Tumor endothelial marker 1 (Tem1) functions in thegrowth and progression of abdominal tumorsAkash Nanda*†‡, Baktiar Karim‡§, Zhongsheng Peng§, Guosheng Liu§, Weiping Qiu§, Christine Gan*†,Bert Vogelstein*†¶�, Brad St. Croix**, Kenneth W. Kinzler*†�, and David L. Huso*§�

*The Sidney Kimmel Comprehensive Cancer Center, †Ludwig Center for Cancer Genetics and Therapeutics, §Department of Comparative Medicine, and¶Howard Hughes Medical Institute, Johns Hopkins Medical Institutions, Baltimore, MD 21205; and **Tumor Angiogenesis Section, Mouse Cancer GeneticsProgram, National Cancer Institute, National Institutes of Health, Frederick, MD 21702

Contributed by Bert Vogelstein, December 31, 2005

Tumor endothelial marker 1 (Tem1; endosialin) is the prototypicalmember of a family of genes expressed in the stroma of tumors. Toassess the functional role of Tem1, we disrupted the Tem1 gene inmice by targeted homologous recombination. Tem1�/� mice werehealthy, their wound healing was normal, and tumors grew nor-mally when implanted in s.c. sites. However, there was a strikingreduction in tumor growth, invasiveness, and metastasis aftertransplantation of tumors to abdominal sites in mice withoutfunctional Tem1 genes. These data indicate that the stroma cancontrol tumor aggressiveness and that this control varies withanatomic site. Therefore, they have significant implications for themechanisms underlying tumor invasiveness and for models thatevaluate this process.

endosialin � metastasis � stroma � tumor invasiveness � angiogenesis

Tumor angiogenesis is required for the growth of humantumors to clinically relevant sizes (1–4). Global analysis of

altered gene expression patterns in endothelial cells from humancolorectal cancer tissues identified a series of genes termedtumor endothelial markers (TEMs) (5). These genes were foundto be overexpressed in the endothelium of both human andanimal tumors (5–7). TEM1 is the prototype of this class, with itsmRNA exhibiting the highest differential expression betweennormal and human tumor endothelium. Interestingly, after itsdiscovery as a TEM, TEM1 was found to encode endosialin, theantigen recognized by a monoclonal antibody that had beendeveloped nearly a decade earlier and shown to specifically reactwith human tumor endothelium and fibroblast-like cells withinthe tumor stroma (8, 9). Endosialin is a 165-kDa single-passtransmembrane glycoprotein that has been classified as a C typelectin-like membrane receptor. It has multiple extracellulardomains consisting of three EGF-like domains, a sushi-likedomain, and a C lectin-like domain (6, 9).

In addition to the importance of TEM gene products aspotential therapeutic targets (10, 11), the study of their func-tional role(s) may provide new insights into angiogenesis. To-ward this goal, we generated a mouse model with a targeted nullmutation in the Tem1 gene. We demonstrate that Tem1 is notrequired for angiogenesis during fetal development, postnatalgrowth, or wound healing. However, optimal growth, invasion,and metastasis of tumors implanted in abdominal sites of adultmice were found to critically depend on Tem1 expression. Thesedata demonstrate that Tem1 plays a critical role in determiningtumor behavior in clinically relevant anatomical sites.

ResultsGeneration and Characterization of Tem1 Knockout (KO) Mice. Atargeting vector was designed to knock out the Tem1 gene byhomologous recombination in mouse ES cells (Fig. 1A). PCRwas used to distinguish homozygous (KO), heterozygous (HET),and WT mouse genotypes (Fig. 1B). Tem1 has been reported tobe expressed during embryonic development but at very lowlevels in adults (6). Using RT-PCR, we detected Tem1 expression

in embryonic liver, skin, and colon late in gestation (embryonicday 15.5) of WT but not KO mice (Fig. 1C). Null mice of bothsexes were phenotypically normal, and crossing HET male andfemale mice resulted in WT, HET, and null offspring at theexpected Mendelian ratios and litter sizes (Fig. 2A). There wereno differences in growth rates between WT and KO mice (Fig.2B). Additionally, no abnormalities were apparent in any organsystem after complete histopathological analysis of fetal, new-born, and adult mouse tissues. These experiments demonstratedthat Tem1 disruption did not cause any lethal embryonic defectsor interfere with development or fertility.

s.c. Vascularization. We next assessed s.c. vascularization in KOand WT mice in a wound healing assay. During healing of alesion inflicted with a 6-mm punch biopsy, there were no

Conflict of interest statement: Under a licensing agreement between The Johns HopkinsUniversity and Genzyme, technologies related to serial analysis of gene expression and theTEMs were licensed to Genzyme for commercial purposes, and B.V., K.W.K., and B.S.C. areentitled to a share of the royalties received by the university from the sales of the licensedtechnologies. The serial analysis of gene expression technology is freely available toacademia for research purposes. K.W.K. is a consultant to Genzyme. The university andresearchers (B.V. and K.W.K.) own Genzyme stock, which is subject to certain restrictionsunder university policy. The terms of these arrangements are being managed by theuniversity in accordance with its conflict of interest policies.

Abbreviations: KO, knockout; LLC, Lewis lung carcinoma; TEM, tumor endothelial marker.

‡A.N. and B.K. contributed equally to this work.

�To whom correspondence may be addressed. E-mail: vogelbe@welch.jhu.edu, kinzlke@jhmi.edu, or dhuso@jhmi.edu.

© 2006 by The National Academy of Sciences of the USA

Fig. 1. Generation of Tem1 KO mice. (A) The strategy for targeted disruptionof the single exon of the Tem1 gene. (B) PCR of genomic DNA. (C) RT-PCR ofmRNA from the indicated mouse embryonic tissues. The lane marked Mcontains a 250-bp marker fragment.

www.pnas.org�cgi�doi�10.1073�pnas.0511306103 PNAS � February 28, 2006 � vol. 103 � no. 9 � 3351–3356

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020

significant differences in the rate of wound closure between KOand WT mice (Fig. 3A). Histopathological examination of thewound sites also revealed no morphological differences duringwound healing in these mice. To examine the process of angio-genesis in wounds, the vasculature was stained for PECAM-1 orwith rabbit anti-human von Willebrand factor (Factor VIII).Stained vessels were quantified in histological sections madefrom the wound site at the time of closure (14 days). Nomorphological differences between KO and WT mice wereapparent at the wound sites. In addition, there were no differ-ences in the number or size of vessels formed at the site of thewounds (Fig. 3B).

We next compared the growth of s.c. tumors in KO and WTmice. There were no differences in the growth rate of the Lewislung carcinoma (LLC) tumor cell line (Fig. 3C) or the S180sarcoma cell line (data not shown) after s.c. injection into KOand WT mice. We conclude that skin wound healing and thegrowth of s.c. tumors does not depend on Tem1 expression.

Abdominal Orthotopic Xenografts in Immunodeficient Mice. Becausethe TEM1 transcript was first identified by virtue of its overex-pression in the endothelium of colorectal tumors, we comparedthe growth of HCT116 human colorectal cancer cells as ortho-topic xenografts in an immunodeficient, nude mouse back-ground. Tumor fragments derived from the HCT116 cell linewere surgically attached to the serosal surface of the largeintestines of KO and WT nude mice. Over a 19-day period oftumor growth, KO mice survival was 89% compared to 11% for

WT mice (Fig. 4A). In addition, WT mice had a much higherincidence of peritoneal carcinomatosis (89% vs. 33%) and livermetastases (33% vs. 0) than KO mice (Fig. 4B). Accordingly,total tumor volume was significantly larger in WT compared toKO animals (P � 0.018).

Tumor Growth at Abdominal Sites in Immunocompetent Mice. Thegrowth of implanted tumors at abdominal sites was compared inimmunocompetent KO and WT mice by using the LLC cell line.In the first set of experiments, tumor fragments were surgicallyimplanted into the livers of KO and WT mice as described in

Fig. 2. Effects of Tem1 KO on reproduction and growth. (A) Mendelian ratio.Males and females that were each heterozygous for Tem1 KO were crossedand genotypes of the offspring born were determined. (B) Growth rates ofmice at the indicated ages were determined. The solid and broken linesrepresent WT and mutant mice, respectively. Six mice were used in each group,and means and SD are indicated.

Fig. 3. Effects of Tem1 KO on s.c. vascularization. (A) Rate of wound closure.At least three mice were used for each time point and means and SD areindicated. (B) Vessel number and vessel size at wound site after 14 days. Fivemice were used for each genotype. (C) Rate of tumor growth after s.c. injectionof LLC cells. The solid and broken lines represent WT and mutant mice,respectively. Three mice of each genotype were used, and means and SD areindicated.

3352 � www.pnas.org�cgi�doi�10.1073�pnas.0511306103 Nanda et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020

Methods. Three weeks after surgery, the tumors in WT mice hadgrown to an average volume that was more than twice that oftumors implanted into the livers of KO mice (Fig. 5A, P � 0.041).In a second set of experiments, tumor fragments were surgicallyimplanted onto the serosal surface of the large intestine. Thetumors in WT mice grew to more than four times the averagetumor volume compared to similarly implanted tumors in KOmice (Fig. 5B, P � 0.0002). The pattern of growth was strikinglydifferent in WT vs. KO mice. The tumors in WT mice grewspherically around the entire colon, encircling it within a largetumor mass (Fig. 5 C and D). In contrast, the tumors in KO micegrew only laterally, generating flattened lesions along the serosalsurface (Fig. 5 C and D). Histologically, tumors from WT micenot only expanded spherically but also extensively invaded thefull thickness of the intestine, including the epithelial layer liningthe intestinal lumen (Fig. 5E). Growth of the tumors in KO miceabruptly ceased when cells reached the lamina propria, com-pletely failing to invade the inner layers of the intestine (Fig. 5E).

Vascularization of Implanted Tumors. As expected, Tem1 expressionwas detected in the tumors in WT animals beginning at 2 weeksafter transplantation in WT mice but could not be detected inKO mice (Fig. 6A). In situ hybridization performed at 3 weeksof tumor growth revealed that VE-cadherin-positive blood ves-sels were present in the tumor stroma of both WT and KO mice(Fig. 6B). In contrast, Tem1 expression was detected in cellslining vessels and in fibroblast-like cells in the stroma of tumors

grown in WT mice but in neither cell type of Tem1 KO mice(Fig. 6B).

We quantified the number of vessels of various sizes in sections

Fig. 4. Growth of HCT116 orthotopic xenografts in nu�nu mice. (A) Survivalcurve after implantation of HCT116 tumor fragments onto the serosal surfaceof the large intestine. Nine mice of each genotype were assessed. (B) Incidenceof peritoneal carcinomatosis and liver metastasis. Nine mice of each genotypewere evaluated.

Fig. 5. Growth of abdominal tumors in immunocompetent mice. (A) Aver-age volume of tumors 3 weeks after implantation into the liver. Six mice ofeach genotype were assessed, and means and SD are shown (P � 0.04). (B)Average volume of tumors 3.5 weeks after implantation onto the serosalsurface of the colon. Six mice of each genotype were assessed. and means andSD are shown (P � 0.0002). (C) Orthotopic tumors (arrows) growing on theserosal surface of the large intestine of WT or KO recipient mice, 3.5 weeksafter implantation. (D and E) Low and high power views of the same tumorsafter histopathologic processing. The tumor cells in the WT mice invaded theintestinal epithelial cells surrounding the lumen (arrows), whereas those inthe KO mice did not cross the muscularis propria (arrow). (Scale bars: 100 �m.)

Nanda et al. PNAS � February 28, 2006 � vol. 103 � no. 9 � 3353

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020

of tumors that had been implanted on the serosal surface ofcolons in immunocompetent mice. For this purpose, formalin-fixed sections were stained by immunohistochemistry for CD31antigen (PECAM-1), a marker of vascular endothelium (Fig.6C). Vessels of 50–100 �m diameter and those of �100 �m

diameter were more numerous in tumors from WT than in thosefrom KO mice, although this difference was not statisticallysignificant (Fig. 6D). However, there was a statistically signifi-cant increase in the number of smaller vessels (�50 �m diam-eter) in tumors from KO mice compared to those from WT mice(Fig. 6D). There was no qualitative or quantitative differences intumors grown in KO or WT mice when assessed by immuno-histochemistry with antibodies to alpha-smooth muscle actin(data not shown). This marker for pericyte coverage of vesselsincreases with vessel maturation.

DiscussionThe results reported above demonstrate that Tem1 plays animportant functional role in experimental tumor progression.KO of Tem1 dramatically changed tumor growth patterns,altered vascularization, decreased bulk growth, prevented localinvasion, and reduced metastasis. Interestingly, these alterationsin tumor growth were not accompanied by any apparent abnor-malities during early development or in wound healing. Theysuggest that tumors are more than unhealed wounds and thatspecific host-associated factors like Tem1 play a very specific rolein tumor stroma formation (12). The fact that dramatic effectswere observed in abdominal but not s.c. tumors demonstratesthat the associated processes were organ-site specific. It may berelevant that normal angiogenic factors can also be expressed inan organ site-specific fashion (13–18). One of the most importantimplications of the work described here is that conventionalanimal models for studying tumorigenesis, employing s.c. inoc-ulation of tumor cells, may not be optimal for assessing thephysiologic parameters that govern tumor cell growth andinvasion in their natural environment (19, 20).

Specific factors from stromal cells that influence tumor pro-gression, invasion, and metastasis are only beginning to come tolight (21–26). Recent examples include the CXCL14 andCXCL12 chemokines that are overexpressed in tumor myoepi-thelial cells and myofibroblasts, respectively, and bind to recep-tors on epithelial cells and enhance their proliferation, migra-tion, and invasion (27, 28). TEM1 expression is presumablyconfined to the cell surface and, thereby, would be expected tomediate its effects through direct cell to cell interactions ratherthan via a paracrine manner.

Our results are consistent with previous studies showing thatTem1 is expressed predominantly on cells lining blood vesselsbut could also be detected on fibroblast-like stromal cells (6,8, 29). The in situ hybridization techniques we used in micecannot discriminate between endothelial cells and pericytes.Similarly, we cannot distinguish whether the effects of Tem1disruption are due to the absence of Tem1 expression in bloodvessels vs. their absence in fibroblast-like stromal cells. Theinteresting possibility exists, therefore, that the effects ontumors are primarily due to the absence of Tem1 in thefibroblastic stroma. According to this hypothesis, the increasednumber of small blood vessels that are observed in the stromaof tumors in Tem1 KO mice represent the response to anabnormal microenvironment (21–26). We speculate that in-teraction between tumor cells and endosialin normally inducesthe expression of other proteins in the tumor cells or stromathat facilitate tumor invasion and growth. In the absence ofTem1, a different gene expression program is called into play,one that is not conducive to invasion but is angiogenic. In analternate hypothesis, the effects of TEM1 on vessels and tumorgrowth may be exclusively dependent on TEM1 expression invessels. According to this model, TEM1 would be required forthe efficient maturation of vessels within tumors. In Tem1 KOmice, vessels would fail to mature efficiently, leading todecreased numbers of medium and large vessels and a com-pensatory increase in small vessels. This abnormal angiogenicresponse could dampen tumor growth and invasion. Regard-

Fig. 6. Angiogenesis in abdominal tumors. (A) RT-PCR analysis of Tem1expression in allograft tumors of the large intestine at the indicated timesafter implantation. The lane marked M contained size markers. (B) In situhybridization using probes for VE-cadherin and Tem1 in the tumors. (C)Immunohistochemistry of CD31� vessels in an abdominal allograft tumorimplanted on the serosal surface of the large intestine of a Tem1 KO mouse.(Scale bars: Left, 100 �m; Right, 15 �m.) (D) Vessel number and size. Four miceof each genotype were evaluated, and means and SD are shown (P � 0.024 forvessels �50 �m).

3354 � www.pnas.org�cgi�doi�10.1073�pnas.0511306103 Nanda et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020

less of which model holds true, the data indicate that thestroma can control tumor aggressiveness, and, importantly,that this control can vary with anatomic site.

Our results raise several questions that can be addressed infuture studies. Cell-specific KO of the Tem1 gene should be ableto determine whether the effects of Tem1 on tumor growth aredue to its expression in fibroblastic stromal cells rather thanvascular cells. Identification of the receptor for Tem1, and itsintracytoplasmic signaling pathways, will obviously be crucial tofurther understanding the mechanisms through which Tem1affects tumor invasiveness.

MethodsTargeting Construct and Generation of Tem1 Null Mice. GenomicDNA clones were isolated from a bacteriophage library gener-ated from 129SvEv mice (Taconic Farms). A targeting vector(pKO Scrambler 907, Stratagene) was constructed in which theneo (reverse orientation) and diphtheria toxin genes were in-cluded for positive and negative selection, respectively. A 5.8-kb5� genomic homology arm and a 3.4-kb 3� genomic homologyarm were used to target the single exon Tem1 gene. Electropo-ration of 129SvEv ES cells, selection of colonies, and microin-jection into 129SvEv blastocysts to generate chimeric mice wasperformed by using conventional methods (inGenious TargetingLaboratory, Stony Brook, NY). Appropriate targeting by 5� and3� homology arms was confirmed by PCR. Chimeric mice weremated to C57Bl6 to select for germline transmission and gen-erate heterozygotes. Mouse colonies were generated by using amating scheme devoid of parent-offspring and sibling matings tomaintain a mixed genetic background. All experiments exceptthose involving xenografts were carried out on mice of this mixedB6�129 background. For xenograft experiments, B6�129 micewere crossed with outbred nude mice (Charles River Laborato-ries) to generate colonies of null and WT Tem1 nude mice. Allanimal experimental procedures were approved by the Institu-tional Animal Care and Use Committee at The Johns HopkinsUniversity.

Genotyping of Tem1 Mutant Mice by Genomic PCR. Genotyping wascarried out by PCR of DNA isolated from tail clippings(GenElute, Sigma). The primers used were as follows: 5�-AGCGCATGCTCCAGACTGC-3� and 5�-CGTCTGGTTT-GAGTTAGGGG-3� to detect the mutant allele and 5�-TCTACCCTCAGCTACCCCC-3� and 5�- GCTTTGGTGGTC-GATATGTG-3� to detect the WT allele.

Expression of Tem1 by RT-PCR. Skin, liver, and colon from embry-onic day 15.5 fetuses from WT mice and KO mice were used toassess expression of Tem1 mRNA during development. TotalRNA was extracted by TRIzol (Invitrogen). Total RNA fromeach sample was denaturated at 68°C for 5 min, reverse tran-scribed into cDNA (Superscript First, Invitrogen), and thenamplified by PCR. The forward and reverse primers for Tem1were 5�- TATCCAGACCTGCCTTTTGG-3� and 5�-GGTATC-CCCAGGATCAAGGT-3�, and for GAPDH, were 5�-CG-GAGTCAACGGATTTGGTCGTAT-3� and 5�-AGCCT-TCTCCATGGTGGTGAAGAC-3�, respectively. PCR productswere evaluated by agarose gel electrophoresis and at least threeRNA preparations from different animals were used to confirmall results. A similar RT-PCR assay was used to evaluate Tem1expression over time in LLC tumors transplanted to the colonsof WT and KO mice.

Body Growth and Wound Healing. To determine the growth rate,mice were weighed twice a week from 5 days to 6 weeks of agein animals of both sexes. To assess wound healing, 7-week-oldfemale null and WT female mice were anesthetized and the leftf lank was shaved and scrubbed. A wound was created in the flank

with a 6.0-mm diameter uni-punch disposable biopsy instrument(Premier, Plymouth Meeting, PA). Wound diameters were mea-sured with an electronic digital caliper (Fisher Scientific) everyother day for 14 days in three mice of each genotype. At 2, 6, 10,and 14 days, mice were euthanized and tissue collected and fixedin 10% buffered formalin for histological assessment and bloodvessel counting.

s.c. Tumor Growth. Suspensions of 5 � 105 LLC cells or S180sarcoma cells in 0.3 ml PBS were injected s.c. into the left f lanksof 6-week-old female Tem1 null and WT mice. The LLC andS180 cells have previously been shown to be ‘‘universal’’ donorcells for various mouse strains by virtue of their low levels ofMHC expression (30, 31). After 8, 11, 13, 15, and 18 days, tumordiameters were measured with an electronic digital caliper.

Tumor Fragment Implantation. Tumor fragments were preparedvia s.c. injection of 1 � 105 LLC cells in the flank of 1-month-oldWT mice. After 20 days, the mice were euthanized and thetumors were removed and placed in ice-cold PBS. Six- to7-week-old Tem1 null and WT B6�129 female mice were used asrecipients. The mice were anesthetized with isoflurane, and theabdomen was scrubbed for surgery. A small incision was madeparallel to the linea alba, and the cecum was exteriorized. Asingle fragment of tumor of �1 mm3 in size was implanted on thetop of the serosal surface. A surgical suture was used to attachthe tumor to the serosal wall without penetrating the inner layersof the intestine. The organ was returned to the abdominal cavity,and the abdominal wall was closed. For transplantation tohepatic sites, an incision was made in the ventral left upperabdominal quadrant. The left lobe of the liver was carefullyexposed, and a blunt needle was used to puncture the livercapsule. A single fragment of tumor of �0.5 mm3 in size was thenplaced in the hole made by the needle. After hemostasis, theabdomen and skin were closed. Both intestinal and intrahepatictransplantation surgeries were repeated twice by using six ani-mals each time for each group. Complete necropsies and histo-logical examination were performed after 21 days in mice witheither type of transplantation. Abdominal tumor volumes weremeasured by using the formula (��6) � length � width � height.Along with the tumors, the colon, liver, lung, and other organswere placed in buffered 10% formalin, processed, sectioned, andeither stained for hematoxylin�eosin or used for in situ hybrid-ization or immunohistochemistry.

Histology, Immunohistochemistry, Vessel Quantification, and in SituHybridization. To quantify vascularization, 5-�m-thick sections oftransplanted tumor or tissue from the experimental skin woundswere deparaffinized in xylene and rehydrated through gradedalcohols. Five mice were used for each group. Endogenoushydrogen peroxide was quenched with peroxidase for 10 min.Slides were then incubated with blocking solution [10% bovineserum (Vector Laboratories) in PBS] for 60 min at roomtemperature to saturate nonspecific protein-binding sites. Theslides were incubated with primary antibody [goat anti-PECAM-1 (sc-1506; Santa Cruz Biotechnology) at 1:50; rabbitanti-human von Willebrand factor (vWF; AB7356; Chemicon) at10 �g�ml in PBS; or mouse anti-smooth muscle actin (A2547;Sigma)] for 1 h at room temperature. After three washes with0.1% Tween 20 in PBS, sections were incubated for 30 min withappropriate secondary antibody diluted in blocking solution.Slides were developed with ABC reagent and diaminobenzidine.Photographs of histological sections were taken with a digitalcamera (Nikon Dx1200). Number and diameter of all bloodvessels in the tumor or wound bed area were determined in 10high power fields (�400). In situ hybridization was performed onparaffin sections by using a modification of a protocol describedin ref. 5. A detailed protocol is available upon request.

Nanda et al. PNAS � February 28, 2006 � vol. 103 � no. 9 � 3355

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020

Statistics. Data are presented as the mean and SD [GraphPad(San Diego) PRISM software]. The numbers of mice per groupare noted in the figure legends, and all experiments wererepeated at least twice with similar results. The Mann–Whitney Test was used to compare body weights, and P values

were determined by using the Student two-tailed t test unlessotherwise indicated.

This work was supported by National Institutes of Health GrantsCA57345, CA62924, GM07184, GM07309, and RR00171.

1. Iruela-Arispe, M. L. & Dvorak, H. F. (1997) Thromb. Haemostasis 78, 672–677.2. Kerbel, R. S. (2000) Carcinogenesis 21, 505–515.3. Folkman, J. & Kalluri, R. (2004) Nature 427, 787.4. Carmeliet, P. & Jain, R. K. (2000) Nature 407, 249–257.5. St. Croix, B., Rago, C., Velculescu, V., Traverso, G., Romans, K. E., Mont-

gomery, E., Lal, A., Riggins, G. J., Lengauer, C., Vogelstein, B. & Kinzler,K. W. (2000) Science 289, 1197–1202.

6. Carson-Walter, E. B., Watkins, D. N., Nanda, A., Vogelstein, B., Kinzler, K. W.& St. Croix, B. (2001) Cancer Res. 61, 6649–6655.

7. Davies, G., Cunnick, G. H., Mansel, R. E., Mason, M. D. & Jiang, W. G. (2004)Clin. Exp. Metastasis 21, 31–37.

8. Rettig, W. J., Garin-Chesa, P., Healey, J. H., Su, S. L., Jaffe, E. A. & Old, L. J.(1992) Proc. Natl. Acad. Sci. USA 89, 10832–10836.

9. Christian, S., Ahorn, H., Koehler, A., Eisenhaber, F., Rodi, H. P., Garin-Chesa,P., Park, J. E., Rettig, W. J. & Lenter, M. C. (2001) J. Biol. Chem. 276,7408–7414.

10. Marty, C., Langer-Machova, Z., Sigrist, S., Schott, H., Schwendener, R. A. &Ballmer-Hofer, K. (June 10, 2005) Cancer Lett., 10.1016�j.canlet.2005.04.029.

11. Nanda, A. & St. Croix, B. (2004) Curr. Opin. Oncol. 16, 44–49.12. Dvorak, H. F. (1986) N. Engl. J. Med. 315, 1650–1659.13. LeCouter, J., Kowalski, J., Foster, J., Hass, P., Zhang, Z., Dillard-Telm, L.,

Frantz, G., Rangell, L., DeGuzman, L., Keller, G. A., et al. (2001) Nature 412,877–884.

14. LeCouter, J., Lin, R., Frantz, G., Zhang, Z., Hillan, K. & Ferrara, N. (2003)Endocrinology 144, 2606–2616.

15. Cleaver, O. & Melton, D. A. (2003) Nat. Med. 9, 661–668.16. Oh, P., Li, Y., Yu, J., Durr, E., Krasinska, K. M., Carver, L. A., Testa, J. E. &

Schnitzer, J. E. (2004) Nature 429, 629–635.

17. Ruoslahti, E., Duza, T. & Zhang, L. (2005) Curr. Pharm. Des. 11, 3655–3660.

18. Invernici, G., Ponti, D., Corsini, E., Cristini, S., Frigerio, S., Colombo, A.,Parati, E. & Alessandri, G. (2005) Exp. Cell Res. 308, 273–282.

19. Sikder, H., Huso, D. L., Zhang, H., Wang, B., Ryu, B., Hwang, S. T., Powell,J. D. & Alani, R. M. (2003) Cancer Cell 4, 291–299.

20. Perk, J., Iavarone, A. & Benezra, R. (2005) Nat. Rev. Cancer 5, 603–614.21. Tlsty, T. D. & Hein, P. W. (2001) Curr. Opin. Genet. Dev. 11, 54–59.22. Bissell, M. J., Kenny, P. A. & Radisky, D. C. (2005) Cold Spring Harbor Symp.

Quant. Biol. 70, 1–14.23. Bhowmick, N. A., Neilson, E. G. & Moses, H. L. (2004) Nature 432, 332–337.24. Hu, M., Yao, J., Cai, L., Bachman, K. E., van den Brule, F., Velculescu, V. &

Polyak, K. (2005) Nat. Genet. 37, 899–905.25. Shekhar, M. P., Pauley, R. & Heppner, G. (2003) Breast Cancer Res. 5, 130–135.26. Debnath, J. & Brugge, J. S. (2005) Nat. Rev. Cancer 5, 675–688.27. Allinen, M., Beroukhim, R., Cai, L., Brennan, C., Lahti-Domenici, J., Huang,

H., Porter, D., Hu, M., Chin, L., Richardson, A., et al. (2004) Cancer Cell 6,17–32.

28. Orimo, A., Gupta, P. B., Sgroi, D. C., Arenzana-Seisdedos, F., Delaunay, T.,Naeem, R., Carey, V. J., Richardson, A. L. & Weinberg, R. A. (2005) Cell 121,335–348.

29. MacFadyen, J. R., Haworth, O., Roberston, D., Hardie, D., Webster, M. T.,Morris, H. R., Panico, M., Sutton-Smith, M., Dell, A., van der Geer, P., et al.(2005) FEBS Lett. 579, 2569–2575.

30. Alfaro, G., Lomeli, C., Ocadiz, R., Ortega, V., Barrera, R., Ramirez, M. &Nava, G. (1992) Vet. Immunol. Immunopathol. 30, 385–398.

31. Plaksin, D., Gelber, C. & Eisenbach, L. (1992) Int. J. Cancer 52, 771–777.

3356 � www.pnas.org�cgi�doi�10.1073�pnas.0511306103 Nanda et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

3, 2

020