delineationofacellularhierarchyinlungcancerrevealsan ...tary fig. s1 and data not shown). together...

Tumor and Stem Cell Biology

Delineation of aCellular Hierarchy in LungCancer Reveals anOncofetal Antigen Expressed on Tumor-Initiating Cells

Marc Damelin1, Kenneth G. Geles1, Maximillian T. Follettie1, Ping Yuan2, Michelle Baxter1,Jonathon Golas1, John F. DiJoseph1, Maha Karnoub1, Shuguang Huang1, Veronica Diesl1,Carmen Behrens2, Sung E. Choe3, Carol Rios1, Janet Gruzas1, Latha Sridharan1, Maureen Dougher1,Arthur Kunz1, Philip R. Hamann1, Deborah Evans1, Douglas Armellino1, Kiran Khandke1,Kimberly Marquette3, Lioudmila Tchistiakova3, Erwin R. Boghaert1, Robert T. Abraham1,Ignacio I. Wistuba2, and Bin-Bing S. Zhou1

AbstractPoorly differentiated tumors in non–small cell lung cancer (NSCLC) have been associated with shorter patient

survival and shorter time to recurrence following treatment. Here, we integrate multiple experimental modelswith clinicopathologic analysis of patient tumors to delineate a cellular hierarchy in NSCLC. We show that theoncofetal protein 5T4 is expressed on tumor-initiating cells and associated with worse clinical outcome inNSCLC. Coexpression of 5T4 and factors involved in the epithelial-to-mesenchymal transition were observed inundifferentiated but not in differentiated tumor cells. Despite heterogeneous expression of 5T4 in NSCLCpatient–derived xenografts, treatment with an anti-5T4 antibody–drug conjugate resulted in complete andsustained tumor regression. Thus, the aggressive growth of heterogeneous solid tumors can be blocked bytherapeutic agents that target a subpopulation of cells near the top of the cellular hierarchy. Cancer Res; 71(12);4236–46. �2011 AACR.

Introduction

Non–small cell lung cancer (NSCLC) is the leading cause ofcancer-related deaths, due to high incidence and lack ofeffective therapies (1). NSCLC has heterogeneous pathologiesand molecular profiles, which presents challenges to mole-cular targeting and treatment (2–4). The degree of tumor celldifferentiation is an independent prognostic factor for clinicaloutcome, as less differentiated tumors are associated withincreased risk of death in both squamous cell carcinoma andadenocarcinoma (5). Strikingly, less differentiated tumors arealso associated with increased risk of recurrence after resec-tion (5). The molecular and cellular bases for these associa-tions are not understood.

Investigation of the cellular heterogeneity in tumors hasencouraged new frameworks, such as the cancer stem cell

hypothesis, for understanding the cellular hierarchy as itrelates to disease progression and for designing therapeuticstrategies (6–8). These experiments have shown that capa-cities for self-renewal and differentiation vary among cells in agiven tumor. The types and relative proportions of various cellsubpopulations in solid tumors are likely major determinantsof disease progression and therapeutic responsiveness incancer patients. In many cases, only certain subpopulationsof cancer cells from a tumor can initiate tumor growth inimmunodeficient mice. Herein, these cells are referred to astumor-initiating cells (TIC); they are also known in the lit-erature as cancer stem cells and cancer-initiating cells (6–8).TICs have been isolated, enriched, or identified in tumors ofbrain, breast, colon, head and neck, lung, melanocytes, pan-creas, prostate, and the hematopoietic system, typically withantibodies against cell surface markers (9). In general, thetumors grown from TICs recapitulate the cellular heteroge-neity of the original tumor, which underscores the develop-mental plasticity of many TICs and the differentiationhierarchy in many tumor tissues.

Tumor cell subpopulations and their distinct roles in dis-ease progression can be studied using a variety of comple-mentary approaches (8). The isolation of defined cellsubpopulations from tumors, xenografts, and established celllines based on cell surface markers allows for an in-depthmolecular analysis of rather homogeneous subpopulations.The culturing of cancer cells in defined serum-free media and/or 3-dimensional (3D) matrices has been shown to maintainkaryotype, transcriptome, and signaling that are reflective ofthe tissue of origin (10, 11). Patient-derived xenografts (PDX)

Authors' Affiliations: 1Oncology Research Unit, Pfizer Inc., Pearl River,New York; 2Department of Pathology, The University of Texas, MDAnderson Cancer Center, Houston, Texas; and 3Global BiotherapeuticTechnologies, Pfizer Inc., Cambridge, Massachusetts

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

M. Damelin and K.G. Geles contributed equally to this work.

Corresponding Author: Marc Damelin, Pfizer Inc., 401 N. MiddletownRd., Bldg. 200-4618, Pearl River, NY 10965. Phone: 845-602-7985; Fax:845-602-5557; E-mail: [email protected] or Ken Geles. Phone:845-602-8233; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-10-3919

�2011 American Association for Cancer Research.

CancerResearch

Cancer Res; 71(12) June 15, 20114236

Research. on January 25, 2021. © 2011 American Association for Cancercancerres.aacrjournals.org Downloaded from

Published OnlineFirst May 3, 2011; DOI: 10.1158/0008-5472.CAN-10-3919

http://cancerres.aacrjournals.org/

are established directly from freshly resected tumors andexhibit rich architectures that reflect the original patienttumors. PDXs are complex yet tractable preclinical modelsthat can be used to study cellular heterogeneity and evaluatethe efficacy of therapeutic agents.Here, we apply these emerging frameworks and technolo-

gies to delineate a cellular hierarchy in NSCLC and apply thenew knowledge to drug discovery. By integrating novel experi-mental models of NSCLC with the clinicopathologic analysisof patient tumors, we identify the 5T4 oncofetal antigen as amarker of undifferentiated, highly tumorigenic cells and ofworse clinical outcome. We show preclinical efficacy of ananti-5T4 antibody–drug conjugate (ADC) and suggest that thistherapeutic strategy could be applied to other surface markersexpressed on TICs.

Materials and Methods

Isolation and serum-free culture of human lungtumor cellsHuman lung tumor tissue (sample 87426A1) was obtained

from Asterand in accordance with appropriate consent pro-cedures, washed in PBS, and minced. Fragments were disag-gregated with collagenase III (250 U/mL) at 37�C for 2 to 3 hrs.Partially digested tissue fragments were passed through steelmesh screens. Cells were sieved through a 40-mm cell strainer(BD Falcon) and treated with red blood cell lysis buffer(Roche). Dead cells were excluded by magnetic separation(Miltenyi Biotec). The culture was established and maintainedin serum-free bronchial epithelial cell growthmedium (BEGM;Lonza), which was changed 50% every other day.

Air–liquid interface culturesMillicell 1-mm PET hanging cell culture inserts (Millipore)

were seeded with 2.5 � 105 cells from 87426A1 culture. After1 to 2 days, the upper and lower chambers were rinsed withPBS, and CnT-23 medium (CELLnTec) containing 50 nmol/Lretinoic acid and 1 mmol/L CaCl2 was added only to thelower chamber, leaving cells exposed to the air. Liftedcultures were fed every other day with fresh medium orharvested in Buffer RLT (Qiagen) for RNA isolation or TBS/0.5% NP-40 for immunoblot analysis. For gene expressionprofiling, "undifferentiated" values are averages of triplicatesamples, and "differentiated" values, due to limited sample,were averaged from differentiation time points of 8, 16, and24 days.

Animal studiesAll animal studies were approved by the Pfizer Pearl River

Institutional Animal Care and Use Committee according toestablished guidelines. To assess tumorigenic potential ofsorted cells, the cells were implanted in 50% Matrigel orwhere noted, 50% growth factor–reduced Matrigel (BD Bios-ciences) subcutaneously between the shoulder blades.60257A1 cells were implanted in NOD-SCID (severe combinedimmunodeficient) mice and tumor incidence was recordedafter 9 weeks. For H460T and HCC2429, 100 sorted cells wereimplanted per nu/nu mouse. Tumors were measured at least

once a week with tumor volume¼ 0.5 � width2 � length. TheP values are based on 2-sided Student's t test. Efficacy studiesare described in Supplementary Information.

Clinical samplesThis study was approved by the MD Anderson Institutional

Review Board. Tissue specimens were collected between 1997and 2002 from 320 lung cancer tumors (204 adenocarcinomasand 116 squamous cell carcinomas), and selected specimenswere used for construction of tissue microarrays. The primaryantibodies used for immunohistochemistry were mousemonoclonal anti-CD24 (Clone SN3b; LabVision), rabbit poly-clonal anti-CD44 (Sigma), and mouse monoclonal anti-5T4(MAB4975, R&D Systems). The statistical analysis was basedon membrane expression levels as described in Supplemen-tary Data.

Additional Materials and Methods can be found in theSupplementary Data.

Results

A cell line model reveals 5T4 expression in TICsTo identify a model of tumor heterogeneity in NSCLC, we

screened several cell lines for heterogeneous expression of cellsurface markers that had previously been used to enrich orisolate TICs (6, 9). Staining of the H460T cell line with anti-CD24 and anti-CD44 antibodies revealed distinct subpopula-tions of cells with a stable distribution over many passages inculture (Fig. 1A and data not shown). The delineation betweenCD24�/lo and CD24hi was readily apparent; the threshold forCD44 was based on the isotype control. To determine whetherthese subpopulations had different tumorigenic capacities, weisolated cells of each subpopulation by fluorescence-activatedcell sorting (FACS) and implanted them subcutaneously intoimmunocompromised mice. CD24�/loCD44hi cells formedlarge tumors rapidly, whereas CD24hiCD44hi cells exhibitedslower tumor growth (P < 0.001; Fig. 1B). Tumors fromCD24�/loCD44hi cells were also larger than tumors fromCD24�/loCD44lo cells (P < 0.02; Supplementary Fig. S1A). Invitro, CD24�/loCD44hi cells migrated more efficiently thanCD24hiCD44hi cells and exhibited multipotency in that theygave rise to less tumorigenic CD24hiCD44hi cells (Supplemen-tary Fig. S1 and data not shown). Together these resultsindicated that the CD24�/loCD44hi combination of surfacemarkers enriched for TICs in H460T. Similar results wereobserved for another NSCLC cell line, HCC2429 (data notshown).

To establish the clinical relevance of these results, weexamined CD24 and CD44 expression in patient samples.By immunohistochemistry, CD24�/loCD44hi membraneexpression was associated with less differentiated tumorhistology (P ¼ 0.0456). CD24�/loCD44hi membrane expressionwas also associated with worse overall survival in malesquamous cell carcinoma (P ¼ 0.0310; SupplementaryFig. S2A). In addition, analysis of published mRNA data (12)revealed that CD24�/loCD44hi was associated with worseoverall survival in adenocarcinomas (P ¼ 0.0501; Supplemen-tary Fig. S2B). Taken together, these results showed that the

A Cellular Hierarchy in NSCLC

www.aacrjournals.org Cancer Res; 71(12) June 15, 2011 4237




TIC markers identified in H460T and HCC2429 were also poorprognosis markers in NSCLC, which indicates the clinicalrelevance of the cell line models.

From a therapeutic perspective, it is not feasible to targetthe CD24�/loCD44hi combination of markers. Therefore, wesought to identify novel TIC markers in NSCLC that could bepotential therapeutic targets. We generated gene expressionprofiles from triplicate samples of sorted CD24�/loCD44hi andCD24hiCD44hi cells from H460T (Supplementary Fig. S1E). The5T4 (TPBG) gene had significantly higher expression inthe more tumorigenic CD24�/loCD44hi cells than in theCD24hiCD44hi cells (Fig. 1C and Supplementary Table S1).Consistent with this result, in NSCLC clinical samples, 5T4mRNA expression was significantly higher in CD24�/loCD44hi

versus "Not CD24�/loCD44hi" tumors (ref. 12; P ¼ 0.0293;Supplementary Fig. S2C).

5T4 is a single-pass transmembrane protein that is prefer-entially expressed in many tumor types compared with nor-mal tissues (13) and has been explored as a cancer target.During embryonic development, 5T4 functions in epithelial-to-mesenchymal transition (EMT) and cell migration, at leastin part by interacting with CXCR4 and/or the actin cytoske-

leton (14–17). Given the oncofetal characteristics of 5T4 andits increased expression in the tumorigenic CD24�/loCD44hi

subpopulation of H460T, we focused on 5T4 as a potential TICmarker in NSCLC.

Immunofluorescence studies in H460T revealed that stain-ing for 5T4 and CD24 were mutually exclusive, consistent withthe gene expression data, and moreover that most if not allCD24�/lo cells expressed 5T4 (Fig. 1D). This result wasextended by flow cytometric (Supplementary Fig. S3) andimmunoblot analysis (Fig. 1E).

On the basis of the expression of 5T4 in normal progenitorcells during embryonic development (18), we hypothesizedthat 5T4 expression is associated with a less differentiatedstate and that inducing differentiation of tumor cells wouldreduce 5T4 expression. Sorted cells were treated withretinoic acid and subjected to immunoblot analysis.Indeed, 5T4 expression was dramatically reduced in treatedCD24�/loCD44hi cells (Fig. 1E).

To test tumorigenicity based on 5T4 expression, we sortedcells from a clonal line of H460T and implanted them intomice. As predicted, 5T4hi cells were more tumorigenic than5T4�/lo cells (P < 0.03; Fig. 1F), which indicated that 5T4 can

103

1,400 120

100

80

60

40

20

0

350

300

250

150

50

06 10 14

Day postimplant18 22

200

100

1,200

1,000

800 *P < 0.001

*P < 0.03

*

*CD24–/IoCD44hi

CD24–/Io

CD44hi

CD24–/Io

CD44hi

CD24hi

CD44hi

CD24hi

CD44hi

CD24hiCD44hi

5T4hi

5T4–/Io

600

400

200

00

Veh

icle

Ret

inoi

c ac

id

Veh

icle

Ret

inoi

c ac

id

5 10 15Day postimplant

20 25 30

A

D E

5T4

5T4 CD24 β-Actin

F

B C 5T4 (TPBG)

102

CD

44

Tum

or v

olum

e (m

m3 )

Tum

or v

olum

e (m

m3 )

mR

NA

sig

nal i

nten

sity

101

100

100

5.6

22.6 71.3

0.5

101 102

CD24

103

Figure 1. Oncofetal antigen 5T4 is identified in a cell line model of TICs. A, distinct populations of H460T were revealed by flow cytometry withanti-CD24 and anti-CD44 antibodies. B, CD24�/loCD44hi and CD24hiCD44hi cells were sorted and implanted subcutaneously into mice. Values indicatemean� SEM. C, 5T4 mRNA signal intensities on gene microarrays probed with samples from sorted CD24�/loCD44hi and CD24hiCD44hi cells. Values indicatemean � SD. D, H460T cells were fixed, stained with anti-CD24-Qdot585 and anti-5T4-Qdot705, and imaged with a 40� objective. E, immunoblot analysis onsorted cells that were treated with retinoic acid or vehicle. F, tumor growth of 5T4hi- and 5T4�/lo-sorted cells. H460T clonal line CD24N-26 was used asdescribed in the text. A representative experiment is shown. Values indicate mean � SEM.

Damelin et al.

Cancer Res; 71(12) June 15, 2011 Cancer Research4238




independently enrich for TICs in this system. On the basis ofall of these results, we sought to further establish 5T4 as anovel TIC antigen by using PDXs.

5T4 enriches for tumorigenicity in a PDXWe established NSCLC PDX lines by implanting fragments

of freshly resected tumors (squamous cell carcinoma) sub-cutaneously in immunocompromised mice. Each line waspropagated by explanting xenograft fragments into naiveanimals and was maintained exclusively in vivo. In eachPDX line, the histology of the xenograft resembled that ofthe original tumor (data not shown).To determine whether 5T4 expression enriched for tumori-

genic capacity in PDXs, 5T4hi and 5T4�/lo cells were sortedfrom dissociated 60257A1 PDXs and implanted into animals.Cells from this PDX line exhibited a wide continuous range of5T4 expression levels, and the highest and lowest 15% wereassigned 5T4hi and 5T4�/lo (Fig. 2A). 5T4hi cells yielded sub-stantially higher tumor incidence than 5T4�/lo cells in 3independent experiments (Table 1). On the basis of limitingdilution analysis, 5T4hi cells were about 30-fold more tumori-genic than 5T4�/lo cells when implanted in 50% growthfactor–reduced Matrigel (Table 1). To determine whethertumorigenicity was dependent on growth factors, as reportedpreviously (19), the experiment was repeated 2 additionaltimes in regular Matrigel. The tumorigenicity of 5T4�/lo cellswas slightly enhanced in these conditions but was still sig-nificantly lower than that of 5T4hi cells (Table 1). Tumors thatgrew from sorted 5T4hi cells showed heterogeneous 5T4expression (Fig. 2B). Thus, the results in the PDX are con-sistent with those from H460T and show that 5T4hi expressioncan enrich for TICs in NSCLC.To uncover molecular differences between 5T4hi and

5T4�/lo cells, we generated gene expression profiles from cellssorted on 3 independent occasions from 60257A1. Strikingly,the Clara cell secretory protein (CCSP, also known as uter-oglobin, SCGB1A) was expressed 20-fold higher in 5T4hi cellsthan in 5T4�/lo cells (Fig. 2C). CCSP has been characterized asa marker of lung stem cells in normal tissue and adenocarci-noma (20). In addition, several genes associated with tumor-igenesis, invasion, and stem cells were expressed at higherlevels in 5T4hi cells including carcinoembryonal antigensCEACAM6 and CEACAM1, cathepsin S (CTSS), gelsolin(GSN), and interleukin (IL) 8 (Fig. 2C). In contrast, low levelsof CD133 were comparable in both populations (Supplemen-tary Fig. S4A). In summary, the gene expression profilesdistinguished the 5T4hi and 5T4�/lo populations and offeredpotential explanations for their different tumor-initiatingcapacities.

5T4 and markers of EMT are expressed in cells near thetop of the tumor hierarchyImmunohistochemical analysis of our PDXs revealed a

range of 5T4 expression levels and patterns. Three of 6PDX lines exhibited heterogeneous expression of 5T4, andin 2 of the 3, 5T4 expression was coupled to tumor architec-ture in that expression was higher at the tumor–stromainterface (Fig. 3A and B and Supplementary Fig. S4B). Notably,

in 37622A1, a pattern similar to 5T4 was observed for vimen-tin, a marker of EMT, and Ki67, a proliferation marker(Fig. 3A). Coimmunostaining confirmed that 5T4hi cells weretypically positive for Ki67 in this PDX line (Fig. 3B).

When a serum-free culture was established from the37622A1 PDX (Fig. 3C), the cells exhibited uniformly high5T4 expression, whereas the PDX exhibited a broader expres-sion range (Fig. 3D). Because the culture media supportsgrowth of undifferentiated cells, this result was consistentwith the hypothesis that 5T4 is highly expressed on TICs.Strikingly, when the cultured cells were implanted into ani-mals, the resulting tumors exhibited heterogeneous 5T4expression (Fig. 3E and Supplementary Fig. S4C). This resultsuggested that 5T4hi cells gave rise to 5T4�/lo cells duringtumor growth by a mechanism that potentially requirestumor–stroma interaction or other microenvironmental cues.

To further investigate the cellular hierarchy in NSCLC, wedeveloped a novel 3Dmodel of lung cancer differentiation. In aserum-free culture established directly from freshly resectedtumor (sample 87426A1), cells could either be maintained inthe undifferentiated state submerged in stem cell media or beinduced to differentiate at the air–liquid interface with reti-noic acid but notably without the use of serum (Fig. 4A). Theair–liquid interface is considered a physiologic environmentfor lung cells and used to study differentiation of immortalizedbronchial epithelial cells (21). The 87426A1 culture efficientlyformed a 3D multilayered epithelium on exposure to the air–liquid interface for 18 days (Fig. 4B).

Immunoblot analysis showed that 5T4 levels were high inthe undifferentiated state and decreased quickly and drama-tically on differentiation (Fig. 4C). This result was analogous toretinoic acid treatment of sorted H460T cells (Fig. 1E).

To obtain a global view of the cellular hierarchy in NSCLC,gene expression profiles were generated from undifferentiatedand differentiated cells. Consistent with the above results, 5T4expression decreased and CD24 expression increased duringdifferentiation (Fig. 4D and E). These data were confirmed byflow cytometry (Supplementary Fig. S5). The profiles revealedstriking patterns of genes involved in EMT and angiogenesis.The EMT markers vimentin, fibronectin, Slug, and Twist wereexpressed at significantly higher levels in undifferentiated cells(Fig. 4D); in contrast, the epithelial differentiation markersmucin and involucrin as well as transglutaminase (TGM) andcytokeratins were expressed at significantly higher levels indifferentiated cells (Fig. 4E). The coexpression of 5T4 andvimentin in undifferentiated cells was consistent with thesimilar expression patterns of 5T4 and vimentin in 37622A1PDX (Fig. 3A). Interestingly, angiogenesis factors, includingVEGF-A, VEGF-C, PDGF-A (platelet-derived growth factor),PGF (placental growth factor), and uPA (urokinase-type plas-minogen activator), were expressed at significantly higherlevels in undifferentiated cells (Fig. 4F). These results arenoteworthy given the higher 5T4 expression at the tumor–stroma interface in 2 of our 6 PDX lines.

An unbiased meta-analysis revealed several gene signatureswith significant expression changes during differentiation(Supplementary Table S2). Gene sets with higher expressionin undifferentiated cells included signatures of poor clinical






3,000CCSP/SCGB1A1 CEACAM6 CEACAM1

12,000 200

100

150

50

0

10,000

8,000

6,000

4,000

2,000

0

2,000

2,500

A

B

C

1,500

500

700

600500

400

300

200

100

0

0

1,000

3,000

2,000

2,500

1,500

500

0

1,000

300

200

250

150

50

0

100

mR

NA

mR

NA

5T4hi

CTSS IL8 GSN

5T4–/Io

5T4hi 5T4–/Io 5T4hi 5T4–/Io 5T4hi 5T4–/Io

5T4hi 5T4–/Io 5T4hi 5T4–/Io

100 μm

250,000

200,000

100,000

FS

C-A 150,000

50,000

250,000

200,000

100,000

150,000

50,000

15.1

5T4–/Io

5T4

5T4hi

14.7

0 102 103

Control

104 105 0 102 103 104 105

Figure 2. 5T4 can enrich for TICsin NSCLC. A, 5T4 expression inthe 60257A1 PDX. See Table1 for tumor incidence of the cellsubpopulations. B,immunohistochemistry of 5T4 intumors that grew from 5T4hi cellssorted from 60275A1. C, mRNAsignal intensities on genemicroarrays probed with samplesfrom 5T4hi or 5T4�/lo cells sortedfrom PDXs on 3 independentoccasions. Values indicate themean � SD.

Damelin et al.





Table 1. Tumor incidence in animals implanted with 5T4hi or 5T4�/lo cells sorted from 60257A1 PDXs in3 independent experiments

No. of cells Growth factor–reduced Matrigel(Expt. 1)

Matrigel (Expt. 2; Expt. 3)

5T4hi cells 5T4�/lo cells 5T4hi cells 5T4�/lo cells

500 10% (1/10) n.d.1,500 20% (2/10) 0% (0/10)5,000 62% (5/8) 0% (0/10) 60% (12/20; 6/10) 20% (4/20; 2/10)15,000 n.d. 11% (1/9)

Abbreviation: n.d., not determined.

5T4A

B

C

Patient sample 37622A1100

80

60

Isotype controlTumor (PDX)Culture (SF)

40Cou

nt

20

0 102 103

5T4104 105

PDX

SF culture

Reimplant into animals

D E

5T4

Ki67

Ki67 DNA Merge

Vimentin

Figure 3.Heterogeneous expression of 5T4 in NSCLC PDXs. A, immunohistochemistry on serial sections of 37622A1 PDX stained for 5T4, Ki67, and vimentin.Scale bar, 100 mm. B, confocal images of coimmunostaining of 5T4 and Ki67 in 37622A1 PDX. Scale bar, 100 mm. C, schematic of the establishmentof an adherent serum-free (SF) culture from 37622A1 PDX. D, analysis of live human epithelial antigen–positive cells in the PDX (red) and the PDX-derivedculture (blue) compared with isotype control (gray). E, immunohistochemistry of 5T4 in a tumor xenograft that grew from the SF culture.






prognosis, stem cells, oncogenic signaling, and developmentalsignaling. Gene sets with higher expression in differentiatedcells included signatures of good clinical prognosis, differen-tiated tumors, and differentiated cells. Taken together, theseresults show a cellular hierarchy in a primary NSCLC cultureand support the concept that less differentiated tumor cellscan have a more malignant phenotype.

Correlation between 5T4 expression andclinicopathologic features and disease outcomes

We sought to determine whether 5T4 expression in NSCLCtumors correlated with clinicopathologic features and diseaseoutcomes. In a panel of surgically resected tumors, 5T4expression levels and patterns varied, and each sample wasassigned an overall score. The distribution of 5T4 expression

Undifferentiated

A

B

C

D

E

F

Days

5T4

Ponceau S

5T4 (TPBG)

INVOLUCRIN

VEGF-A VEGF-C PDGF-A PGF uPA

TGM1 MUCIN1 KERATIN4 CD24

SLUG TWIST1 FIBRONECTIN VIMENTIN1,500 4,000

3,000

2,000

1,000

0

6,000

4,500

3,000

1,500

0

3,000 750

500

250

0

2,000

1,000

0

300

200

100

0

600 5,000

4,000

3,000

2,000

1,000

0

400

200

0

400 10,000

8,000

6,000

4,000

2,000

300

200

100

0 0

10,0002,000

1,500

1,000

500

0

8,000

6,000

4,000

2,000

0

8,000

6,000

4,000

2,000

0

800

600

400

200

0

8,000

6,000

4,000

2,000

0

500mR

NA

mR

NA

mR

NA

0Undiff Diff

Undiff Diff

Undiff Diff Undiff Diff Undiff Diff Undiff Diff Undiff Diff

Undiff Diff Undiff Diff Undiff Diff Undiff Diff

Undiff Diff Undiff Diff Undiff Diff Undiff Diff

1,000

0 1 2 4 10 21

Differentiated

Figure 4. 5T4 and markers of EMTare associated with theundifferentiated state. A,schematic of serum-free cultureunder conditions that maintain theundifferentiated state or inducedifferentiation at the air–liquidinterface. B, micrographs of the87426A1 culture in theundifferentiated state (left) andafter differentiation for 18 days(right). Scale bars, 100 mm. C, 5T4immunoblot analysis duringdifferentiation. D–F, mRNA signalintensities on gene microarraysprobed with samples ofundifferentiated (undiff) anddifferentiated (diff) cells. Valuesrepresent mean � SD.

Damelin et al.





according to clinicopathologic characteristics is shown inSupplementary Table S3. Notably, membranous 5T4 expres-sion was higher in poorly and moderately versus well-differ-entiated tumors (P¼ 0.0005 and P¼ 0.0197, respectively, withoverall ANOVA yielding P ¼ 0.0022; see SupplementaryTable S4). In addition, high membranous 5T4 expressionwas associated with shorter survival time (P ¼ 0.0283,Fig. 5A; and P ¼ 0.078 in multivariate analysis, SupplementaryTable S5) and shorter time to recurrence (P ¼ 0.0495, Fig. 5B;and P ¼ 0.034 in multivariate analysis, SupplementaryTable S5). When the cohort was stratified by 5T4hi and5T4�/lo, there was no difference in the impact of patienttreatment on clinical outcome (not shown). These observa-tions are consistent with our experimental results that indi-cate that 5T4 is associated with less differentiated, moretumorigenic cells in NSCLC.

An ADC is specifically cytotoxic to 5T4-expressing cellsTaken together, these results provided the rationale for

targeting 5T4-expressing cells with an ADC even in tumorswith heterogeneous 5T4 expression. Unconjugated anti-5T4monoclonal antibody has not exhibited any in vitro activity orin vivo efficacy (22). To generate the ADC, anti-5T4 antibody

(13), previously shown to internalize into cells (22), wasconjugated to the enediyne natural product calicheamicin,a DNA-damaging agent (ref. 23; Fig. 6A). To show specificityfor 5T4-expressing cells, in vitro experiments were carried outon sorted H460T cells. In a proliferation assay, the 5T4-enriched population was greater than 10-fold more sensitiveto the ADC than the 5T4-low population (Fig. 6B). In a colonyformation assay in which cells were treated with drug andthen seeded at low density, the 5T4-enriched population wassimilarly greater than 10-fold more sensitive to the ADC thanthe 5T4-low population (Fig. 6C). In contrast, no difference insensitivities was observed when the 2 populations were trea-ted with calicheamicin alone, anti-CD22 calicheamicin ADC,camptothecin, 5-fluoruracil, or ionizing radiation (data notshown). These results showed the specificity of the anti-5T4ADC.

Regression of tumors with heterogeneous 5T4expression with an anti-5T4 ADC

We hypothesized that a therapeutic agent that targets andeliminates 5T4hi-expressing TICs could ultimately causeregression of tumors despite heterogeneous expression of5T4 within the tumor. The stable chemical linker betweenantibody and drug restricts the release of calicheamicin tocells that internalize the ADC (24). We assessed the efficacy ofan anti-5T4 ADC on the growth of 2 PDX lines with hetero-geneous 5T4 expression. In both lines, 5T4 expression waspredominant at the tumor–stroma interface, but the overall5T4 expression level was higher in 37622A1 than 60274A1(Supplementary Fig. S4B and D). 37622A1 tumors harbored aK-ras G13V mutation, whereas 60274A1 tumors had wild-typeK-ras (Fig. 6D and E).

PDX tumors were treated with anti-5T4 ADC, anti-CD33ADC or vehicle; the anti-CD33 ADC served as a negativecontrol because these PDX lines do not express CD33 (datanot shown). In 37622A1, treatment with anti-5T4 ADCregressed the tumors, and no regrowth was observed even3 months after the last dose; in contrast, treatment with anti-CD33 ADC or vehicle did not inhibit tumor growth (Fig. 6D).Similarly, in 60274A1, treatment with anti-5T4 ADC regressedthe tumors despite the lower 5T4 expression in this line(Fig. 6E). Treatment with calicheamicin (not conjugated toan antibody) has not shown any significant impact on tumorgrowth (22, 24, 25).

In contrast to the efficacy observed with anti-5T4 ADC,treatment of both PDXs with cisplatin at the maximumtolerable dose regressed tumors only transiently and thetumors regrew after treatment was completed (Fig. 6E andSupplementary Fig. S6). These results highlight the superiorlong-term efficacy of an ADC that targets TICs as comparedwith a conventional chemotherapeutic.

Discussion

5T4 and the cellular hierarchy in NSCLCUsing an integrated strategy, we have delineated a cellular

hierarchy in NSCLC and identified 5T4 as a clinically relevantmarker that can enrich for TICs. We found that 5T4 expression

1.0

A

B

0.8

0.6

0.4

Sur

vivi

ng p

ropo

rtio

n0.

20.

0

0 1 2Years

3

P = 0.0283

5T4hi

5T4–/lo

4 5

1.0

0.8

0.6

0.4

Rec

urre

nce-

free

pro

port

ion

0.2

0.0

0 1 2Years

3

P = 0.0495

5T4hi

5T4–/lo

4 5

Figure 5. 5T4 expression in NSCLC is associated with worse clinicaloutcome. A, 5T4 expression in NSCLC is associated with worse overallsurvival in patients. B, 5T4 expression in NSCLC is associated with shortertime to recurrence in patients.






was associated with high tumorigenic capacity and the undif-ferentiated state; these results parallel the expression of 5T4 incycling progenitor cells during embryonic development (18).The association of 5T4 expression with EMT in tumors,observed in the 87426A1 serum-free culture and the37622A1 PDX, mirrors the association of 5T4 with EMT innormal development (15). Our results are consistent with arecent study that described the relationship between EMT andstem cell phenotypes in normal and cancerous tissues (26).

Expression of 5T4 has been associated with advanceddisease and/or poor prognosis in colorectal, ovarian, andgastric cancers (27–29). We found that 5T4 is associated withless differentiated tumor histology, shorter time to recurrence,

and worse overall survival in NSCLC. Interestingly, althoughCD133 was used to isolate NSCLC TICs (30, 31) and wasexpressed in drug-tolerant PC9 cells (32), its expression inNSCLC was not associated with patient survival (33). On thebasis of the published observation that CD133þ cells arequiescent (30) and our observation in a different model that5T4hi cells are proliferating, we speculate that CD133þ and5T4hi might define different TIC populations even in the sameNSCLC tumor—analogous to the emerging evidence for dis-tinct stem cell populations (e.g., quiescent vs. active/prolifer-ating, in normal tissues; ref. 34). The clinical relevance of the5T4 marker suggests that 5T4hi TICs contribute directly to oneor more aspects of disease progression in NSCLC. For

100

50

MT

S (

% o

f con

trol

)Tu

mor

vol

ume

(mm

3 )Tu

mor

vol

ume

(mm

3 )

Tum

or v

olum

e (m

m3 )

100

50

No.

of c

olon

ies

(% o

f con

trol

)

Anti-5T4 ADC

A

D

E

B C

K-ras G13V

PDX 37622A1

PDX 60274A1

K-ras wild type

10–3

3,000

2,500

2,000

1,000

1,500

500

3,5002,000

1,500

500

0

1,000

2,500

2,000

1,500

500

0

1,000

3,000

00 10 20

*

* * * * * * * * * **

* * * *

30 40 50 60 70 80 90 100

0 10 20 30 40 50Day after staging

0 10 20 30 40Day after staging

60 70 80 90 100

110 120

10–2

5T4 enriched

5T4 low

5T4 enriched

5T4 low

10–1 10–5 10–4 10–3 10–2 10–1 11Anti-5T4–calicheamicin (ng/mL) Anti-5T4–calicheamicin (ng/mL)

Key

Vehicle

Anti-5T4 conjugate

Anti-CD33 conjugate

Cisplatin

* P < 0.05

10 100 1,000

Figure 6. Long-term regression of PDXs on treatment with an anti-5T4 ADC. A, schematic of an anti-5T4 calicheamicin conjugate. B, the 5T4-enrichedpopulation from H460T is sensitive to anti-5T4–calicheamicin conjugate in a cell growth assay. Crosshairs indicate IC50 values of free calicheamicin. C, the5T4-enriched population from H460T is sensitive to anti-5T4–calicheamicin conjugate in a clonogenic assay. Crosshairs indicate IC50 values of freecalicheamicin. D and E, long-term regression of PDXs by treatment with anti-5T4 ADC but not cisplatin. Animals were dosed on days 1, 5, and 9 after staging.Values indicate mean � SEM. *, time points with statistically significant differences in tumor size. See also Supplementary Figure S6.

Damelin et al.





instance, 5T4hi TICs may contribute to metastasis, as thatprocess might originate when cells in a primary epithelialtumor undergo EMT (35).The morphologic heterogeneity observed in NSCLC and

other epithelial tumors is likely paralleled by a continuum oftumor-initiating capabilities. In other words, the term "TIC"refers to a quantitative trait that is influenced not only bycancer cell autonomous signaling but also by cellular context(e.g., neighboring malignant cells and the tumor microenvir-onment). For instance, the tumorigenicity of defined cellpopulations can depend on parameters such as mouse strainand the use of Matrigel (19, 36). We found a similar effect inthat the 5T4�/lo cells yielded higher tumor incidence whenimplanted with regular Matrigel as opposed to growth factor–reduced Matrigel. Furthermore we observed 2 examples ofcontext-dependent TIC behavior: the uniformly 5T4hi cells inthe PDX-derived culture gave rise to tumors with heteroge-neous 5T4 expression (Fig. 3) and the uniformly 5T4hi cells inthe tumor-derived culture gave rise to cells with dramaticallyreduced 5T4 expression under differentiation conditions invitro (Fig. 4). Xenografts and serum-free cultures can becomplementary models to study tumor cell plasticity, hier-archy, and stromal interaction.

Therapeutic targeting of a TIC antigen5T4 is an oncofetal protein as defined by expression during

embryonic development and in cancer but little expression innormal adult tissues. Recent work showed the importance ofanother oncofetal protein, L1CAM, in glioblastoma TICs (37).Unconjugated antibodies directed at TIC surface markersmight affect TIC survival, differentiation, and/or interactionwith stroma, but many of them are expected to have limitedsingle-agent activity in the clinic. Our data provide preclinicalproof of concept for targeting TICs with ADCs. Strikingly,efficacy was observed in KRAS wild-type and KRAS mutanttumors, consistent with the ADCmechanism of action. Futurestudies could determine whether the anti-5T4 ADC strategy inNSCLCmight transcend KRAS status, in contrast to anti-EGFR(epidermal growth factor receptor) therapies (38).The efficacy of a TIC-directed therapy will likely depend on

the proportion and properties of the targeted TICs. Even if

quiescent or slow-growingTICs are destroyed or differentiated,tumors with proliferating progenitor-like cells might notregress in the short term (8); on the other hand, immediatetherapeutic impact might be achieved if we eliminate prolif-erating TICs that contribute to both pathologic damage anddisease progression. The dramatic, antigen-specific activity ofthe anti-5T4 ADC might be explained by the proliferative stateof 5T4hi TICs, but further studies are needed to fully understandthe mechanism, which might also involve a bystander effect(39) and/or drug-induced 5T4 expression (e.g., as observed in60274A1 PDX following cisplatin treatment; our unpublisheddata). Because of the continuous nature of 5T4 expression thatwe observed in somemodels, we cannot exclude the possibilitythat the anti-5T4 ADCmight target both TICs and non-TICs insome tumors, which would be therapeutically advantageous.

In summary, our study delineates a cellular hierarchy inNSCLC, characterizes the undifferentiated highly tumorigeniccell subpopulation, and shows the application of this knowl-edge to the development of anticancer therapeutics. Charac-terization of 5T4 as a TIC marker in NSCLC suggests anexplanation for the long-term efficacy of the anti-5T4 ADCdespite heterogeneous 5T4 expression in the tumors. Futurestudies could uncover additional molecules that, even ifexpressed in only a subpopulation of cells in the tumor,constitute appealing therapeutic targets against aggressiveheterogeneous diseases.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

The authors thank Carolyn Discafani, Lisa McNeil, Sharon Yang, Vincent Lau,Christine Rizzo, Justin Grindley, David Fruhling, Kim Arndt, Fred Immermann,Tianhui Zhou, and Andrew Vlahos for technical support and/or comments onthe manuscript; Pfizer Worldwide Comparative Medicine for vivarium support;and John Minna for providing HCC2429.

The costs of publication of this article were defrayed in part by the paymentof page charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received October 26, 2010; revised March 2, 2011; accepted March 22, 2011;published OnlineFirst May 3, 2011.

References1. Molina JR, Yang P, Cassivi SD, Schild SE, Adjei AA. Non-small cell

lung cancer: epidemiology, risk factors, treatment, and survivorship.Mayo Clin Proc 2008;83:584–94.

2. Herbst RS, Heymach JV, Lippman SM. Lung cancer. N Engl J Med2008;359:1367–80.

3. Sica G, Yoshizawa A, Sima CS, Azzoli CG, Downey RJ, Rusch VW,et al. A grading system of lung adenocarcinomas based on histologicpattern is predictive of disease recurrence in stage I tumors. Am JSurg Pathol 2010;34:1155–62.

4. Sato M, Shames DS, Gazdar AF, Minna JD. A translational view of themolecular pathogenesis of lung cancer. J ThoracOncol 2007;2:327–43.

5. Sun Z, Aubry MC, Deschamps C, Marks RS, Okuno SH, Williams BA,et al. Histologic grade is an independent prognostic factor for survivalin non-small cell lung cancer: an analysis of 5018 hospital- and712 population-based cases. J Thorac Cardiovasc Surg 2006;131:1014–20.

6. Lobo NA, Shimono Y, Qian D, Clarke MF. The biology of cancer stemcells. Annu Rev Cell Dev Biol 2007;23:675–99.

7. Reya T, Morrison SJ, ClarkeMF,Weissman IL. Stem cells, cancer, andcancer stem cells. Nature 2001;414:105–11.

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

9. Visvader JE, Lindeman GJ. Cancer stem cells in solid tumours:accumulating evidence and unresolved questions. Nat Rev Cancer2008;8:755–68.

10. Debnath J, Brugge JS. Modelling glandular epithelial cancers in three-dimensional cultures. Nat Rev Cancer 2005;5:675–88.

11. Lee J, Kotliarova S, Kotliarov Y, Li A, Su Q, Donin NM, et al. Tumorstem cells derived from glioblastomas cultured in bFGF and EGFmoreclosely mirror the phenotype and genotype of primary tumors than doserum-cultured cell lines. Cancer Cell 2006;9:391–403.






12. Shedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, GeraldWL, et al. Gene expression-based survival prediction in lung adeno-carcinoma: a multi-site, blinded validation study. Nat Med 2008;14:822–7.

13. Hole N, Stern PL. A 72 kD trophoblast glycoprotein defined by amonoclonal antibody. Br J Cancer 1988;57:239–46.

14. Carsberg CJ, Myers KA, Stern PL. Metastasis-associated 5T4 antigendisrupts cell-cell contacts and induces cellular motility in epithelialcells. Int J Cancer 1996;68:84–92.

15. EasthamAM, Spencer H, Soncin F, Ritson S, Merry CL, Stern PL, et al.Epithelial-mesenchymal transition events during human embryonicstem cell differentiation. Cancer Res 2007;67:11254–62.

16. Southgate TD, McGinn OJ, Castro FV, Rutkowski AJ, Al-Muftah M,Marinov G, et al. CXCR4 mediated chemotaxis is regulated by 5T4oncofetal glycoprotein in mouse embryonic stem cells. PLOS One2010;5:e9982.

17. Awan A, Lucic MR, Shaw DM, Sheppard F, Westwater C, Lyons SA,et al. 5T4 interacts with TIP-2/GIPC, a PDZ protein, with implicationsfor metastasis. Biochem Biophys Res Commun 2002;290:1030–6.

18. Barrow KM,Ward CM, Rutter J, Ali S, Stern PL. Embryonic expressionof murine 5T4 oncofoetal antigen is associated with morphogeneticevents at implantation and in developing epithelia. Dev Dyn2005;233:1535–45.

19. Quintana E, Shackleton M, Sabel MS, Fullen DR, Johnson TM, et al.Efficient tumour formation by single human melanoma cells. Nature2008;456:593–8.

20. Reynolds SD, Malkinson AM. Clara cell: progenitor for the bronchiolarepithelium. Int J Biochem Cell Biol 2010;42:1–4.

21. Vaughan MB, Ramirez RD, Wright WE, Minna JD, Shay JW. A three-dimensional model of differentiation of immortalized human bronchialepithelial cells. Differentiation 2006;74:141–8.

22. Boghaert ER, Sridharan L, Khandke KM, Armellino D, RyanMG,MyersK, et al. The oncofetal protein, 5T4, is a suitable target for antibody-guided anti-cancer chemotherapy with calicheamicin. Int J Oncol2008;32:221–34.

23. Hamann PR, Hinman LM, Hollander I, Beyer CF, Lindh D, Holcomb R,et al. Gemtuzumab ozogamicin, a potent and selective anti-CD33antibody-calicheamicin conjugate for treatment of acute myeloidleukemia. Bioconjug Chem 2002;13:47–58.

24. Hamann PR, Hinman LM, Beyer CF, Lindh D, Upeslacis J, Flowers DA,et al. An anti-CD33 antibody-calicheamicin conjugate for treatment ofacute myeloid leukemia. Choice of linker. Bioconjug Chem 2002;13:40–6.

25. DiJoseph JF, Armellino DC, Boghaert ER, Khandke K, Dougher MM,Sridharan L, et al. Antibody-targeted chemotherapy with CMC-544: a

CD22-targeted immunoconjugate of calicheamicin for the treatmentof B-lymphoid malignancies. Blood 2004;103:1807–14.

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

27. Naganuma H, Kono K, Mori Y, Takayoshi S, Stern PL, Tasaka K, et al.Oncofetal antigen 5T4 expression as a prognostic factor in patientswith gastric cancer. Anticancer Res 2002;22:1033–8.

28. Starzynska T, Marsh PJ, Schofield PF, Roberts SA, Myers KA, SternPL. Prognostic significance of 5T4 oncofetal antigen expression incolorectal carcinoma. Br J Cancer 1994;69:899–902.

29. Wrigley E, McGown AT, Rennison J, Swindell R, Crowther D, Star-zynska T, et al. 5T4 oncofetal antigen expression in ovarian carci-noma. Int J Gynecol Cancer 1995;5:269–74.

30. Bertolini G, Roz L, Perego P, Tortoreto M, Fontanella E, Gatti L, et al.Highly tumorigenic lung cancer CD133þ cells display stem-like fea-tures and are spared by cisplatin treatment. Proc Natl Acad Sci U S A2009;106:16281–6.

31. Eramo A, Lotti F, Sette G, Pilozzi E, Biffoni M, Di Virgilio A, et al.Identification and expansion of the tumorigenic lung cancer stem cellpopulation. Cell Death Differ 2008;15:504–14.

32. Sharma SV, Lee DY, Li B, Quinlan MP, Takahashi F, Maheswaran S,et al. A chromatin-mediated reversible drug-tolerant state in cancercell populations. Cell 2010;141:69–80.

33. Salnikov AV, Gladkich J, Moldenhauer G, Volm M, Mattern J, Herr I.CD133 is indicative for a resistance phenotype but does not representa prognostic marker for survival of non-small cell lung cancer patients.Int J Cancer 2010;126:950–8.

34. Li L, Clevers H. Coexistence of quiescent and active adult stem cells inmammals. Science 2010;327:542–5.

35. Thiery JP. Epithelial-mesenchymal transitions in tumour progression.Nat Rev Cancer 2002;2:442–54.

36. Quintana E, Shackleton M, Foster HR, Fullen DR, Sabel MS, JohnsonTM, et al. Phenotypic heterogeneity among tumorigenic melanomacells from patients that is reversible and not hierarchically organized.Cancer Cell 2010;18:510–23.

37. Bao S, Wu Q, Li Z, Sathornsumetee S, Wang H, McLendon RE, et al.Targeting cancer stem cells through L1CAM suppresses gliomagrowth. Cancer Res 2008;68:6043–8.

38. De Luca A, Normanno N. Predictive biomarkers to tyrosine kinaseinhibitors for the epidermal growth factor receptor in non-small-celllung cancer. Curr Drug Targets 2010;11:851–64.

39. Ginestier C, Liu S, Diebel ME, Korkaya H, Luo M, Brown M, et al.CXCR1 blockade selectively targets human breast cancer stem cellsin vitro and in xenografts. J Clin Invest 2010;120:485–97.

Damelin et al.





2011;71:4236-4246. Published OnlineFirst May 3, 2011.Cancer Res Marc Damelin, Kenneth G. Geles, Maximillian T. Follettie, et al. an Oncofetal Antigen Expressed on Tumor-Initiating CellsDelineation of a Cellular Hierarchy in Lung Cancer Reveals

Updated version

10.1158/0008-5472.CAN-10-3919doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2011/05/03/0008-5472.CAN-10-3919.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/71/12/4236.full#ref-list-1

This article cites 39 articles, 5 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/71/12/4236.full#related-urls

This article has been cited by 14 HighWire-hosted articles. Access the articles at:

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

SubscriptionsReprints and

[email protected] at

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

Permissions

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

.http://cancerres.aacrjournals.org/content/71/12/4236To 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-10-3919

http://cancerres.aacrjournals.org/content/suppl/2011/05/03/0008-5472.CAN-10-3919.DC1

http://cancerres.aacrjournals.org/content/71/12/4236.full#ref-list-1

http://cancerres.aacrjournals.org/content/71/12/4236.full#related-urls

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

mailto:[email protected]

http://cancerres.aacrjournals.org/content/71/12/4236


delineationofacellularhierarchyinlungcancerrevealsan ...tary fig. s1 and data not shown). together...

Documents