Identification of Metastasis-Associated Receptor Tyrosine Kinases

in Non–Small Cell Lung Cancer

Carsten Muller-Tidow,1Sven Diederichs,

1Etmar Bulk,

1Thorsten Pohle,

2Bjorn Steffen,

1

Joachim Schwable,1Sylvia Plewka,

1Michael Thomas,

1Ralf Metzger,

3Paul M. Schneider,

3

Christian H. Brandts,1Wolfgang E. Berdel,

1and Hubert Serve

1

1Department of Medicine, Hematology/Oncology; 2Department of Medicine, Gastroenterology, University of Munster,Munster; and 3Department of Visceral and Vascular Surgery, University of Cologne, Cologne, Germany

Abstract

Development of distant metastasis after tumor resection isthe leading cause of death in early-stage non–small cell lungcancer (NSCLC). Receptor tyrosine kinases (RTK) areinvolved in tumorigenesis but only few RTKs have beensystematically studied in NSCLC. Here, we provide quantita-tive real-time reverse transcription-PCR expression data ofall RTKs (n = 56) in primary tumors of 70 patients withearly-stage (I-IIIA) NSCLC. Overall, 33 RTKs were expressedin at least 25% of the patients. Several RTKs weresignificantly expressed higher in tumors that ultimatelymetastasized. The hazard risk for metastasis developmentin stage I/II disease was increased at least 3-fold for tumorswith high expression levels of insulin receptor, neurotrophictyrosine receptor kinase 1, epidermal growth factor receptor,ERBB2, ERBB3, platelet-derived growth factor receptor BBB,fibroblast growth factor receptor 1, or leukocyte tyrosinekinase. Relative risks were reduced 3-fold by expression ofEPHB6 or DKFZ1. Three members of the epidermal growthfactor receptor family were associated with a high risk ofmetastasis, emphasizing the validity of our data. High ERBB3expression was significantly associated with decreasedsurvival. Taken together, our genome-wide RTK expressionmap uncovered the previously unknown value of severalRTKs as potential markers for prognosis and metastasisprediction in early-stage NSCLC. The identified RTKsrepresent promising novel candidates for further functionalanalyses. (Cancer Res 2005; 65(5): 1778-82)

Introduction

Non–small cell lung cancer (NSCLC) ranks among the mostfrequent cancers in the world. Its mortality is only slightly lower

than its incidence indicating that most patients die from NSCLC

despite current therapy approaches. Only patients whose tumorscan be completely resected have a significant chance of cure.

Tumor resection is restricted to patients with stage I and II and

sometimes stage III disease. However, even patients with verysmall tumors often develop distant metastasis. It is unclear

whether the metastatic potential of individual tumors developsover time or whether the basic genetic program of the primarytumor predetermines the metastatic capability (1). Whereas bothconcepts seem reasonable, recent data from our lab and others’indicate that a metastatic program is inherent to tumors that dometastasize early (2, 3).Several approaches have been used to dissect the mechanisms

that determine the metastatic potential (2, 4–7). We describe anovel approach to identify metastasis-related genes and theirpotential use for diagnostic purposes. Based on the knowledgethat a few receptor tyrosine kinases [RTK; e.g., epidermal growthfactor receptor (EGFR) and ERBB2] are known to play animportant role in solid tumor metastasis (8–10), we reasoned thatother RTKs might also be important. Besides EGFR and ERBB2,only few RTKs have been studied for their involvement inmetastasis. Recently, we established a genome-wide RTK profilingapproach by real-time quantitative reverse transcription-PCR (RT-PCR; ref. 11). Here, we provide data on the role of RTKs in themetastasis of early-stage NSCLC tumors. In addition to the knownmetastasis-associated genes EGFR and ERBB2, several RTKspreviously not known to be associated with the metastaticprocess were identified as strong predictors for the developmentof metastasis in early-stage NSCLC.

Materials and Methods

Non–Small Cell Lung Cancer Patient Samples. Primary tumor

specimens were obtained at the time of initial surgery for early-stage

NSCLC (12). Histologic and survival data were published previously (2, 7).Patients with stage I/II disease had their tumors resected and did not

receive additional treatment, whereas stage IIIA patients were irradiated

following surgery. A long follow-up period of at least 36 months ensured

the correct classification of metastasizing and non-metastasizingpatients.

RNA Isolation and cDNA Preparation. The tumor samples were

pathohistologically analyzed for the percentage of tumor cells. Onlytumor biopsies with at least 70% cancer cells were used. RNA was

isolated using TRIzol reagent (Invitrogen, San Diego, CA). A total of 1 AgRNA from each sample was reverse-transcribed using oligo-d(T) primer

and Moloney murine leukemia virus reverse transcriptase (Clontech, PaloAlto, CA).

Primer and Probe Design. Sequence information was obtained from

Genbank and previously published data (13). In brief, primers and probeswere designed to span exon-exon junctions and to be outside of the

conserved kinase domain using Primer Express software (Applied

Biosystems, Foster City, CA). The resulting primer and probe sequenceswere verified by alignment. The GAPDH probe was labeled with 5V-VIC and

3V-TAMRA. All other probes were labeled 5V-FAM and 3V-TAMRA (Euro-

GenTec, Seraing, Belgium). Reliability of PCR amplification and detection

was verified on serial dilutions of standard cDNA before analysis of patientsamples.

Note: C. Muller-Tidow and S. Diederichs contributed equally to this work.Requests for reprints: Carsten Muller-Tidow, Department of Medicine A,

Hematology/Oncology, University of Munster, Domagkstr. 3, 48149 Munster,Germany. Phone: 49-251-835-6229; Fax: 49-251-835-2673; E-mail: muellerc@uni-muenster.de.

I2005 American Association for Cancer Research.

Cancer Res 2005; 65: (5). March 1, 2005 1778 www.aacrjournals.org

Research Article

Semiautomated Analysis of Gene Expression by Real-timeQuantitative Reverse Transcription-PCR. A semiautomated setup

(Tecan Genesis RP150 automated pipetting system) was established for

reliable and rapid RT-PCR analysis of 384 wells in parallel. The PCR

reaction mixture contained 600 nmol/L of each primer and 200 nmol/L

probe in a final volume of 22.5 AL. PCR conditions were 50jC/10 s, 95jC/10 min, and 40 cycles of 95jC/15 s and 60jC/min in a real-time PCR

machine (ABI PRISM 7900 Sequence Detector, TaqMan). Expression levels

of RTKs or the housekeeping gene GAPDH were quantitated using a

fluorescence-based detection method described previously (11, 14, 15).

Relative gene expression levels were calculated using standard curves

generated by serial dilutions of a cDNA mixture by the SDS 2.0 software.

Expression levels for each gene and each sample were divided by the

GAPDH expression level. Several genes were independently analyzed twice

and reproducibility was always excellentwith a correlation coefficient r > 0.95

and P < 0.001.

Statistical Analysis. Statistical analyses were carried out using SPSS

11.0. Expression differences were tested for significance using Mann-

Whitney U and Kruskal-Wallis analysis. Cross-table analyses and

hazard risks were calculated using Fisher’s exact test. For survival

analyses, Cox regression analyses and Kaplan-Meier tests were done

and significance was calculated using the log-rank test. All tests were

executed two sided and the error level was set at 5% (P < 0.05). Owing

to this study designed to identify novel relevant RTKs in the met-

astatic process, Bonferroni or Sidak posthoc analyses were not done toavoid type II errors despite the increased risk of type I errors (16).

Therefore, statistical values should be considered of exploratory

significance.

ResultsAnalysis of Receptor Tyrosine Kinase Expression by Quan-

titative Real-time Reverse Transcription-PCR. From the humangenome project, a total of 56 RTKs have been identified. We haverecently analyzed the expression patterns of all RTKs in a widevariety of human cancers (11). These analyses established vastdifferences in the expression of RTKs in different types of humantumors.In the current study, we analyzed the association of RTKs with

clinical features in patients with early-stage NSCLC. According toquantitative real-time RT-PCR results, nine RTKs were expressed inmore than 75% of all NSCLC tumors, 18 RTKs were expressed inmore than 50% of the patients, and 33 RTKs were expressed in atleast 25% of the tumors (Fig. 1A). RTK expression patterns ofindividual tumors showed a high degree of variability (Fig. 1B).The different histologic NSCLC subtypes differed in RTK

expression levels: adenocarcinomas expressed higher levelsof RON (P = 0.02), c-KIT (P = 0.04), ERBB2 (P = 0.001), and ERBB3

Figure 1. RTK expression in NSCLC. A, RTK expression was analyzed by quantitative real-time RT-PCR. The bar diagram indicates the percentage of NSCLCsamples (n = 70) with detectable mRNA expression for each RTK. B, mRNA expression levels of all RTKs in 70 NSCLC samples grouped according to theirhistologic subtype. Blue, samples without detectable expression. Expression levels of RT-PCR positive samples are indicated in log scale.

Metastasis-Associated RTKs in Non–Small Cell Lung Cancer

www.aacrjournals.org 1779 Cancer Res 2005; 65: (5). March 1, 2005

(P < 0.001), whereas large cell carcinomas expressed the highest

levels of fibroblast growth factor receptor 1 (P = 0.03). EPHB6

levels were high in squamous cell carcinomas (P = 0.009) and

expression of neurotrophic tyrosine receptor kinase 2 was notably

absent in adenocarcinomas (P = 0.01). In our study, patients with

stage I or II disease expressed significantly higher levels of c-KIT

(P = 0.04) and fibroblast growth factor receptor 1 (P = 0.047) in

comparison to stage IIIA patients. Expression levels of three

RTKs, MER (P = 0.01), EPHB2 (P = 0.04), and EPHB6 (P = 0.09,

not significant), differed with regard to the smoking history of

the patients.

Receptor Tyrosine Kinases Predict Metastasis and Survivalin Early-stage Non–Small Cell Lung Cancer. The association

between RTK expression and metastasis or survival was analyzed

in all tumors (n = 62) and separately in stage I/II NSCLCtumors (n = 44). For these analyses, patient samples were split

into two groups depending on the expression level of each RTK.

To avoid a bias in data analysis, the median of the analyzedsamples was used as a cutoff value for each RTK. Samples

expressing levels higher than the median were regarded as

‘‘high’’ whereas samples with expression levels below the median

were regarded as ‘‘low.’’ For RTKs with less than half of thesamples showing significant expression levels, samples with

detectable expression were regarded as ‘‘high.’’ We used this

dichotomized data to analyze the risk of patients with high levelexpression of a certain RTK to metastasize. The hazard ratio

was calculated by cross-table analysis and is shown for stage I/II

patients (Fig. 2).Two RTKs were associated with reduced risk of metastasis

(hazard ratio less than 0.3) when expressed in the primary tumor:DKFZ1 or EPHB6 (Fig. 2). DKFZ1 expression was detected in 12of 44 (27%) stage I/II NSCLC patients, whereas EPHB6 wasexpressed in 18 of 44 patients (41%). Stage I/II patients withtumors expressing either DKFZ1 or EPHB6 had an excellentprognosis with only 2 of 23 (8.7%) patients developing distantmetastasis compared with 8 of 21 (38.1%) in patients with tumorsexpressing neither RTK (Fig. 3A). Rates in metastasis developmentwere reflected in Kaplan-Meier plots indicating far better survivalfor patients with tumors expressing either one or both RTKs(DKFZ1 and EPHB6). This result was consistent if NSCLC tumorsfrom all stages were included (n = 62, P = 0.02) or if only tumorsfrom stage I/II were analyzed (n = 44, P = 0.02). The meansurvival differed significantly between 71 months (expression ofeither DKFZ1 or EPHB6) versus 52 months (no expression ofeither RTK; n = 62). In a multivariate Cox regression analysisincluding age, sex, smoking status, tumor grade, and histology, theexpression of either DKFZ1 or EPHB6 emerged as an independentprognostic factor for survival (P = 0.02 for n = 62; P = 0.018 forn = 44 stage I/II only; Fig. 3B).High expression of eight RTKs was associated with an at least

thrice increased risk of distant metastasis (Fig. 2). High expressionlevels of the insulin receptor indicated a more than 7-fold increasedrisk to develop metastasis (95% CI 1.3-40.2, P = 0.03) for patientswith stage I/II disease. The neurotrophic tyrosine receptor kinase 1/nerve cell growth factor receptor 1 predicted a 5.6-fold increasedrisk of metastasis (CI 1.2-26.1, P = 0.03) when overexpressed (Fig. 2).Patients not expressing either of these RTKs did not suffer distantmetastasis (0 of 17 = 0.0%), whereas patients expressing either RTKfrequently developed metastases (10 of 27 = 37.0%; Fig. 4A).In addition, the overall survival of patients expressing neither of

these two RTKs was significantly better in Kaplan-Meier and Coxregression analyses (P = 0.033 and P = 0.014, respectively)independent of tumor stage (Fig. 4B).In accordance with published data (17), EGFR and ERBB2 were

among the genes predictive of metastasis (Fig. 2). In addition, theEGFR family member ERBB3 was closely associated withmetastasis. High expression of either EGFR (33.3% metastasisversus 13.0% metastasis in the low expressor group) or ERBB2(33.3% versus 13.0%) or ERBB3 (33.3% versus 10.0%) indicated anincreased fraction of metastasizing tumors whereas ERBB4 did notdisplay this effect (Fig. 4A). In Kaplan-Meier survival analysis, stageI/II patients with high expression of ERBB3 showed significantlyshorter survival (P = 0.044; Fig. 4B).

Discussion

In this study, we provide a detailed map of RTK expression inNSCLC. Our systematic approach corroborates the importance ofRTKs already known to be involved in NSCLC but also indicatesthat other thus far neglected RTKs might have an important rolethere. One focus of our study was to reveal their potentialinvolvement in metastasis. In recent years, we and others have used

Figure 2. Relative hazard to develop distant metastasis based on RTKexpression. Tumors were subdivided into a high or low expressing groupfor each RTK. The relative risk to develop distant metastasis-associated withhigh expression of the respective RTK was calculated for stage I and II patients(n = 44). The dashed lines indicate the threshold for either 3-fold reduced or3-fold increased risk of metastasis.

Cancer Research

Cancer Res 2005; 65: (5). March 1, 2005 1780 www.aacrjournals.org

various techniques to identify genes that are associated with therisk of distant metastasis in NSCLC revealing novel genes linked tometastasis (2, 7, 12). Several groups have tried educated guessapproaches and have identified promising metastasis markers (e.g.,refs. 5, 6, 18). The best known example is the EGFR whichconstitutes an important target for directed therapies (19). Owingto RTKs often being at the mechanistic basis of pathogeneticprocesses in cancer, we decided to study the expression of all RTKsin NSCLC metastasis. We chose quantitative real-time RT-PCRbecause of its high specificity and unparalleled sensitivity.First, expression levels of eight genes were associated with an

increased risk of metastasis. Among these, three EGFR familymembers were found. Because EGFR and ERBB2 are well knownto be associated with metastasis (17), these findings provideevidence for the reliability and validity of our approach. Our datafurther indicate that ERBB3, a ‘‘kinase dead’’ RTK that formsheterodimers with EGFR and ERBB2 (20, 21), is closely linked toNSCLC metastasis. RTKs can exert oncogenic functions either dueto overexpression or activating mutations (e.g., ref. 22). In theEGFR, a paradigm for an overexpressed oncogene due to, forexample, gene amplification (23), mutations were recently

identified and closely associated with patients’ response to

gefitinib (24, 25). None of the other screened RTKs containedmutations (25) whereas mutations of ERBB2 in lung cancer have

recently been found (26). Despite the clinically relevant EGFR

mutations, overexpression of EGFR and possibly of other RTKs

remains a dominant mechanism within tumor pathogenesis due

to high EGFR expression found in almost all patients whereas

mutations were relatively rare.In addition to EGFR family members, several RTKs were newly

identified to confer an increased risk of metastasis. The insulinreceptor and its related growth factors and receptors are wellknown to be associated with tumor cell growth (27) but apotential role in NSCLC metastasis has not been consideredbefore. The neurotrophic tyrosine receptor kinase 1/nerve cellgrowth factor receptor 1 receptor is overexpressed in neuro-blastoma but associates with a rather good prognosis in thistumor entity (28). In human lung cancer cell lines, nerve growthfactor signaling stimulates clonal growth (29). We conclude thatoverexpression of RTKs could not only contribute to oncogenictransformation but might also enhance metastatic properties.

Figure 3. DKFZ1 and EPHB6 expression associates with reduced risk ofmetastasis and increased survival. A, early-stage NSCLC patients (stage I/II,n = 44) expressing neither DKFZ1 nor EPHB6 exhibited the highest risk ofdistant metastasis (38.1%) whereas patients expressing both RTKs wereassociated with a much lower incidence of metastasis (8.7%). B, inKaplan-Meier and Cox regression analyses (see text), expression of DKFZ1or EPHB6 was significantly associated with a survival benefit for allpatients (n = 62) or specifically for stage I/II patients (n = 44).

Figure 4. Insulin receptor, neurotrophic tyrosine receptor kinase 1, and ERBB3expression associates with increased risk of metastasis and decreasedsurvival. A, expression of insulin receptor, neurotrophic tyrosine receptor kinase1, or one EGFR family member (EGFR, ERBB2, or ERBB3) in early-stageNSCLC correlated with an increased proportion of patients developing distantmetastasis. B, in all analyzed NSCLC patients, expression of either insulinreceptor or neurotrophic tyrosine receptor kinase 1 contributed to significantlyimpairedoverall survival in Kaplan-Meier and Cox regression analyses. ForERBB3, a similar correlation was found for early-stage NSCLC patients.

Metastasis-Associated RTKs in Non–Small Cell Lung Cancer

www.aacrjournals.org 1781 Cancer Res 2005; 65: (5). March 1, 2005

Taken together, our findings indicate that besides the well-known EGFR family, several other RTKs could play a role in NSCLCmetastasis. The RTK expression differences can serve as markersfor prognosis and metastasis prediction. This RTK expression mapwill guide future analyses to uncover the detailed functions of RTKsin NSCLC pathogenesis and metastasis. Our data provide the basisfor additional basic studies in vitro and in vivo .

AcknowledgmentsReceived 9/17/2004; revised 11/29/2004; accepted 12/23/2004.

Grant support: Research in our lab is supported by grants from the WilhelmSander-Foundation (2001.086.2), DFG (Heisenberg program Mu1328/3-1) and GermanCancer Aid (Deutsche Krebshilfe 10-2155-Mu3).

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

References1. Muller-Tidow C, Diederichs S, Thomas M, Serve H.Genome-wide screening for prognosis-predicting genesin early-stage non–small-cell lung cancer. Lung Cancer2004;45:S145–50.

2. Diederichs S, Bulk E, Steffen B, et al. S100 familymembers and trypsinogens are predictors of distantmetastasis and survival in early-stage non–small celllung cancer. Cancer Res 2004;64:5564–9.

3. Ramaswamy S, Ross KN, Lander ES, Golub TR. Amolecular signature of metastasis in primary solidtumors. Nat Genet 2003;33:49–54.

4. Bhattacharjee A, Richards WG, Staunton J, et al.Classification of human lung carcinomas by mRNAexpression profiling reveals distinct adenocarcinomasubclasses. Proc Natl Acad Sci U S A 2001;98:13790–5.

5. Gemma A, Takenaka K, Hosoya Y, et al. Alteredexpression of several genes in highly metastaticsubpopulations of a human pulmonary adenocarcino-ma cell line. Eur J Cancer 2001;37:1554–61.

6. Wigle DA, Jurisica I, Radulovich N, et al. Molecularprofiling of non–small cell lung cancer and correlationwith disease-free survival. Cancer Res 2002;62:3005–8.

7. Ji P, Diederichs S, Wang W, et al. MALAT-1, a novelnoncoding RNA, and thymosinh4 predict metastasisand survival in early-stage non–small cell lung cancer.Oncogene 2003;22:6087–97.

8. Fontanini G, Vignati S, Bigini D, et al. Epidermalgrowth factor receptor (EGFr) expression in non–smallcell lung carcinomas correlates with metastatic involve-ment of hilar and mediastinal lymph nodes in thesquamous subtype. Eur J Cancer 1995;31A:178–83.

9. Brandt BH, Roetger A, Dittmar T, et al. c–erbB–2/EGFR as dominant heterodimerization partners deter-mine a mitogenic phenotype in human breast cancercells. Faseb J 1999;13:1939–49.

10. Saucier C, Papavasiliou V, Palazzo A, Naujokas MA,Kremer R, Park M. Use of signal specific receptor

tyrosine kinase oncoproteins reveals that pathwaysdownstream from Grb2 or Shc are sufficient for celltransformation and metastasis. Oncogene 2002;21:1800–11.

11. Muller–Tidow C, Schwable J, Steffen B, et al. High–throughput analysis of genome–wide receptor tyro-sine kinase expression in human cancers identifiespotential novel drug targets. Clin Cancer Res 2004;10:1241–9.

12. Muller-Tidow C, Metzger R, Kugler K, et al. CyclinE is the only cyclin-dependent kinase 2–associatedcyclin that predicts metastasis and survival in earlystage non–small cell lung cancer. Cancer Res 2001;61:647–53.

13. Robinson DR, Wu YM, Lin SF. The protein tyrosinekinase family of the human genome. Oncogene2000;19:5548–57.

14. Heid CA, Stevens J, Livak KJ, Williams PM. Real timequantitative PCR. Genome Res 1996;6:986–94.

15. Gibson UE, Heid CA, Williams PM. A novel methodfor real time quantitative RT-PCR. Genome Res 1996;6:995–1001.

16. Perneger TV. What’s wrong with Bonferroni adjust-ments. BMJ 1998;316:1236–8.

17. Wells A. Tumor invasion: role of growth factor-induced cell motility. Adv Cancer Res 2000;78:31–101.

18. Marchetti A, Tinari N, Buttitta F, et al. Expression of90K (Mac-2 BP) correlates with distant metastasis andpredicts survival in stage I non–small cell lung cancerpatients. Cancer Res 2002;62:2535–9.

19. Cappuzzo F, Gregorc V, Rossi E, et al. Gefitinib inpretreated non–small-cell lung cancer (NSCLC): anal-ysis of efficacy and correlation with HER2 andepidermal growth factor receptor expression in locallyadvanced or metastatic NSCLC. J Clin Oncol 2003;21:2658–63.

20. Guy PM, Platko JV, Cantley LC, Cerione RA,Carraway KL III. Insect cell-expressed p180erbB3

possesses an impaired tyrosine kinase activity. ProcNatl Acad Sci U S A 1994;91:8132–6.

21. Holbro T, Beerli RR, Maurer F, Koziczak M,Barbas CF III, Hynes NE. The ErbB2/ErbB3 hetero-dimer functions as an oncogenic unit: ErbB2requires ErbB3 to drive breast tumor cell prolifera-tion. Proc Natl Acad Sci U S A 2003;100:8933–8.

22. Mizuki M, Schwable J, Steur C, et al. Suppression ofmyeloid transcription factors and induction of STATresponse genes by AML-specific Flt3 mutations. Blood2003;101:3164–73.

23. Tidow N, Boecker A, Schmidt H, et al. Distinctamplification of an untranslated regulatory sequence inthe egfr gene contributes to early steps in breast cancerdevelopment. Cancer Res 2003;63:1172–8.

24. Lynch TJ, Bell DW, Sordella R, et al. Activatingmutations in the epidermal growth factor receptorunderlying responsiveness of non–small-cell lung can-cer to gefitinib. N Engl J Med 2004;350:2129–39.

25. Paez JG, Janne PA, Lee JC, et al. EGFR mutations inlung cancer: correlation with clinical response togefitinib therapy. Science 2004;304:1497–500.

26. Stephens P, Hunter C, Bignell G, et al. Lung cancer:intragenic ERBB2 kinase mutations in tumours. Nature2004;431:525–6.

27. Yu H, Rohan T. Role of the insulin-like growth factorfamily in cancer development and progression. J NatlCancer Inst 2000;92:1472–89.

28. Kogner P, Barbany G, Dominici C, Castello MA,Raschella G, Persson H. Coexpression of messengerRNA for TRK proto-oncogene and low affinity nervegrowth factor receptor in neuroblastoma with favorableprognosis. Cancer Res 1993;53:2044–50.

29. Oelmann E, Sreter L, Schuller I, et al. Nerve growthfactor stimulates clonal growth of human lung cancercell lines and a human glioblastoma cell line express-ing high-affinity nerve growth factor binding sitesinvolving tyrosine kinase signaling. Cancer Res 1995;55:2212–9.

Cancer Research

Cancer Res 2005; 65: (5). March 1, 2005 1782 www.aacrjournals.org