Cadmium Malignantly Transforms Normal Human Breast Epithelial Cells into a Basal-like Phenotype

Background Breast cancer has recently been linked to cadmium exposure. Although not uniformly supported, it is hypothesized that cadmium acts as a metalloestrogenic carcinogen via the estrogen receptor (ER). Thus, we studied the effects of chronic exposure to cadmium on the normal human breast epithelial cell line MCF-10A, which is ER-negative but can convert to ER-positive during malignant transformation. Methods Cells were continuously exposed to low-level cadmium (2.5 μM) and checked in vitro and by xenograft study for signs of malignant transformation. Transformant cells were molecularly characterized by protein and transcript analysis of key genes in breast cancer. Results Over 40 weeks of cadmium exposure, cells showed increasing secretion of matrix metalloproteinase-9, loss of contact inhibition, increased colony formation, and increasing invasion, all typical for cancer cells. Inoculation of cadmium-treated cells into mice produced invasive, metastatic anaplastic carcinoma with myoepithelial components. These cadmium-transformed breast epithelial (CTBE) cells displayed characteristics of basal-like breast carcinoma, including ER-α negativity and HER2 (human epidermal growth factor receptor 2) negativity, reduced expression of BRCA1 (breast cancer susceptibility gene 1), and increased CK5 (cytokeratin 5) and p63 expression. CK5 and p63, both breast stem cell markers, were prominently overexpressed in CTBE cell mounds, indicative of persistent proliferation. CTBE cells showed global DNA hypomethylation and c-myc and k-ras overexpression, typical in aggressive breast cancers. CTBE cell xenograft tumors were also ER-α negative. Conclusions Cadmium malignantly transforms normal human breast epithelial cells—through a mechanism not requiring ER-α—into a basal-like cancer phenotype. Direct cadmium induction of a malignant phenotype in human breast epithelial cells strongly fortifies a potential role in breast cancer.

Breast cancer is a common disease and a lead ing cause of cancer deaths in women (Bray et al. 2004;Parkin et al. 2005). However, the etiology of breast cancer remains incom pletely defined. Evidence indicates that both endocrine and environmental factors play mechanistic roles in female breast cancer (Bray et al. 2004), and estrogenic hormones are implicated as major determinants of breast cancer risk (Bernstein 2002;Bray et al. 2004). Endogenous estrogens impact nor mal breast growth and development, increas ing proliferation of critical cell populations, whereas exogenous, pharmaco logic estrogens and xeno estrogens likely contribute to accu mulated breast cancer risk (Bernstein 2002;Bray et al. 2004). However, classical estrogens alone cannot account for all cases of human breast cancer (Coyle 2004).
Cadmium is a toxic metal and common environmental contaminant [International Agency for Research on Cancer (IARC) 1993; Waalkes 2003]. A human lung carcinogen, cadmium has several other target sites in rodents, including tissues considered endocrine sensitive (IARC 1993;Waalkes 2003). Recent data indicate that human cadmium exposure may be associated with female breast cancer (McElroy et al. 2006), although this initial, hypothesisforming work does not allow for establishment of definitive causality. There are no corollary data showing carcino genic activity for cadmium in female rodent mammary tis sue, but it can cause mammary gland prolifera tion in mice (Johnson et al. 2003). Additional studies, including in vitro cancer model studies, are critical to clarify any role for cad mium in this important and deadly disease.
Cadmium probably acts in all stages of the oncogenic process, and acts through multiple, nonexclusive mechanisms such as oxidative stress, oncogene activation, apoptotic by pass, and altered DNA methylation (Waalkes 2003). Recently, it was proposed that cad mium acts as a metallo estrogen via inter actions with estrogen receptorα (ERα), stimulat ing downstream estrogenrelated processes (GarciaMorales et al. 1994;Johnson et al. 2003;Stoica et al. 2000). This has led to fears that cadmium could act as an xeno estrogen in estrogenrelated cancers such as breast cancer (Darbre 2006). It is suspected that a critical early event in many breast cancers is consti tutive activation of the ERα (Zhang et al. 2005). Data indicating that human cadmium exposure may be a risk factor in breast cancer (McElroy et al. 2006) support a concern but do not actually address mechanism. Indeed, the theory that cadmium is metallo estrogenic has not been fortified by actual data associ ating it with acquired malignant phenotype in vivo, such as breast tumors in animals, or in vitro, such as malignantly transformed breast cells. Other researchers have found that cadmium lacks strong estrogenic activity or actually inhibits ER (Le Guével et al. 2000;Silva et al. 2006). We found little evidence of ERα activation in vivo or in vitro by cad mium (Coppin JF, Waalkes MP, unpublished data). It is evident that cadmium can act through various non-estrogenrelated mech anisms, and several mechanisms can occur simultaneously. Further, breast cancer is not always a disease that is absolutely estrogen dependent (Coyle 2004).
Given the importance of female breast can cer, the emergence of data indicating that cad mium may be a risk factor and the unresolved proposal that this could occur through a met alloestrogenic mechanism both clearly indicate that additional research is needed, including research using in vitro carcino genesis model systems. Thus, it was our goal to investigate the role of ER in a cell model of cadmiuminduced breast cancer. We examined the malignant Background: Breast cancer has recently been linked to cadmium exposure. Although not uniformly supported, it is hypothesized that cadmium acts as a metalloestrogenic carcinogen via the estrogen receptor (ER). Thus, we studied the effects of chronic exposure to cadmium on the normal human breast epithelial cell line MCF-10A, which is ER-negative but can convert to ER-positive during malignant transformation. Methods: Cells were continuously exposed to low-level cadmium (2.5 µM) and checked in vitro and by xenograft study for signs of malignant transformation. Transformant cells were molecularly characterized by protein and transcript analysis of key genes in breast cancer. results: Over 40 weeks of cadmium exposure, cells showed increasing secretion of matrix metalloproteinase-9, loss of contact inhibition, increased colony formation, and increasing invasion, all typical for cancer cells. Inoculation of cadmium-treated cells into mice produced invasive, metastatic anaplastic carcinoma with myoepithelial components. These cadmium-transformed breast epithelial (CTBE) cells displayed characteristics of basal-like breast carcinoma, including ER-α negativity and HER2 (human epidermal growth factor receptor 2) negativity, reduced expression of BRCA1 (breast cancer susceptibility gene 1), and increased CK5 (cytokeratin 5) and p63 expression. CK5 and p63, both breast stem cell markers, were prominently over expressed in CTBE cell mounds, indicative of persistent proliferation. CTBE cells showed global DNA hypomethylation and c-myc and k-ras over expression, typical in aggressive breast cancers. CTBE cell xenograft tumors were also ER-α negative. conclusions: Cadmium malignantly transforms normal human breast epithelial cells-through a mechanism not requiring ER-α-into a basal-like cancer phenotype. Direct cadmium induction of a malignant phenotype in human breast epithelial cells strongly fortifies a potential role in breast cancer.

Materials and Methods
Cells and cell culture. MCF10A cells, derived from normal human breast epithe lium and immortalized but non tumorigenic (Soule et al. 1990), were grown in a base medium (MEGM Bullet Kit; Cambrex, East Rutherford, NJ), with all additives supplied in the kit except cholera toxin. Cultures were incubated at 37°C in 5% CO 2 in a humidi fied atmosphere and passed weekly. Cells were exposed continuously to 2.5 µM cadmium (CdCl 2 ; purity 99%; Sigma, St. Louis, MO) for up to 40 weeks. We used cultures grown in cadmiumfree medium as passagematched controls. Once malignant transformation was established by formation of xeno graft tumors, they were desig nated cadmiumtransformed breast epithelial (CTBE) cells.
We used untreated MCF7 human breast cancer cells as positive controls for ERα and ERβ protein and SKBR3 human breast can cer cells (Chrestensen et al. 2007) as positive controls for HER2 (human epidermal growth factor receptor 2) protein.
Matrix metalloproteinase-9. As an indica tion of malignant phenotype, secreted matrix metalloproteinase9 (MMP9) activity was assessed as described (BenbrahimTallaa et al. 2005) during cadmium exposure. Activity was measured in conditioned media by zymo graphic gels, and quantitation was based on control values set at 100%.
In vitro invasion. We examined the effect of cadmium on in vitro invasive ability using a modified Boyden blindwell chamber assay (Tokar and Webber 2005). Data were based on control cells set at 100%.
Colony formation. We assessed effects of chronic cadmium exposure on cellular ability to form colonies when plated in soft agar as described by Tokar and Webber (2005). Data were normalized to control cells set to 100%.
Xenograft tumorigenicity. Animal care was provided in accordance with the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources 1996). The animals were treated humanely and with regard for alleviation of suffering. Mice were housed under conditions of con trolled temperature, humidity, and light cycle.
For the xenograft study [National Cancer Institute (NCI)Frederick Animal Facility], 1 × 10 6 control cells or chronic cadmium treated (40 weeks) cells were inoculated bilaterally under the renal capsules (50 µL/ capsule) of 10 female nude (NCrnu) mice (NCIFrederick) per celltreatment group. Mice were palpated twice daily for signs of tumors and killed when tumors developed or at 6 months after inoculation. A complete necropsy was performed, and obvious tumors, both kidneys, and all abnormal tissues were fixed in 10% buffered formalin. Tissues were embedded in paraffin, sectioned, stained with hematoxylin and eosin (H&E), and analyzed by light microscopy.
We determined gene expression at the tran script level by reverse transcriptionpolymerase chain reaction (RTPCR) as described previ ously (BenbrahimTalla et al. 2005). The data were normalized to individual βactin level and  Global DNA methylation. We deter mined global DNA methylation by the methyl acceptance assay, as described previously (BenbrahimTallaa et al. 2005), at 0, 20, and 40 weeks of cadmium exposure.
Tumor ER-α. We used CTBE cellgenerated xeno graft tumors for immuno histochemical analysis of ERα protein.
For a positive control, we used an ERαpositive human breast tumor paraffin block (PanTomics, Richmond, CA). We used poly clonal antibody against human ERα as the primary antibody (at a dilution of 1:1,000) and a streptavidinconjugated secondary antibody. Antibody binding was visualized with an avid inbiotinperoxidase kit (VECTASTAIN Elite ABC Kit; Vector Laboratories, Burlingame, CA) with diaminobenzidine as the chromagen and hematoxylin as a nuclear counterstain. As a control, the primary antibodies were omitted. Statistical analysis. All data except tumor incidence are presented as mean ± SE from three or more independent samplings. Significance was determined by Student's ttest, by analysis of variance followed by Dunnett's multiple comparison test, or by Fisher exact test as appropriate, with p ≤ 0.05 considered significant.

Cadmium-exposed breast cells acquire a cancer phenotype.
We assessed the ability of chronic, lowlevel cadmium to induce transformation in the MCF10A ERnegative human breast epithelial cell line using various measure ments including MMP9, an enzyme secreted to degrade the extra cellular matrix and facili tate tumor cell invasion. A marked, progres sive increase in the secretion of active MMP9 occurred with cadmium exposure ( Figure 1A). By 40 weeks of exposure, cadmiumexposed cells also started forming cell mounds when confluent ( Figure 1B); this mounding indicates a loss of contact inhibition, which allows cells to continue to divide and form multiple hori zontal layers, a common occurrence with can cer cells. Although mounding was common in cadmiumtreated cells, it was seldom observed in control cells. Cadmiumtreated cells even formed mounds when subconfluent, which we did not observe in controls (data not shown). Cadmium also markedly increased colony for mation in soft agar by 40 weeks of exposure ( Figure 1C), which is typical of cancer cells and is thought to reflect anchorageindependent growth of tumorinitiating/cancer stem cells (Stingl et al. 2006;Tokar and Webber 2005). Invasive ability was also greatly increased by 40 weeks of cadmium exposure ( Figure 1D), a common characteristic of cancer cells. Cadmium-exposed breast cells acquire a malignant phenotype. Compelling evidence that cadmium had triggered a malignant pheno type was provided when malignant tumors formed in mice that had been inoc ulated under the renal capsule with cells chronically exposed to cadmium (40week exposure) (Figure 2A). The CTBE cells pro duced highly aggressive carcinoma within as little as 1 month. No tumors arose after inoculation with control cells. CTBE cells produced highly malignant, invasive, ana plastic carcinoma with myo epithelial com ponents containing epithelial, mesenchymal, and undifferentiated cells ( Figure 2B). CTBE cell tumors showed metastatic potential, as exemplified by a metas tasis to a regional lymph node ( Figure 2C). Invasive carcino mas make up approximately 85% of all diag nosed human breast cancers, and regional node metastasis is common with aggressive breast tumors.

CTBE cells have basal-like malignant breast tumor characteristics.
Various breast cancer phenotypes have been defined based on molecular pathology, including the myo epithelial basallike carcinoma of the breast that is characterized as ERnegative and HER2negative with increased expression of CK5 and p63 (Fadare and Tavassoli 2007;Yehiely et al. 2006). Indeed, the ERnegative MCF10A cells remained negative for ERα and ERβ protein when they became CTBE cells ( Figure 3A). ERα and ERβ proteins were undetectable in CTBE cells com pared with a positive control breast cancer cell line (MCF7). Also, genes downstream of ERα driven by estrogens were not acti vated by chronic cadmium in CTBE cells, including pS2 (data not shown). Control and CTBE cells also showed no HER2 protein ( Figure 3A) compared with an HER2positive breast cancer line (SKBR3). In contrast, MT, which is over expressed in ERnegative breast cancers, was relatively low in control cells but markedly increased in CTBE cells ( Figure 3B).  Further, MCF10A cells are considered to have normal BRCA1 function (You et al. 2004), yet BRCA1 expression was suppressed in CTBE cells ( Figure 3C). Both CK5 and p63 were overexpressed in CTBE cells ( Figure 4A). CK5 and p63 are considered stem cell markers in breast tissue, and p63 may act as an oncogene. Foci of cell mounding, common in CTBE cells but rare in controls, indicate cells with loss of contact inhibition that maintain active proliferation. When we assessed foci for p63 and CK5, we found little or no expression in an uncom mon foci from control cells ( Figure 4B), but the much more commonly occurring CTBE cell mound foci showed intense expression for both p63 and CK5 protein in association with the mound (Figure 4C).

CTBE cells acquire characteristics of aggressive malignant breast cancer cells.
Compared with control cells, CTBE cells showed marked increases of both K-ras ( Figure 5A) and c-myc ( Figure 5B), oncogenes that are commonly over expressed in aggressive breast cancers (Eckert et al. 2004;Jamerson et al. 2004). In breast cancers, DNA hypo methylation decreases progressively as tumor grade worsens (Agrawal et al. 2007), and CTBE cells showed a significant and progres sive increase in global DNA hypomethylation with cadmium exposure ( Figure 5C).
Xenograft tumors remain ER-α negative. The remarkable cellular expansion in going from the tissue culture environment to a xeno graft tumor could provide a stimulus for acquired expression of genes not seen in vitro, such as ERα. However, analysis of CTBE induced xeno graft tumors showed minimal ERα protein in the tumor cells ( Figure 6A) compared with strong nuclear staining in a human breast carcinoma known to be ERα positive ( Figure 6B). A lymph node metasta sis from a CTBEformed tumor also showed minimal ERα protein ( Figure 6C).
Aromatase in CTBE cells. Cadmium may have indirectly provided MCF10A cells with estrogen via increased aromatase activ ity. However, transcript analysis indicated that CTBE cells showed no higher levels  (145 ± 34% of control; n = 3) than passage matched control cells (100 ± 21%).

Discussion
Environmental factors may account for a large portion of human breast cancers, perhaps approaching 60% (Coyle 2004). Established risk factors such as exogenous estrogens account for a significant portion of this risk (Bernstein 2002;Bray et al. 2004) but do not explain the remainder (Coyle 2004). The increasing incidence and geographic variation of human breast cancer has begun to focus attention on the etiologic potential of other environmental factors (Coyle 2004). Cadmium, a common environ mental pollutant, may be such a factor (McElroy et al. 2006) and is noteworthy as a biologically persistent and cumulative metal (IARC 1993;Waalkes 2003). Unusually high levels of cadmium are found in human breast tissue, perhaps indicating specific binding (Antilia et al. 1996), although inter individual levels vary widely. Cadmium was linked to human breast cancer in a recent population based, case-control study (McElroy et al. 2006). Based on urinary cadmium, both breast cancer risk and tumor aggressiveness increased with increasing exposure (McElroy et al. 2006). This is consistent with our data, where cad mium in vitro both induced malignant trans formation and produced highly aggressive cells, as the molecular pheno type of the CTBE cells equates to a cancer of poor prognosis (Fadare and Tavassoli 2007;Yehiely et al. 2006). The direct triggering of an acquired malignant phe notype by cadmium in human breast epithelial cells strongly supports the emerging epidemio logic data indicating a role for cadmium in human breast cancer (McElroy et al. 2006). This present study demonstrated that ERnegative human breast epithelial cells undergo transformation with chronic cad mium exposure. However, CTBE cells remain ERnegative after acquisition of malignant phenotype in vitro and even after production of xenograft tumors in vivo. Indeed, cadmium produced an apparent basallike breast cancer phenotype, including ER negativity, HER2 negativity, reduced BRCA1 expression, and increased expression of p63 and CK5, all note worthy charac teristics of basallike human breast cancer pheno type (Fadare and Tavassoli 2007;Liu et al. 2008;RibeiroSilva et al. 2005;Yehiely et al. 2006). Basallike breast cancers clinically show both poor relapsefree and poor survival rates (Fadare and Tavassoli 2007;Yehiely et al. 2006), and the anaplas tic xenograft tumors formed with CTBE cells are consistent with an aggressive tumor with poor prognosis. One mechanism proposed for cadmium is that it acts through actions at ERα that mimic estrogens, thereby chroni cally activating pathways that predispose to estrogenrelated cancer (GarciaMorales et al. 1994;Johnson et al. 2003;Stoica et al. 2000). The MCF10A cells used in this study can be treated in various ways to undergo transforma tion with the emergence of stimulated ERα expression as the probable basis for the malig nant conversion (Shekhar et al. 1998;Zhang et al. 2005). MCF10A cells can show acti vation of genes not seen in basallike breast cancer phenotypes, such as HER2, with acquired malignant potential (Li et al. 2004). MCF10A cells are fully capable of reversing their ER negativity during acquired malignant phenotype (Shekhar et al. 1998;Zhang et al. 2005). A key early event in estrogendependent breast cancers is activation of ERα, which can occur with MCF10A cell transforma tion (Zhang et al. 2005). Yet, in our model, MCF10A cells were ERnegative at the onset, remained so in vitro after cadmiuminduced malignant transformation, and continued to be ERnegative after forming xeno graft carci nomas. Cadmium has a variety of possible car cinogenic mechanisms, but from this work it appears unlikely that metallo estrogenic actions through ER were a major factor. Nonetheless, it is possible that cadmium may have metallo estrogenic effects in some instances, and a recent epidemiologic study associated dietary cadmium with endometrial cancer, another site considered estrogenrelated (Akesson et al. 2008). However, assumption of mechanism without clear and compelling data may be unwarranted with carcinogens like cadmium, which clearly has multiple possible mechanisms (IARC 1993;Waalkes 2003).
Several studies have shown that human breast tissue concentrates cadmium, and this is exaggerated in cancerous tissue (Antila et al. 1996;Ionescu et al. 2006;Rydzewska et al. 2004;Strumylaite et al. 2008). The metalbinding protein MT avidly binds cad mium and likely accounts for its long tissue residence time (Cherian et al. 2003). In humans, breast tumor MT over expression is associated with a poorer prognosis (Jin et al. 2004). A remarkably clear correlation exists in breast tumors between MT over expression and poor ER expression (El Sharkawy and Farrag 2008), indicating that increased MT may be another basallike phenotype marker. Tissues often accumulate cadmium associated with MT (Cherian et al. 2003). Thus, whatever mechanisms may operate in the breast, this MT over production would toxico kinetically favor cadmiuminduced tumor formation by enhancing the metals accumulation.  Both p63 and CK5 expression were mark edly increased in CTBE cells. CK5 and p63 are both considered basallike breast carcinoma markers (Fadare and Tavassoli 2007;Yehiely et al. 2006) and markers for breast stem cells (Boecker and Buerger 2003;RibeiroSilva et al. 2005). It appears that p63 functions to pre serve adult breast stem cells, facilitating replica tion and regeneration, possibly by restricting proliferation from an undifferentiated state (RibeiroSilva et al. 2005). Similarly, CK5 positive cells likely represent undifferentiated adult stem cells with potential to differentiate into glandular or myoepithelial cells (Boecker and Buerger 2003). CTBE cells show increased expression of both p63 and CK5, particularly in areas of cell mounding (active proliferation), indicating the over production of stemlike cells that have lost appropriate differentiation capacity during malignant transformation. An emerging hypothesis is that breast stem cells are critical targets of carcinogens and that blocked differentiation is likely a major pathway to can cer (Dontu et al. 2005). The fact that CTBE cells over express stem cell markers and produce a poorly differentiated, aggressive anaplastic xeno graft carcinoma is consistent with this hypothesis. Reduced expression of BRCA1 also strongly correlates with over expression of both CK5 and p63 (RibeiroSilva et al. 2005). BRCA1 is considered to be a breast tumor sup pressor gene, and reduced expression or loss of function is associated with ERnegative basal type breast cancers (Liu et al. 2008;Ribeiro Silva et al. 2005). Accumulating data indicates that BRCA1 regulates stem/progenitor cell fate in the breast (Liu et al. 2008;RibeiroSilva et al. 2005), and loss of function or suppressed BRCA1 expression may lead to dys regulated stem cell selfrenewal or differentiation lead ing to basaltype breast carcinomas (Liu et al. 2008). Loss of BRCA1 function can cause the accumulation of genetically unstable breast stem cells, providing critical targets for further carcinogenic events (Liu et al. 2008). Thus, CTBE cells showed both p63 and CK5 over expression together with a significant loss of BRCA1 expression and ER negativity, all con sistent basal breast cancer phenotype (Liu et al. 2008;RibeiroSilva et al. 2005), which may indicate a loss of differentiation capacity during the acquisition of basal malignant phenotype.
Further studies are needed to elucidate the mechanisms by which cadmium may cause breast cancer. However, in the present study, cadmium malignantly transformed a breast epithelial cell, producing various molecular hallmarks of a basallike breast cancer, includ ing ER negativity. Thus, actions for cadmium as a metallo estrogen in this case are unlikely. It is possible that cadmium acted by produc ing altered DNA methylation status, thereby altering expression of key genes, including oncogenes, as seen in prior work with other cell transformation systems (Qu et al. 2005;Takiguchi et al. 2003). It also appears that cadmium transformation distorted stem cell population dynamics, a common occurrence in onco genesis (Wicha et al. 2006). Defining the exact mechanism of action for cadmium in the present case will require additional work. Regardless of the precise mechanism, the direct triggering of malignant phenotype by cadmium in human breast epithelial cells unambiguously supports a role for the metal in human breast cancer.