Rat models of 17β-estradiol-induced mammary cancer reveal novel insights into breast cancer etiology and prevention

Abstract

Numerous laboratory and epidemiologic studies strongly implicate endogenous and exogenous estrogens in the etiology of breast cancer. Data summarized herein suggest that the ACI rat model of 17β-estradiol (E2)-induced mammary cancer is unique among rodent models in the extent to which it faithfully reflects the etiology and biology of luminal types of breast cancer, which together constitute ~70% of all breast cancers. E2 drives cancer development in this model through mechanisms that are largely dependent upon estrogen receptors and require progesterone and its receptors. Moreover, mammary cancer development appears to be associated with generation of oxidative stress and can be modified by multiple dietary factors, several of which may attenuate the actions of reactive oxygen species. Studies of susceptible ACI rats and resistant COP or BN rats provide novel insights into the genetic bases of susceptibility and the biological processes regulated by genetic determinants of susceptibility. This review summarizes research progress resulting from use of these physiologically relevant rat models to advance understanding of breast cancer etiology and prevention.

INTRODUCTION

Estrogens exert their diverse biological actions in the human breast and other target tissues by binding to and activating ESR1 [estrogen receptor (ER) alpha] and ESR2 (ER beta), two ligand-activated transcription factors that are encoded by distinct genes (14). Upon binding estrogen, ESR1 and ESR2 recruit an assortment of coregulatory proteins to specific DNA sequence motifs termed estrogen responsive elements and thereby regulate gene transcription in a manner that is specific to cell type and physiological context. ESR1 and ESR2 also exert ligand-independent actions on gene expression as a consequence of posttranslational modifications, for example phosphorylation at specific amino acid residues by an assortment of protein kinases. Studies of estrogen action in genome-edited mice and rats indicate that Esr1 is the primary mediator of estrogen action in the mammary gland (14). This conclusion is further supported by the observed lack of breast development in a young human female who harbors a homozygous missense mutation in ESR1 that reduces the ability of the ESR1 protein to bind hormone and activate gene transcription (119). In addition to acting through ESR1 and ESR2, estrogens can act through a membrane-associated G protein-coupled estrogen receptor (GPER) (118). Although the functions of GPER in the mammary gland/breast are not fully defined, studies using genome-edited mouse models do not support a major role for Gper in the regulation of mammary gland development by estrogens (111).

Numerous laboratory and epidemiologic studies strongly implicate endogenous and exogenous estrogens in the etiology of breast cancer (79, 80, 165). For example, multiple intervention studies demonstrated that selective estrogen receptor modulators, which block estrogen signaling through ESR1 and ESR2, and aromatase inhibitors, which block estrogen biosynthesis, reduce by ~50% the incidence of breast cancer in high-risk cohorts (45–47, 58, 162). Because these studies and others provide overwhelming evidence that estrogens contribute to breast cancer development, the International Agency for Research on Cancer and the National Toxicology Program of the National Institute of Environmental Health Sciences have categorized steroidal estrogens as known carcinogens in humans (72a, 117). These studies and a multitude of others also strongly suggest that ESR1 plays a central role in mediating the actions of estrogens in breast cancer development. However, the molecular mechanisms through which estrogens contribute to breast cancer development remain poorly defined. For this reason, our group and others have focused effort on development and characterization of rat models of 17β-estradiol (E2)-induced mammary cancer. The purpose of this article is to review research progress resulting from use of these physiologically relevant rat models to identify the mechanisms through which estrogens, together with interacting genetic and environmental factors, influence mammary cancer development, the hope being that the knowledge revealed will advance our ability to reduce the incidence of breast cancer in humans.

ACI RAT MODEL OF E2-INDUCED MAMMARY CANCER

Published data summarized herein suggest that the ACI rat model of E2-induced mammary cancer is unique among rodent models in the manner in which it faithfully reflects the etiology and biology of luminal types of breast cancer, which together constitute ~70% of all breast cancers. First described by our group in 1997, the ACI model is the only rodent model in which E2 delivered continuously at physiological levels that are normally observed during the periovulatory phase of the menstrual cycle or pregnancy in humans induces mammary cancer to an incidence of virtually 100% (139). Development of mammary cancers in E2-treated ACI rats is preceded by rapid induction of diffuse lobuloalveolar hyperplasia with subsequent appearance of focal regions of atypical hyperplasia, ductal carcinoma in situ, and invasive mammary carcinoma (Fig. 1) (35, 64, 84). The E2-induced mammary cancers highly express Esr1, Esr2, progesterone receptor (Pgr), and GATA binding protein 3 (Gata3), a transcription factor that functions during development of the mammary luminal epithelium (64, 76, 126, 163). Each of these molecular phenotypes is also exhibited by luminal breast cancer subtypes in humans (63, 73, 90).

Fig. 1.

Development and progression of 17β-estradiol (E2)-induced mammary cancer in ACI rats. The schematic illustrates the time following initiation of E2 treatment to appearance of lobuloalveolar hyperplasia (<7 days), focal regions of atypical hyperplasia (FRAH) (~70 days), and mammary carcinoma (median latency to palpable cancer is 125 days), including ductal carcinoma in situ (DCIS) and locally invasive carcinoma. Hematoxylin and eosin (H&E)-stained mammary tissues from virgin female ACI rat (A) and E2-treated ACI rats (B–E) illustrating E2-induced lobuloalveolar hyperplasia (B), FRAH (C), comedo-type DCIS (D), and invasive carcinoma with desmoplasia (E).

Genome instability is a hallmark of breast cancers in humans, and distinct types and/or patterns of somatic genetic alterations have been associated with specific breast cancer subtypes and clinical outcomes (67, 85, 107, 113). For example, ~85% of breast cancers exhibit aneuploidy, defined as a deviation from a normal diploid complement of chromosomes. Similarly, ~70% of mammary cancers induced in ACI rats by E2 exhibit aneuploidy with associated nonrandom patterns of gains and/or losses of specific chromosomes; i.e., chromosome instability (3, 64, 87, 88). By contrast, mammary cancers induced in rats by 7,12-dimethylbenz[a]anthracene (DMBA), an agent commonly used to induce mammary cancer in rat models, remain diploid (3, 26, 87, 128). Analysis of a panel of 28 mammary cancers induced by E2 in ACI rats using comparative genomic hybridization (CGH) revealed chromosome instability in 64% of the cancers examined (3). On average, 2.9 discernable somatic copy number alterations (CNAs) were observed in those tumors exhibiting chromosome instability, and the majority of the observed somatic changes (87%) were gains or losses of whole chromosomes. The most frequently observed CNA (10/28 tumors) was loss of rat chromosome 5 (RNO5), and this CNA was frequently observed in association with loss of RNO20 (7/10 tumors)(Fig. 2). No chromosome instability was observed when 15 mammary cancers induced in SS/JrHSDMcwi rats by DMBA were similarly evaluated by CGH (2). Together, these data further attest to the relevance of the ACI rat model of E2-induced mammary cancer to breast cancer in humans and further suggest that E2 and DMBA induce mammary cancer through distinct mechanisms or induce distinct types of mammary cancer. Examination of data from The Cancer Genome Atlas revealed that several of the most frequently deleted segmental CNAs in luminal breast cancers are orthologous to loci that reside on RNO5 or RNO20, the two chromosomes that are most frequently lost in mammary cancers induced by E2 in ACI rats (20). These frequently deleted CNAs include chromosome regions 9p21.3 (orthologous to RNO5), 1p36.11 (RNO5), 6q14.3 (RNO5 and RNO20), and 6q21 (RNO5 and RNO20). Moreover, several segmental CNAs that are frequently gained in luminal breast cancers are orthologous to rat chromosomes that are gained, but never lost, in mammary cancers induced by E2 in ACI rats, including 8p11.23 (RNO16), 8q24 (RNO7), 12q15 (RNO7), and 17q12 (RNO10). CGH analyses were also performed on an additional panel of 71 mammary cancers induced by E2 in F1 or congenic rats generated in crosses between susceptible ACI rats and resistant Copenhagen (COP) or Brown Norway (BN) rats. These analyses reinforced the data generated from analyses of mammary cancers from ACI rats and further indicated that the constitution of the germline influences the pattern of somatic CNAs observed within the mammary cancers induced by E2, a finding that has subsequently been established for human breast cancers (3, 107).

Fig. 2.

Comparative genomic hybridization (CGH) analysis of mammary carcinoma induced by 17β-estradiol in ACI rat. A: metaphase image showing the combined hybridization signals from test (green fluorescence) and reference (red fluorescence) DNA. B: reverse image of DAPI-stained metaphase chromosomes including the results of the chromosome identification analysis. C: the average (n = 10 metaphase cells) fluorescence ratio (FR) curves from the analysis of mammary tumor designated ACI-353. For 3 chromosomes, RNO5, RNO15, and RNO20, the FR curves are significantly displaced to the left, showing that there is a loss of copy number for these chromosome in this tumor. The red numerals under each chromosome ideogram indicate the chromosome number followed by the number of metaphase chromosomes evaluated for that chromosome in the illustrated mammary tumor. The curve next to the Y chromosome is irrelevant in this analysis, since the test DNA and the reference DNA are both from female rats. A–C are from Adamovic et al. (3). D: the ACI-353 tumor was further evaluated by array CGH as part of a subsequent study, confirming loss of RNO5, RNO15, and RNO20 in this tumor.

MECHANISMS UNDERLYING ESTROGEN-INDUCED MAMMARY CARCINOGENESIS

It is well established that tamoxifen, a selective ER modulator, is effective in preventing development of breast cancer in high-risk cohorts, strongly suggesting the involvement of ER-dependent pathways in breast cancer etiology (45–47). Concurrent treatment of ACI rats with E2 and tamoxifen results in dramatically fewer tumors than observed in rats treated with E2 alone, indicating that mammary cancer development in E2-treated ACI rats similarly occurs in large part through an ER-dependent mechanism(s) (86, 89, 143). Surgical oophorectomy, selective estrogen receptor modulators, and aromatase inhibitors serve as endocrine-based therapies for estrogen-dependent, luminal type, breast cancers. Mammary cancers induced by E2 in ACI rats rapidly regress upon withdrawal of administered E2 or upon treatment with tamoxifen, indicating that E2 is required for sustained cancer cell proliferation and survival (64, 126). Because studies of rat and mouse gene knockout models indicate that Esr1 is the primary mediator of E2 action in the rodent mammary gland (43, 101, 127), we conjecture that Esr1 is the ER isoform required for development, growth, and progression of mammary cancer in E2-treated ACI rats.

In addition to E2, multiple naturally occurring and synthetic estrogens induce mammary cancer development in ACI rats, including diethylstilbestrol, 17alpha-ethinylestradiol, estrone, equilin (an equine estrogen), estradiol-17beta-fatty acid esters (stearate and palmitate), and zeranol (38, 39, 70, 95, 110, 167). Diethylstilbestrol was widely used in the 1940s through 1960s, in the belief that treatment would inhibit spontaneous abortion, and has subsequently been associated with increased breast cancer risk in exposed mothers as well as their female offspring (28, 157). 17Alpha-ethinylestradiol is used in many oral contraceptive formulations, and estrone and equilin are major constituents of widely used postmenopausal hormone regimens, including Premarin and Prempro. Use of hormonal formulations containing these estrogens for contraception or alleviation of menopausal symptoms has been associated with increased risk of breast cancer in numerous studies (12, 23, 49, 56, 99, 124). Estradiol-17beta-fatty acid esters are naturally occurring metabolites of E2 whose relevance to breast cancer etiology in humans is not known. Zeranol is a nonsteroidal estrogen that is produced by specific fungal species, can be present in fungal-contaminated grains, and is used in the United States and elsewhere to promote growth of livestock, possibly leading to consumption by humans. Each of these agents acts as an agonist upon Esr1, further supporting a role of ER-mediated pathways in the etiology of mammary cancer in ACI rats.

Studies of human cohorts and rodent models indicate that progesterone is also an important hormonal contributor to breast cancer development (16, 24, 25, 97, 124). Among the most informative studies is the prospective, randomized, Women’s Health Initiative, which demonstrated that women using postmenopausal hormone regimens containing conjugated equine estrogens plus medroxyprogesterone acetate exhibit an approximate 30% increased risk of developing invasive breast cancer relative to women given a placebo (24, 124). Multiple observations suggest that induction of mammary cancer in ACI rats by E2 requires progesterone and Pgr. First, treatment of ovariectomized ACI rats with E2 alone fails to induce mammary cancer development, whereas tumors develop in ovariectomized rats treated with E2 plus progesterone (13, 139). Second, concurrent treatment of ACI rats with E2 and the Pgr antagonist mifepristone results in many fewer mammary tumors compared with rats treated with E2 alone (15). Third, the preneoplastic and neoplastic lesions induced by E2 in the mammary glands of ACI rats express a high level of Pgr (64). Taken together, these data indicate that the mechanism(s) leading to mammary cancer development in ACI rats requires the actions of both E2 and progesterone acting through their respective nuclear receptors.

Estrogens may also contribute to breast cancer development through mechanisms that are independent of ER action. One hypothetical ER-independent mechanism involves metabolism of estrogens through catecholestrogen intermediates to reactive estrogen quinones that adduct purine bases in DNA and generate mutagenic lesions through formation and subsequent misrepair of apurinic sites (21). This hypothesis is based on evidence from in vitro studies and studies of Esr1 knockout mouse models and is further supported by observations of elevated relative levels of estrogen-purine adducts in urine or serum of women with breast cancer or at high risk of developing breast cancer compared with women at low risk (21, 130, 164–166). However, those studies that have directly tested this hypothesis in rat models have not yielded supporting evidence. Specifically, continuous treatment of ACI rats with the catecholestrogens 2-hydroxyestradiol, 4-hydroxyestradiol, or 4-hydroxyestrone, the latter two of which are proposed to be direct precursors to the putative ultimate carcinogens estradiol-3,4-quinone and estrone-3,4-quinone, respectively, did not induce development of mammary cancer (158). Moreover, intramammary injection of estrone-3,4-quinone did not induce mammary cancer development in outbred Crl:CD (SD) rats (41). Thus, a causal role for estrogen quinones in mammary cancer etiology remains to be established in an in vivo model.

GENETIC BASES OF SUSCEPTIBILITY TO BREAST CANCER IN HUMANS AND E2-INDUCED MAMMARY CANCER IN RAT MODELS

Many factors contribute to breast cancer development. It is estimated that genetic factors contribute to 25–30% of overall breast cancer risk within the population, with environmental factors contributing to the remainder of population risk. A small fraction of genetic risk within the population is attributable to rare but moderately to highly penetrant mutations in a small number of extensively studied genes including BRCA1, BRCA2, and TP53 (115). By contrast, the majority of genetic risk appears to result from the combined actions of numerous common variants whose individual effects are weakly penetrant. Approximately 160 such breast cancer risk variants have been localized in genome-wide association studies (GWASs) (4, 5, 19, 30, 40, 48, 51–54, 92–94, 96, 140, 159). The sites (i.e., cell type) and mechanisms (i.e., target genes and biological processes influenced) of action of these common risk-determining variants remain unknown. Because of this lack of knowledge regarding the mechanisms through which common risk variants influence breast cancer development and the perception that their individual effects are small and perhaps not modifiable, their importance is often viewed with skepticism within the breast cancer research community. To illustrate the importance of these common variants, we calculated the impact of a representative risk-associated variant with a minor, risk-increasing, allele (T) frequency of 0.2035 and an observed effect size of 1.16 (i.e., GWAS odds ratio) using established methods (11, 140). The increased 5 yr risk for those individuals that are heterozygous or homozygous for the risk-associated minor allele, relative to individuals homozygous for the major allele, is clearly apparent by the rightward shift in the risk distribution curves (Fig. 3). Moreover, the calculated population attributable risk associated with this common variant indicates that 16,072 (6.4%) of the anticipated 252,710 cases of invasive breast cancer in the US female population in 2017 would not occur if the presumed causal effects of this allele could be eliminated. By contrast, the calculated population attributable risk associated with germline mutations in BRCA1, with a mutant allele frequency of 0.06% and an effect size of 9.08, indicates that 1,218 (0.48%) of the anticipated cases would not occur in the absence of these mutations (7).

Fig. 3.

Modeling the impact of a common breast cancer-associated variant on population risk. The Breast Cancer Surveillance Consortium (BCSC) Risk Calculator (11) is based on data from 661,068 women aged 35–74 yr of all racial/ethnicity groups enrolled in the BCSC cohort during 1996–2002; these underlying data were collected and shared by the National Cancer Institute-funded BCSC (HHSN261201100031C). The calculated average 5 yr absolute risk of breast cancer in this cohort = 1.31%. The hypothetical impact of a representative risk-associated SNP allele [rs2284378, risk-associated allele (T) frequency = 0.2035, associated odds ratio = 1.16, P = 1.1×10⁻⁸] was modeled by assigning genotype at this single nucleotide polymorphism randomly across the women in the BCSC cohort. The calculated average absolute risk in women homozygous for the risk associated (T/T) allele = 1.71%; risk in heterozygotes (T/C) = 1.48%; risk in women homozygous for the nonrisk-associated (C/C) allele = 1.29%. A list of the BCSC investigators and procedures for requesting BCSC data for research purposes are provided at: https://breastscreening.cancer.gov/.

To begin to define the genetic bases of susceptibility to E2-induced mammary cancer, we compared development of mammary cancer in multiple inbred rat strains (62, 131, 136, 138, 139, 150). ACI (AxC 9335, Irish) rats were chosen for study because of the groundbreaking research performed by Dunning and Curtis (36–38), who developed the ACI strain and were among the first to recognize that susceptibility to induction of mammary and other cancers is highly heritable. COP rats were studied because the ACI strain was developed from an intercross between COP and AUG (August) rats (69). DA rats were studied because it is believed within the rat genetics community that the DA strain was derived from an intercross between ACI and COP rats, a putative relationship that is strongly supported by a genome-wide evaluation of simple sequence length polymorphisms between the ACI, COP, and DA rat strains (155). BN rats were studied because they are genetically distant from ACI rats and served as the reference strain for the rat genome project (55, 155). Latency to appearance of palpable mammary cancer, the number of independent mammary cancers per rat, and incidence within the treated population were quantified in these studies. The resulting data indicated that ACI rats are the most highly susceptible to E2-induced mammary cancer among the four strains examined (Fig. 4). Mammary cancer incidence in E2-treated ACI rats approached 100% within 200 days of initiation of E2 treatment, and most ACI rats exhibited multiple, grossly discernable, mammary carcinomas. The majority of DA rats also developed mammary cancer when treated with E2, but only a single DA rat exhibited more than one cancer, indicating that DA rats are susceptible to E2-induced mammary cancer, but less so than ACI rats. Few COP rats developed mammary cancer, and those that did so rarely exhibited more than a single tumor. Interestingly, BN rats were refractory to induction of mammary cancer by E2 within the time interval examined in these studies. Together, these data indicate that the four rat strains examined span the full susceptibility continuum, where risk of developing mammary cancer in response to E2 treatment ranges from 0% for BN to virtually 100% for ACI. It is important to note that mammary cancers were not observed in sham-treated control rats from any of the four strains over the time course examined, indicating the importance of continuous elevation of circulating E2 to levels normally observed in humans during the periovulatory phase of the menstrual cycle or during pregnancy, which is a window of high susceptibility to breast cancer (133, 139, 150).

Fig. 4.

Strain differences in susceptibility to 17β-estradiol-induced mammary cancer. Data summarized in this figure illustrate that ACI rats are most highly susceptible to E2-induced mammary cancer. By contrast, genetically related COP rats are resistant and BN rats are virtually refractory to induction of mammary cancer by E2. DA rats, which appear to have been generated through an intercross between ACI and COP rats, exhibit a susceptible phenotype that is intermediate to that of ACI and COP rats. Data for ACI, BN, and COP rats are adapted from (27, 83, 131). Data for DA are unpublished. ACI, n = 126; BN, n = 13; COP, n = 7; DA, n = 13.

Reciprocal intercrosses were performed between susceptible ACI rats and resistant COP or BN rats to define the manner in which susceptibility segregates in the resulting F1 and F2 progeny. In both pairs of intercrosses the large phenotypic variations exhibited by the F2 populations, relative to that exhibited by the parental strains, suggested that susceptibility to E2-induced mammary cancer behaves as a complex (i.e., multigene) genetic trait (62, 131, 138). Moreover, the F1 populations generated in each pair of intercrosses [e.g., (ACI×COP)F1 and (COP×ACI)F1, female in each cross designated first] exhibited similar susceptibilities. These latter data strongly suggest that susceptibility to E2-induced mammary cancer in the rat models is neither mediated nor influenced by milk-borne mammary tumor viruses passed from a susceptible ACI dam to her offspring, in contrast to what has been documented in specific inbred mouse strains (123). Linkage analyses were performed on the different F2 populations to map the quantitative trait loci (QTL) that influence susceptibility to mammary cancer. Estrogen-induced mammary cancer 1 (Emca1) and Emca2 were mapped to rat chromosomes 5 (RNO5) and RNO18, respectively, in the reciprocal intercrosses between susceptible ACI rats and resistant COP rats (62). Emca3 through Emca9 were mapped to RNO2, RNO7, RNO3, RNO4, RNO6, RNO5, and RNO18, respectively, in the intercrosses between ACI rats and resistant BN rats (131, 136). The linkage data indicated that COP or BN alleles at all of the Emca loci except Emca7 reduced susceptibility to E2-induced mammary cancer; by contrast, BN alleles at Emca7 increased susceptibility (62, 131). Congenic rat strains were subsequently developed that harbor alleles from resistant COP or BN rats solely across an individual Emca locus introgressed onto the ACI genetic background (27, 132, 136). Characterization of these congenic strains confirmed the existence of the mapped Emca QTL as well as the direction of its actions on susceptibility to E2-induced mammary cancer.

Multiple GWAS were published subsequent to the mapping of Emca1 through Emca9, allowing for the first time a direct comparison between the rat Emca QTL and breast cancer risk loci mapped in human populations. Interestingly, this comparison indicated the peak logarithm of the odds ratio region for several of the Emca loci is orthologous to a region of the human genome that has more recently been associated with breast cancer risk in one or more GWAS (27, 129, 132). The physical relationships between the rat and human loci are illustrated in Table 1. Moreover, three of the Emca loci are orthologous to loci that harbor variants that influence breast mammographic density, an established biomarker of breast cancer risk (27). Together, these genetic data further illustrate the relevance of these rat models of E2-induced mammary cancer to breast cancers in humans. We are continuing to use the power of genetically defined congenic and knockout rat models to further localize the genetic determinants of mammary cancer susceptibility and to elucidate the mechanisms through which the causal variants residing within specific Emca loci influence mammary cancer development.

Table 1.

Physical relationships between Emca mammary cancer susceptibility loci in rats and GWAS breast cancer risk loci in humans

Locus	Marker	Proximal Marker	Distal Marker	Phenotype(s)	Reference	Orthologous Risk Locus	Lead SNP in Locus	Reported Gene(s)	Orthologous Region in Rat, Mb	Reference
Emca1	D5Rat30	D5Rat53*	D5Rat57*	incidence	62	1p34.2	rs4233486	intergenic	139.55	94
	133.78 Mb	102.95 Mb	154.73 Mb	latency		1p34.1	rs1707302	PIK3R3,	135.07	94
								LOC101929626
						9p21.3	rs1011970	CDKN2A, CDKN2B	107.90	92, 94, 159
						9q31.2	rs865686	KLF4, ACTL7A, RAD23B	72.28	48, 92, 93
Emca2	D18Rat21	D18Rat27*	D18Rat43*	incidence	62	5q22.1	rs6882649	NREP	26.19	94
	43.54 Mb	19.03 Mb	68.44 Mb	latency
Emca3	D2Rat16	D2Rat251†	D2Mgh3†	multiplicity	27, 131	5p13.3	rs2012709	intergenic	62.05	92, 94
	43.64 Mb	6.2 Mb	63.5 Mb			5p12	rs4415084	intergenic	51.17	48
						5q11.1	rs35951924	intergenic	49.00	94
						5q11.2	rs889312	MAP3K1	43.45	5, 40, 92, 93, 159
						5q14.2	rs7707921	ATG10	19.85	92, 94
						5q14.3	rs10474352	ARRDC3	8.73	19, 94
Emca4	D7Rat19	D7Rat44*	D7Rat15*	latency	27, 131	8q22.3	rs514192	intergenic	76.04 Mb	94
	101.92 Mb	69.40 Mb	111.04 Mb	multiplicity		8q23.1	rs12546444	ZFPM3	79.50	94
						8q23.3	rs13267382	LINC00536	90.79	92, 94
						8q24.13	rs58847541	intergenic	98.15	94
						8q24.21	rs13281615	intergenic	101.78	40, 92, 93, 94
Emca5	D3Rat114	D3Rat80†	D3Rat3†	latency	27, 131	2q13	rs71801447	BCL2L11	120.76	94
	155.26 Mb	36.60 Mb	169.7 Mb	multiplicity		2q31.1	rs2016394	METAP1D, DLX1, DLX2	58.08	93, 94
						20p12.3	rs16991615	MCM8	125.48	94
						20q11.22	rs2284378	RALY, EIF2S2, ASIP	150.33	140
Emca6	D4Rat103	D4Rat14*	D4Rat202*	latency	27, 131	3p26.1	rs6762644	ITPR1, EGOT	140.44	92, 93, 94
	84.48 Mb	43.41 Mb	155.47 Mb	multiplicity		3p13	rs6805189	FOXP1	131.37	94
						7p15.1	rs17156577	CREB5	83.39	94
						7q32.3	rs4593472	FLJ43663	58.42	92, 94
						7q35	rs720475	ARHGEF5, NOBOX	72.73	92, 93, 94
Emca7	D6Rat22	D6Rat68*	D6Rat81*	multiplicity	27, 131	2p25.1	rs113577745	GRHL1	43.79	94
	76.00 Mb	2.54 Mb	111.72 Mb			2p24.1	rs12710696	intergenic	35.57	30, 51, 94
						2p23.3	rs6725517	ADCY3	28.57	94
						2p23.2	rs67073037	WDR43	23.47	30
						14q13.3	rs2236007	PAX9, SLC25A21	77.61	92, 93, 94
						14q24.1	rs999737	RAD51B, RAD51L1	103.11	92, 93, 94, 96
Emca8	D5Rat95	D5Rat134*	D5Rat37*	latency	31	1p34.2	rs4233486	intergenic	139.55	94
	128.93 Mb	51.42 Mb	147.49 Mb	multiplicity		1p34.1	rs1707302	PIK3R3, LOC101929626	135.07	94
						9p21.3	rs1011970	CDKN2A, CDKN2B	107.90	92, 94, 159
						9q31.2	rs865686	KLF4, ACTL7A, RAD23B	72.28	48, 92, 93
						9q33.1	rs1895062	ASTN2	81.29	94
Emca9	D18Rat30	D18Rat133†	D18Rat89†	multiplicity	27	5q22.1	rs6882649	NREP	26.19	94
	5.83 Mb	3.4 Mb	62.9 Mb			18q11.2	rs527616	intergenic	6.69	92, 93, 94
						18q12.1	rs36194942	CDH2	8.15	96

Marker column shows nearest LOD (logarithm of the odds ratio) peak for QTL and its physical location (Rnor_v.6.0). Lead SNP in locus shows the SNP that achieved genome wide significance (P value ≤ 5×10⁻⁸) and exhibited smallest P value. Orthologous Region in Rat shows coordinate on indicated rat chromosome (Rnor_v.6.0) that is orthologous to region harboring risk-associated SNP. GWAS, genome-wide association study; QTL, quantitative trait locus; SNP, single nucleotide polymorphism.

^*Markers define QTL 95% confidence interval.

^†Markers define QTL congenic interval.

LINKING GENETIC DETERMINANTS OF SUSCEPTIBILITY TO THE BIOLOGY OF THE NORMAL AND NEOPLASTIC MAMMARY GLAND

Approaches for studying the biology of breast cancer risk variants in human populations are limited by many factors, including genetic heterogeneity, lack of access to tissues for molecular and biochemical studies, and ethical considerations. The utility of human cell lines for studying breast cancer risk variants is restricted by lack of knowledge regarding the sites of action of the variants (i.e., cell or tissue types) as well as the likelihood of interactions between multiple variants, and/or interactions between variants and environmental factors. Physiologically relevant rat models overcome these limitations by allowing one to study the actions of risk variants in a uniform genetic background, evaluate interactions between variants, control for influences of the environment, and to extend, when necessary, studies to tissues and organ systems beyond the mammary gland.

To begin to define the functional relationships between the Emca variants and mammary gland biology, we extensively evaluated the manner in which the mammary glands of susceptible ACI and resistant BN rats respond to E2 (35). The data from this study revealed a set of molecular and cellular phenotypes in normal mammary gland that: 1) are rapidly induced by E2; 2) differ quantitatively between E2-treated ACI and BN rats; and 3) are heritable. Phenotypes that are prominent in mammary glands of E2-treated ACI rats, but are absent in BN rats, include a high rate of proliferation within the luminal epithelium, which results in a “dense” epithelium (i.e., high number of epithelial cells per unit volume) that is apparent upon quantitative evaluation of mammary gland whole mounts and thin sections as well as upon radiographic evaluation (Fig. 5). Phenotypes that are present in treated BN rats, but are virtually absent in ACI rats, include pronounced luminal ectasia, defined as grossly dilated mammary ducts containing secreted milk proteins, and associated changes in the adjacent collagenous extracellular matrix (ECM). Our unpublished data indicate that several of these histologically discernable phenotypes are controlled by genetic variants residing within specific Emca loci, suggesting that these phenotypes reflect biological processes that underlie susceptibility to E2-induced mammary cancer.

Fig. 5.

Identification of phenotypes in normal mammary gland that are associated with susceptibility or resistance to 17β-estradiol-induced mammary cancer. All images are derived from sham-treated (Ctrl) or 3 wk E2-treated (E2) ACI or BN rats. Detailed information on methods and quantitative data for the different phenotypes are presented in (35). Row 1: carmine stained mammary gland whole mounts illustrate dramatic induction of lobuloalveolar hyperplasia in E2-treated ACI rats, with modest induction of this phenotype in treated BN rats. Row 2: H&E-stained sections illustrate increased mammary epithelial cell number per field (i.e., high epithelial density) in E2-treated ACI rats compared with treated BN rats. Luminal ectasia is apparent in mammary glands of E2-treated BN rats. Row 3: immunostained mammary tissues illustrate dramatic induction of luminal epithelial cell proliferation in E2-treated ACI rats, relative to treated BN rats. Cell nuclei were identified by staining DNA with 4',6-diamidino-2-phenylindole (DAPI, blue); basal epithelial cells were identified by immunostaining for cytokeratin 5 (K5, green); luminal epithelial cells were identified by immunostaining for cytokeratin 8 (K8, red); cells transiting S phase in 4 h window preceding sacrifice were identified by immunostaining for BrdU (yellow). Row 4: immunostained mammary tissues illustrate expression of milk proteins (green), with prominent luminal ectasia present in E2-treated BN rats, but absent in treated ACI rats. Rows 5 and 6: Sirius red-stained tissues illustrate prominent ECM surrounding ectatic ducts of E2-treated BN rats that is not apparent in treated ACI rats. Row 7: second harmonic imaging microscopy illustrates prominent collagenous stroma associated with mammary ducts in E2-treated BN rats that is less apparent in treated ACI rats. Mammary glands from sham-treated ACI and BN rats did not differ discernably from one another in any of the examined phenotypes.

Gene expression profiles have been generated for whole mammary glands of E2-treated ACI and BN rats. Evaluation of these data revealed differential expression of multiple genes that may influence luminal epithelial cell proliferation, luminal ectasia, and ECM deposition observed upon comparison of these rat strains (35). Expression of Pgr, Wnt4, Tnfsf11 (RankL), Areg, and Gata3 was higher in mammary glands of E2-treated ACI rats compared with identically treated BN rats (Fig. 6). Each of these genes encodes a protein product that plays a defined role in mammary gland development. For example, progesterone acts through Pgr to stimulate lobuloalveolar development during pregnancy (18, 29, 72), whereas Wnt4 and Tnfsf11 function downstream of Pgr in promoting lobuloalveolar development and are requisite paracrine mediators of the actions of progesterone in the regulation of mammary stem cell number (9, 17, 44, 78). Spp1 is among the set of genes that was highly expressed in mammary glands of E2 treated BN rats relative to ACI rats. Spp1 encodes a secreted phosphoprotein (aka osteopontin) that is highly expressed in the mammary gland during lactation and involution and is also more highly expressed in mammary glands of parous mice and rats relative to nulliparous controls (31, 104, 116, 135, 160). Inhibition of Spp1 expression in the luminal epithelium of the mouse mammary gland inhibits lobuloalveolar development, expression of genes encoding milk proteins, and milk production (105). BN alleles at Emca8 decrease susceptibility to E2-induced mammary cancer, decrease expression of Pgr, and increase expression of Spp1, suggesting functional relationships between Emca8 and genes that control epithelial cell proliferation and differentiation (132).

Fig. 6.

Genes involved in regulation of mammary gland development are differentially expressed in estradiol-treated ACI and BN rats. Gene expression in normal mammary gland from ACI and BN rats was evaluated following 12 wk of E2 treatment using Affymetrix Rat 230.2 GeneChips. Unsupervised clustering was performed using a subset of the genes that were differentially expressed between E2-treated susceptible ACI rats and treated resistant BN rats.

MODIFICATION OF MAMMARY CANCER DEVELOPMENT IN ACI RATS BY DIETARY FACTORS

As noted above, genetic factors are estimated to contribute to 25–30% of overall breast cancer risk with environmental factors (i.e., nongenetic) contributing to the remaining 70–75% of population risk. Substantial evidence suggests that diet is a significant environmental determinant of breast cancer risk, and considerable attention has been focused on identifying specific dietary factors that influence breast cancer risk in humans as well as mammary cancer development in rodent models. However, only a few strong associations between a specific dietary factor and breast cancer risk have been established. A recent report from the Continuous Update Project based on extensive review and evaluation of published epidemiologic data provides strong evidence that consumption of alcohol increases risk of breast cancer in premenopausal and postmenopausal women. This report also provides limited evidence to suggest consumption of nonstarchy vegetables or consumption of foods that contain carotenoids may lower risk of breast cancer in both premenopausal and postmenopausal women (29a).

The impact of ethanol consumption on E2-induced mammary tumorigenesis has not been evaluated in the ACI rat model. However, the impact of specific plant-derived dietary constituents on induction of mammary cancer in ACI rats by E2 has been evaluated in multiple studies. Phenotypes examined include latency to appearance of palpable cancer, tumor incidence, tumor number per animal, mean tumor volume, and/or a variety of biochemical and molecular parameters. Multiple dietary constituents have been observed to inhibit E2-induced mammary tumorigenesis in ACI rats including blueberries, black raspberries, and jamun berries (freeze-dried/powdered in diet) (6, 8, 75, 121); mixed gamma-tocopherols (0.3% in diet) (32); vitamin C (1% in drinking water) (91); and the common food preservative butylated hydroxyanisole (0.7% in diet) (144, 146). Other dietary constituents were observed to inhibit development of E2-induced mammary cancer when administered from subcutaneous implants, including curcumin (10), ellagic acid (102, 161), and resveratrol (147). Consumption of soy-derived isoflavones yielded paradoxical effects on different mammary cancer phenotypes: tumor incidence and multiplicity were reduced, whereas latency to appearance of palpable cancer was shortened, and total tumor volume was increased (98). Interestingly, consumption of quercetin, a plant-derived flavonol, shortened latency to appearance of palpable mammary cancer in E2-treated ACI rats without impacting tumor incidence or multiplicity (145).

A common property of the different dietary factors evaluated in the studies reviewed here is their abilities to function as antioxidants, and several of these studies demonstrated actions of a specific dietary constituent on biomarkers related to oxidative stress (32, 33, 91, 141, 142, 144, 148). One mechanism suggested to underlie the protective actions of those dietary factors that inhibited mammary tumorigenesis is induction of expression of nuclear factor erythroid 2-like 2 (NFE2L2; aka NRF2) and downstream genes that encode enzymes that eliminate reactive oxygen species [e.g., superoxide dismutase 3 (SOD3), NAD(P)H quinone dehydrogenase 1 (NQO1) and catalase] or repair the 8-hydroxyguanine lesions generated by reactive oxygen species [e.g., 8-oxoguanine DNA glycosylase (OGG1)]. Although these studies collectively support the hypothesis that estrogens may contribute to mammary cancer development through generation of oxidative stress, the mechanisms through which, and the cell types in which, E2 induces oxidative stress and possible downstream genotoxicity remain to be elucidated. In this regard, it is often suggested that the estrogens estradiol and estrone are metabolized to estrogen quinones coupled with generation of reactive oxygen species from redox cycling between their respective quinone and semiquinone forms (50, 74, 130, 166). However, it remains to be established whether redox cycling between estrogen quinones/semiquinones is a source of oxidative stress in vivo and whether or not this putative source of reactive oxygen species contributes to development of breast cancer in humans or rodent models. Because mitochondria are the prevalent source of reactive oxygen species, we suggest that the oxidative stress that appears to occur in the mammary gland upon chronic treatment of ACI rats with E2 arises within the mitochondria. Production within the mitochondria of superoxide, conversion of superoxide to hydrogen peroxide (H₂O₂), metabolism of H₂O₂ to H₂O, and release of H₂O₂ from mitochondria are regulated processes that are strongly influenced by the metabolic state of the cell as well as by genetic, epigenetic, and environmental factors (103, 125). H₂O₂ also functions as an important signaling molecule through its ability to posttranslationally modify enzymes, transcription factors, and other proteins that contain reactive cysteine residues and thereby regulate cell proliferation, cell fate determination, and other biological processes relevant to breast cancer etiology (22). Extensive data from studies of rodent models indicate dietary energy restriction extends lifespan, inhibits development of multiple cancer types, and regulates mitochondrial function, including production and/or release of reactive oxygen species from mitochondria. A 40% restriction of dietary energy consumption markedly inhibits induction of mammary cancer in E2-treated ACI rats; latency to appearance of palpable cancer was prolonged, incidence was reduced, tumor burden was reduced, and proliferation of mammary luminal epithelial cells was reduced in animals maintained on the energy-restricted diet relative to freely fed controls (65, 66, 149). We postulate that dietary energy restriction may phenocopy the inhibitory actions of dietary antioxidants on development of E2-induced mammary cancers by impacting production or release of mitochondrial-derived reactive oxygen species.

In addition to the studies cited above that demonstrate inhibitory actions of specific dietary constituents on induction of mammary cancer in ACI rats by E2, two studies have evaluated the impact of dietary constituents on induction of mammary cancer in ACI rats by synthetic estrogens. Retinyl acetate, a widely used vitamin A supplement, partially inhibits induction of mammary cancer by 17alpha-ethinylestradiol when provided in the diet (412,000 IU/kg diet) (70). As noted above, consumption of foods that contain carotenoids, some of which are precursors to vitamin A, may lower risk of breast cancer in both premenopausal and postmenopausal women (29a). Caffeine, administered (1 or 2 mg/ml) in drinking water, dramatically inhibits induction of mammary cancer in ACI rats treated with the synthetic estrogen diethylstilbestrol (114). Dose-dependent inhibitory actions of caffeine were observed on latency to appearance of palpable mammary cancer, mammary cancer incidence, and tumor multiplicity. Some, but not all, retrospective and prospective epidemiologic studies suggest that consumption of caffeine/coffee may reduce breast cancer risk (42, 77, 109). Together, the studies cited herein illustrate the value of the ACI rat model for defining mechanisms through which dietary and other environmental factors modify breast cancer risk.

GENETIC BASES UNDERLYING OTHER ESTROGEN-INDUCED BIOLOGICAL RESPONSES AND PATHOLOGIES

Prolactin-producing pituitary adenomas (i.e., prolactinoma) are a common tumor type in humans, and estrogens have been implicated in their etiology (57, 68, 81, 100, 120, 134). Moreover, prolactin has been suggested to contribute to breast cancer development in humans and rodent models (108, 156). ACI rats uniformly develop pituitary lactotroph hyperplasia and/or adenoma when treated for long durations with estrogens, including E2 and diethylstilbestrol (71, 139, 153). These pituitary tumors result in marked enlargement of the pituitary gland and hyperprolactinemia and can lead to undesired morbidity and mortality in long-term studies. COP rats, which are resistant to E2-induced mammary cancer, also develop lactotroph hyperplasia/adenoma when treated with estrogens, albeit to a quantitative lesser extent than ACI rats (150–152). Estrogen-treated DA rats, which develop mammary cancer in response to E2 treatment, and BN rats, which are refractory to E2-induced mammary cancer, are both highly resistant to induction of pituitary lactotroph hyperplasia/adenoma (151). This observed lack of concordance between susceptibility to mammary cancer and susceptibility to pituitary lactotroph hyperplasia/adenoma strongly suggests that pituitary tumor-associated hyperprolactinemia is not a primary driving force in development of E2-induced mammary cancers in these rat models. Further supporting this conjecture are reports indicating that caffeine and dietary energy restriction each dramatically inhibit induction of mammary cancers in estrogen-treated ACI rats but do not inhibit induction of pituitary lactotroph hyperplasia/adenoma and associated hyperprolactinemia (66, 114). Studies were performed to define the genetic bases of the differing susceptibilities of the ACI, COP and BN rat strains to estrogen-induced pituitary lactotroph hyperplasia/adenoma (82, 83, 137, 152, 154). Multiple Estrogen-induced pituitary tumor (Ept) QTL were mapped in intercrosses between ACI and COP or BN rats. Interestingly, these Ept QTL exhibit little physical overlap with the Emca QTL that harbor genetic determinants of susceptibility to E2-induced mammary cancer, indicating that, for the most part, distinct sets of genetic variants determine susceptibility to these two estrogen-induced neoplasms. Based on these genetic data, a novel rat strain, ACWi, that retains the unique susceptibility of ACI rats to E2-induced mammary cancer but lacks the morbidity and mortality resulting from pituitary lactotroph hyperplasia/adenoma and associated hyperprolactinemia was developed through selectively breeding COP alleles at two of the major Ept QTL onto the ACI genetic background (34). The ACWi strain appears to be particularly well suited for studies focused on evaluating dietary and pharmacologic-based strategies for breast cancer prevention.

In some inbred rat strains, continuous estrogen treatment induces uterine inflammation (i.e., pyometritis), which when severe can result in morbidity and mortality in long-term studies. This phenotype is common in treated BN rats and severe in treated DA rats but is not observed in treated ACI or COP rats. Moreover, this phenotype is heritable in crosses between BN rats and ACI rats. QTL that harbor variants that influence development of pyometritis were mapped to RNO5 in crosses between susceptible BN rats and resistant ACI and F344 rats (59, 112).

It has long been recognized that estrogens act through ER-dependent mechanisms to regulate B and T lymphocyte homeostasis. Regulation of lymphopoiesis occurs at least in part through inhibition of thymocyte proliferation within the thymic cortex and corticomedullary boundary (60). This biological response to estrogens is rat strain specific; for example, male BN rats exhibit a more robust antiproliferative response to diethylstilbestrol relative to male ACI rats. Estrogen-induced thymic atrophy 1 (Esta1) was mapped to RNO10, and Esta2 and Esta3 were mapped to distinct regions of RNO2 in reciprocal intercrosses between ACI and BN rats (61). These Esta QTL did not influence induction by diethylstilbestrol of pituitary lactotroph hyperplasia/adenoma in the F2 progeny from these same intercrosses. These data, together with data presented above, clearly indicate that the biological actions of estrogens in different cell and tissue types are usually rat strain specific and behave as complex genetic traits controlled by distinct sets of genetic variants.

CONCLUDING STATEMENTS AND UNANSWERED QUESTIONS

The studies summarized herein indicate that the ACI rat model of E2-induced mammary cancer is highly relevant to understanding the etiology and progression of breast cancer in humans, most particularly the luminal breast cancer subtypes. Mammary cancer development in ACI rats is driven by sustained elevation of circulating E2 and requires the actions of progesterone; both of these steroid hormones are strongly implicated in development of breast cancers in humans. Moreover, published and emerging data indicate that the Emca QTL that influence development of E2-induced mammary cancer in the rat models are orthologous to breast cancer risk loci identified in GWASs of human populations, strongly suggesting that the genes and biological processes that determine susceptibility to breast cancer are conserved between the rat and human species. In addition to these shared endocrine and genetic bases of tumorigenesis, the cancers that develop in E2-treated ACI rats exhibit somatic genome changes that mirror those most frequently observed in luminal breast cancers, further illustrating the relevance of the rat models to humans. Recent whole exome and whole genome sequence analyses of large numbers of breast cancers provide a catalog of somatic mutations that likely drive development of the different breast cancer subtypes (20, 106). No such catalog of somatic mutations has been generated for cancers induced in the rat models by E2. Therefore, the full extent to which the human and rat breast cancer somatic mutation spectra overlap is not known.

GWASs have localized more than 160 common breast cancer risk-associated genetic variants within the human genome. As noted above, published and emerging data indicate that a subset of these mapped risk loci are orthologous to the Emca loci that harbor genetic determinants of susceptibility to E2-induced mammary cancer in the rat models, strongly suggesting that the genetic bases of susceptibility to breast cancer are conserved across species. At this time, the causal genetic variants within the mapped risk loci that influence breast cancer development are not known for either human or rat. Because the causal genetic variant(s) that resides within a specific risk locus may differ between individuals in the human population and probably differ between humans and rats, it is less important to identify the causal variant than it is to identify the evolutionarily and functionally conserved genetic element, be it gene or regulatory, and downstream biological process through which the causal variant exerts its effect. The rat models of E2-induced mammary cancer described herein are ideally suited to this task.

One novel and significant insight emanating from studies of the rat models is that the mammary glands of susceptible and resistant rats exhibit dramatic differences in their biological responses to E2 that are revealed as multiple quantifiable molecular and cellular phenotypes. Moreover, data emerging from ongoing studies indicate that these strain-specific responses are heritable and controlled by genetic variants that reside within the Emca loci, suggesting these differences in responsiveness to E2 are directly related to differences in susceptibility to E2-induced mammary cancer. Based on these findings, we hypothesize that a significant fraction of heritable risk of breast cancer in humans results from genetic differences in how specific cell populations within the breast respond to estrogens and progestins. Because the susceptibility- and resistance-associated phenotypes identified in the rat models can be quantified by histological and/or molecular analyses, it may be possible to evaluate the relationships between expression of these phenotypes in human breast and breast cancer risk. The rat models are ideally suited for guiding comparative studies in humans to firmly establish a causal relationship between genetically controlled responsiveness to estrogens and progestins and development of breast cancer.

Another novel and significant insight provided by studies of the rat models of E2-induced mammary cancer is that the constitution of the germline influences the somatic genetic aberrations observed in the cancers. Although this is currently best established for CNAs, this relationship between the germline and somatic genomes probably extends to other types of somatic changes including missense mutations and structural variation. This relationship implies that common germline determinants of susceptibility shape the spectrum of somatic alterations required for cancer to develop. Because the somatic changes collectively influence the biological behavior of the cancer, a better understanding of the relationship between the germline and somatic genomes should ultimately enhance our understanding of the mechanisms that drive development and progression of the different breast cancer subtypes and lead to improved strategies for using genomic information to guide treatment of breast cancers.

A third insight gained from the rat models is that development of E2-induced mammary cancer is strongly modified by a variety of environmental factors, including consumption of antioxidants, caffeine, and energy (i.e., total calories consumed). One implication of this observation is that the actions of the genetic variants that make the ACI rat uniquely susceptible to mammary cancer can be overridden by environmental factors. The mechanisms responsible for modification of E2-induced mammary tumorigenesis by these particular environmental factors are not fully understood but appear to be related, at least in part, to generation of reactive oxygen species and oxidative stress. Full elucidation of these mechanisms and their underlying gene-environment interactions will enhance our understanding of the roles of estrogens in breast cancer etiology and may lead to improved strategies for preventing breast cancer.

Many environmental factors, including diet, exogenous hormones, and hormone mimetics, influence breast cancer risk in humans and rodent models. Data summarized herein indicate that risk of breast cancer is also determined by many common genetic variants. Because many environmental factors and genetic variants, as well as interactions between these factors and variants, influence breast cancer development, it is challenging to use genetic and environmental exposure data to accurately predict individual risk. Based on our observations that multiple genetic determinants of susceptibility to E2-induced mammary cancer converge to regulate specific susceptibility- and resistance-associated phenotypes, we hypothesize that the myriad of factors that influence breast cancer development do so by acting upon a relatively small number of biological processes in normal breast. If this hypothesis is correct, then quantification of well-defined susceptibility- and resistance-associated phenotypes (i.e., function-based biomarkers) in normal breast may provide the means to more accurately predict risk.

Each of the inbred rat models of susceptibility and resistance to E2-induced mammary cancer described herein is thought to be representative of a single individual within the genetically diverse human population. The heterogeneity exhibited by these strains in their biological responsiveness to E2 and their relative susceptibility to E2-induced mammary cancer can be exploited by performing mechanistic studies to define how estrogens and other hormones contribute to mammary cancer development and how different environmental factors influence mammary gland biology and cancer development in both susceptible and resistant models. Because of the high degree of relevance of these models to the etiology of breast cancer, the information generated in studies using these models is likely to be translatable to humans.

GRANTS

Support for the studies described herein was provided by National Institutes of Health Grants R01-CA-077876, R01-CA-204320, P30-CA-014520, and U01-ES-026127; Susan Komen for the Cure Grant KG081343; and Department of Defense Breast Cancer Research Program Grant W81XWH-11-1-0175.

DISCLOSURES

The authors have no conflicts of interest to disclose.

AUTHOR CONTRIBUTIONS

J.D.S. conceived and designed research; J.D.S., K.L.D., and A.C.C. analyzed data; J.D.S., K.L.D., and A.C.C. interpreted results of experiments; J.D.S., K.L.D., and A.T.-D. prepared figures; J.D.S. drafted manuscript; J.D.S., K.L.D., A.C.C., and A.T.-D. edited and revised manuscript; J.D.S., K.L.D., A.C.C., and A.T.-D. approved final version of manuscript; K.L.D. and A.C.C. performed experiments.