Breast Cancer

To identify genes involved in breast cancer, polymerase chain reaction-selected cDNA subtraction was utilized to construct a breast cancer-subtracted library. Differential screening of the library isolated the growth factor-inducible immediate-early geneCyr61, a secreted, cysteine-rich, heparin binding protein that promotes endothelial cell adhesion, migration, and neovascularization. Northern analysis revealed that Cyr61was expressed highly in the invasive breast cancer cell lines MDA-MB-231, T47D, and MDA-MB-157; very low levels were found in the less tumorigenic MCF-7 and BT-20 breast cancer cells and barely detectable amounts were expressed in the normal breast cells, MCF-12A. Univariate analysis showed a significant or borderline significant association between Cyr61 expression and stage, tumor size, lymph node positivity, age, and estrogen receptor levels. Interestingly, expression of Cyr61 mRNA increased 8- to 12-fold in MCF-12A and 3- to 5-fold in MCF-7 cells after 24- and 48-h exposure to estrogen, respectively. Induction of Cyr61mRNA was blocked by tamoxifen and ICI182,780, inhibitors of the estrogen receptor. Stable expression of Cyr61 cDNA under the regulation of a constitutive promoter in MCF-7 cells enhanced anchorage-independent cell growth in soft agar and significantly increased tumorigenicity and vascularization of these tumors in nude mice. Moreover, overexpression of Cyr61 in MCF-12A normal breast cells induced their tumor formation and vascularization in nude mice. In summary, these results suggest that Cyr61 may play a role in the progression of breast cancer and may be involved in estrogen-mediated tumor development.

Breast cancer is the most common form of malignancy and the second-leading cause of cancer-related death among women in the United States. The nature of the cellular and molecular changes that lead to breast cancer remains poorly defined. Several of the aberrant tumor suppressor genes that have been identified, include BRCA1, BRCA2 (1,2), and p53 (3); however, they are silenced or mutated in only a fraction of breast cancers. Oncogenes associated with breast cancers include myc, CCND1, and Her2 (4 -8), but only 15-30% of invasive breast cancers show increased expression of these genes.
We set out to isolate differentially expressed genes in human breast cancer. Suppression subtractive hybridization (SSH) 1 and differential screening (9) identified genes highly expressed in the carcinoma cell line MDA-MB-231 and either absent or minimally expressed in the normal breast cell line MCF-12A. Cyr61 was one of the genes isolated from the screening of the subtracted cDNA library. This gene codes for a growth factorinducible, immediate-early gene first identified in murine fibroblasts (10). Cyr61 is a secreted, cysteine-rich, heparin-binding protein that associates with the extracellular matrix. Purified Cyr61 protein has been reported to mediate cell adhesion, stimulate chemotaxis, augment growth factor-induced DNA synthesis, enhance cell survival, and induce angiogenesis in vivo (10 -12). Because these characteristics may foster the progression of breast cancer, we studied this gene in detail and found it to be highly expressed in some invasive breast cancer cell lines and 36% of primary breast tumors. Furthermore, characterization of the oncogenic activity of Cyr61 demonstrated that forced expression of Cyr61 enhanced MCF-7 cell growth in soft agar and promoted tumor growth in both normal breast and breast cancer cells in nude mice.

SSH and Differential Screening
SSH was performed by using the PCR-Select cDNA subtraction kit (CLONTECH). Tester double-stranded cDNA was synthesized from 2 g of poly(A) ϩ RNA isolated from the breast cancer cell line, MDA-MB-231; and driver cDNA was made from 2 g of RNA from the normal breast cell line, MCF-12A. The subtracted library was differentially screened with 32 P-labeled probes synthesized as first-strand cDNA from tester and driver. The differential clones were picked and confirmed by Northern analysis.

Proteins and Antibodies
Recombinant human Cyr61 protein was purified from an Escherichia coli host strain (BL21) programmed for synthesis of the Cyr61 protein via pGEX-5X-2 expression vector (Amersham Pharmacia Biotech). Anti-Cyr61 antibodies were prepared from polyclonal rabbit antisera raised against a GST-Cyr61 fusion protein. Monoclonal antibodies to CD31 were obtained from DADO Corp. (Carpinteria, CA).

RNA Preparation and Northern Analysis
Total RNA was isolated from cell lines and patient tissue by using TRIzol reagent (Life Technologies, Inc.) according to the standard protocol. Cyr61 cDNA probe was labeled with [ 32 P]dCTP by using a random primer (Life Technologies, Inc.). Total cellular RNA was separated on 1.2% formaldehyde-agarose gels and was immobilized on a Hybond-N ϩ membrane by standard capillary transfer and UV cross-linking. The membrane was hybridized with the Cyr61 probe by standard protocol and was rehybridized with a 32 P-labeled glyceraldehyde-3-phosphate dehydrogenase cDNA to confirm equal loading of the samples.

Cell Transfection and Soft Agar Assays
The expression vector pcDNA61 was constructed by placing fulllength human Cyr61 cDNA into the pcDNA3.1 eukaryotic expression vector containing the neomycin gene under the control of the same promoter (Invitrogen). The constructs were transfected into MCF-12A and MCF-7 cells by using LipofectAMINE, and transfectants were selected for G418 resistance (400 and 450 g/ml, respectively). The selected clones were confirmed by Northern analysis. For clonogenic assay, cells were plated into 24-well flat-bottomed plates using a twolayer soft agar system with a total of 1 ϫ 10 3 cells/well in a volume of 400 l/well, as described previously (13). After 14 days of incubation, the colonies were counted and measured. All experiments were done at least three times using triplicate plates per experimental point.

Cell Migration Assays
Cell migration assays were performed according to the protocol from Chemicon (Temecula, CA). 5 ϫ 10 4 cells were added to the top of each modified Boyden chamber (10-m thickness and 8-m pores) containing polycarbonate membranes (6.5-mm diameter) coated on the underside of the membrane with 10 g/ml vitronectin and with the lower chamber containing 500 l of migration buffer (medium with 0.5% bovine serum albumin). Cells were allowed to migrate to the underside of the top chamber for 4 -8 h. The migratory cells attached to the bottom surface of the membrane were stained with 0.1% crystal violet in 0.1 M borate, pH 9.0, and 2% ethanol for 20 min at room temperature. The stained cells were extracted by using extraction buffer (Chemicon). The number of migratory cells per membrane was determined by absorbance at 550 nm.

Tumorigenicity Assay
Stably transfected MCF-12A/61 and MCF-12A/V cells (1.0 ϫ 10 6 cells/flank) and MCF-7/61 and MCF-7/V cells (5 ϫ 10 4 cell/flank) were injected subcutaneously into 8-week-old female nude mice. Each animal was injected at two sites in the flanks. The resulting tumors were measured once a week, and tumor volume (mm 3 ) was calculated by using the standard formula: length ϫ width ϫ height ϫ 0.5236. Tumors were harvested 6 weeks after injection and individually weighed before fixation. Data were presented as both tumor volume (mean Ϯ S.D.) and tumor weight (mean Ϯ S.D.). Statistical analysis was performed with software (GraphPad, San Diego, CA) using the Student's t test.

Real-time Quantitative PCR
Quantitative PCR analysis was performed using the TaqMan PCR Core Reagent kit (PE Biosystems). cDNA of breast cancer samples were diluted, and real-time PCR was performed following the protocol. Her2/ neu-specific primers were 5Ј-ACAGTGGCATCTGTGAGCTG and 5Ј-CCCACGTCCGTAGAAAGGTA. The TaqMan probe for Her2/neu was 5Ј-CCAGCCCTGGTCACCTACAACACAG. ␤-Actin was used for normalization, and ␤-actin-specific primers were 5Ј-GATCATTGCTCCTC-CTGAGC and 5Ј-ACTCCTGCTTGCTGATCCAC. The TaqMan probe for ␤-actin was 5Ј-CTCGCTGTCCACCTTCCAGCAGAT.

Statistical Analysis
Univariate-Chi square methods were used to compare stage, tumor size category, and node status category in Cyr61-positive versus Cyr61negative individuals. Age, ER levels, and PgR levels were compared using t tests and Wilcoxon rank sum tests. The kappa statistic was used to assess concordance among stage, tumor size category, and node status and is reported with its standard error.
Multivariate-The simultaneous relationship between the six predictors and Cyr61 was modeled using classification tree methods. Comparison of stage versus tumor size showed that these two variables are proxies for each other (kappa ϭ 0.94 Ϯ 0.05, observed agreement ϭ FIG. 1. Northern blots of differentially expressed genes isolated from breast cancer cells by PCR-selected subtractive hybridization. Northern blots demonstrate representative clones isolated using suppression subtractive hybridization and differential screening which are differentially expressed. The cDNAs used as probes on Northern blots were isolated from the subtracted cDNA library. The same blots were hybridized with glyceraldehyde-3-phosphate dehydrogenase cDNA to confirm similar loading. MCF-12A is a normal breast cell line and MDA-MB-231 is a breast cancer cell line. 42/44 ϭ 95%). Therefore, tumor size was not included as a candidate in the multivariate analysis. The concordance between stage and lymph node status was only moderate (kappa ϭ 0.54 Ϯ 0.12). A logistic regression analysis was also carried out but gave poor results and is therefore not reported.

Cyr61 Is Highly Expressed in Some Breast Cancer Cell Lines and Primary
Breast Tumors-Using a differential screening technique (see "Materials and Methods"), we identified 36 genes highly expressed in breast cancer cell line MDA-MB-231 as compared with normal breast cell line MCF-12A (data not shown). The genes each displayed 6-fold or greater expression in MDA-MB-231 than in MCF-12A, as determined using Northern blots and densitometric analysis (Fig. 1). Cyr61, a growth factor-inducible immediate-early gene, is one of these differentially expressed genes. Cyr61 expression was examined in a panel of normal breast and breast cancer cell lines. Northern analysis showed that Cyr61 mRNA was prominently expressed   To determine the pattern of Cyr61 expression in primary breast tumors, RNAs were isolated from quick-frozen breast samples obtained at initial surgery from 44 individuals with breast cancer (Table I). Each breast carcinoma and matching normal breast tissue was confirmed histologically. Expression of Cyr61 was easily detectable in sixteen of 44 (36%) primary breast cancer samples (Table I), but levels were negligible in normal breast tissues as shown by Northern analysis (Fig. 2B, representative autoradiogram of Northern blot).
We analyzed the Her2/neu status for the clinical samples by performing real-time PCR using Her2/neu-specific primers. Breast cancer cell lines BT-474 and MCF-7 were used as high expressor and low expressor controls, respectively. Her2/neu was highly expressed in 7 of 16 (44%) Cyr61-positive samples compared with 6 of 28 (21%) Cyr61-negative samples, suggesting Cyr61 expression is positively correlated with Her2/neu expression (Table I).
Univariate analysis (Tables II and III) showed either significant or borderline significant association between breast cancer stage, tumor size, lymph node status, ER levels in the primary tumor as well as age at onset of disease compared with whether the primary tumor expressed Cyr61 on Northern blot. For analysis of stage, only 6 of 28 (21%) of women having stages I and II were positive for Cyr61; in contrast, 10 of 16 (63%) of women having stages IIIA, IIIB, or IV expressed Cyr61 (p Ͻ 0.006). Similarly, only 6 of 28 (21%) individuals with primary breast cancer size of either 1A (Յ2 cm in diameter) or 2A (Յ5 cm) had tumors that were Cyr61-positive; in comparison, 10 of 16 (63%) of individuals with a stage of either 3A or 4B (Ն5 cm) breast cancer had primary tumors that expressed Cyr61 (p ϭ 0.006). Furthermore, of the 17 patients who were lymph nodenegative, only two were Cyr61-positive (12%) compared with 14 Cyr61-positive tumors among the 27 individuals who were lymph node-positive (52%, p ϭ 0.01). The median and mean ages of patients who were Cyr61-negative, were 52 and 53 (Ϯ10 S.D.), respectively, compared with a median and mean age of 66 and 64 (Ϯ11.7 S.D.) in those who were Cyr61-positive (p ϭ 0.003). Tumors that were Cyr61-negative had a mean ER score of 40 fmol/g (Ϯ69 S.D.), whereas those that were Cyr61-positive had a mean ER score of 102 fmol/g (Ϯ110 S.D.) (p ϭ 0.03). No statistical difference in the mean progesterone (PgR) values was noted in those that were Cyr61-negative (mean 94.56 fmol/g) versus those that were Cyr61-positive (mean 129.5 fmol/g) (p ϭ 0.55).
Our classification tree multivariate model predicted that, if the patient at initial diagnosis was less than 57 years of age and had a breast cancer with an ER of less than 24 fmol/g, Cyr61 would be negative (%Cyr61-positive, 0/17 ϭ 0%). If the individual was less than 57 years old and had an ER greater than 24 fmol/g, the model predicted that Cyr61 would be positive (%Cyr61-positive, 4/4 ϭ 100%). If age was greater than 57 and the stage was I or II, the model predicted that Cyr61 would be negative (3/10 ϭ 30% Cyr61-positive); if the age of the patient was greater than 57 and had either stage IIIA or IIIB, Cyr61 was predicted to be positive (9/9 (100%) were Cyr61positive). This model has an observed sensitivity of 81% and observed specificity of 86% and gave an overall observed correct classification of 84% correctly classified. According to the tree model, when predicting who is Cyr61-positive, ER is only important in younger women under the age of 58, whereas stage is only important in older women over the age of 57. In summary, the model posits that the proportion who have Cyr61positive breast cancer increases with age, stage, and ER level.
Expression of Cyr61 Is Modulated through the Estrogen Re-  ceptor Pathway in MCF-12A and MCF-7 Cells-Previous studies have shown that Cyr61 is inducible in the uterus by estrogen treatment in ovariectomized rats (14). Furthermore, the correlation of Cyr61 expression in ER ϩ breast cancers observed in our experiments suggested a potential interaction between the estrogen receptor pathway and expression of Cyr61. To determine whether expression of Cyr61 was regulated by estrogen and estrogen blockade, the estrogen-responsive normal breast cells, MCF-12A, and breast cancer cells, MCF-7, were harvested at different times after estrogen and antiestrogen treatment. Expression of Cyr61 mRNA was induced at 16 h after estradiol (10 Ϫ9 M) treatment and reached maximum (8-to 12-fold) levels at 48 h and started to decrease slightly after 72 h in the MCF-12A cells (Fig. 3A). Induction of expression of with pcDNA61 containing either full-length cDNA of Cyr61 (MCF-12A/61 and MCF-7/61) or empty vector pcDNA3.1 (MCF-12A/V and MCF-7/V) as control. As expected, Cyr61 was highly expressed in the MCF-12A/61 and MCF-7/61 but not in the MCF-12/V and MCF-7/V transfected cells as shown by Northern studies (data not shown) and Western analysis (Fig. 4A). The MCF-7/61 cells expressing the Cyr61 vector formed significantly more colonies in soft agar (mean, 2.2-Ϯ 0.6-fold more colonies; p Ͻ 0.05) than MCF-7/V cells harboring the empty vector or MCF-7 control cells (Fig. 4B). The MCF-7/61 colonies also were substantially larger than the MCF-7/V and MCF-7 colonies (data not shown). The results indicated that forced expression of Cyr61 promoted anchorage-independent clonogenic growth of MCF-7 cells. Neither MCF-12A/V nor MCF-12A/61 formed colonies after 4 weeks.
To assess the effects of estrogen and tamoxifen on anchorageindependent growth of MCF-7 cells, clonogenic proliferation of MCF-7/V and MCF-7/61 in soft agar containing either estradiol (10 Ϫ9 M) or estradiol and tamoxifen (10 Ϫ7 M) was evaluated. Estrogen treatment significantly (p Ͻ 0.05) enhanced colony formation of both MCF-7/V and MCF-7/61 cells, and tamoxifen blocked the estrogen-stimulated colony formation in both of these cell types (Fig. 4C).
Cyr61 Promotes Tumor Growth and Vascularization in Nude Mice-On the basis of our in vitro studies indicating that overexpression of Cyr61 promotes anchorage-independent clonogenic proliferation in soft agar and cell migration, we investigated the effect of expression of Cyr61 on tumor development and neovascularization by comparing tumor formation of MCF-12/V and MCF-12A/6 cells, as well as MCF-7/V and MCF-7/61 cells, in nude mice. These cells were injected subcutaneously into 8-week-old nude mice, and tumor growth was measured once a week. Tumors from the normal breast cells expressing Cyr61 (MCF-12A/61) first became apparent 3 weeks after injection, and all of the mice developed tumors ranging from 0.6 to 1.4 g at 6 weeks after injection (Figs. 6A, 6B, and 7A). In contrast, the control mice that received MCF-12A/V cells containing empty vector remained tumor-free even at 12 weeks after injection (Fig. 7A). In the other experimental group, the MCF-7/61 cells expressing Cyr61 at a high level developed tumors with a significantly shorter latency (p Ͻ 0.05) and marked increased tumor growth compared with the tumors from the control MCF-7/V cells (Fig. 6, A and B). All 16 tumors that developed from MCF-7/61 cells appeared within 3 weeks of subcutaneous injection of the cells, whereas only eight tumors from 16 separate injections of the MCF-7/V developed during the entire 6 weeks of observation.
To examine whether the tumorigenic ability of Cyr61 in vivo was associated with angiogenic activity, tumors that developed from both MCF-7/61 and MCF-7/V cells were analyzed histochemically using antibody against CD31. Immunohistochemical analysis demonstrated robustly increased blood vessel density in MCF-7/61 tumors compared with those from the MCF-7/V controls (Fig. 7B). The tumors that developed from both MCF-7/61 and MCF-12A/61 had a similar histologic appearance. Both were classified as high grade infiltrating ductal carcinomas (Bloom-Richardson Grade III). The tumors had marked nuclear pleomorphism, absence of tubule formation, and a high mitotic count (12-15 per high power field). In contrast, the control MCF-7 tumors were moderately well differentiated infiltrating ductal carcinomas (Bloom-Richardson grade I) with tubule formation in greater than 75% of the tumor and less than one mitosis was observed per high power field.
Several lines of evidence support a role for CCN molecules in tumorigenesis. Although elevated expression of chicken nov mRNA was consistently found in all MAV1 (myeloblastosisassociated virus 1)-and MAV2-induced avian nephroblastomas (19), the human homolog of chicken nov (novH) is mainly overexpressed in tumors of predominantly stromal origin such as Wilms' tumors (25). Consistent with its profibrotic properties, CTFG is overexpressed in pancreatic cancers (26), mammary tumors (27), and melanomas (28). WISP-1 is strongly expressed in the fibrovascular stroma of breast tumors developing in Wnt-1 transgenic mice (22). Recent studies showed that the forced overexpression of WISP-1 in normal rat kidney fibroblasts (NRK-49F) was sufficient to induce their transformation (29).
In the present study, Cyr61 was found to be prominently expressed in the highly tumorigenic breast cancer cell lines such as MDA-MB-231, MDA-MB-157, MDA-MB-436, and T47D, but the levels were low in the less tumorigenic cells, MCF-7 and BT-20. It was barely detectable in normal breast cell lines MCF-10A and MCF-12A ( Fig. 2A). Furthermore, Cyr61 was expressed in 36% primary breast tumor samples (16/44) and was undetectable in normal breast tissues (samples N1 and N15) as shown by Northern analysis (Fig. 2B). After submission of our study, Tsai et al. (30) also noted that Cyr61 was expressed in about 30% of breast cancers. We found that expression of Cyr61 in breast tumor samples correlated with ER positivity and lymph node involvement. The latter is routinely used as a prognostic and predictive marker in the clinical management of breast cancer (Table I). These results suggest that Cyr61 is closely associated with the malignant phenotype in breast cancer and may serve as a marker of potential progression of the cancer. Statistical analysis of the 44 fresh tumors showed that tumors that were at a more advanced stage at diagnosis, with large primary tumors, that expressed Her-2/neu, and that had lymph node involvement, were more likely to express Cyr61. The duration of follow-up is too short to know if the prognosis of these individuals parallels their Cyr61 expression. This becomes more complex, because Cyr61 expres-sion also correlated with ER expression in the primary tumors. The ER-positive breast tumors are responsive to estrogen blockade, and ER expression in the breast cancer is a good prognostic indicator. Thus, we find a paradox that cyr61 expression is frequently associated with advance disease at diagnosis, but these tumors are often ER-positive.
Another interesting finding of this study is that expression of Cyr61 in breast cell lines can be modulated by estrogen and antiestrogen (Fig. 3). Estradiol markedly increased the expression of Cyr61 in both normal and breast cancer cells in a time-dependent manner. Thus, induction of Cyr61 is not specific for transformation of breast cells. Nevertheless, Cyr61 may play a role in tumor progression, because overexpression of Cyr61 in the normal breast cells (MCF-12A) allowed them to form tumors in nude mice. Fresh breast tumors that expressed Cyr61 had a greater level of expression of ER (mean, 102 fmol/g) than those tumors that did not express Cyr61 (mean, 40 fmol/g). Because we have not found an estrogen response element in the promoter region of the human Cyr61 (data not shown), we speculate that estrogen acts through one or more primary estrogen-responsive genes whose product regulates the expression of Cyr61. Consistent with this hypothesis is the finding that estradiol required a lag of 8 -16 h to stimulate the visible accumulation of Cyr61 mRNA in the breast cells.
The role of Cyr61 in breast tumor growth was evaluated in several experimental tumor models. The forced expression of Cyr61 in MCF-7 cells (MCF-7/61) markedly stimulated their anchorage-independent cell growth in soft agar and significantly enhanced their tumorigenicity and vascularization in vivo (Fig. 6). MCF-7/61 cells that highly expressed Cyr61 developed larger and more vascularized tumors in nude mice (Fig.  7B) and had a much shorter latency in their development of tumors than did the MCF-7/V cells containing the empty vector. This is consistent with an earlier observation showing that the gastric adenocarcinoma cell line RF-1 became tumorigenic when induced to express Cyr61 (31). Furthermore, we found that the overexpression of stably transfected Cyr61 in the normal breast cell line MCF-12A, which does not normally express Cyr61, resulted in tumor formation in nude mice (Fig. 6, A and  B). Taken together, these results suggest that prominent ex-  Cyr61 expressor). B, immunohistochemical analysis demonstrating robustly increased blood vessel density in MCF-7/61 tumors (left) in nude mice compared with those from the MCF-7/V controls (right) when immunostained with anti-CD31 antibodies.
pression of Cyr61 may facilitate transformation of breast tissue.
Recently, WISP-1, another CNN family member closely related to Cyr61, was found to be a Wnt-1-and ␤b-cateninresponsive oncogene (29). Transfected and overexpressed WISP-1 in normal rat kidney fibroblast cells (NRK-49F) induced their morphological transformation, accelerated cell growth, enhanced saturation density in vitro, and permitted the formation of tumors in nude mice. Considering that Cyr61 has four identical structural domains and is closely related to WISP-1, it might also be involved in the Wnt-1 and ␤b-catenin pathways, especially those that enhance tumor development and progression.
Previous studies have suggested that Cyr61 is involved in angiogenesis through its interaction with the ␣ v ␤ 3 integrin (31)(32)(33). This could provide insight into how it might enhance tumorigenesis. Angiogenesis, the formation of new capillaries from pre-existing blood vessels, is required for growth of solid tumors. This process is complex, encompassing migration and proliferation of endothelial cells and tube formation. It is regulated by many factors (34), which may include Cyr61 by promoting migration of human microvessel endothelial cells and inducing neovascularization via ␣ v ␤ 3 integrin (31). Integrins comprise a large family of cell surface receptors that enhance cellular adhesion to the extracellular matrix, which can exert profound control over cells. The effects of the matrix are primarily mediated by integrins. Integrin signals are involved in different cellular activities, including cell migration, proliferation, and survival, as well as diverse biological processes, including embryogenesis, angiogenesis, immune response, and tumor metastasis (35)(36)(37). Integrin ␣ v ␤ 3 may have an important role in tumor vascularization and progression (38). Antagonists of integrin ␣ v ␤ 3 significantly decrease tumor vascularization and size (39). The interaction of Cyr61 with integrin ␣ v ␤ 3 provides a molecular mechanism to account for the angiogenic and tumorigenic activities that occurred in our breast cancer model system.