal pathways (17),and to guide their metastatic dissemination in a human pan-creatic tumor model (6).

In the present study, through the analysis of a variety of lungcancer patient–derived xenografts (PDX; ref. 18) and clinicalsamples, we phenotypically characterize and functionally provethe existence of a novel subset of CICs, regulated by the micro-environment, driving lung tumor dissemination and metastasisand we identify CXCR4 signaling as a potential target for noveltherapeutic strategies.

Materials and MethodsList of antibodies

Antibodies used for FACSanalysiswere anti-humanPECD133/1, FITC CD326 (EpCAM) (Miltenyi Biotech); APC CD187(CXCR4; BD Pharmingen); anti-mouse PerCP-eFluor 710 MHCclass I, anti-human PerCP-eFluor 710 CD31 (eBioscience); andanti-human PC5 CD34, PC5 CD45 (Beckman Coulter). Anti-human FITC cytokeratin (CK3-6H5, Miltenyi Biotech) was usedfor immunofluorescent staining. For IHC analyses, followingantibodies were used: anti-human CD133/1 (AC133, MiltenyiBiotech), CXCR4 (CXC-4000-RM, Biotrend), keratin, Pan Ab-1(Clone AE1/AE3, Dako), CD34 (clone QbndN/10, Neomarkers),and low-molecular-weight cytokeratins (clone AE1/AE3/PCK26,VentanaMedical Systems). Detailed protocols for flow cytometry,immunofluorescence, and IHC staining are in SupplementaryMaterials and Methods.

Tissue dissociation and cultureClinical specimens were obtained from a consecutive series of

97 consenting patients (17 also had matching lymph nodesavailable). The protocol was approved by the Internal Reviewand the Ethics Boards of the Fondazione IRCCS IstitutoNazionaleTumori of Milan.

The protocol for primary tumors, lymph nodes, PDXs, andmurine lungs tissue dissociation was already described (15).Culture method to obtain cancer tissue–originated spheroids(CTOS) and murine lung tissue spheroids was adapted fromKondo and colleagues (19). Protocols are detailed in Supplemen-tary Materials and Methods.

Cell lines and treatmentsAll tumor cell lineswere cultured inRPMIwith 10%FBSat 37�C

under 5% CO2. The LT73 primary cell line was derived from asurgical specimen of a male patient with lung adenocarcinoma.TheH460 cell line (lung large cell carcinoma)was purchased fromATCC and authenticated by DNA short tandem repeat (STR)profiling.

To induce EMT, cell cultures were treated with TGFb (5 ng/mL;Peprotech) for 5 up to 15 days.

CTCE-9908 inhibitor and corresponding scramble controlpeptide were kindly provided by British Canadian BioSciencesCorp. and used for in vitro cell culture treatment at 10 mg/mLfor 30 minutes at 37�C in the appropriate growing medium(RPMI-1640 plus 1% FBS or in stem cell medium), before FACSanalysis or in in vitro invasion assays, detailed in SupplementaryMaterials and Methods.

Animal studiesIn vivo experiments were approved by the Ethics Committee for

Animal Experimentation of the Fondazione IRCCS IstitutoNazio-nale dei Tumori, according to institutional guidelines.

PDX were established from primary lung cancers from patientsundergoing surgical resection, who gave their informed consentand expanded as previously described (18).

For in vivo combination therapy, cisplatin (Teva Pharma) wasadministered i.v. at 5 mg/kg every 7 days for 3 weeks, startingwhen tumors were palpable (�50mm3). CTCE-9908was admin-istered i.p. (50 mg/kg) once every 5 days after cisplatin admin-istration for 3 weeks.

A human-in-micemodel of bonemetastasis was established, aspreviously reported (20). A bone metastasis assay was performedby injecting s.c. close to human bone implant 1� 105 cells sortedfrom PDXmodels for CD133 and CXCR4 or by intracardiac (i.c.)route, injecting in the left ventricle 1 � 104 sorted cells in 100 mLphysiologic solution. See also Supplementary Materials andMethods.

Integrated strategies to detect disseminating tumor cells inmurine lung

To set up the most robust strategy to detect disseminatingtumor cells (DTC) inmurine lungs, intravenous injection of scalardoses of H460 cells was performed in SCID mice (n ¼ 2 mice foreach dose) and lungs were analyzed 2 hours after injection byFACS, real-time PCR, and IHC.

For FACS analysis, injected tumor cells were identified usinga negative gating strategy to exclude 7-aminoactinomycin D(7-AADþ) dead cells and mouse H2Kþ cells that was able tospecifically detect as few as 103 single tumor cells in murine lungs(Supplementary Fig. S1A and Supplementary Materials andMethods).

For detection of tumor cells in murine lungs by real-time PCR,analysis of human-specific b2 microglobulin (B2M) housekeep-ing gene expression was performed (Supplementary Fig. S1B andSupplementary Materials and Methods).

IHC analysis on formalin-fixed, paraffin-embedded (FFPE)lung tissue for pan-cytokeratins (CK) expression was able torobustly identify single cells when injected at 104. To detect lowernumber of single cells, at least 5 serial sections of FFPE lung tissue,cut at 1.5-mm thickness, had to be analyzed (SupplementaryFig. S1C).

ResultsThe subset of CD133þCXCR4þ lung CICs is endowed withincreased seeding and metastatic potential

To verify the presence of a metastatic subset in lung CICs, weanalyzed by FACS 78 primary human non–small cell lung cancer(NSCLC) for CD133 and CXCR4 expression: double-positivecells accounted for 0.3% of total tumor cells (median value range,0%–8%) and represented 20% of CD133þ cells (median value

Identification of Lung Cancer Metastasis-Initiating Cells

www.aacrjournals.org Cancer Res; 75(17) September 1, 2015 3637




E LT111 CD133+/CXCR4+ metastasis

CXCR4

CD

133

Mouse H2K

**C

D13

3+C

XC

R4+

(%)

Primarytumors

Lymphnodes

A

C

Lung invasion

CD133–/CXCR4–CD133+/CXCR4+ CD133+/CXCR4–

0

1

2

4

5

LT73

H460

LT111-PDX

×40

×40

×40

Area of invasion

×40 ×40

6 Months after injection

******

D

Area of lung invasion (%)

***

** **

**

Tum

or c

ells

in th

e lu

ng (%

)

B 3 Months after injection

Tum

or c

ells

in th

e lu

ng (%

)

***

CD133+

CXCR4+CD13

3+CXCR4–

CD133–

CXCR4–

CD133+

CXCR4+

CD133+CXCR4+

CD133+CXCR4–

CD133–CXCR4–

H460

0 20 40 60

LT73

LT111

CD133+

CXCR4–CD13

3–CXCR4–

0.00 0.0

1.5

3.0

4.5

6.0

7.5

0.02

0.04

0.06

0.08

CD

133

CX

CR

4

Figure 1.CD133þCXCR4þ cells have enhanced metastatic potential. A, percentages of CD133þCXCR4þ cells in paired primary tumors and lymph node metastases (n ¼ 17).P ¼ 0.01. B, percentage of human tumor cells detected by flow cytometry in murine lungs 3 months and 6 months after i.v. injection of CD133þCXCR4þ,CD133þCXCR4�, CD133�CXCR4� cells. n ¼ 13 mice per group at 3 months and n ¼ 7 mice per group at 6 months. �� , P � 0.001. C, representative images ofhematoxylin and eosin histologic sections of murine lungs scanned with Aperio ePathology System, 6 months after injection of CD133þCXCR4þ, CD133þCXCR4�,and CD133�CXCR4� cells sorted from LT111-PDX, LT73, and H460. Areas of tumor invasion for each model are shown on the right at higher magnification(�40); yellow arrows, representative infiltrating metastatic cells; green arrows, polymorphonuclear leukocytes associated with tumor foci. D, invasion of murinelungs was calculated by scanning four different layers of each histologic section in C and quantifying the percentage of invaded areas for each single lungwith ImageScope Software. Bars are the mean percentage of invaded area � SEM. � , P � 0.05; �� , P � 0.01. E, representative FACS analysis of LT111 CD133þ

CXCR4þcell–generated metastasis, 6 months after injection; tumor cells are identified as mouse H2K� (left dot plot) and analyzed for CD133 and CXCR4expression (right dot plot). IHC staining confirming CD133 and CXCR4 positivity is shown on the right.

Bertolini et al.

Cancer Res; 75(17) September 1, 2015 Cancer Research3638




range, 0%–100%). Interestingly, when primary tumors werecompared with matched lymph node metastasis (n ¼ 17), thefrequency of CD133þ cells coexpressing CXCR4 was significantlyhigher in metastases (3-fold enrichment, P ¼ 0.01; Fig. 1A).

To dissect the metastatic potential of different subpopulationsof lung CICs, CD133þCXCR4þ and CD133þCXCR4� cells sortedfrom different PDXs (n ¼ 5), from a primary cell line freshlyestablished from a lung adenocarcinoma sample (Lung Tumor73, LT73) and from H460 cells were injected in the tail vein ofSCID mice. CD133�CXCR4� non–CICs were used as control. Todetect and monitor during time tumor cells in murine lungs, weset up and validated a method on the basis of flow cytometricanalysis, not relying on specific human antigens expressionand supported by qRT-PCR and IHC analysis (SupplementaryFig. S1A–S1C). Threemonths after injection, rare tumor cells weredetectable in murine lungs and were consistently higher in thegroups injected with CD133þCXCR4þ (2-fold increase þ/þ vs.þ/�; 5-fold increase þ/þ vs. �/�, P ¼ 0.001; Fig. 1B). After anextended period of time (6 months), CD133þCXCR4þ cells,sorted from one PDX model (LT111) and LT73 and H460 celllines, showed their full metastatic potential and greatest ability tocolonize murine lungs compared with other injected groups asreveled by both FACS analysis (4.5-fold increase, P¼ 0.0003þ/þvs. þ/�; 6.4 fold increase, P ¼ 0.0006 þ/þ vs. �/�; Fig. 1B) andhistologic analysis (Fig. 1C): software-assisted quantification ofmacroscopic areas of tumor invasion clearly demonstrated thehigher degree of infiltration of lung parenchyma of double-positive cells compared with moderate and slight degree ofinfiltration of single-positive and double-negative cells (P ¼0.0003þ/þ vs.þ/�; P¼ 0.0004þ/þ vs.�/�; Fig. 1D). A relevantinflammatory infiltrate, mainly consisting of polymorphonuclearleukocytes, was also observed associatedwith tumor foci (Fig. 1C)possibly explaining the long latency periods before metastaticcells proliferation. The phenotype of injected double-positivecells remained unchanged during early colonization steps, imply-ing that their survival advantage enables them to support metas-tasis initiation (Fig. 1E).

To confirm the high metastatic properties of CD133þCXCR4þ

cells and their selective advantage in seeding different organs, weexploited an innovative model of humanized mice implantedwith human bone (20, 21), the second more frequent metastaticsite ofNSCLC. The implantedhumanbone, detected in themouseflank by X-rays examination, was viable as indicated by thepresence of bone marrow cells, mineralized areas, stromal cells,and neovascularization (Fig. 2A). Unsorted tumor cells andCD133þCXCR4þ, CD133þCXCR4�, and CD133�CXCR4� sub-sets were isolated from two PDXs (LT59 and LT111) and injectedboth s.c. and intracardially in mice bearing bone, to modeldifferent aspects of bone seeding. CD133þCXCR4þ cells wereable to colonize human bone generating metastatic lesion afterboth s.c. and i.c. injectionwithhigh efficiency (65%,9of 14mice),conversely CD133þCXCR4� and CD133�CXCR4� cells did not(Fig. 2B and Supplementary Table S1). Enhanced bone

CD133+/CXCR4+ lung metastasis

0

100O. T

h. (m

m)

200

300

400

500

*

Others CD133+/CXCR4+

Osteoid thickness

E

×60×10

B

Osteocytes

Vessels

Stromal cells

×40

A

Humanbone

×60×10

CD133+/CXCR4+ bone metastasis

DC

×20

CD133+/CXCR4– cells

×20

CD133+/CXCR4+ cells

×60

CD

34

Figure 2.CD133þCXCR4þ cells efficiently initiate bone metastasis. A, left, X-rayimaging of mouse implanted with human bone fragment in the flank. Right,hematoxlyin and eosin on human implanted bone showing viable osteocytes,stromal cells, human vessels (top). IHC staining for human CD34þ (bottom)confirmed the presence of human vessels (magnification, �40). B, IHCstaining for low molecular cytokeratins of bone metastasis derived fromCD133þCXCR4þ cells (magnification, �10) and inset at higher magnification(�60). C, trichrome staining of human implanted bone in mice injected withCD133þCXCR4� (top) and CD133þCXCR4þ (bottom) showing the presence

of new collagen fibers (stained in blue). D, mean values � SD of the osteoidthickness of humanbone inCD133þCXCR4þ cell-injectedmice comparedwithall other groups (n¼ 4). � , P� 0.05. E, hematoxlyin and eosin staining of lungof mouse bearing human bone after s.c. injection of CD133þCXCR4þ cellsshowing massive lung metastatization (magnification, �10; left) and inset oftumor cells at �60 magnification (right).






A

0

CD133+

CXCR4+

CD133

+ /CXCR4

+

CD133+

CXCR4+

CD133

+ /CXCR4

+

CD133+

CXCR4+

CD133

+ /CXCR4

+

0.51

1.52

2.53

Pos

itive

cel

lsfo

ld c

hang

e

H460 tumorsCTRLCISPTCISPT + CTCE-9908

C

0

20

40

60

80

LT128 LT111 LT59 LT138

Inva

ding

cel

ls/fi

eld

SDF-1

*

*

* *

CTRL CISPT CISPT+CTCE-9908E

×4×4×4

0.0

5.0

10.0

15.0

% P

ositi

ve c

ells

(lung

met

asta

ses)

H460 CTRLCISPTCISPT+CTCE-9908

*

F

*

0.0

10.0

20.0

30.0

% P

ositi

ve c

ells

(lung

DTC

s)

LT111-PDX CTRLCISPTCISPT+CTCE-9908

** **

**

B

Scramble–1

0123

10152025

CTCE-9908C

D13

3+C

XC

R4+

cel

ls in

inva

ding

fra

ctio

n (r

elat

ive

frequ

ency

) *

0.01

0.1

1

10

CXCR4 OCT4 NANOG SOX2

mR

NA

exp

ress

ion

leve

ls

rela

tive

to u

ntre

ated

CTR

L

H460 tumorsCISPT CISPT+CTCE-9908

D

0

10

20

30

40

50

LT128 LT111 LT59 LT138

Inva

ding

cel

ls/fi

eld

FBS NC

Scramble

**

*

G

NCScrambleCTCE-9908

CTCE-9908

CK

s

Figure 3.CXCR4 inhibition prevents metastatic dissemination of chemoresistant CD133þCXCR4þ cells. A, invasion assay performed on PDX cells treated with CTCE-9908or scramble peptide and chemoattracted with SDF-1 (left) or FBS (right). After 72 hours, invaded cells were counted in four random fields of the inserts; data,mean value � SD. � , P < 0.05. B, relative frequency of CD133þCXCR4þ cells in invading cells fraction, recovered after treatment with CTCE-9908 and scramblepeptide, compared with PDX total population (n ¼ 4 models in duplicate experiments). � , P < 0.05. C, fold change in CD133þ and CXCR4þ cells in treatedH460 xenografts compared with control, evaluated by FACS. Data are mean % � SD. n ¼ 4 per group. D, qRT-PCR for stemness genes expression in cisplatin orcisplatin þ CTCE-9908–treated H460 xenografts. Untreated control tumors were used as calibrator. E, representative IHC staining for human cytokeratins ofH460 lung metastases in control, cisplatin, and cisplatin þ CTCE-9908–treated groups. Magnification, �10. Percentage of CD133þ and CXCR4þ cells in H460 lungmetastases (F) and in LT111 DTCs (G) in control and treated groups, as evaluated by flow cytometry. Data, mean value � SD. n ¼ 4 mice per group. � , P < 0.05;�� , P � 0.01; �� , P � 0.001.

Bertolini et al.





remodeling, revealed by large areas of new bone apposition andan increased osteoid thickness due to metastasis formation, wasevident in mice injected with CD133þCXCR4þ cells comparedwith all other groups (P < 0.05; Fig. 2C and D). Notably, CD133þ

CXCR4þ cells injected both s.c. and i.c. in mice carrying humanbone implants also showed significantly higher ability to dissem-inate and initiate lung metastasis, with extensive replacement ofnormal lung parenchyma with cancer cells (Fig. 2E), comparedwith CD133þCXCR4� and CD133�CXCR4� cells or unsortedcells (CD133þCXCR4þ vs. others groups, P¼ 0.007; Supplemen-tary Table S1).

Altogether, these results prove the high dissemination andcolonization potential of CD133þCXCR4þ cells and their abilityto support metastasis development at distant organs.

Inhibition of the CXCR4 pathway prevents metastatic spread ofchemoresistant CD133þCXCR4þ cells

To block the migration of the CD133þCXCR4þ metastaticsubset, we tested the CXCR4 inhibitor CTCE-9908, a small pep-tide analogue of SDF-1, previously shown to be effective inpreventing metastatic dissemination in different cancer models(22–25).

In vitro CTCE-9908 treatment could block invasion of PDX-derived tumor cells (n ¼ 4 PDX models) induced by both SDF-1and FBS (Fig. 3A). Interestingly, we also verified that the invadingfraction of PDXs cells attractedwith a nonspecific signal (FBS)was10-fold enriched in CD133þCXCR4þ cells compared with totalpopulation and that CTCE-9908 inhibitor was able to specificallyblock their invasion (4-fold change; Fig. 3B), thus pointing atCXCR4 axis as a pivotal mediator of lung CIC invasion.

To test the efficacy of CXCR4 inhibition in a clinically relevanttherapeutic setting, we treated in vivo metastatic H460 xenograftswith CTCE-9908 in combination with standard chemotherapy.We previously reported that cisplatin treatment, although effec-tive in reducing tumor size, induces enrichment of chemoresistantCICs in the residual tumor (15). Combination treatment withCTCE-9908 had a similar efficacy in inhibiting subcutaneoustumor growth as cisplatin (Supplementary Fig. S2A), but it wasable to prevent the relative increase of CD133þCXCR4þ cells andconcomitant upregulation of stemness genes induced by chemo-therapy alone (Fig. 3C and D).

FACS analysis of H460 lung metastases in control mice dem-onstrated a 10-fold enrichment for CD133þCXCR4þ cells com-pared with matched s.c. xenograft (P < 0.05), confirming theirpivotal role also in the spontaneous metastasis development(Supplementary Fig. S2B).

Cisplatin treatment primed the dissemination of theCD133þCXCR4þ subset (2-fold change) inducing a boost inmetastasis formation (1.8-fold change) compared with controlgroup (Fig. 3E and F). Combination therapy with CXCR4inhibitor was able to counteract cisplatin effect, drasticallyreducing metastatic formation (2-fold decrease) and concom-itantly decreasing the subsets of CD133þ and CD133þCXCR4þ

Figure 4.CD133þCXCR4þ and EpCAM� cells are crucial in early steps of colonization. A,time course analysis of i.v. injected LT111, LT73, and H460 tumor cells. Arrow

on the left indicates increasing metastatic behavior of tested tumor cells.Percentages of tumor cells (left, y-axis) and relative fold increase inCD133þCXCR4þ and cells CD133þEpCAMþ subsets are evaluated by FACSanalysis ofmurine lungs. Data representmean�SEM.n¼ 2mice at given timepoint. B, qRT-PCR analysis for stemness genes expression of H460 cells,2 hours, 1 and 3weeks after i.v. injection inmurine lung. H460parental cell linewas used as calibrator.






EpCAM

CD

133

Tum

or c

ells

in th

e lu

ng (%

)

CD

133+

CX

CR

4+E

pCA

M−

(%)

Total cells

Invadingcells

**A

×20

×40

×20

×40

PDX DTCs-tumorE FD

LT59-CTOSLT59-DTCs

CX

CR

4C

EpCAM

LT59 DTCs-tumorLT59 CTOS-tumor

92.7%85.5%

0.6% 9.8%

C

Lung spheroids cytospin

CK

s

mB2mEndogenous control

hB2MLung spheroids

hB2M Murine lung

BCD133+ cells in

invading cells fractionCD133+ cells in total PDX cells

18%0.6%

CX

CR

4

CX

CR

4

EpCAM EpCAM

0.0001

0.0100

1.0000

100.0000

LT138DTCs

LT73DTCs

LT187DTCs

LT59DTCs

LT128DTCs

mR

NA

exp

ress

ion

leve

ls

rela

tive

to C

TOS

-PD

Xs

CDH1 FN1 SNAI2

0.1

1.0

10.0

100.0

1000.0

10000.0

LT138DTCs

LT73DTCs

LT187DTCs

LT59DTCs

LT128DTCs

mR

NA

exp

ress

ion

leve

ls

rela

tive

to C

TOS

-PD

Xs

OCT4 NANOG SOX2

Tumor cells in the lung:12.3%

Tumor cells in the lung:0.3%

LT 59-CTOS LT59-DTCs

7AAD/H2K mouse7AAD/H2K mouse

H

10.2% CTOS-tumor

G

DTCs-tumor

**

CD133 CD133

FSC

-H

EpCAM

CX

CR

4C

0100200300400500600700800900

0 10 20 30 40 50 60 70Day

TW

(m

g) 5/6

4/6

0

200

400

600

800

1K

0

200

400

600

800

1K

FSC

-H::

FSC

-hei

ght

100 101 102 103 104

FSC

-H

100 101 102 103 104

100

101

102

103

104

100 101 102 103 104

0.0

0.2

0.4

0.6

0.8

1.0

35

25

15

51

0

1.000 E-2

1.000 E-1

1.000 E+1Amplification plot

ΔRn 1.000

10 15 20Cycle

25 30 35 40 45

Figure 5.DTCs are enriched in CD133þCXCR4þEpCAM� CICs and generate tumors with distinctive features. A, percentages of the CD133þCXCR4þEpCAM� subset withinCD133þ cells in total PDX cells and in the corresponding invading cells fraction recovered after the invasion assay, evaluated by FACS as shown on theright. n ¼ 5 PDXs. P ¼ 0.01. B, representative image of lung spheroids culture and immunofluorescence staining for human CKs on lung spheroids cytospins,identifying single human DTC (top). Representative qRT-PCR amplification plot (bottom) showing increased expression of human B2M gene in lung spheroids (redcircle) compared with lung tissue from PDX-bearing mice (green circle). (Continued on the following page.)

Bertolini et al.





cells in lung metastasis compared with cisplatin-treated group(2.6- and 2.3-fold change, respectively; Fig. 3E and F andSupplementary Fig. S2C).

Next, we also proved in a highly disseminating PDX model(LT111) that in vivo combination treatment with CTCE-9908 hada modest effect on tumor growth inhibition (Supplementary Fig.S2D) but was able to prevent tumor spread associated withcisplatin treatment (1.5-fold change) and to efficiently reducethe fraction of disseminated CD133þCXCR4þ CICs to murinelungs (4.7-fold change; Fig. 3G and Supplementary Fig. S2E).

We additionally verified in cell lines that CXCR4 is also fre-quently expressed (mean, 70%; range, 50%–90%) in anothersubset of CICs previously identified as chemoresistant(CD133þABCG2þ; ref. 15) and that CXCR4 inhibition could alsoprevent enrichment of CD133þCXCR4þABCG2þ induced bychemotherapy (data not shown).

Our results suggest that combination treatment with CXCR4inhibitor could specifically impair dissemination of chemoresis-tant and metastatic CD133þCXCR4þ cells.

CD133þCXCR4þ and EpCAM� CICs have survival advantage atectopic sites in early steps of colonization

To examine the role of CD133þCXCR4þ cells in different stepsof the metastatic cascade, we injected i.v. 1 � 106 unsortedlung tumor cells endowed with different metastatic behavior(LT111-PDX cells, weakly metastatic; LT73 cell line, moderatelymetastatic; H460 cell line, highly metastatic), and we monitoredin time course experiment which tumor cell subsets were prefer-entially selected to drive specific steps ofmetastasis development.

Flow cytometric analysis of murine lung 2 hours after injectionconfirmed the presence of tumor cells that maintained the samephenotype of original tumor cells (Supplementary Fig. S3A). After1 week, the number of tumor cells drastically decreased (20-fold)with a concomitant enrichment for theCD133þCXCR4þ subset inthe surviving tumor fraction (Fig. 4A). We also detected in LT73and LT111 surviving cells a significant increase for a fraction ofCD133þ CICs negative for the expression of epithelial antigenEpCAM compared with parental injected tumor cells (18-foldincrease), whereas the parental H460 cell line does not express atall EpCAM marker (Fig. 4A). Surviving LT111 cells remained inmurine lung, without proliferating, for up to 3months and stablyshowed enrichment for CD133þCXCR4þ and CD133þEpCAM�

subsets (Fig. 4A). Conversely, during metastatic growth of LT73and H460 cells, both subsets of CD133þCXCR4þ and CD133þ

EpCAM� cells, diminished and reverted to the original percentageof the parental cell linewhenmetastases reachedmacroscopic size(Fig. 4A and Supplementary Fig. S3B). Expression of stemnessgenes followed the same trend of CD133þCXCR4þ cells in thedifferent phases of metastatic development (Fig. 4B).

In summary, these findings confirm the primary role ofCD133þCXCR4þ CICs in metastasis process, due to a superior

survival advantage and colonization ability, and reveal the exis-tence of the CD133þEpCAM� subset strongly enriched in earlysteps of colonization.

The subset of CD133þCXCR4þEpCAM� CICs sustains lungcancer dissemination and is endowed with high tumorigenicand metastatic potential

To further investigate the involvement of the CD133þCXCR4þ

and EpCAM� subsets in a preclinical model of spontaneoustumor dissemination, we analyzed lungs of PDX-bearing SCIDmice.

While generally no histologic metastases were evident, FACSanalysis was able to detect a subset ofDTCs in 9 of 11 PDXmodels(n¼ 6 adenocarcinoma; 1 large cell carcinoma, LCC; 2 squamouscell carcinoma models, SCC), in a percentage (range, 0.001%%–

0.01%) consistent with that estimated from human B2M expres-sion by qRT-PCR (data not shown).

Lung DTCs were already detectable at early stages of subcuta-neous tumor development and remained stable during tumorgrowth (Supplementary Fig. S4A). Notably, immunophenotyp-ing of lung DTCs revealed a constant increase for the fraction ofCD133þCXCR4þ CICs (P ¼ 0.04) and in particular a predomi-nant enrichment for the subset of CD133þEpCAM� cells (P ¼0.007) compared with parental subcutaneous tumors (Supple-mentary Fig. S4A and S4B), confirming evidence obtained in i.v.time course experiments.

Moreover, an in vitro invasion assay on dissociated PDX cells(n ¼ 5) proved that CD133þCXCR4þEpCAM� cells representedthe CIC subset more significantly enriched in the invadingfraction compared with total population (90-fold, P ¼ 0.01;Fig. 5A).

To functionally characterize the subset of lung DTCs enrichedin disseminating CICs, we adapted an in vitro culture methoduseful to enrich for tumor cells (19); from 6 different PDXs(4 adenocarcinoma, 1 SCC, 1 LCC), we generated cultures ofCTOS and corresponding lung tissue spheroids, highly enrichedin DTCs compared with freshly dissociated lung tissue (range, 70-to 145-fold increase) as also confirmed by qRT-PCR analysis forhB2M expression and by staining for human cytokeratins (Fig. 5Band Supplementary Fig. S4C).

Real-time analysis of lung spheroids cultures showed a strongupregulation for stemness genes and for a panel of EMT-relatedgenes in DTCs compared with matching PDX-CTOS, subtendingan enrichment for the subset of CD133þ stem–like cells endowedwith mesenchymal features (i.e., EpCAM�; Fig. 5C).

To functionally test tumorigenic properties of DTCs, we s.c.injected in NOD-SCID mice 1 � 105 lung spheroids dissociatedcells, containing approximately 200DTCs and1�105PDX-CTOSdissociated cells generated from LT59 PDX. DTCs were able tooriginate tumors that closely resembled histology and phenotypeof original PDX (Fig. 5D) but contained a larger amount of

(Continued.) Mouse B2m endogenous control gene is equally expressed between samples. C, relative quantification of stemness genes (left) and EMT-associatedgenes expression (right) in lung DTCs compared with corresponding PDX-CTOS, used as calibrators (n ¼ 2 independent experiments). D, hematoxylin andeosin of PDX and corresponding DTC tumor, derived from s.c. injection of lung spheroids (magnification,�20 and�40). E, flow cytometric analysis of LT59 tumorfrom CTOS and DTCs indicating the enrichment for the CXCR4þEpCAM� subset (red quadrant, lower dot plots) within gated CD133þ cells (upper dot plots).F, growth curves of secondary tumors from s.c. injection of 1� 105 LT59-CTOS and LT59-DTCs primary tumors cells. Data are the averages of tumors weight� SD.n ¼ 6 mice per group. Tumor take is reported. G, percentage of lung disseminated tumor cells evaluated by FACS in mice bearing secondary LT59-CTOS andLT59-DTC tumors. n ¼ 4 mice per group. � , P < 0.05. H, representative FACS analysis of lungs of mice injected i.v. with 1 � 105 LT59-CTOS cells andLT59-lung spheroid cells, enriched in DTCs. Metastases from DTCs (mouse H2K� cells) were analyzed for CD133 and EpCAM expression.






BA

LT73 parental

LT73EMT

****

CD133

CD133+CXCR4+EpCAM−CD133+CXCR4+

LT73parentall

LT73EMT

Rel

ativ

e %

of

CD

133+

sub

sets

F

H

**

Ctrl+CAF-CM1+CAF-CM2+CAF-CM3+CAF-CM4

LT73 CD133– cell line

C

*

CTOS CTOS-TGFβ

0.8% 3.4%

CD133 CD133

EDLT73

CD133

LT73 CD133– cell line

Untreated5 Weeks after

TGFβ withdrawal

CD133 CD133

0.17% 0.14%0.00%

CD133

0.14%

EpCAMEpCAMEpCAMEpCAMEpCAM

9.2%0.0001%

I

CtrL CAF-CM3

CAF-CM4

CAF-CM1

CAF-CM2

Rel

ativ

e %

of c

ell s

ubse

ts

CtrL

J

CAF-CM3

CAF-CM4

CAF-CM1

CAF-CM2

Lung dissemination

Tum

or c

ells

in th

e lu

ng (%

)

CD

133+

(%)

0.1

10.0

1,000.0

Rel

ativ

e m

RN

A e

xpre

ssio

n

CAF

0

010

20

30

4080

120

LT73-EMTMSC

CAF-CM1

CAF-CM2

CAF-CM3

CAF-CM4

Ctrl

CD

133+

(%)

LT73 CD133-

0.0

CD133CD133+CXCR4+EpCAM–

0.51.01.52.02.53.03.5

mR

NA

exp

ress

ion

leve

lsre

lativ

e to

LT7

3CD

133–

Ctrl SNAI2

FN1VIMENTIN

G

+ TGFβ

Figure 6.Microenvironment stimuli generate disseminating CICs. A, percentage of CD133þ cells evaluated by FACS in the parental LT73 cell line and in LT73-EMT cells.n ¼ 10 independent experiments. P ¼ 0.0001. B, relative mRNA expression of EMT-associated genes and stemness genes in LT73-EMT cells, MSC, and CAFs.CAF cells were used as calibrator for analysis. (Continued on the following page.)

Bertolini et al.





CD133þCXCR4þEpCAM� cells compared with parental xeno-graft (Fig. 5E). Moreover, cells from DTC tumors maintained aslightly increased expression of EMT-related genes comparedwithCTOS tumors and exclusively possessed the ability to proliferatein vitro as spheres in anchorage-independent condition (Supple-mentary Fig. S4D and S4E).

Remarkably, in vivo s.c. injection of cells from DTC tumorsdemonstrated an enhanced ability to generate secondarytumors (Fig. 5F) and higher lung dissemination ability com-pared with CTOS tumor (disseminated tumor cells: 0.6% � SD0.1 vs. 0.09% � SD 0.03; Fig. 5G).

Finally, intravenous injection of LT59 lung spheroids dis-sociated cells (200 DTCs) in NOD-SCID mice generated lungmetastases, which again showed an increased fraction CD133þ

CXCR4þEpCAM� cells, whereas up to 1 � 105 LT59 cells andLT59-CTOS cells failed to initiate metastasis (Fig. 5H).

Our evidence proves the existence of an invasive and highlytumorigenic CD133þCXCR4þEpCAM� subset that specificallydrives lung tumor dissemination.

Microenvironment stimuli modulate the subset of CD133þ

CXCR4þEpCAM� lung disseminating CICsThe mesenchymal properties displayed by CD133þ

CXCR4þEpCAM� disseminating CICs, lead us to investigate thepossible involvement of EMT activation in their generation/mod-ulation (26).

Induction of EMT through TGFb treatment of the LT73 tumorcell line (LT73-EMT) evaluated as morphologic changes, modu-lation of EMT-related genes, and increased invasive potential(Supplementary Fig. S5A–S5C), caused a 10-fold increase in thefraction of CD133þ CICs (P ¼ 0.0001) with a concomitantupregulation of stemness gene expression (Fig. 6A and Supple-mentary Fig. S5D). We also verified that LT73-EMT cells showedan expression pattern for EMT and stemness-related genes moresimilar to mesenchymal stem cells (MSC) and then to differen-tiated adult mesenchymal cells (i.e., fibroblasts derived fromprimary lung tumors; Fig. 6B).

Increase of CD133þ CICs in LT73-EMT cells was functionallyproved by enhanced in vivo tumorigenic potential and ensuingtumors stablymaintained an augmented content of CD133þ cellscompared with LT73 parental tumors (Supplementary Fig. S5Eand S5F).

Examination of CD133þ CIC subsets in LT73-EMT cellsdemonstrated a 3.8-fold increase (P ¼ 0.04) for metastaticCD133þCXCR4þ cells that was mainly driven by the expansionof CD133þCXCR4þEpCAM� subpopulation (13.5-fold increase,P ¼ 0.02; Fig. 6C).

In PDXs and primary CTOS cultures (n ¼ 5), EMT induced byTGFb was able to enrich for CD133þCXCR4þEpCAM� dissemi-nating CICs (15-fold change, P ¼ 0.05; Fig. 6D and Supplemen-

tary Fig. S5G), associated with an increased in vitro invasivenessand upregulation of EMT and stemness genes (SupplementaryFig. S5H and S5I).

To evaluate whether the CIC enrichment was due to anexpansion of the pre-existing CD133þ pool or to de novogeneration, we depleted CD133þ cells by FACS from the LT73cell line, creating the LT73-CD133� line that stably maintaineda CD133� phenotype during in vitro culturing (Fig. 6E). TGFbtreatment of LT73-CD133� cells induced the generation of afraction of CD133þ cells (mean � SD, 0.14%� 0.05%), similarto that observed in LT73 parental cells (mean � SD, 0.09% �0.02%), that however exhibited a lower expression of EpCAMcompared with endogenous CD133þ cells (Fig. 6E). Upon TGFbwithdrawal, LT73-CD133� cells slowly reverted to an epithelialphenotype, whereas the fraction of de novo generated mesen-chymal CD133þ cells persisted, indicating a stable reprogram-ming (Fig. 6E and Supplementary Fig. S5L and S5M). Consis-tently, tumors derived from s.c. injection of LT73-CD133� EMTappeared with a shorter latency period than LT73-CD133� cellsand maintained a similar fraction of de novo generated CD133þ

cells (Supplementary Fig. S5N and S5O).To verify whether physiologically relevant tumor microenvi-

ronment stimuli could show similar proficiency in generatinglung CICs, LT73-CD133� cells were cocultured with conditionedmedium (CM) from cancer-associated fibroblasts (CAF) isolatedfrom primary human lung tumors. A variable but constantlydetectable fraction of de novo CD133þ cells was observed afterexposure to different CMs, mirrored by a parallel modulation ofEMT genes (Fig. 6F and G) indicating that also stimuli frommicroenvironment are able to endow tumor cells with mesen-chymal traits associated with CIC generation.

Because stromal SFD-1 has been proven to induce EMTactivation in different tumor types (27, 28), we explored therole of SDF-1 as a possible mediator of CAF effects on tumorcells. In line with reported evidence, we confirmed that in vitrotreatment of the LT73 cell line with SDF-1 resulted in upregula-tion of EMT genes (Supplementary Fig. S5P). Treatment ofLT73-CD133� cells with CMs and CXCR4 inhibitor CTCE-9908 demonstrated a partial prevention of EMT genes modu-lation compared with CMs alone, thus confirming that SDF-1represents one of the CAF-secreted factors that mediate EMTactivation in tumor cells (Fig. 6G).

CM-generated CD133þ CICs were functionally associated withan enhanced in vivo tumorigenicity compared with untreatedCD133� cells (Fig. 6H), and interestingly, the ensuing xenograftsdisplayed a CD133 fraction reflecting the proficiency of CAF-CMsto create CICs in vitro (Fig. 6I). Notably, the relative percentages ofde novo generated CD133þCXCR4þEpCAM� cells detected in CM-treated LT73-CD133� tumors correlated with their disseminatingpotential to murine lungs (Fig. 6I and J).

(Continued.) C, fold change of CD133þCXCR4þ andCD133þCXCR4þEpCAM� subsetswithin CD133þ cells in LT73 parental and LT73-EMT cell lines, detected by FACS.Mean � SD. n ¼ 10 independent experiments. � , P < 0.05. D, representative FACS analysis for one primary CTOS culture (left) and after TGFb treatment (right),showing the enrichment for CD133þ cells (top dot plots) and the relative increase in the CD133þCXCR4þEpCAM� subset (red quadrant, bottom dot plots).E, representative FACS analysis for CD133 expression in the LT73 cell line, in the LT73-CD133� cell line, untreated and treated with TGFb, and 5 weeks afterTGFb withdrawal. Expression of EpCAM within gated CD133þ cells is shown (bottom dot plots). F, mean % � SD of de novo-generated CD133þ cells in theLT73-CD133� cell line cultured with different CAF-CM. n ¼ 3 independent analysis. G, real-time PCR from EMT gene expression in the LT73-CD133� cell lineafter treatments with CMs alone or in combination with CTCE-9908. Untreated cells were used as calibrator. H, growth curves of tumors generated by s.c. injectionof 1 � 105 LT73-CD133� cells treated with CAF-CMs. Tumor weights (mg) reported are the mean � SD. n ¼ 8 per groups. � , P < 0.05. I, flow cytometric analysis oftumors in H for CD133þ cells and relative content of the CD133þCXCR4þEpCAM� subset. Data, mean � SD. n ¼ 4 per group. J, percentage of disseminatedcells detected by FACS in lungs of mice bearing tumors from LT73-CD133� cells treated with CAF-CMs. Mean � SD. n ¼ 4 per group. � , P < 0.05.






Our data demonstrate a dynamic phenotype of lung CICs andshow that stimuli from tumor microenvironment primarily gen-erate the subset of CD133þCXCR4þEpCAM� CICs functionallyassociated with tumor dissemination.

Disseminating CICs can be detected in lymph nodes of lungcancer patients and correlate with poor clinical outcome

Comparison of lymph node metastases and matched syn-chronous primary NSCLC tumors (n ¼ 17) reveled a 20-foldenrichment in the fraction of CD133þCXCR4þEpCAM� cellsin metastases (P ¼ 0.006; Fig. 7A), confirming in a clinical

setting our preclinical data for the existence of CD133þ

CXCR4þEpCAM� CICs related to tumor dissemination andmetastasis initiation.

To evaluate the potential clinical relevance of disseminatingCICs, we analyzed, by FACS, 78 surgically resected primaryNSCLC for expression of CD133, CXCR4, and EpCAM. As wepreviously reported, the overall frequency of CD133þ cells wasgenerally low with a median value of 1.3% (intertertile range,0.7%–2%). A subset of CD133þCXCR4þEpCAM� cells wasdetected in 50% of cases, independently from the total amountof EpCAMþ tumor cells. The CD133þCXCR4þEpCAM� fraction

Figure 7.Disseminating CICs in primary NSCLC correlate with poor prognosis. A, percentages of CD133þCXCR4þEpCAM� cells in paired primary tumors and lymphnodemetastasis (n¼ 17; P¼ 0.006) detected by FACS analysis; right, tumor cells are identified as CD45�CD34�CD31� cells and relative percentages of the CD133þ

CXCR4þEpCAM� subset are evaluated within gated CD133þ cells. B, Kaplan–Meier estimates of OS and DFS for 78 patients with NSCLC accordingto the presence of disseminating CICs (CD133þCXCR4þEpCAM�) within primary tumors.

Bertolini et al.





accounted for 12.5% of total CD133þ cells (range, 0.4%–100%).Therewas no correlation among the total amount ofCD133þ cellsand clinical variables, including age, sex, histologic subtype,grade, or stage (Supplementary Table S2). The frequency ofmigrating CICs (CD133þCXCR4þEpCAM�) within the CD133pool was more frequent in SCCs (P ¼ 0.017). Interestingly, bothCD133þ and CD133þCXCR4þEpCAM� content was associatedwith the presence of chronic obstructive pulmonary disease (P ¼0.025 and P ¼ 0.011, respectively), indicating a possible role ofthe lung microenvironment in modulating stemness and dissem-inating properties. More importantly, while the total amount ofCD133þ cells didnot correlatewith clinical outcome, thepresenceof the CD133þCXCR4þEpCAM� subset was negatively correlatedwith both disease-free survival (DFS, P ¼ 0.033) and overallsurvival (OS, P¼ 0.055; Fig. 7B). The association of disseminatingCICs with DFS remained significant also in subclass analyseswhen only nonsquamous tumors were analyzed (P¼ 0.038) andin a multivariate model even after adjustment for stage [P¼ 0.03;HR, 2.25; 95% confidence interval (CI), 1.05–4.82]. Particularly,in tumors with low content of CD133þ cells (�2%), the presenceof CD133þCXCR4þEpCAM� metastatic CICs was associatedin multivariate analysis with both shorter DFS (P ¼ 0.03; HR,3.34; 95% CI, 1.09–10.18) and OS (P ¼ 0.05; HR, 3.10, 95% CI,1.00–9.59), thus representing a risk factor for tumor relapse andpoor prognosis in patients with NSCLC.

DiscussionAlthough metastasis has generally been considered a mul-

tistep process requiring acquisition of specific genetic altera-tions providing tumor cells with crucial advantages (29),plasticity is gaining attention as a fundamental trait formetastasis-prone tumors, and the identification of definitephenotypes linked to this process appears crucial for itsunderstanding (30).

Here, we demonstrate that the subset of CD133þCXCR4þ lungCICs drives tumor cells seeding and metastasis initiation in lungcancer. The high metastatic potential of CD133þCXCR4þ cellsand their enhanced ability to seed and colonize different organs(i.e., lung and bone) is substantiated in experimental lung metas-tasis aswell as in anhumanizedmurinemodel of bonemetastasis:interestingly, in such an experimental model, the presence of a"human pre-metastatic niche" represented by implanted boneis conductive for generation of overt metastases in a consid-erable shorter time than in an experimental metastasis assay,stressing the importance of proficient niche to support anddictate tumor development and the challenge of recapitulatingthe human microenvironment in mouse (31–33).

Next, we prove for the first time that CXCR4 targeting is able tocounteract the chemotherapy-induced metastatic spread of che-moresistant fractions of CD133þCXCR4þ and possibly also ofCD133þABCG2þCICs (15) thatmostly express CXCR4 receptors,pointing at combination therapy with CXCR4 inhibitor as anattractive novel strategy to improve neoadjuvant and adjuvanttherapy for NSCLC.

Time course analysis of phenotypic changes in i.v. injectedtumor cells at different steps of metastatic cascade revealed inthe first phase of colonization an enrichment for a subset oflung CD133þCXCR4þ cells negative for EpCAM; similarenrichment was observed in spontaneous early disseminatedtumor cells detected in PDX preclinical models. These data are

in line with evidence in different tumor types for an earlytumor metastatic spread guided by cells endowed with mes-enchymal and stem-like features (11, 12, 34). Besides thephenotypic characterization, we functionally prove that lungDTCs, surviving in ectopic murine soil without metastaticoutgrowth, are indeed highly tumorigenic and metastaticwhen injected s.c. with Matrigel or i.v. in NOD/SCID mice,proving the metastasis initiation potential of the specificsubset of CD133þCXCR4þEpCAM� cells, as well as highlight-ing once more the need of a favorable metastatic niche todictate tumor growth (32, 33, 35).

Remarkably, tumors generated from DTCs, as well as patient'slymph nodes metastases, are enriched in the CD133þ

CXCR4þEpCAM� CIC subset, demonstrating in a clinical setting,the relevance of this subpopulation in metastasis formation.

The mesenchymal features of disseminating tumor cells aswell as the well-known link between EMT activation andacquisition of invasiveness and stem-like features promptedus to investigate the possible involvement of EMT in theirgeneration (7, 36, 37). Our data indeed confirmed that EMTtriggering generates CD133þ cells and, more interestingly, wedemonstrated for the first time that stimuli from tumor micro-environment, specifically those from CAF, could dictate de novocreation of the CD133þCXCR4þEpCAM CIC subset directlylinked to EMT induction and acquisition of in vitro and in vivoincreased dissemination. Similar data proving the capability ofmicroenvironmental stimuli to convert progenitor cells intocancer stem cells with metastatic properties are reported incolon cancer (10).

Moreover our finding showing that CAF-secreted SDF-1 pro-motes EMT activation and likely the acquisition of stemnessfeatures in lung tumor cells, in line with evidence reported incolorectal cancer (27), strengthens the potential therapeutic use ofCXCR4 inhibitor to prevent both dissemination of metastaticCICs and their generation through the impairment of tumor–stroma cross-talk.

The clinical relevance of our findings was demonstrated bythe fact that the frequency of CD133þCXCR4þEpCAM� withinthe CIC pool in primary tumors represents an independentindicator of poor prognosis in surgically resected NSCLCpatients, suggesting that dissection of the biologic complexityof primary tumors will be crucial to improve clinical manage-ment of lung cancer.

The prospective to identify the fraction of metastatic CD133þ

CXCR4þ CICs negative for EpCAM expression in circulatingtumor cells from patients with lung cancer holds therefore greatinterest and implications from both a prognostic and therapeuticpoint of view, as recently highlighted inpatientswithbreast cancerwhere isolated EpCAM� circulating tumor cells showed highinvasiveness and ability to generate brain and lung metastases(38–40).

In conclusion, we provide evidence for the existence of novelhuman lung disseminating CICs regulated by microenvironmen-tal signals able to complete all the steps of the metastatic cascadeand related to clinical recurrence. We also propose CXCR4 target-ing as a novel therapeutic option in neoadjuvant and adjuvantsetting to counteract the dissemination of chemoresistant andmetastatic lung CICs.

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






Authors' ContributionsConception and design: G. Bertolini, F. Andriani, R. Ferracini, L. Mariani,I. Roato, G. Sozzi, L. RozDevelopment of methodology: G. Bertolini, L. D'Amico, M. Milione,R. Ferracini, L. RozAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): G. Bertolini, L. D'Amico, M. Moro, L. Gatti,M. Tortoreto, M. Milione, U. PastorinoAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): G. Bertolini, L. D'Amico, E. Landoni, P. Perego,R. Miceli, F. Andriani, M. Milione, L. Mariani, G. Sozzi, L. RozWriting, review, and/or revision of the manuscript: G. Bertolini, L. D'Amico,M. Moro, E. Landoni, R. Miceli, F. Andriani, L. Mariani, U. Pastorino, I. Roato,G. Sozzi, L. RozAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): E. Landoni, D. Wong, R. Caserini, M. TortoretoStudy supervision: G. Sozzi, L. Roz

AcknowledgmentsThe authors are thankful for the gift of the CTCE-9908 compound and

technical support of British Canadian BioSciences Corporation. They also thank

Dr. Claudio Tripodo for valuable scientific advice and helpful suggestions andStefano Piccolo for critical reading of this article.

Grant SupportThis work is funded by grants from Associazione Italiana Ricerca Cancro

(IG9227 and IG13403 to L. Roz, IG10096 to G. Sozzi, and ED12162 toL. Roz and G. Sozzi), Compagnia di San Paolo di Torino (G. Sozzi),European Community Integrated Project 037665 CHEMORES (G. Sozziand P. Perego), European Community Seventh Framework Programme(FP7/2007-2013) under grant agreement no. HEALTH-F2-2010-258677(Collaborative Project CURELUNG to L. Roz) and Italian Ministry of Health:Ricerca Sanitaria Finalizzata e Giovani Ricercatori 2009 (GR 2009- 1584485to I. Roato).

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 December 23, 2014; revised April 30, 2015; accepted June 5, 2015;published OnlineFirst July 3, 2015.

References1. Siegel R, NaishadhamD, Jemal A. Cancer statistics, 2013. CA Cancer J Clin

2013;63:11–30.2. ClarkeMF,Dick JE, Dirks PB, EavesCJ, JamiesonCH, JonesDL, et al. Cancer

stem cells–perspectives on current status and future directions: AACRWorkshop on cancer stem cells. Cancer Res 2006;66:9339–44.

3. Visvader JE, Lindeman GJ. Cancer stem cells in solid tumours: accumu-lating evidence and unresolved questions. Nat Rev Cancer 2008;8:755–68.

4. Zhou BB, Zhang H, Damelin M, Geles KG, Grindley JC, Dirks PB. Tumour-initiating cells: challenges and opportunities for anticancer drug discovery.Nat Rev Drug Discov 2009;8:806–23.

5. Brabletz T, Jung A, Spaderna S, Hlubek F, Kirchner T. Opinion: migratingcancer stem cells - an integrated concept of malignant tumour progression.Nat Rev Cancer 2005;5:744–9.

6. Hermann PC, Huber SL, Herrler T, Aicher A, Ellwart JW, Guba M, et al.Distinct populations of cancer stem cells determine tumor growth andmetastatic activity in human pancreatic cancer. Cell Stem Cell 2007;1:313–23.

7. Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY, et al. Theepithelial-mesenchymal transition generates cells with properties of stemcells. Cell 2008;133:704–15.

8. Polyak K, Weinberg RA. Transitions between epithelial and mesenchymalstates: acquisition of malignant and stem cell traits. Nat Rev Cancer 2009;9:265–73.

9. Scheel C, Eaton EN, Li SH, Chaffer CL, Reinhardt F, Kah KJ, et al. Paracrineand autocrine signals induce and maintain mesenchymal and stem cellstates in the breast. Cell 2011;145:926–40.

10. TodaroM, Gaggianesi M, Catalano V, Benfante A, Iovino F, BiffoniM, et al.CD44v6 is a marker of constitutive and reprogrammed cancer stem cellsdriving colon cancer metastasis. Cell Stem Cell 2014;14:342–56.

11. RhimAD,Mirek ET, Aiello NM,Maitra A, Bailey JM,McAllister F, et al. EMTand dissemination precede pancreatic tumor formation. Cell 2012;148:349–61.

12. WengD, Penzner JH, SongB, Koido S, Calderwood SK,Gong J.Metastasis isan early event in mouse mammary carcinomas and is associated with cellsbearing stem cell markers. Breast Cancer Res 2012;14:R18.

13. Creighton CJ, Li X, Landis M, Dixon JM, Neumeister VM, Sjolund A, et al.Residual breast cancers after conventional therapy displaymesenchymal aswell as tumor-initiating features. Proc Natl Acad Sci U S A 2009;106:13820–5.

14. Singh A, Settleman J. EMT, cancer stem cells and drug resistance: anemerging axis of evil in the war on cancer. Oncogene 2010;29:4741–51.

15. Bertolini G, Roz L, Perego P, TortoretoM, Fontanella E,Gatti L, et al. Highlytumorigenic lung cancer CD133þ cells display stem-like features and arespared by cisplatin treatment. Proc Natl Acad Sci U S A 2009;106:16281–6.

16. Domanska UM, Kruizinga RC, Nagengast WB, Timmer-Bosscha H, Huls G,de Vries EG, et al. A reviewonCXCR4/CXCL12 axis in oncology: no place tohide. Eur J Cancer 2013;49:219–30.

17. Jung MJ, Rho JK, Kim YM, Jung JE, Jin YB, Ko YG, et al. Upregulationof CXCR4 is functionally crucial for maintenance of stemness indrug-resistant non-small cell lung cancer cells. Oncogene 2013;32:209–21.

18. Moro M, Bertolini G, Tortoreto M, Pastorino U, Sozzi G, Roz L. Patient-Derived Xenografts of Non-Small Cell Lung Cancer: resurgence of an oldmodel for investigation of modern concepts of tailored therapy and cancerstem cells. J Biomed Biotechnol 2012;2012:568567.

19. Kondo J, EndoH,OkuyamaH, IshikawaO, IishiH, TsujiiM, et al. Retainingcell-cell contact enables preparation and culture of spheroids composed ofpure primary cancer cells from colorectal cancer. Proc Natl Acad Sci U S A2011;108:6235–40.

20. Roato I, Caldo D, Godio L, D'Amico L, Giannoni P, Morello E, et al. Boneinvading NSCLC cells produce IL-7: mice model and human histologicdata. BMC Cancer 2010;10:12.

21. Kuperwasser C, Dessain S, Bierbaum BE, Garnet D, Sperandio K, GauvinGP, et al. A mouse model of human breast cancer metastasis to humanbone. Cancer Res 2005;65:6130–8.

22. Huang EH, Singh B, Cristofanilli M, Gelovani J, Wei C, Vincent L, et al. ACXCR4 antagonist CTCE-9908 inhibits primary tumor growth and metas-tasis of breast cancer. J Surg Res 2009;155:231–6.

23. Drenckhan A, Kurschat N, Dohrmann T, Raabe N, Koenig AM, Reichelt U,et al. Effective inhibition of metastases and primary tumor growth withCTCE-9908 in esophageal cancer. J Surg Res 2013;182:250–6.

24. Porvasnik S, SakamotoN, Kusmartsev S, Eruslanov E, KimWJ, CaoW, et al.Effects of CXCR4 antagonist CTCE-9908 on prostate tumor growth. Pros-tate 2009;69:1460–9.

25. Richert MM, Vaidya KS, Mills CN, Wong D, Korz W, Hurst DR, et al.Inhibition of CXCR4 by CTCE-9908 inhibits breast cancer metastasis tolung and bone. Oncol Rep 2009;21:761–7.

26. Hollier BG, Evans K, Mani SA. The epithelial-to-mesenchymal transitionand cancer stem cells: a coalition against cancer therapies. J MammaryGland Biol Neoplasia 2009;14:29–43.

27. Hu TH, Yao Y, Yu S, Han LL, Wang WJ, Guo H, et al. SDF-1/CXCR4promotes epithelial-mesenchymal transition and progression of colorectalcancer by activation of the Wnt/beta-catenin signaling pathway. CancerLett 2014;354:417–26.

28. Li X, Li P, Chang Y, XuQ,WuZ,MaQ, et al. The SDF-1/CXCR4 axis inducesepithelial-mesenchymal transition in hepatocellular carcinoma. Mol CellBiochem 2014;392:77–84.

29. Nguyen DX, Massague J. Genetic determinants of cancer metastasis.Nat Rev Genet 2007;8:341–52.

Bertolini et al.





30. MarjanovicND,Weinberg RA, Chaffer CL. Cell plasticity and heterogeneityin cancer. Clin Chem 2013;59:168–79.

31. Psaila B, Lyden D. The metastatic niche: adapting the foreign soil. Nat RevCancer 2009;9:285–93.

32. Malanchi I, Santamaria-Martinez A, Susanto E, Peng H, Lehr HA, DelaloyeJF, et al. Interactions between cancer stem cells and their niche governmetastatic colonization. Nature 2012;481:85–9.

33. Oskarsson T, Acharyya S, Zhang XH, Vanharanta S, Tavazoie SF,Morris PG,et al. Breast cancer cells produce tenascin C as a metastatic niche compo-nent to colonize the lungs. Nat Med 2011;17:867–74.

34. Baccelli I, Schneeweiss A, Riethdorf S, Stenzinger A, Schillert A, Vogel V,et al. Identification of a population of blood circulating tumor cells frombreast cancer patients that initiates metastasis in a xenograft assay. NatBiotechnol 2013;31:539–44.

35. OskarssonT, Batlle E,Massague J.Metastatic stem cells: sources, niches, andvital pathways. Cell Stem Cell 2014;14:306–21.

36. Kurrey NK, Jalgaonkar SP, Joglekar AV, Ghanate AD, Chaskar PD,Doiphode RY, et al. Snail and slug mediate radioresistance and che-moresistance by antagonizing p53-mediated apoptosis and acquiring astem-like phenotype in ovarian cancer cells. Stem Cells 2009;27:2059–68.

37. Kong D, Banerjee S, Ahmad A, Li Y, Wang Z, Sethi S, et al.Epithelial to mesenchymal transition is mechanistically linked withstem cell signatures in prostate cancer cells. PLoS ONE 2010;5:e12445.

38. Zhang L, Ridgway LD, Wetzel MD, Ngo J, Yin W, Kumar D, et al. Theidentification and characterization of breast cancer CTCs competent forbrain metastasis. Sci Transl Med 2013;5:180ra48.

39. Pantel K, Alix-Panabieres C, Riethdorf S. Cancer micrometastases. Nat RevClin Oncol 2009;6:339–51.

40. Thiery JP, Lim CT. Tumor dissemination: an EMT affair. Cancer Cell 2013;23:272–3.






2015;75:3636-3649. Published OnlineFirst July 3, 2015.Cancer Res Giulia Bertolini, Lucia D'Amico, Massimo Moro, et al. Correlate with Poor Prognosis

Initiating Cells Sustain Tumor Dissemination and−Lung Cancer −/EpCAM+/CXCR4+Microenvironment-Modulated Metastatic CD133

Updated version

10.1158/0008-5472.CAN-14-3781doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2015/07/03/0008-5472.CAN-14-3781.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/75/17/3636.full#ref-list-1

This article cites 40 articles, 7 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/75/17/3636.full#related-urls

This article has been cited by 4 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://cancerres.aacrjournals.org/content/75/17/3636To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-14-3781

http://cancerres.aacrjournals.org/content/suppl/2015/07/03/0008-5472.CAN-14-3781.DC1

http://cancerres.aacrjournals.org/content/75/17/3636.full#ref-list-1

http://cancerres.aacrjournals.org/content/75/17/3636.full#related-urls

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

mailto:[email protected]

http://cancerres.aacrjournals.org/content/75/17/3636


Documents

Microenvironment-Modulated Metastatic LungCancer ... · 3636–49. 2015 AACR. Introduction Despite novel therapeutic strategies, survival rates for lung cancer have not greatly improved