The molecular signature that defines tumor microvasculature will likely provide clues as to how vascular-dependent tumor proliferation is regulated. Using purified endothelial cells, we generated a database of gene expression changes accompanying vascular proliferation in invasive breast cancer. In contrast to normal mammary vasculature, invasive breast cancer vasculature expresses extracellular matrix and surface proteins characteristic of proliferating and migrating endothelial cells. We define and validate the up-regulated expression of VE-cadherin and osteonectin in breast tumor vasculature. In contrast to other tumor types, invasive breast cancer vasculature induced a high expression level of specific transcription factors, including SNAIL1 and HEYL, that may drive gene expression changes necessary for breast tumor neovascularization. We demonstrate the expression of HEYL in tumor endothelial cells and additionally establish the ability of HEYL to both induce proliferation and attenuate programmed cell death of primary endothelial cells in vitro. We also establish that an additional intracellular protein and previously defined metastasis-associated gene, PRL3, appears to be expressed predominately in the vasculature of invasive breast cancers and is able to enhance the migration of endothelial cells in vitro. Together, our results provide unique insights into vascular regulation in breast tumors and suggest specific roles for genes in driving tumor angiogenesis.

A critical role has been established linking unchecked microvascular proliferation and tumor growth. One of the most studied cancers in relation to neovascularization, breast cancer has provided a paradigm for the role of angiogenesis in cancer (1). Defining the gene expression alterations associated with tumor-driven neovascularization may yield therapeutically important targets for cancer intervention. In recent years there have been technological advances allowing large-scale expression profiling of cancer. Such profiling, including microarray (2) and serial analysis of gene expression (3, 4), along with the recent sequencing of the human genome, have advanced our current understanding of the molecular pathways involved in cancer progression. Such studies would be helpful in breast cancer, because there is a lack of molecular markers that can allow for an accurate prediction of response to specific therapies or more precisely determine whether a tumor is likely to metastasize to distant organs (5, 6). Heterogeneity in primary breast tumors, containing widely varying quantity and makeup of surrounding stroma, have made it difficult to provide common gene expression profiles that could more precisely predict the invasive capacity of a tumor and determine whether specific treatments will be effective therapies.

Bulk tissue expression profiling may mask differential expression in specific cell types of the tumor, as recent findings reveal genetic alterations that occur in cells surrounding epithelial tumors (reviewed in ref. 7). Moreover, using whole tumors may also mask the profiles of metastatic epithelial cells, because only a small proportion may have an underlying metastatic potential (8). It is becoming evident that stromal cells and the extracellular matrix interact with tumor epithelium to influence cancer progression. Such an influence is evidenced by the fact that tumor cells grow and metastasize best at their orthotopic site (compared with ectopic sites), associated with marked differences in angiogenesis (reviewed in ref. 9). Because the stromal microenvironment is important (10, 11, 12), it is necessary to study the molecular consequences associated with the cross-talk between cell types to gain a more comprehensive understanding of tumor progression. Many genes have been found to be aberrantly expressed in tumor epithelium and more recently in the surrounding stroma (13), yet there is little information on the gene expression alterations that occur in the breast vascular endothelium that may ultimately promote angiogenesis and provide a route for tumor cell dissemination into the circulation.

Enhanced angiogenesis is associated with an increased risk of metastasis and poor prognosis in breast cancer (14, 15). Neo-angiogenesis is also required at the metastatic site allowing micrometastases to grow into macrometastatic lesions. Therefore, angiogenesis within metastases is a very desirable therapeutic target considering the mortality associated with distant metastasis in breast cancer. Profiling expression changes that occur in the vasculature of breast cancer will provide insight into the mechanisms underlying tumor vascular growth and also reveal attractive targets for antiangiogenic therapies.

Serial analysis of gene expression technology is a powerful technology that has been used for expression profiling of both specific cell types (16, 17) and bulk tumors, including primary breast tumors (18). Serial analysis of gene expression is an open gene expression platform providing analysis of the entire transcriptome with a quantitative, digital output. To date, expression profiling from cancer-associated, pure vascular-specific cells has been limited to a serial analysis of gene expression application on a single normal and tumor endothelial cell preparation from colon (16) and several normal brain and malignant brain tissues (19). These studies demonstrated the ability to define both tumor-specific endothelial genes and normal endothelial genes. It is noteworthy that the genes discovered to be growth or tumor-specific could be generally classified as extracellular matrix components or surface proteins likely to play a role in adhesion or cell-cell interaction. Transcription factors and other classes of upstream regulators were generally lacking. This finding suggested that the molecular events driving tumor angiogenesis were either solely dependent on extracellular events or that tumor-specific transcription factors were at too low a level to be evaluated with the expression platform used. Moreover, the extent to which circulating endothelial cells and vasculogenesis contribute to the overall expression changes is unclear. Here, we report our findings on serial analysis of gene expression analysis of purified endothelial cells from freshly resected specimens of two invasive breast cancers and one normal reduction mammoplasty. The gene expression profiles derived in our current study define unique profiles for vascular gene expression in breast tumorigenesis.

Immunopurification of Specific Cell Populations.

Fresh resected tissue (normal mammoplasty or primary breast tumor tissue) was obtained from the Johns Hopkins Hospital Surgical Pathology Division with approval from the Institutional Review Board and processed immediately. The surgically resected samples were kept at room temperature for no more than 30 minutes after resection. Tissue was collected from surgical pathology on ice. Digestion of the minced tissue was carried out at 37°C up to a maximum of 90 minutes, depending on the speed of tissue disruption, which varied with each tissue. Tissue was digested with collagenase A at 37°C to yield a single cell suspension. This was followed by a number of negative selections including red blood cells lysis (NH4Cl) and removal of penetrating immune cells including monocytes, lymphocytes, and macrophages (anti CD-14 and CD-45 DynaBeads, Dynal Biotech, Brown Deer, WI). Epithelial and endothelial cells were then positively selected using the Dynal magnetic bead-linked antibody method essentially as described (16, 20). Epithelial cells were removed before selection with P1H12 Dynal immunobeads. Immunostaining was then performed using von Willebrand factor and 4′, 6-diamidino-2-phenylindole nuclear staining to confirm the purification of endothelial cells. Under these conditions the secondary antibody bead binding to P1H12 surrounding the cells also allowed for additional confirmation of cell identity. Immunostaining revealed that the immunopurification technique yielded endothelial cell purity of >95% (data not shown).

Serial Analysis of Gene Expression Analysis.

RNA was extracted from ∼20,000 purified cells. LongSAGE analysis (21) was performed on RNA from the endothelium of 2 breast tumor samples and 1 normal mammoplasty tissue, yielding ∼50,000 tags (Table 1). All of the tag frequency calculations are normalized to exactly 50,000 tags. For comparative analysis to the colon standard serial analysis of gene expression tag libraries, longSAGE tag counts were aggregated based on common 10 base sequence tags. Genes that corresponded to the tags were then determined, and the fold difference between tags from normal breast endothelium and breast tumor endothelium were calculated. To provide for a conservative estimate of tumor-induced and normal-induced genes, a ratio was calculated using the minimum tumor serial analysis of gene expression tag number (for tumor-induced to normal-induced ratios) or the maximal tumor serial analysis of gene expression tag number (for normal-induced to tumor-induced ratios). Data derived from normal brain endothelial cells and glioma endothelial cells were used as comparators to define breast-specific markers (19).

In situ Hybridization/Immunohistochemistry.

For in situ hybridization, a mixture of riboprobes for each gene was generated yielding products of 450 to 550 bases as described (16). Briefly, riboprobes were generated from an amplification of the DNA sequence of interest, involving incorporation of a T7 primer in an initial PCR reaction followed by use of T7 polymerase for in vitro transcription. The in vitro transcription also included labeling of transcripts with digoxigenin (Roche Diagnostics, Indianapolis, IN). Paraffin-embedded sections of normal breast tissue, primary breast cancers of various stages, and tissue arrays were obtained from the Johns Hopkins Hospital Surgical Pathology Division and used for in situ hybridization and immunohistochemistry. For in situ hybridization, individual riboprobes were initially used to determine those that gave the best signal with minimal background DNA binding, and these were pooled into a mixture to be used for all of the subsequent hybridizations for each gene studied. Protocols were as described previously, including deparaffinization and fixation of tissues and pretreatment for access to target nucleic acid sequence, followed by addition of digoxigenin-labeled riboprobe overnight (16). Addition of horseradish peroxidase-anti-digoxigenin (Dako) and two rounds of tyramide amplification involved the addition of biotinyl tyramide (Dako) that binds in the vicinity of horseradish peroxidase. Signal was visualized using the 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium alkaline phosphatase system (involving the addition of a Dako AP-antibiotin). Some were counterstained with nuclear fast red.

In vitro Functional Assays.

Adenovirus expressing HEYL and PRL3 were constructed and determined to viably infect human microvascular endothelial cells. Both constructs expressed the gene of interest to a level at least 15-fold higher than empty vector control. The effect of HEYL overexpression on human microvascular endothelial cell proliferation was assessed by infecting human microvascular endothelial cells for 18 hours at a multiplicity of infection of 200. Cell number was determined at days 0, 2, and 3. The migration effects of adenoviral-expressing PRL3 were assessed by infecting 50,000 human microvascular endothelial cells for 48 hours with a multiplicity of infection of 300. Cells were plated in the top well containing basal media and migrated toward the bottom chamber containing 5% fetal bovine serum as attractant. Migrated cells were read after 24 hours using cell titer glow (Promega, Madison, WI). Terminal deoxynucleotidyl transferase-mediated nick end labeling assays were used to assess the apoptotic potential of human umbilical vein endothelial cells infected with HEYL or control adenovirus. Twenty-five thousand human umbilical vein endothelial cells were transfected with AD-GFP or AD-HEYL and cultured for 3 days in serum-free medium. Apoptosis was detected using In Situ Cell Death terminal deoxynucleotidyl transferase-mediated nick end labeling detection kit, TMR red (Roche).

Quantitative PCR.

ABI Prism 7900 detection system was used according to the manufacturer’s specifications. Data output is shown as log2 values relative to the mean of the 10 normal breast tissue RNAs used. Data were normalized to either 18 s rRNA or to the median of a set of preselected ubiquitous endothelial cell marker genes. These genes were defined as those having small variance in previously derived pure endothelial cell serial analysis of gene expression data. The genes included in the set are: Claudin 5, ROBO4, TIE2, Hevin, Sox18, and von Willebrand factor. Gene-specific PCRs were performed in duplicate on two separate occasions with similar results. Probe sequences used for gene-specific amplification were as follows: KDR: TCCCAGGCTGCACCCATGGC, TEM7: CTTTGCCTATAAAGAGATCCCTATGTCTGTCCCGG, PRL3: TTTGACGATGGGGCGCCCCCGC, HEYL: ACGGCGTCGAGACCGCATCA, Neuritin: CCGCAGGGCCTGGACGACAA, and TEM1: CGCTGGCTGTCGACGGCTACCTGTGCCAGTT.

PRL3 Expression in Serial Analysis of Gene Expression Libraries.

PRL3 expression (serial analysis of gene expression tag = taggtcagga) was determined for breast-specific cells or tissue in both public serial analysis of gene expression database resources (CGAP) and our own serial analysis of gene expression database. Library data from our database included B1 and B2; normal bronchial epithelial cells; and tumor cell lines B5, B6, B7, B8, B9, and B10. Additional tumor lines included 21-PT, 21-MT, MDA-467, SKBR3, BT-474, and MDA-231; BEC1: BECT1, BEC2: BECT2 and BEC3; BECN1 (this study); BEC5; bone metastasis epithelial cells, BEC6; and PCR amplified normal breast epithelial cells. All of the other library information can be found on the internet.3

Data Generation and Confirmation of Cell Purity.

We performed longSAGE analysis on endothelial cells isolated from 1 normal breast sample and 2 invasive ductal carcinoma samples. A summary of the samples used, the approximate number of endothelial cells purified, and the serial analysis of gene expression tags accumulated are summarized in Table 1 A. Each library was sequenced to just over 50,000 tags. All of the additional tag calculations are based on normalization of tag counts to exactly 50,000. Epithelial cell isolates were also collected but used solely to define relative cell specificity by PCR (see below).

The premise for our study is that pure endothelial cell populations can be derived from surgical samples of both normal and malignant breast tissue. The protocol used for the isolation of pure endothelial cells uses a combination of negative and positive immunoselections. Immunopurification of specific cell types requires evidence that the final cell preparations are essentially pure. We confirmed the purity of our endothelial cell populations by analyzing the serial analysis of gene expression tag frequency for genes known to be selectively expressed in specific subpopulations of cells. Genes specific for endothelial, epithelial, and hematopoietic cells were analyzed in the serial analysis of gene expression libraries for the 3 samples (Table 1 B). Genes with preferential expression in vascular endothelium including von Willebrand factor, CD31, CD34, and hevin were moderately and uniformly expressed in all 3 of the samples. Genes specific to epithelium (keratin 8 and 6A) or hematopoietic cells (CD14 and CD45) had negligible expression within the constructed serial analysis of gene expression libraries with the total absence of corresponding tags in most cases. Furthermore, von Willebrand factor and CK18 reverse transcription-PCR analysis performed on cDNA generated from the endothelial cell populations demonstrated little contamination of epithelial cells within the endothelial cell populations (data not shown). This confirms that the samples used were essentially pure endothelial cells. However, it is reasonable to assume that a small percentage of cells are derived from endothelial cell-associated pericytes.

Genes Expressed in Invasive Breast Cancer Endothelium.

Genes that are overexpressed in cancer offer attractive therapeutic targets, especially if expressed at relatively low levels in normal tissues. Table 2,A represents genes that were expressed at least 6-fold higher (conservatively calculated by taking the tag ratio of the lower-expressing tumor sample over the tag frequency in the normal sample) in breast tumor endothelium when compared with the endothelial cells isolated from normal mammoplasty tissue. Table 2 A also shows the tumor induction ratio of the genes in a previously examined colon tumor study (16).

As expected, genes encoding proteins involved in extracellular matrix function (collagens, MMP9, ADAMTS4, and TIMP1) were transcriptionally active in tumor vasculature compared with normal breast vasculature (Table 2,A). Most of these extracellular matrix-encoded genes were similarly up-regulated in colon tumor vasculature (Table 2,A). Unexpectedly, we identified the transcription factors SNAIL1 and HEYL as being induced in breast tumor vasculature. Although HEYL was also observed to be transcriptionally up-regulated in colon tumor vasculature (tag: ggttgttgcg), the induction relative to normal colon vasculature was moderate. In contrast, both SNAIL1 and HEYL were induced at least 10-fold in breast tumor vasculature relative to normal breast vasculature. Five genes could be recognized as being induced ≥10-fold in breast tumor endothelial cells and induced at least 5-fold higher in breast than that observed in both colon and brain tumor endothelial cells (see serial analysis of gene expression analysis in Materials and Methods; Table 2 B). Two of these five genes included SNAIL1 and HEYL.

Also unexpected was the finding that PRL3 was highly induced in breast tumor vasculature, because this gene was shown previously to be expressed primarily in epithelial tumor cells (17). It is noteworthy that although both HEYL and PRL3 serial analysis of gene expression tags demonstrate a differential induction in tumor endothelial cells over normal endothelial cells, the specific tag within the genes that is induced differs between colon and breast endothelial cell libraries (Table 2 C). For both HEYL and PRL3, the most differential tag frequencies observed for colon endothelial cells derives from a 3′ extended form of the transcripts. These extended transcripts are based on gene prediction algorithms incorporating all of the available expressed sequence tag data. Thus, within the colon endothelial cell data, the recognized 3′ ultimate tags for HEYL and PRL3 show limited or no tumor induction, respectively. It remains unclear why there is this differential transcript detection between colon and breast endothelial cells.

Finally, the robust induction of the cell-cell interaction protein VE-cadherin was unique to breast tumor vasculature, with no induction in colon tumor vasculature (Table 2 A) or brain tumor vasculature (data not shown).

Decreased Gene Expression in Invasive Breast Cancer.

The concerted reduction or absence of expression of genes in tumor vasculature as compared with normal vasculature may reveal genes that function to suppress tumor and/or vascular growth. With this in mind, we sorted the data to reveal genes that showed little or no expression in tumor vasculature compared with normal breast vasculature. A striking down-regulation of numerous genes was observed (Table 3). Genes expressing secreted proteins may lend themselves to direct therapeutic intervention and are likely to play a role in extracellular matrix stabilization or cell adhesion. Particularly noteworthy is the observed down-regulation of both lysyl oxidase-like 1 and lysyl oxidase. Members of the lysyl oxidase gene family have been implicated in the regulation of tumor growth, albeit with highly contrasting results from different studies (22). The involvement of lysyl oxidase genes in extracellular matrix formation and repair may have implication for regulating the plasticity of tumor vasculature (23).

We were interested in learning which tumor-repressed, vascular genes were conserved in their reduced gene expression across different tumor types. The most highly conserved gene down-regulated in tumor vasculature is the neuritin gene (NRN1 and CPG15), exhibiting a 9-fold reduction in breast tumors (Table 3) and a 6- and 4-fold reduction in colon and glioma tumors, respectively (data not shown). Neuritin encodes a glycosylphosphatidylinositol-anchored protein that has been demonstrated to affect neurite growth in vitro(24).

Confirmation of Gene Expression Alteration in Normal, Ductal Carcinoma In situ, and Invasive Ductal Carcinoma Tissue.

The expression patterns for selected genes observed in the serial analysis of gene expression library were confirmed by reverse transcription-PCR (Fig. 1,A). In addition to the purification of endothelial cells from normal and tumor breast tissue, we also immunopurified the adjacent epithelial cells. RNA derived from both endothelial cells and epithelial cells were tested for expression of SNAIL1, HEYL, PRL3, and VE-cadherin. As expected, SNAIL1 was expressed in both the endothelium and the epithelium of invasive cancer. The expression of SNAIL1 in tumor epithelium was reported previously, but its expression in breast cancer endothelium is a novel finding (25, 26). HEYL and PRL3 expression was negligible in both cell types in normal mammoplasty tissue. In invasive breast cancer, the expression of HEYL and PRL3 was observed predominantly in the endothelial cell population, with relatively minimal expression in tumor epithelial cells. As noted previously, VE-cadherin was seen to be expressed at low levels in the endothelium of normal breast tissue. This was also the case using reverse transcription-PCR where expression was detectable in normal endothelium but was present at significantly higher levels in the tumor endothelium (Fig. 1 A).

We additionally validated the observed overexpression of selected genes by performing quantitative real-time PCR on a panel of normal breast tissue (n = 10) and breast tumor tissue RNAs (n = 20; Fig. 1,B). The tumor vasculature-induced genes PRL3 and TEM7 showed elevated expression (>2 fold) in 63% and 37% of the tumors studied, respectively, when the data were normalized to vascular content within the tissue samples (see Materials and Methods). The prevalence of PRL3 overexpression in breast tumor tissues was reduced to 40% of the tumors when expression was normalized to 18 s rRNA, suggesting that there is contribution to PRL3 expression from nonvascular sources (data not shown). Of the genes tested, the HEYL gene showed the most dramatic induction in breast tumor vasculature, demonstrating an 83% tumor induction (Fig. 1,B). We also confirmed a pattern of tumor down-regulation of neuritin in breast cancer (46% of tumors <2-fold down-regulated), consistent with our primary expression data (Fig. 1 B). Neither TEM1 nor KDR demonstrated a robust induction in breast tumor samples relative to normal breast tissue controls.

As an extensive database of breast serial analysis of gene expression data exists both from our work and within the pubic domain, we determined the expression of PRL3 across all of the breast serial analysis of gene expression studies to date. Expression of PRL3 is clearly elevated in tumor endothelial cells relative to normal endothelial cells, bulk tissue samples, and breast tumor cell lines (Fig. 1,C), lending additional support to a preferential expression of PRL3 in breast tumor vasculature relative to epithelial cells. The PRL3 tag was observed at a frequency of 11 tags to 50,000 in one public normal breast endothelial cell library (Br-N-Endothelial). However, these data were excluded from Fig. 1 C, because it was observed that this library showed significant expression of both epithelial and hematopoietic-specific genes, in obvious contrast to other fresh tissue-derived endothelial cell libraries.

In situ hybridization was used to localize the expression of gene transcripts in tumor vasculature. For all of the in situ hybridization experiments, KDR (VEGFR2) was used as a control to localize the endothelial cells within the vasculature. This allowed confirmation of vascular-specific staining with the probes of interest. A HEYL mixture was used to probe for the expression of HEYL in normal tissue and tissue derived from a patient sample of invasive ductal carcinoma (Fig. 2). When comparing staining in the normal tissue with that of H&E and KDR, it is evident that the HEYL probe does not stain the endothelium in normal breast tissue vasculature. In contrast, the invasive breast cancer section demonstrates strong labeling by the HEYL-specific probes paralleling the staining seen with the endothelial-specific marker KDR. Additional interrogation of a human breast tissue array demonstrates conclusive labeling of HEYL within invasive carcinoma (data not shown).

A similar trend was also observed when in situ hybridization was performed using a mixture of PRL3 riboprobes. Similar to HEYL, there was an absence of PRL3 RNA expression in the normal tissues analyzed, whereas clear, endothelial-specific expression was observed in the blood vessels of invasive carcinoma (Fig. 2). It is important to note that expression appeared to be limited to vascular cells with a lack of binding to the epithelium or other stromal components. There was also a lack of binding to endothelium surrounding areas of ductal carcinoma in situ, indicating that these two genes, HEYL and PRL3, are only expressed in invasive carcinoma (data not shown).

In situ hybridization analysis for VE-cadherin showed a light staining in normal tissues that was enhanced in invasive carcinoma (Fig. 3,A). Because a comparison between staining of different tissues is not quantitative, this was additionally analyzed by using a single patient sample that contained areas of normal tissue, ductal carcinoma in situ, and invasive ductal carcinoma on the same section. Fig. 3 B illustrates VE-cadherin, staining as slightly positive in the normal area, becoming stronger in the vasculature surrounding the ductal carcinoma in situ and strongest in areas of invasive ductal carcinoma, mimicking the exact pattern of both KDR and CD31 binding. Therefore, VE-cadherin expression increases as the tumor becomes more invasive. The rim of microvessels around ductal carcinoma in situ is characteristic of high-grade ductal carcinoma in situ.

Immunohistochemical Analysis of Osteonectin Expression.

Because there was a high-affinity, paraffin-reactive antibody to osteonectin available, immunohistochemistry was used to confirm the enhanced expression of this gene in breast tumor vasculature. Fig. 4 shows the results of two different cases containing either normal and invasive breast cancer or ductal carcinoma in situ and invasive ductal carcinoma on the same paraffin-embedded section of two patient samples. As observed with PRL3 and HEYL, there was an absence of expression (in this case protein expression) in normal tissue and ductal carcinoma in situ, whereas strong staining was observed in the endothelium and stromal fibroblasts of invasive breast cancer. The staining of the vascular endothelium was confirmed by comparing staining patterns with that of CD31 antibody (data not shown). Our results suggest that osteonectin protein is expressed to progressively higher levels as the tumor progresses from ductal carcinoma in situ to invasive carcinoma.

This finding was pursued in greater detail by examining tissue microarrays containing breast carcinoma, hematogenous metastases, and normal breast tissue. Immunohistochemistry for osteonectin was performed, and all of the spots were scored as negative, weakly positive, moderate positive, or strong positive. Among 11 cases with benign breast lobules, the endothelium and stromal fibroblasts were completely negative in 2 cases, focally weakly positive in 5 cases, weakly positive in 2 cases, and focally moderately positive in 2 cases. In contrast, all 20 of 20 primary carcinomas are positive for osteonectin. In the 20 primary carcinomas, the endothelium and stromal fibroblasts of 9 were moderately positive, 9 were moderately to strongly positive, and 2 were strongly positive. Only in 3 of the 20 primary carcinomas were the epithelial cells themselves labeled, 1 weakly and 2 moderately. Among hematogenous metastases, staining for osteonectin was more frequent and stronger. The stroma of 11 were moderately positive, 5 were moderately to strongly positive, 2 were focally strongly positive, and 8 were strongly positive. Only in 5 of 26 metastatic carcinomas were the epithelial cells themselves labeled for osteonectin, all of which did so weakly. Thus, osteonectin expression is found primarily in the endothelial cells and stromal fibroblasts and rarely in the carcinoma cells of both invasive tumors and their hematogenous metastases. Furthermore, the expression of osteonectin is higher in hematogenous metastasis compared with invasive breast cancer.

Neuritin Expression Is Decreased in Invasive Breast Cancer.

We next wanted to validate the expression of the gene found to represent the most conserved, tumor-repressed vascular gene, NRN1. Using a riboprobe specific for the NRN1 RNA transcript, binding was evident in the vasculature of normal mammoplasty tissue, as verified by the binding patterns of KDR (Fig. 5). The vascular-specific binding of NRN1 probe was abolished in invasive breast cancer tissue (Fig. 5).

Functional Assessment of Breast Tumor-Induced Vascular Genes.

We next addressed the functional implications of overexpressing specific breast tumor endothelial cell genes. Adenovirus expressing either HEYL or PRL3 was used to infect human microvascular endothelial cells. The transfected cells were assessed for their ability to affect migration, proliferation, and tube formation relative to cells infected with empty vector or green fluorescent protein expressing adenoviral control. The overexpression of HEYL showed a clear and reproducible enhancement of human microvascular endothelial cells proliferation (Fig. 6,A). The functional implications for HEYL overexpression were specific to cellular proliferation, as HEYL overexpression had no effect on human microvascular endothelial cell tube formation or migration (data not shown). Similar results were observed for human umbilical vein endothelial cells (data not shown). The overexpression of HEYL also showed a protective effect on human umbilical vein endothelial cell apoptosis (Fig. 6,B). Adenovirus-expressing green fluorescent protein-infected human umbilical vein endothelial cells showed a large amount of apoptosis when cells were starved for serum. In contrast, the overexpression of HEYL resulted in healthy cells after 3 days and little apoptosis (Fig. 6 B). Untransfected human umbilical vein endothelial cells essentially all succumbed to apoptosis (data not shown).

The overexpression of PRL3 in human microvascular endothelial cells showed an effect on cellular migration and not on proliferation (Fig. 6 C). PRL3 overexpression had a modest effect on tube formation, increasing the quantity of tubes as compared with control-infected cells (data not shown). These results suggest separate and specific roles for individual breast tumor vascular genes.

To identify genes that may regulate tumor-directed angiogenic growth we have used serial analysis of gene expression to generate transcriptomes of endothelial cells from normal and malignant breast tissue. The work presented here demonstrates the utility of cell-specific purification linked to comprehensive gene expression analysis and expands on the limited quantitative gene expression studies performed thus far on pure, tissue-derived endothelial cells. Among the genes significantly induced in breast tumor endothelium were a number of transcription factors including HEYL and SNAIL1 and the tyrosine phosphatase PRL3, all of which potentially have numerous gene targets that may contribute to angiogenesis and tumor progression.

HEYL was identified recently as a basic helix-loop-helix transcription factor, a family of factors known as key regulators of embryonic development or differentiation (27). It has been shown that HEYL can be activated by constitutively active forms of Notch receptors, making Notch receptors upstream regulators of HEYL expression (28). The role of enhanced expression of HEYL in the breast tumor endothelium is an interesting one that needs to be studied at a functional and molecular level to determine the role of the induction of this gene on angiogenesis and tumorigenesis. The fact that HEYL and other members of the Notch pathway have limited expression in adulthood makes this gene a potential target for therapy due to the reduced likelihood of systemic toxicity.

There have been a number of publications recently regarding the role of the SNAIL1 family of zinc-finger transcription factors in tumorigenesis. Of particular interest is the role of SNAIL1 in epithelial-mesenchymal transition that could promote the invasive capacity of epithelium in breast cancer. The transition from epithelial to mesenchymal cells allows for enhanced migratory and invasive properties, and, hence, such a transition may contribute to tumor progression (reviewed in ref. 29). Our finding that SNAIL1 is induced in tumor endothelium, additional to its enhanced expression in breast tumor epithelium (26, 30), has not been reported previously and suggests a role for SNAIL1 in the transcriptional regulation of genes important in angiogenesis. Our study has demonstrated that SNAIL1 expression is enhanced 10-fold in breast tumor endothelium, whereas expression is absent in the endothelium of colon and brain carcinoma.

PRL3 is a proposed tyrosine phosphatase with a COOH-terminal prenylation motif that allows its association with the plasma membrane. This gene was reported recently to have a role in colorectal cancer metastasis (17) where it was found to be expressed at high levels in tumor cells of metastatic tumors, with a significantly lower level expression in the vasculature. Our results suggest a shift in the cell-specific expression of PRL3 in breast cancer versus colon cancer. We observed a 6-fold induction in breast tumor endothelium by serial analysis of gene expression yet an apparent absence or low expression in surrounding epithelium by in situ hybridization. There is limited information on the role of this gene and its molecular targets, yet as a phosphatase it may have important roles in cell signaling. It has been shown recently that PRL3 promotes cell motility, invasion, and metastasis of Chinese hamster ovary cells (31).

A gene found to be expressed at low levels in normal endothelium but at higher levels in breast tumor endothelium was VE-cadherin. This adherence molecule is localized to the interendothelial cell junction and has an important role in maintaining endothelial permeability. Previous studies support a role of VE-cadherin in angiogenesis and tumor growth when there is active vessel growth (32). Antibodies directed toward VE-cadherin inhibit angiogenesis and modulate endothelial permeability (33, 34). Moreover, dominant-negative mutants of VE-cadherin inhibit endothelial growth (35). Our results demonstrating an enhanced expression of VE-cadherin as disease progresses is consistent with previous findings suggesting a role for VE-cadherin in angiogenesis as opposed to vasculogenesis (reviewed in ref. 36).

For a tumor cell to invade the stroma and enter the circulation, it has to cross the extracellular matrix. This process requires proteinases (such as matrix metalloproteinase proteins) or the alteration of the extracellular matrix architecture. Osteonectin, which was found to be induced at least 7-fold in breast tumor endothelium, has a role in the latter. The fact that this protein was only expressed in endothelium of invasive breast cancer and not in ductal carcinoma in situ may support a role for this gene in altering extracellular matrix properties during tumor cell invasion and/or during angiogenesis. Osteonectin is a bone-matrix protein that has been previously found induced in mammary and other cancers, including prostate (37). Tumor-promoting effects seem to be specific for prostate and breast to date, as studies in ovarian cancer have conversely shown that osteonectin expression is associated with decreased endothelial proliferation and apoptotic induction (38). In normal tissue, osteonectin modulates cell-extracellular matrix interactions during tissue remodeling; regulates extracellular secretion of extracellular matrix components including regulation of the transendothelial flux of macromolecules; and is also involved in cell differentiation, cell migration, and angiogenesis. In prostate and breast cancer cells osteonectin enhances matrix metalloproteinase activity and promotes invasion and specific metastasis to bone in vivo. Osteonectin does not, however, stimulate tumor growth or promote invasion of cells that are not metastatic to bone (39). This is an important finding, because bone metastasis is prevalent in both types of cancer and is associated with patient mortality. Therefore, osteonectin may have a role in distant metastasis, although its expression does not have any prognostic significance in studies undertaken thus far. Past studies are limited in suggesting a role for osteonectin in breast tumor endothelium. In fact, studies have focused on the role of osteonectin expression in tumor epithelium. In contrast, our immunohistochemistry experiments using single sections as well as tissue arrays show that the endothelium is the major site of expression of osteonectin in most tumors, whereas osteonectin was detectable in a very small proportion of the carcinoma cells within a small subset of tumors. It will be of interest to determine the role of osteonectin expression in the vasculature and whether such expression can predict the metastatic outcome of breast cancer.

A number of genes with decreased expression in tumor endothelium were identified. We identified the gene NRN1 as the most conserved, tumor-repressed vascular marker when we looked across colon, breast, and brain cancer vascular transciptomes. The neuritin gene was originally identified due to its induction by neural activity, being a downstream effector of activity-induced neurite outgrowth (40). In the context of neuronal regulation, neuritin serves to promote growth as a membrane-bound, GPI-anchored protein (24). It has become clear recently that several gene products seem to share roles in neurogenesis and angiogenesis (41). Studies are ongoing to assess the functional significance of NRN1 expression in regulating angiogenesis.

This work provides novel insights into the genes that are altered in human breast cancer vasculature, suggesting roles in angiogenesis, tumor growth, and invasion. Although several common patterns in gene expression were observed in breast tumor vasculature compared with colon tumor vasculature, clear differences suggest unique signatures for tissue-specific tumor vasculature. Additional work will define the roles for these genes in driving tumor angiogenesis and vasculogenesis.

Fig. 1.

Expression of selected breast tumor vascular markers. A, RNA from cells immunopurified from normal breast tissue and that from invasive breast cancer was reverse transcribed. Reverse transcription-PCR was then performed using primers specific for SNAIL1, HEYL, PRL3, and VE-cadherin. The 36B4 gene (ribosomal protein) was amplified as a control that should be expressed in all samples. B, quantitative real-time PCR for KDR, TEM7, PRL3, HEYL, Neuritin, and TEM1 in a panel of normal and invasive ductal carcinoma breast tumor samples. Expression was normalized to a set of epithelial cell-specific markers as described in Materials and Methods. The values are log2 representations relative to the mean of the normal samples. Normal breast tissue RNA (tissue 1 to 10), invasive ductal carcinoma breast tissue (tissues 11 to 40). Values are shown for average tumor to normal >2 fold (Avg. T/N) and percent tumor over normal >2 fold (% T/N). C, PRL3 expression in serial analysis of gene expression libraries. Abundance of the PRL3 serial analysis of gene expression tag was determined for in-house and public breast serial analysis of gene expression data using the tag taggtcagga. Library data from our database included B1 and B2; normal bronchial epithelial cells; and tumor cell lines B5, B6, B7, B8, B9, B10. Additional tumor lines included 21-PT, 21-MT, MDA-467, SKBR3, BT-474, and MDA-231; BEC1: BECT1, BEC2: BECT2 and BEC3; BECN1 (this study); BEC5; bone metastasis epithelial cells, BEC6; and PCR amplified normal breast epithelial cells. All other library information can be found on the internet.3

Fig. 1.

Expression of selected breast tumor vascular markers. A, RNA from cells immunopurified from normal breast tissue and that from invasive breast cancer was reverse transcribed. Reverse transcription-PCR was then performed using primers specific for SNAIL1, HEYL, PRL3, and VE-cadherin. The 36B4 gene (ribosomal protein) was amplified as a control that should be expressed in all samples. B, quantitative real-time PCR for KDR, TEM7, PRL3, HEYL, Neuritin, and TEM1 in a panel of normal and invasive ductal carcinoma breast tumor samples. Expression was normalized to a set of epithelial cell-specific markers as described in Materials and Methods. The values are log2 representations relative to the mean of the normal samples. Normal breast tissue RNA (tissue 1 to 10), invasive ductal carcinoma breast tissue (tissues 11 to 40). Values are shown for average tumor to normal >2 fold (Avg. T/N) and percent tumor over normal >2 fold (% T/N). C, PRL3 expression in serial analysis of gene expression libraries. Abundance of the PRL3 serial analysis of gene expression tag was determined for in-house and public breast serial analysis of gene expression data using the tag taggtcagga. Library data from our database included B1 and B2; normal bronchial epithelial cells; and tumor cell lines B5, B6, B7, B8, B9, B10. Additional tumor lines included 21-PT, 21-MT, MDA-467, SKBR3, BT-474, and MDA-231; BEC1: BECT1, BEC2: BECT2 and BEC3; BECN1 (this study); BEC5; bone metastasis epithelial cells, BEC6; and PCR amplified normal breast epithelial cells. All other library information can be found on the internet.3

Close modal
Fig. 2.

In situ hybridization of HEYL and PRL3. in situ hybridization of the HEYL and PRL3 genes were performed on sections of normal breast tissue (N) and a sample of invasive breast cancer (T). Serial sections were H&E stained, and in situ hybridization was also carried out using the endothelial-specific KDR riboprobe as a control. The purple staining in KDR probed sections represents the position of the endothelial cells (vasculature). ×40 magnification.

Fig. 2.

In situ hybridization of HEYL and PRL3. in situ hybridization of the HEYL and PRL3 genes were performed on sections of normal breast tissue (N) and a sample of invasive breast cancer (T). Serial sections were H&E stained, and in situ hybridization was also carried out using the endothelial-specific KDR riboprobe as a control. The purple staining in KDR probed sections represents the position of the endothelial cells (vasculature). ×40 magnification.

Close modal
Fig. 3.

In situ hybridizations for VE-cadherin expression. A, mixtures of KDR and VE-cadherin riboprobes were hybridized to a breast tissue array containing normal (N) and tumor (T) sections. Arrays were also H&E stained to determine the presence of normal or tumor epithelium and, hence, the location of tumor endothelium. B, a breast tumor sample that contained regions of normal breast (A), ductal carcinoma in situ (DCIS; B), and invasive ductal carcinoma (IDC; C) on the same section was stained with H&E, KDR, or VE-cadherin riboprobes or with CD31 antibodies. Results were visualized ×40.

Fig. 3.

In situ hybridizations for VE-cadherin expression. A, mixtures of KDR and VE-cadherin riboprobes were hybridized to a breast tissue array containing normal (N) and tumor (T) sections. Arrays were also H&E stained to determine the presence of normal or tumor epithelium and, hence, the location of tumor endothelium. B, a breast tumor sample that contained regions of normal breast (A), ductal carcinoma in situ (DCIS; B), and invasive ductal carcinoma (IDC; C) on the same section was stained with H&E, KDR, or VE-cadherin riboprobes or with CD31 antibodies. Results were visualized ×40.

Close modal
Fig. 4.

Immunohistochemistry of osteonectin. An osteonectin antibody was used to probe different samples of breast cancer, one containing regions of normal breast tissue and invasive cancer on the same section (N and T) and the other having areas of ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) on the same slide. ×40 magnification.

Fig. 4.

Immunohistochemistry of osteonectin. An osteonectin antibody was used to probe different samples of breast cancer, one containing regions of normal breast tissue and invasive cancer on the same section (N and T) and the other having areas of ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) on the same slide. ×40 magnification.

Close modal
Fig. 5.

In situ hybridization of neuritin. Serial sections from a human breast tissue array were probed with KDR as an endothelial specific control or with a mixture of riboprobes specific for the neuritin gene. Both normal samples (N) and invasive carcinoma (T) were also H&E stained to locate blood vessels.

Fig. 5.

In situ hybridization of neuritin. Serial sections from a human breast tissue array were probed with KDR as an endothelial specific control or with a mixture of riboprobes specific for the neuritin gene. Both normal samples (N) and invasive carcinoma (T) were also H&E stained to locate blood vessels.

Close modal
Fig. 6.

In vitro assays on overexpressing HEYL and PRL3 endothelial cells. A, adenovirus expressing HEYL was used to infect human microvascular endothelial cells, and the ability of the human microvascular endothelial cells to proliferate was assessed at 0, 2, and 3 days. Green fluorescent protein-expressing adenovirus (GFP) or untransfected cells (non) were used as controls. GM, endothelial growth medium. B, terminal deoxynucleotidyl transferase-mediated nick end labeling assay detecting differential apoptosis between AD-HEYL and AD-GFP infected human umbilical vein endothelial cells. Bright field, GFP, and TMR red (apoptosis) detection fields are shown. The AD-HEYL virus used in this study also expresses GFP. C, adenoviral-expressing PRL3 was used to infect human microvascular endothelial cells, and the cells were assayed for migration activity. Empty vector (EV) or untransfected human microvascular endothelial cells were used as controls. BM/BM, background migration of cells in the absence of a serum stimulus. BM/BM +5% FBS (fetal bovine serum), cells migrating toward the serum attractant; bars, ±SD.

Fig. 6.

In vitro assays on overexpressing HEYL and PRL3 endothelial cells. A, adenovirus expressing HEYL was used to infect human microvascular endothelial cells, and the ability of the human microvascular endothelial cells to proliferate was assessed at 0, 2, and 3 days. Green fluorescent protein-expressing adenovirus (GFP) or untransfected cells (non) were used as controls. GM, endothelial growth medium. B, terminal deoxynucleotidyl transferase-mediated nick end labeling assay detecting differential apoptosis between AD-HEYL and AD-GFP infected human umbilical vein endothelial cells. Bright field, GFP, and TMR red (apoptosis) detection fields are shown. The AD-HEYL virus used in this study also expresses GFP. C, adenoviral-expressing PRL3 was used to infect human microvascular endothelial cells, and the cells were assayed for migration activity. Empty vector (EV) or untransfected human microvascular endothelial cells were used as controls. BM/BM, background migration of cells in the absence of a serum stimulus. BM/BM +5% FBS (fetal bovine serum), cells migrating toward the serum attractant; bars, ±SD.

Close modal

Grant support: Johns Hopkins SPORE Grant P50 CA88843 (S. Sukumar).

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

Note: B. S. Parker’s current address: Department of Research, Peter MacCallum Cancer Centre, Melbourne, Australia. A. Bardelli’s current address: The Oncogenomics Center, Institute for Cancer Research and Treatment, University of Torino Medical School, Torino, Italy.

Requests for reprints: Stephen L. Madden, Genzyme Corporation, 5 Mountain Road, Framingham, MA 01701. Phone: 508-270-2175; E-mail: steve.madden@genzyme.com

3

Internet address: http://www.ncbi.nlm.nih.gov/SAGE/index.cgi?cmd = libsearch.

Table 1

Sample summary

A. Endothelial samples used for SAGE analysis
SampleDiseaseIdentityCell numberSAGE tags
Endo 1 Tumor BEC T1 30,000 51,000 
Endo 2 Tumor BEC T2 20,000 52,000 
Endo 3 Normal BEC N1 40,000+ 52,000 
A. Endothelial samples used for SAGE analysis
SampleDiseaseIdentityCell numberSAGE tags
Endo 1 Tumor BEC T1 30,000 51,000 
Endo 2 Tumor BEC T2 20,000 52,000 
Endo 3 Normal BEC N1 40,000+ 52,000 
B. Confirmation of cell identity
Gene descriptionUG IDSpecificityBEC T1BEC T2BEC N1HFMDA-453
von Willebrand factor Hs.110802 EC 15 26 16 
Platelet/endothelial cell adhesion molecule (CD31) Hs.78146 EC 13 25 25 
CD34 antigen Hs.367690 EC 15 26 
SPARC-like 1 (mast9, hevin) Hs.75445 EC 113 117 121 
Keratin 8 Hs.242463 Epithelial 124 
CD14 Hs.163867 Hematopoeitic 50 
CD45 Hs.444324 Hematopoeitic 
B. Confirmation of cell identity
Gene descriptionUG IDSpecificityBEC T1BEC T2BEC N1HFMDA-453
von Willebrand factor Hs.110802 EC 15 26 16 
Platelet/endothelial cell adhesion molecule (CD31) Hs.78146 EC 13 25 25 
CD34 antigen Hs.367690 EC 15 26 
SPARC-like 1 (mast9, hevin) Hs.75445 EC 113 117 121 
Keratin 8 Hs.242463 Epithelial 124 
CD14 Hs.163867 Hematopoeitic 50 
CD45 Hs.444324 Hematopoeitic 

NOTE. In B, the number of tags for genes expressed in specific cell types are shown in the tumor endothelial samples and in the normal breast endothelial cells. A hematopoietic SAGE library (HF) and breast cancer cell line SAGE library (MDA-453) are also included. All tag frequencies are per 50,000.

Abbreviations: SAGE, serial analysis of gene expression; EC, endothelial cell.

Table 2

Tumor-selective genes

A. Breast tumor-vascular genes
Long TagBEC T2BEC T1BEC N1Breast T/NColon T/NUG IDDescription
TAAATCCCCACTGGGAC 27 21 21 Hs.151738 Matrix metalloproteinase 9 (gelatinase B) 
GGTTGTTGCGGACATCC 31 20 20 Hs.23823 Hairy/enhancer-of-split HEYL 
CCACGGGATTCTCCTCC 16 15 15 Hs.377812 Homo sapiens clone FLC1492 PRO3121 mRNA 
AACACAGCCTGGGACCA 13 13 13 Hs.278625 Complement component 4A 
ACAGACTGTTAGCCAAG 20 12 12 18 Hs.125036 Tumor endothelial marker 7 precursor 
GCCATTTCTGTGGAGGG 11 10 10 Hs.48029 Snail homolog 1 (Drosophila
GAGCTGGCATAACATTG 13 Hs.394790 Collagen, type IV, α2 
CAGAGATGAATTTATAC 17 45 11 Hs.8997 Heat shock 70kDa protein 1A 
TTAGTGTCGTATTTTGA 18 17 Hs.111779 Secreted protein, acidic, (osteonectin) 
TAATCCTCAAGAAATAA 16 24 Hs.78409 Collagen, type XVIII, α 1 
TAGGTCAGGACCTTGGC 20 16 Hs.43666 Protein tyrosine phosphatase type IVA, member 3 (PRL3) 
AATCTGCGCCTGCGGGG 46 24 Hs.833 Interferon-stimulated protein, 15 kDa 
GAGAGTGTCTGCGGATA 30 22 Hs.5831 Tissue inhibitor of metalloproteinase 1 
GGTGGACACGGATCTGC Hs.111779 Secreted protein, acidic, cysteine-rich (osteonect) 
AATAAAGGCTACAGGCT Hs.179735 Ras homolog gene family, member C 
TACTAGTCCTCTAGAAA Hs.289088 Heat shock 90kDa protein 1, α 
GAGTAACGCGCCCGGCC Hs.122359 Calcium channel, voltage-dependent, α 1H 
TTTCTTAAAGGCATTTT 11 Hs.197114 Serine/arginine repetitive matrix 2 
GTTGTAAAATAAACTTT Hs.7869 ESTs, Highly similar to PLCD_HUMAN 1-acyl-sn-glyce 
GGCAGCCAGAGCTCCAA Hs.75061 Macrophage myristoylated alanine-rich C kinase 
TATGAGGGTAATCGAGA 13 12 Hs.24950 Regulator of G-protein signaling 5 
GATAATTTTTGTTCTTC Hs.6468 HSPC142 protein 
CTGTCCTAGCAGCCACC 12 13 Hs.12956 Tax interaction protein 1 
AGAGGGTGGGAGGGCCC 12 15 Hs.82646 DnaJ (Hsp40) homolog, subfamily B, member 1 
AAGTACGAGGAAAACTT 10 Hs.380824 ESTs, Weakly similar to S55016 protein oaf—fruit 
TGAAATAAAACTCAGTA 11 Hs.9614 Nucleophosmin (nucleolar phosphoprotein B23) 
ACAAGTACTGTATTTTT 17 Hs.76206 Cadherin 5, type 2, VE-cadherin 
TTCTCCCAAATACCGTT 71 69 12 Hs.394790 Collagen, type IV, alpha 2 
TTCACAGATTTGAAAGA Hs.408936 G-protein γ-12 subunit 
TAAAAATGTTTCAAAAA Hs.82085 Serine (or cysteine) proteinase inhibitor, clade E 
A. Breast tumor-vascular genes
Long TagBEC T2BEC T1BEC N1Breast T/NColon T/NUG IDDescription
TAAATCCCCACTGGGAC 27 21 21 Hs.151738 Matrix metalloproteinase 9 (gelatinase B) 
GGTTGTTGCGGACATCC 31 20 20 Hs.23823 Hairy/enhancer-of-split HEYL 
CCACGGGATTCTCCTCC 16 15 15 Hs.377812 Homo sapiens clone FLC1492 PRO3121 mRNA 
AACACAGCCTGGGACCA 13 13 13 Hs.278625 Complement component 4A 
ACAGACTGTTAGCCAAG 20 12 12 18 Hs.125036 Tumor endothelial marker 7 precursor 
GCCATTTCTGTGGAGGG 11 10 10 Hs.48029 Snail homolog 1 (Drosophila
GAGCTGGCATAACATTG 13 Hs.394790 Collagen, type IV, α2 
CAGAGATGAATTTATAC 17 45 11 Hs.8997 Heat shock 70kDa protein 1A 
TTAGTGTCGTATTTTGA 18 17 Hs.111779 Secreted protein, acidic, (osteonectin) 
TAATCCTCAAGAAATAA 16 24 Hs.78409 Collagen, type XVIII, α 1 
TAGGTCAGGACCTTGGC 20 16 Hs.43666 Protein tyrosine phosphatase type IVA, member 3 (PRL3) 
AATCTGCGCCTGCGGGG 46 24 Hs.833 Interferon-stimulated protein, 15 kDa 
GAGAGTGTCTGCGGATA 30 22 Hs.5831 Tissue inhibitor of metalloproteinase 1 
GGTGGACACGGATCTGC Hs.111779 Secreted protein, acidic, cysteine-rich (osteonect) 
AATAAAGGCTACAGGCT Hs.179735 Ras homolog gene family, member C 
TACTAGTCCTCTAGAAA Hs.289088 Heat shock 90kDa protein 1, α 
GAGTAACGCGCCCGGCC Hs.122359 Calcium channel, voltage-dependent, α 1H 
TTTCTTAAAGGCATTTT 11 Hs.197114 Serine/arginine repetitive matrix 2 
GTTGTAAAATAAACTTT Hs.7869 ESTs, Highly similar to PLCD_HUMAN 1-acyl-sn-glyce 
GGCAGCCAGAGCTCCAA Hs.75061 Macrophage myristoylated alanine-rich C kinase 
TATGAGGGTAATCGAGA 13 12 Hs.24950 Regulator of G-protein signaling 5 
GATAATTTTTGTTCTTC Hs.6468 HSPC142 protein 
CTGTCCTAGCAGCCACC 12 13 Hs.12956 Tax interaction protein 1 
AGAGGGTGGGAGGGCCC 12 15 Hs.82646 DnaJ (Hsp40) homolog, subfamily B, member 1 
AAGTACGAGGAAAACTT 10 Hs.380824 ESTs, Weakly similar to S55016 protein oaf—fruit 
TGAAATAAAACTCAGTA 11 Hs.9614 Nucleophosmin (nucleolar phosphoprotein B23) 
ACAAGTACTGTATTTTT 17 Hs.76206 Cadherin 5, type 2, VE-cadherin 
TTCTCCCAAATACCGTT 71 69 12 Hs.394790 Collagen, type IV, alpha 2 
TTCACAGATTTGAAAGA Hs.408936 G-protein γ-12 subunit 
TAAAAATGTTTCAAAAA Hs.82085 Serine (or cysteine) proteinase inhibitor, clade E 
B. Breast-specific, tumor vasculature genes
Short TagColon T/NBrain T/NBreast T/NUnigene IDDescription
GAGCTGGCAT 11 Hs.75617 Collagen, type IV, alpha 2 
AACACAGCCT 13 Hs.170250 Complement component 4A 
GCCATTTCTG 10 Hs.48029 Snail 1 (drosophila homolog) 
GGTTGTTGCG 20 Hs.23823 Hairy/enhancer-of-split HEYL 
TTTGTTAAAA 10 Hs.78054 Pre-mRNA splicing factor similar to S. cerevisiae 
B. Breast-specific, tumor vasculature genes
Short TagColon T/NBrain T/NBreast T/NUnigene IDDescription
GAGCTGGCAT 11 Hs.75617 Collagen, type IV, alpha 2 
AACACAGCCT 13 Hs.170250 Complement component 4A 
GCCATTTCTG 10 Hs.48029 Snail 1 (drosophila homolog) 
GGTTGTTGCG 20 Hs.23823 Hairy/enhancer-of-split HEYL 
TTTGTTAAAA 10 Hs.78054 Pre-mRNA splicing factor similar to S. cerevisiae 
C. PRL3 and HEYL observed tag variants
BEC2 (T)BEC1 (T)BEC3 (N)Colon normalColon tumor
PRL3      
 taggtcagga 20 16 
 ttgggtgaaa (3′ extended) 
HEYL      
 ggttgttgcg 31 20 10 
 cctggttcag (3′ extended) 
C. PRL3 and HEYL observed tag variants
BEC2 (T)BEC1 (T)BEC3 (N)Colon normalColon tumor
PRL3      
 taggtcagga 20 16 
 ttgggtgaaa (3′ extended) 
HEYL      
 ggttgttgcg 31 20 10 
 cctggttcag (3′ extended) 

NOTE. In A, genes upregulated ≥6-fold in breast tumor ECs relative to normal breast ECs. In B, genes demonstrating both a 10-fold induction in breast tumor vasculature and a ≥5-fold induction than observed in colon or brain tumor vasculature. In C, tag frequency for PRL3 and HEYL tag variants observed in breast EC data and colon EC data.

Abbreviations: EC, endothelial cell; EST, expressed sequence tag; T, tumor-induced; N, normal-induced.

Table 3

Breast normal-specific vascular genes

SequenceBEC T2BEC T1BEC N1N/maxTUG IDDescription
CCCGCCAGTGCACAGGG 23 24 Hs.256297 Integrin, α 11 
CACCTCTCATACCCAGG 17 17 Hs.65436 Lysyl oxidase-like 1 
CCACGTGGGCTCATATG 42 15 Hs.179665 Cyclin-dependent kinase inhibitor 1A (p21, Cip1) 
CCCTACCCTGTTACCTT 13 13 Hs.75736 Apolipoprotein D 
GTACTAGTGTTTTTTAG 42 11 Hs.303649 Chemokine (C-C motif) ligand 2 
CAGTCTGAAGATTTTGC 11 11 Hs.306357 Immunoglobulin κ constant 
CTCCCGGAAATCGAGAG 11 11 Hs.103291 Neuritin 1 
CTCTGAGGGGCGACTAG 11 11 Hs.74602 Aquaporin 1 (channel-forming integral protein, 28k) 
GCCAGGTGGCCACTCCA 10 10 Hs.111301 Matrix metalloproteinase 2 (gelatinase A, 72kDa ge 
CATCTTCTAAGATTGCC 10 10 Hs.356487 ESTs, weakly similar to cytokine receptor-like fa 
GGCTGCCCTGGGCAGCC 15 Hs.74566 Dihydropyrimidinase-like 3 
GAGCTCAGTGGAACTTA 15 Hs.374990 CD34 antigen 
CCGTGACTCTGGACTAT Hs.296267 Follistatin-like 1 
CTGTTCTGCACTTTGCT Hs.105689 MSTP031 protein 
AGCCACTGTGTCTGGCC Hs.147313 Similar to CMRF35 antigen precursor (CMRF-35) 
TCGAAGATTCACTGGGA Hs.103395 Hypothetical protein FLJ14146 
AAAGGCATCAGTCCCCC Hs.256297 Integrin, α 11 
CCAGTAACCCCAGCTAC Hs.115175 Sterile α motif and leucine zipper containing 
CCCCCCTGTCCAGCGAG Hs.381214 Hypothetical protein DKFZp434N0650 
GCCTGTACAACCTCAAA 13 Hs.162 Insulin-like growth factor binding protein 2, 36kD 
GGAACACACAGCACAGA Hs.286049 Phosphoserine aminotransferase 
AGACAGTCATTTTTAAC Hs.17109 Integral membrane protein 2A 
ACCCTCAAGAGGAGAAA Hs.119 Wilms’ tumour 1-associating protein 
AAGGAGAACTGAGTGAC Hs.20144 Chemokine (C-C motif) ligand 15 
GACTGCGCGTCTTGCAG 13 Hs.10086 Type I transmembrane protein Fn14 
CTGCCGTGACAACAGGC Hs.75462 BTG family, member 2 
CAGAACAACTATCTTTG Hs.77858 Mesenchyme homeo box 2 (growth arrest-specific hom 
CACGGAGGCGGGGTCAG Hs.320 Xeroderma pigmentosum complementation group C 
AATAAATTCCTTCAACG Hs.76307 Neuroblastoma, suppression of tumorigenicity 1 
CCGCTGATCCACTCTCA Hs.184161 Exostoses (multiple) 1 
CGTATGCCTCAGGCTGC Hs.102267 Lysyl oxidase 
GACCTATCTCTATTGTA Hs.194431 Palladin 
GCTGCTGCGCGCGGGCT Hs.8102 Ribosomal protein S20 
TTCAGGAGGGGCTGGAT Hs.5890 Hypothetical protein FLJ23306 
GAAGTTGCCTTATTGTC Hs.26777 KIAA0843 protein 
TTGATCGAAGTGACTTT Hs.49007 Poly(A) polymerase α 
TTTCCTGTTTAGATAAG Hs.99210 ESTs, weakly similar to hypothetical protein FLJ2 
ACTCCCTCCTTAAAAAG Hs.76230 Ribosomal protein S10 
ACTATTTAAAGCCAGGG Hs.63908 ORM1-like 2 (S. cerevisiae) 
GTCTGCTCCAGGAGCTG 12 Hs.32978 Proprotein convertase subtilisin/kexin type 7 
CAGGAAGAAGTCCTTGG 12 Hs.322704 ESTs, weakly similar to T02670 probable thromboxan 
GTTCTCCCACTGCCCTG Hs.306079 Protein transport protein SEC61 α subunit is of 
TCATTGGCCTTCAGATG Hs.75777 Transgelin 
TAGGTCAACAAACTTGC Hs.183434 ATPase, H+ transporting, lysosomal interacting pro 
TAACCGCAGATAGGAAG Hs.74376 Olfactomedin 1 
TTTCATTTGCTCAGTCG Hs.132821 EST, highly similar to FM02_HUMAN DIMETHYLANILINE 
ACACTTCTTTCAGTATA Hs.83381 Quanine nucleotide binding protein 11 
CTGGAAACGCCGGAACC Hs.1480 Histidine rich calcium binding protein 
GACAGTTGTTTGGAGAA Hs.760 GATA binding protein 2 
GGCCACGTAGCGCGACT Hs.155597 D component of complement (adipsin) 
AAGGATGCGGTAATGGC Hs.83126 TAF11 RNA polymerase II, TATA box binding protein 
GCCAGACCCCTCCCACC Hs.306193 Hypothetical protein LQFBS-1 
TTCTGCAAGACTACTTG Hs.75462 BTG family, member 2 
CAAGGGTAAGAATGAGT Hs.76224 EGF-containing fibulin-like extracellular matrix p 
CAGTGCGTTCAGTGCAA Hs.8302 Four and a half LIM domains 2 
SequenceBEC T2BEC T1BEC N1N/maxTUG IDDescription
CCCGCCAGTGCACAGGG 23 24 Hs.256297 Integrin, α 11 
CACCTCTCATACCCAGG 17 17 Hs.65436 Lysyl oxidase-like 1 
CCACGTGGGCTCATATG 42 15 Hs.179665 Cyclin-dependent kinase inhibitor 1A (p21, Cip1) 
CCCTACCCTGTTACCTT 13 13 Hs.75736 Apolipoprotein D 
GTACTAGTGTTTTTTAG 42 11 Hs.303649 Chemokine (C-C motif) ligand 2 
CAGTCTGAAGATTTTGC 11 11 Hs.306357 Immunoglobulin κ constant 
CTCCCGGAAATCGAGAG 11 11 Hs.103291 Neuritin 1 
CTCTGAGGGGCGACTAG 11 11 Hs.74602 Aquaporin 1 (channel-forming integral protein, 28k) 
GCCAGGTGGCCACTCCA 10 10 Hs.111301 Matrix metalloproteinase 2 (gelatinase A, 72kDa ge 
CATCTTCTAAGATTGCC 10 10 Hs.356487 ESTs, weakly similar to cytokine receptor-like fa 
GGCTGCCCTGGGCAGCC 15 Hs.74566 Dihydropyrimidinase-like 3 
GAGCTCAGTGGAACTTA 15 Hs.374990 CD34 antigen 
CCGTGACTCTGGACTAT Hs.296267 Follistatin-like 1 
CTGTTCTGCACTTTGCT Hs.105689 MSTP031 protein 
AGCCACTGTGTCTGGCC Hs.147313 Similar to CMRF35 antigen precursor (CMRF-35) 
TCGAAGATTCACTGGGA Hs.103395 Hypothetical protein FLJ14146 
AAAGGCATCAGTCCCCC Hs.256297 Integrin, α 11 
CCAGTAACCCCAGCTAC Hs.115175 Sterile α motif and leucine zipper containing 
CCCCCCTGTCCAGCGAG Hs.381214 Hypothetical protein DKFZp434N0650 
GCCTGTACAACCTCAAA 13 Hs.162 Insulin-like growth factor binding protein 2, 36kD 
GGAACACACAGCACAGA Hs.286049 Phosphoserine aminotransferase 
AGACAGTCATTTTTAAC Hs.17109 Integral membrane protein 2A 
ACCCTCAAGAGGAGAAA Hs.119 Wilms’ tumour 1-associating protein 
AAGGAGAACTGAGTGAC Hs.20144 Chemokine (C-C motif) ligand 15 
GACTGCGCGTCTTGCAG 13 Hs.10086 Type I transmembrane protein Fn14 
CTGCCGTGACAACAGGC Hs.75462 BTG family, member 2 
CAGAACAACTATCTTTG Hs.77858 Mesenchyme homeo box 2 (growth arrest-specific hom 
CACGGAGGCGGGGTCAG Hs.320 Xeroderma pigmentosum complementation group C 
AATAAATTCCTTCAACG Hs.76307 Neuroblastoma, suppression of tumorigenicity 1 
CCGCTGATCCACTCTCA Hs.184161 Exostoses (multiple) 1 
CGTATGCCTCAGGCTGC Hs.102267 Lysyl oxidase 
GACCTATCTCTATTGTA Hs.194431 Palladin 
GCTGCTGCGCGCGGGCT Hs.8102 Ribosomal protein S20 
TTCAGGAGGGGCTGGAT Hs.5890 Hypothetical protein FLJ23306 
GAAGTTGCCTTATTGTC Hs.26777 KIAA0843 protein 
TTGATCGAAGTGACTTT Hs.49007 Poly(A) polymerase α 
TTTCCTGTTTAGATAAG Hs.99210 ESTs, weakly similar to hypothetical protein FLJ2 
ACTCCCTCCTTAAAAAG Hs.76230 Ribosomal protein S10 
ACTATTTAAAGCCAGGG Hs.63908 ORM1-like 2 (S. cerevisiae) 
GTCTGCTCCAGGAGCTG 12 Hs.32978 Proprotein convertase subtilisin/kexin type 7 
CAGGAAGAAGTCCTTGG 12 Hs.322704 ESTs, weakly similar to T02670 probable thromboxan 
GTTCTCCCACTGCCCTG Hs.306079 Protein transport protein SEC61 α subunit is of 
TCATTGGCCTTCAGATG Hs.75777 Transgelin 
TAGGTCAACAAACTTGC Hs.183434 ATPase, H+ transporting, lysosomal interacting pro 
TAACCGCAGATAGGAAG Hs.74376 Olfactomedin 1 
TTTCATTTGCTCAGTCG Hs.132821 EST, highly similar to FM02_HUMAN DIMETHYLANILINE 
ACACTTCTTTCAGTATA Hs.83381 Quanine nucleotide binding protein 11 
CTGGAAACGCCGGAACC Hs.1480 Histidine rich calcium binding protein 
GACAGTTGTTTGGAGAA Hs.760 GATA binding protein 2 
GGCCACGTAGCGCGACT Hs.155597 D component of complement (adipsin) 
AAGGATGCGGTAATGGC Hs.83126 TAF11 RNA polymerase II, TATA box binding protein 
GCCAGACCCCTCCCACC Hs.306193 Hypothetical protein LQFBS-1 
TTCTGCAAGACTACTTG Hs.75462 BTG family, member 2 
CAAGGGTAAGAATGAGT Hs.76224 EGF-containing fibulin-like extracellular matrix p 
CAGTGCGTTCAGTGCAA Hs.8302 Four and a half LIM domains 2 

NOTE. Genes upregulated ≥6-fold in normal breast ECs relative to breast tumor ECs.

Abbreviations: EC, endothelial cell; EST, expressed sequence tag.

We thank Dr. Bert Vogelstein for reagents and constant encouragement, Dr. Frank Frassica for freshly resected tissue, and Clarence Lin for pathology support.

1
Leek RD The prognostic role of angiogenesis in breast cancer.
Anticancer Res
2001
;
21
:
4325
-31.
2
Bertucci F, Viens P, Hingamp P, Nasser V, Houlgatte R, Birnbaum D Breast cancer revisited using DNA array-based gene expression profiling.
Int J Cancer
2003
;
103
:
565
-71.
3
Velculescu VE, Zhang L, Vogelstein B, Kinzler KW Serial analysis of gene expression.
Science
1995
;
270
:
484
-7.
4
Nacht M, Ferguson AT, Zhang W, et al Combining serial analysis of gene expression and array technologies to identify genes differentially expressed in breast cancer.
Cancer Res
1999
;
59
:
5464
-70.
5
Dexter DL, Kowalski HM, Blazar BA, Fligiel Z, Vogel R, Heppner GH Heterogeneity of tumor cells from a single mouse mammary tumor.
Cancer Res
1978
;
38
:
3174
-81.
6
Klein CA, Blankenstein TJ, Schmidt-Kittler O, et al Genetic heterogeneity of single disseminated tumour cells in minimal residual cancer.
Lancet
2002
;
360
:
683
-9.
7
Shekhar MP, Pauley R, Heppner G Host microenvironment in breast cancer development: extracellular matrix-stromal cell contribution to neoplastic phenotype of epithelial cells in the breast.
Breast Cancer Res
2003
;
5
:
130
-5.
8
Kang Y, Siegel PM, Shu W, et al A multigenic program mediating breast cancer metastasis to bone.
Cancer Cell
2003
;
3
:
537
-49.
9
Boudreau N, Myers C Breast cancer-induced angiogenesis: multiple mechanisms and the role of the microenvironment.
Breast Cancer Res
2003
;
5
:
140
-6.
10
Haslam SZ, Woodward TL Host microenvironment in breast cancer development: epithelial-cell-stromal-cell interactions and steroid hormone action in normal and cancerous mammary gland.
Breast Cancer Res
2003
;
5
:
208
-15.
11
Liotta LA, Kohn EC The microenvironment of the tumour-host interface.
Nature
2001
;
411
:
375
-9.
12
Wiseman BS, Werb Z Stromal effects on mammary gland development and breast cancer.
Science
2002
;
296
:
1046
-9.
13
Iacobuzio-Donahue CA, Argani P, Hempen PM, Jones J, Kern SE The desmoplastic response to infiltrating breast carcinoma: gene expression at the site of primary invasion and implications for comparisons between tumor types.
Cancer Res
2002
;
62
:
5351
-7.
14
Linderholm B, Lindh B, Tavelin B, Grankvist K, Henriksson R p53 and vascular-endothelial-growth-factor (VEGF) expression predicts outcome in 833 patients with primary breast carcinoma.
Int J Cancer
2000
;
89
:
51
-62.
15
Olewniczak S, Chosia M, Kwas A, Kram A, Domagala W Angiogenesis and some prognostic parameters of invasive ductal breast carcinoma in women.
Pol J Pathol
2002
;
53
:
183
-8.
16
St Croix B, Rago C, Velculescu V, et al Genes expressed in human tumor endothelium.
Science
2000
;
289
:
1197
-202.
17
Saha S, Bardelli A, Buckhaults P, et al A phosphatase associated with metastasis of colorectal cancer.
Science
2001
;
294
:
1343
-6.
18
Porter DA, Krop IE, Nasser S, et al A SAGE (serial analysis of gene expression) view of breast tumor progression.
Cancer Res
2001
;
61
:
5697
-702.
19
Madden SL, Cook BP, Nacht M, et al Vascular gene expression in nonneoplastic and malignant brain.
Am J Pathol
2004
;
165
:
601
-8.
20
Bardelli A, Saha S, Sager JA, et al PRL-3 expression in metastatic cancers.
Clin Cancer Res
2003
;
9
:
5607
-15.
21
Saha S, Sparks AB, Rago C, et al Using the transcriptome to annotate the genome.
Nat Biotechnol
2002
;
20
:
508
-12.
22
Rost T, Pyritz V, Rathcke IO, Gorogh T, Dunne AA, Werner JA Reduction of LOX- and LOXL2-mRNA expression in head and neck squamous cell carcinomas.
Anticancer Res
2003
;
23
:
1565
-73.
23
Smith-Mungo LI, Kagan HM Lysyl oxidase: properties, regulation and multiple functions in biology.
Matrix Biol
1998
;
16
:
387
-98.
24
Nedivi E, Wu GY, Cline HT Promotion of dendritic growth by CPG15, an activity-induced signaling molecule.
Science
1998
;
281
:
1863
-6.
25
Fujita N, Jaye DL, Kajita M, Geigerman C, Moreno CS, Wade PA MTA3, a Mi-2/NuRD complex subunit, regulates an invasive growth pathway in breast cancer.
Cell
2003
;
113
:
207
-19.
26
Blanco MJ, Moreno-Bueno G, Sarrio D, et al Correlation of Snail expression with histological grade and lymph node status in breast carcinomas.
Oncogene
2002
;
21
:
3241
-6.
27
Steidl C, Leimeister C, Klamt B, et al Characterization of the human and mouse HEY1, HEY2, and HEYL genes: cloning, mapping, and mutation screening of a new bHLH gene family.
Genomics
2000
;
66
:
195
-203.
28
Maier MM, Gessler M Comparative analysis of the human and mouse Hey1 promoter: Hey genes are new Notch target genes.
Biochem Biophys Res Commun
2000
;
275
:
652
-60.
29
Vincent-Salomon A, Thiery JP Host microenvironment in breast cancer development: epithelial-mesenchymal transition in breast cancer development.
Breast Cancer Res
2003
;
5
:
101
-6.
30
Cheng CW, Wu PE, Yu JC, et al Mechanisms of inactivation of E-cadherin in breast carcinoma: modification of the two-hit hypothesis of tumor suppressor gene.
Oncogene
2001
;
20
:
3814
-23.
31
Zeng Q, Dong JM, Guo K, et al PRL-3 and PRL-1 promote cell migration, invasion, and metastasis.
Cancer Res
2003
;
63
:
2716
-22.
32
Corada M, Zanetta L, Orsenigo F, et al A monoclonal antibody to vascular endothelial-cadherin inhibits tumor angiogenesis without side effects on endothelial permeability.
Blood
2002
;
100
:
905
-11.
33
Liao F, Li Y, O’Connor W, et al Monoclonal antibody to vascular endothelial-cadherin is a potent inhibitor of angiogenesis, tumor growth, and metastasis.
Cancer Res
2000
;
60
:
6805
-10.
34
Corada M, Liao F, Lindgren M, et al Monoclonal antibodies directed to different regions of vascular endothelial cadherin extracellular domain affect adhesion and clustering of the protein and modulate endothelial permeability.
Blood
2001
;
97
:
1679
-84.
35
Venkiteswaran K, Xiao K, Summers S, et al Regulation of endothelial barrier function and growth by VE-cadherin, plakoglobin, and beta-catenin.
Am J Physiol Cell Physiol
2002
;
283
:
C811
-21.
36
Carmeliet P, Collen D Molecular basis of angiogenesis. Role of VEGF and VE-cadherin.
Ann N Y Acad Sci
2000
;
902
:
249
-62; discussion 2624.
37
Thomas R, True LD, Bassuk JA, Lange PH, Vessella RL Differential expression of osteonectin/SPARC during human prostate cancer progression.
Clin Cancer Res
2000
;
6
:
1140
-9.
38
Yiu GK, Chan WY, Ng SW, et al SPARC (secreted protein acidic and rich in cysteine) induces apoptosis in ovarian cancer cells.
Am J Pathol
2001
;
159
:
609
-22.
39
Jacob K, Webber M, Benayahu D, Kleinman HK Osteonectin promotes prostate cancer cell migration and invasion: a possible mechanism for metastasis to bone.
Cancer Res
1999
;
59
:
4453
-7.
40
Naeve GS, Ramakrishnan M, Kramer R, Hevroni D, Citri Y, Theill LE Neuritin: a gene induced by neural activity and neurotrophins that promotes neuritogenesis.
Proc Natl Acad Sci USA
1997
;
94
:
2648
-53.
41
Bates D, Taylor GI, Minichiello J, et al Neurovascular congruence results from a shared patterning mechanism that utilizes Semaphorin3A and Neuropilin-1.
Dev Biol
2003
;
255
:
77
-98.