Environmental Exposures and Mammary Gland Development: State of the Science, Public Health Implications, and Research Recommendations

Objectives: Perturbations in mammary gland (MG) development may increase risk for later adverse effects, including lactation impairment, gynecomastia (in males), and breast cancer. Animal studies indicate that exposure to hormonally active agents leads to this type of developmental effect and related later life susceptibilities. In this review we describe current science, public health issues, and research recommendations for evaluating MG development. Data sources: The Mammary Gland Evaluation and Risk Assessment Workshop was convened in Oakland, California, USA, 16–17 November 2009, to integrate the expertise and perspectives of scientists, risk assessors, and public health advocates. Interviews were conducted with 18 experts, and seven laboratories conducted an MG slide evaluation exercise. Workshop participants discussed effects of gestational and early life exposures to hormonally active agents on MG development, the relationship of these developmental effects to lactation and cancer, the relative sensitivity of MG and other developmental end points, the relevance of animal models to humans, and methods for evaluating MG effects. Synthesis: Normal MG development and MG carcinogenesis demonstrate temporal, morphological, and mechanistic similarities among test animal species and humans. Diverse chemicals, including many not considered primarily estrogenic, alter MG development in rodents. Inconsistent reporting methods hinder comparison across studies, and relationships between altered development and effects on lactation or carcinogenesis are still being defined. In some studies, altered MG development is the most sensitive endocrine end point. Conclusions: Early life environmental exposures can alter MG development, disrupt lactation, and increase susceptibility to breast cancer. Assessment of MG development should be incorporated in chemical test guidelines and risk assessment.

The trend toward earlier breast develop ment initiation in U.S. girls (Euling et al. 2008) may put them at increased risk of later life out comes such as breast cancer-already the most common cancer in U.S. women (American Cancer Society 2010) and a leading cause of death for U.S. women in midlife . Although many factors, such as nutri tional status and body size, may contribute to maturation trends (Kaplowitz 2008) and breast cancer (Renehan et al. 2008), environmen tal chemicals have been hypothesized to con tribute as well (Birnbaum and Fenton 2003;Brody et al. 2007;Euling et al. 2008). Animal studies demon strate that early life exposure to hormonally active agents can lead to effects on mammary gland (MG) develop ment, impaired lactation, and increased susceptibility to can cer (Fenton 2006). However, the influence of environmental exposures on breast develop ment outcomes is poorly understood, as is the relationship between breast develop ment, lactational deficits, and breast cancer. Few chemicals coming into the marketplace are evaluated for these effects. The findings in ani mal studies raise concerns that perturbations to human breast develop ment may increase the risk for later life adverse effects including lactation impairment, gynecomastia (in males), and breast cancer in either sex.
This review is the result of the Mammary Gland Evaluation and Risk Assessment Workshop held in Oakland, California, USA, on 16-17 November 2009 to review and discuss recent findings of altered MG develop ment after gestational/perinatal expo sure to certain endocrinedisrupting chemi cals (EDCs). Workshop participants included research scientists from multiple disciplines, public health advocates, and risk assessors. Many of the participants are leading interna tionally recognized MG experts.
The goal of the workshop was to improve assessment of MG develop mental end points and their integration into human health risk assessment. Workshop participants discussed current research on the effects of develop mental exposures to EDCs on MG develop ment, the relationship of these effects to later life lactation and cancer outcomes, relative sensitivity of MG develop ment and other develop mental reproductive end points, rele vance of effects in animal models to humans, and MG assessment in current toxicology pro tocols. Data gaps and research recommen dations were identified at the workshop and through interviews with 18 risk assessors and toxicologists. Presentation slides and other workshop materials are available online (Silent Spring Institute 2011). In conjunction with the workshop, experts from seven labora tories in Canada, the United States, and Argentina participated in a roundrobin evaluation of MG whole mounts.

Factors Affecting Mammary Gland Development
The female MG undergoes most of its develop ment postnatally, achieving a fully differentiated state late in pregnancy. This process includes numerous events that can be disrupted by exposure to EDCs. Gestation, puberty, and pregnancy are the critical periods during which EDC exposure may most affect MG develop ment (Fenton 2006). Critical events include mammary bud develop ment in the fetus, exponential epithelial outgrowth during puberty, and the rapid transition to lactational competency that occurs during late pregnancy (Figure 1). These stages occur in both rodents and humans.
Normal MG develop ment. Normal female MG develop ment involves a wellorchestrated sequence of events marked by extensive pro liferation at puberty and by proliferation and differentiation during pregnancy. This process is regulated by hormones, growth factors, and stromal factors and is similar between rodents and humans, although rodent MG develop ment is more completely described (Kleinberg et al. 2009;Medina 2005). Female human MG develop ment begins with budding and branching between 6 and 20 weeks of gesta tion, yielding at birth a primitive gland com posed of ducts ending in ductules. During childhood, MG growth keeps pace with overall body growth; at puberty it accelerates dramatically.
In rodents, the epithelial bud is formed at the site of the nipple around gestation days (GDs) 12-16, and by birth the epithelium has entered the fat pad and formed a ductal tree. The fat pad and mammary epithelium grow at the same pace as the body for the first 2-3 weeks of life, and just before puberty, an exponential growth phase begins. In rodents, and presumably in humans, this phase of duc tal develop ment is characterized by formation of terminal end buds (TEBs), which lead the epithelial extension through the fat pad, leav ing behind a network of branched ducts. After the fat pad is filled, TEBs differentiate into terminal ductal structures, namely, terminal ductal lobular units in humans, lobules and alveolar buds in rats, and terminal ducts in mice. In humans and rodents, additional MG proliferation and regression events occur with each luteal phase of the ovulatory cycle, and at pregnancy there is significant differentia tion of the terminal structures with lobularalveolar develop ment (Kleinberg et al. 2009;Russo 2004a, 2004b). Mammary epithelial growth also occurs in male rats and men, whereas male mice lack mammary epi thelium. Male mice and rats do not normally possess nipples because androgens during ges tation induce regression. Retained nipples in male rats is a charac teris tic effect of pre natal anti androgen exposure (Foley et al. 2001).

Whole mounts and other techniques.
Early life treatment with some hormonally active agents results in altered develop ment of the MG in male and female rodents. Although laboratories vary in their methods for report ing altered MG develop ment, the primary approach has been morphological assess ment of the entire fourth or fifth abdominal MG fat pad mounted flat on a slide, fixed, stained, defatted, and permanently affixed to the slide as a "whole mount." Whole mounts allow an assessment of total and rela tive abundance of mammary terminal ductal structures (i.e., TEBs, terminal ducts, alveolar buds, and lobules), extension of the epithelial cells through the fat pad, and branching pat terns and density at different times during develop ment [e.g., Fenton et al. 2002; see also Supplemental Material,  (Russo and Russo 1978).
Several rodent studies have reported altered MG develop ment after pre natal, neo natal, or Figure 1. Stages of normal rat MG development and effects of environment on subsequent events. Effects of early life EDC exposures can lead to altered developmental programming in the breast and have been reported neonatally, at puberty, and well into adulthood, when effects on lactation or mammary tumorigenesis become evident. The normal morphology and pace of pubertal development are often altered, and these effects can be observed using MG whole-mount preparations. Transient or permanent effects may be due to gene imprinting, altered gene expression, modified endogenous MG signaling, or changes in hormonal milieu. Arrows indicate plausible (black) or more certain (gray) mechanistic pathways. Photomicrographs for early life and puberty were all taken at 16× magnification on a macroscope (adapted from Enoch et al. 2007, with permission from Environmental Health Perspectives); photo micrographs for pregnancy/lactation and adulthood were taken at 10× magnification on a standard microscope (from S.E.F.). Bars = 2 mm.

Puberty Gestation
Altered growth and development; altered carcinogen susceptibility . These studies typically include histo pathological evaluation of MG whole mounts of developing animals and report mor phological features such as branching, extent of growth, and relative proportion of struc tures (e.g., TEBs, lobules, and terminal ducts).

Adulthood Pregnancy/lactation
Other studies report changes in morphology or immuno histochemistry of tissue sections or gene expression in tissue homogenates. Methods and data reporting vary between labo ratories, making it difficult to compare findings across studies. More uniform approaches will facilitate progress; however, unanticipated end points should continue to be reported because this field is still developing.
Morphological changes reflect timing of assessment. Because normal develop ment involves a wellcharacterized, consistent progression of types and ratios of terminal structures and extension through the fat pad, alterations are sometimes reported as acceler ated or delayed develop ment relative to con trols (e.g., Moon et al. 2007). Some agents alter the pace at which differentiation occurs, leading to an increased or decreased number of TEBs depending on timing of assessment. If a peri natally administered EDC causes accelerated develop ment, the number of TEBs in the treated group will be higher than that in vehicletreated controls at weaning [post natal day (PND) 21] because of increased proliferation, but lower in early adulthood (PNDs 45-50) because of accelerated differ entiation, as is seen after exposure to estrogens (Hovey et al. 2005). In the case of an EDC, such as dioxin, that causes delayed develop ment, reduced differentiation leads to a higher number of TEBs in early adulthood and a longer period during which TEBs are present (Brown et al. 1998;Fenton et al. 2002). The number of TEBs present in the gland also depends on the number of ducts in the gland. Therefore, the number of TEBs at a particular time point can be altered by changes to the extent of growth as well as to the pace of dif ferentiation. For example, if the overall num ber of ducts is decreased by an environmental exposure, then the overall number of TEBs in the gland will be decreased compared with those in controls at any time point, as demon strated for perfluorooctanoic acid (PFOA) exposures in mice (White et al. 2009). Some reports have not differentiated between changes in TEB number due to overall gland size and those due to altered develop mental pace. Evaluation at multiple time points and consideration of the total number of terminal ends, as well as the absolute number of TEBs, alleviate this problem and convey the relative number of structures.
Steroid hormones. In one of the first studies of neonatal exposure to estrogen, progester one, or both in mice, Jones and Bern (1979) reported irreversible adult MG effects, includ ing secretory stimulation, dilated ducts, and abnormal lobulo alveolar develop ment. Perinatal treatment with estrogens such as estradiol or diethyl stilbestrol (DES) has been reported to produce accelerated develop ment, characterized by increased pubertal TEB den sity, and to promote ductal prolifera tion dur ing the peri pubertal period in both rats and mice (Fielden et al. 2002;HilakiviClarke et al. 1997;Hovey et al. 2005;Tomooka and Bern 1982;Warner 1976). In addition, Doherty et al. (2010) reported that pre natal DES exposure in mice altered expression in MG of genes that may be important in tumori genesis. Ovariectomy has been reported to diminish or obviate the effect of neo natal ovarian steroids on mouse MG develop ment (Jones and Bern 1979;Mori et al. 1976), and strain differences in sensitivity have also been reported (Mori et al. 1976;Yang et al. 2009). In rats exposed continuously beginning at conception, oral ethinyl estradiol exposure induced ductal hyper plasia in male rat MGs by PND50, and this effect was less appar ent in rats assessed later in life (Latendresse et al. 2009). Thus, morphological changes in MG reflect timing of exposure as well as timing of assessment, and so both of these variables must be considered when comparing results across studies. Supplemental Material, Table 1 (doi:10.1289/ehp.1002864) com piles the methods and findings of studies that have evaluated the effects of hormone, dietary, or chemical exposures during the pre natal, neo natal, or peri pubertal periods on MG develop ment up to 10 weeks. Several addi tional endocrinesensitive end points com monly assessed to indicate relative sensitivity are also included in the table.
In addition, whereas peri natal steroid hor mone exposure alters proliferation and TEB number, peri pubertal exposure that occurs after proliferation has begun affects mainly the differentiation of TEBs into mature struc tures. For example, pubertal DES treatment in rats increased the pace of lobule formation and decreased the number of terminal ducts and TEBs compared with vehicletreated con trols just after puberty (Odum et al. 1999). Prepubertal DES treatment of rats (on PNDs 23-29) resulted in fewer TEBs, termi nal ducts, and alveolar buds, with a concomi tant increase in the more differentiated lobules, overall suggesting a faster differentiation pace (Brown and Lamartiniere 1995). Treatment of post pubertal rodents with steroids or human chorionic gonadotropin increases differentia tion of the MG in a manner thought to mimic develop ment during pregnancy (Russo and Russo 2004b;Sivaraman et al. 1998).
Phytoestrogens. Effects of treatment with phyto estrogens such as genistein are simi lar to those observed after estrogen receptor agonist exposure; peri natal exposure can lead to increased proliferation, and peri pubertal exposure can lead to accelerated differentia tion (reviewed by Warri et al. 2008). For example, HilakiviClarke et al. (1998) and PadillaBanks et al. (2006) showed increased TEBs after peri natal genistein treatment, and Cotroneo et al. (2002) showed acceler ated develop ment in MG after pre pubertal exposure, as indicated by increased TEBs and ductal branching at an early time point, com pared with untreated animals. After gesta tional and lactational genistein exposure, You et al. (2002) observed enhanced glandular dif ferentiation at weaning, and males were more sensitive to the effect than females. Male rats in a multi generational genistein feeding study also showed ductal hyperplasia at PND50, a surprisingly early life stage for these effects (Latendresse et al. 2009 (Brown et al. 1998;Durando et al. 2007;Jenkins et al. 2007). In addition, lategestational treatment with Ziracin, a candidate anti bacterial drug, induced hypoplasia (ducts without any acinar develop ment) in rats (Poulet et al. 2005).
Critical exposure windows and reversibility. Studies of the ubiquitous industrial pollutant dioxin and the highuse herbicide atrazine have investigated critical periods of exposure associ ated with MG effects. Atrazine delayed MG develop ment when exposure occurred around GD17-19 but had less of an effect after earlier 3day windows, and dioxin exposure at GD15, but not after GD19, led to MG under develop ment (Fenton et al. 2002;Rayner et al. 2005). More recent studies on the industrial surfactant PFOA (White et al. 2009) demon strate a simi lar critical period. The heightened sensitivity during this time period is attributed to the formation of the mammary bud and initial branching that occurs during late pregnancy.
volume 119 | number 8 | August 2011 • Environmental Health Perspectives As discussed above, exposure timing and dose influence the pattern of MG changes (Warri et al. 2008).
Although numerous studies have shown persistent effects on the MG, few have evalu ated whether the changes could be revers ible. For example, in utero exposure to dioxin, Ziracin, PFOA, or BPA led to permanent changes in the adult MG (Fenton et al. 2002;Poulet et al. 2005;Vandenberg et al. 2007;White et al. 2009). In contrast, effects of genistein and ethinyl estradiol in male MG appeared to reverse after treatment withdrawal (Latendresse et al. 2009 Doherty et al. (2010) found that in utero exposure of mice to DES or BPA increased protein expression and func tional activity of the histone methyl transferase enhancer of zeste homolog 2 (EZH2) in the MG. EZH2 has been linked to breast cancer risk and epigenetic regulation of tumori genesis. Its upregulation is a potential mechanism through which in utero exposure to these chemicals may produce epigenetic changes leading to increased breast cancer risk (Doherty et al. 2010).
Sex differences. The few studies that have evaluated effects on male MG have indicated that male rats could be more sensitive. For example, one study found altered MG in males, but not females, treated with methoxy clor during gestation (You et al. 2002), and MG effects of genistein and ethinyl estradiol have been reported in males at lower doses than in females (Delclos et al. 2001;Latendresse et al. 2009). Study of sex differences in responsive ness can provide information about mecha nisms of action for the test agents. Although male mice lack mammary epithelia, there are transgenic mouse models in which mammary epithelial growth can be induced in males (Li et al. 2002). Mouse models are needed to study some chemicals, such as PFOA, whose pharmaco kinetics in mice and humans are most similar.
The Organisation for Economic Cooperation and Development (OECD) guidelines for subchronic oral toxicity test ing (OECD 2008) include evaluation of the male, but not female, MG as an optional end point. In some studies using these guidelines, the male MG appears to be among the most sensitive end points evaluated (Okazaki et al. 2001), and at least one such study has found it to be the most sensitive end point in males (Andrews et al. 2002).

Consequences of Altered Mammary Gland Development
Developmental exposures to certain EDCs can lead to MG develop mental effects, lactational deficits, or cancer, but little is known about the relationships between the develop mental and adult end points. The morphologi cal changes in MG develop ment, particularly effects on TEBs, suggest the potential for functional out comes such as lactational insufficiency, altered pubertal timing, pre neoplasia, or increased susceptibility to carcinogens (Fenton 2006 there are data. However, there are data gaps regarding the relation ships among the vari ous MG outcomes because a) only a handful of chemicals have been studied; b) there has not been a standard procedure to assess MG develop mental changes; c) MG assessment in multi generational studies has been limited; and d) few studies include full assessment of dose response.
Carcinogenesis. Reported changes in patterns of breast develop ment in U.S. girls (reviewed by Euling et al. 2008) raise concerns about whether earlier onset of breast develop ment is associated with breast cancer or other adult diseases, because earlier menarche is an established risk factor for breast cancer (Kelsey et al. 1993). Furthermore, studies in humans and rodent models demon strate that hormonal factors that affect MG develop ment also influence susceptibility to carcinogens.
Hormonal factors alter susceptibility to carcinogens. Ovarian, pituitary, and placental hormones, which vary by life stage and with pregnancy events, are important determinants of breast cancer susceptibility in humans and rodents (Russo and Russo 2004b). In both mice and rats treated with chemical carcino gens, hormone withdrawal (ovariectomy) inhibits tumor develop ment, whereas hor mone supplementation increases the incidence of adeno carcinoma. In humans, removal of ovaries by 35 years of age dramatically reduces breast cancer risk (Eisen et al. 2005;Trichopoulos et al. 1972), and anti estrogens are effective in breast cancer treatment and chemoprevention (Vogel et al. 2010).
Susceptibility to carcinogens depends on life stage. The influence of life stage on susceptibility to carcinogen exposure has been demon strated in rats and humans. For example, ionizing radiation is maximally potent as a human breast carcinogen when exposure occurs during childhood or adoles cence (Henderson et al. 2010;Land 1995); this observation is consistent with findings in rodents (Imaoka et al. 2009). The increased tumor response from carcinogen exposure early in life is attributed to the presence of proliferating and undifferentiated structures such as TEBs, which are present during the pubertal mammary epithelial expansion and display elevated DNA synthesis compared with other MG structures (Kleinberg et al. 2009). TEBs are considered the most vulner able MG target structure for carcinogen expo sure (Medina 2007;Russo and Russo 1996). In animals and humans, tumor response from carcinogen exposure is highest when expo sure occurs during adolescence, when TEBs are still abundant (Henderson et al. 2010;Imaoka et al. 2009;Land 1995;Russo and Russo 2004b). As a result, there is concern that exposures to xeno biotics that increase the number or longevity of proliferating TEBs Abbreviations: -, no effect on this end point; Δ, at least one study has reported an association between the exposure and altered outcomes [see details and citations in Supplemental Material, might increase susceptibility to breast cancer (Birnbaum and Fenton 2003;Fenton 2006). After the pubertal growth spurt and through out adult life, it is the terminal ductal struc tures that give rise to breast cancers (Kleinberg et al. 2009;Medina 2007;Russo and Russo 2004b). During pregnancy, differentiation of terminal structures increases, and this dif ferentiation has been hypothesized to account for lower MG sensitivity to carcinogens post pregnancy (Russo and Russo 2004b). Early life exposures to (non carcinogenic) chemicals may affect response to carcinogens in later life. Experimental models involving carcinogen challenge have been used widely to demon strate that hormones and growth factors influence MG develop ment, differentiation, and carcino genesis; these models could read ily be extended to evaluate increased cancer risk from early life environmental exposures. These models have been used, for example, to investigate potential chemo preventive agents that accelerate MG differentiation (e.g., by mimicking pregnancy hormones) and decrease tumor susceptibility (Cotroneo et al. 2002;Kleinberg et al. 2009;Russo and Russo 1996). Rodent models used in these studies include dimethylbenz[a]anthracene (DMBA) and nitroso methyl urea (NMU) challenge in rats and mice, and the mouse mammary tumor virus (MMTV) model. More recently, geneti cally modified mouse models have been used to study mammary tumors that are com parable with human breast tumors in their latency, histo types, and endocrine respon siveness (Cardiff et al. 2000;Kamiya et al. 1995;Medina 2007;Russo and Russo 2004b;Thompson and Singh 2000).
In rodents, early life exposure to hormon ally active agents affects MG tumor formation in carcinogenchallenge models. For example, neonatal estrogen (or androgen) treatment of mice (MMTV model) or rats (DMBA model) induced MG develop mental changes and increased tumors (Lopez et al. 1988;Mori et al. 1976Mori et al. , 1979. In addition, early life exposures to genistein (HilakiviClarke et al. 1998(HilakiviClarke et al. , 1999, alcohol (HilakiviClarke et al. 2004), dioxin (Brown et al. 1998;Desaulniers et al. 2001;Jenkins et al. 2007), and oral BPA (Jenkins et al. 2009) caused increased MG tumor multi plicity and decreased latency after DMBA challenge at PND50. These effects were accompanied by altered MG develop ment observed in whole mounts (genistein, alcohol, dioxin) or altered protein expression (BPA). Lifetime exposures (beginning pre natally) to genistein, ethinyl estradiol, and BPA have been reported to alter MG develop ment and increase incidence of pre neoplastic lesions in the MG, with a stronger effect in early adulthood than at 2 years of age, when MG histopathology is typically performed (Latendresse et al. 2009;Murray et al. 2007;Vandenberg et al. 2008). Shortterm genistein treatment during the peri pubertal period reduces MG tumors after carcinogen chal lenge, whereas peri natal or lifetime exposure seems to increase them, although studies are not consistent (reviewed by Warri et al. 2008). Similarly, both gestational (Brown et al. 1998) and pre pubertal (Desaulniers et al. 2001) dioxin exposure caused increased MG tumors after carcinogen challenge, whereas later life exposure decreased spontaneous MG tumors (Kociba et al. 1978).
In humans, maternal factors that affect the fetal hormone environment also appear to affect later breast cancer risk in daughters, possibly by imprinting the developing MG, thereby altering future tissue responsiveness to hormonal stimulation (e.g., altering estrogen receptor levels or sensitivity) or to geno toxic insult (e.g., by increasing cell proliferation or diminishing differentiation). The hypothe sis that in utero endocrinerelated factors influ ence breast cancer risk of a daughter is sup ported by epidemiology studies that have found a) pre eclampsia associated with reduced breast cancer risk in offspring and b) high birth weight correlated with higher breast can cer risk (Hoover and Troisi 2001;Troisi et al. 2007;Xue and Michels 2007). In addition, there is some evidence that in utero exposure to DES is associated with higher breast cancer risk in women [Palmer et al. (2006); however, Verloop et al. (2010) did not find an associa tion] and with increased MG tumor incidence in rats (Rothschild et al. 1987). Furthermore, the single epidemiologic study of the EDC dichloro diphenyl trichloro ethane (DDT) that used prospective meas ures of adolescent/young adult exposure in relation to breast cancer risk (Cohn et al. 2007) found significant associa tions, whereas many studies in which DDT or its metabolite dichloro diphenyl dichloro ethylene (DDE) were meas ured in older women did not observe an association with breast cancer.
Pregnancy is another critical window cor responding to a time of extensive MG pro liferation and differentiation. DES exposure in pregnant women has been associated with increased breast cancer risk in the mother as well as her daughter (TitusErnstoff et al. 2001). A study of DMBAchallenged rats fed a highfat diet during pregnancy showed an increase in circulating estrogen during pregnancy and increased mammary tumors (HilakiviClarke et al. 1996).
Lactation. The American Association of Pediatrics (AAP) recommends that all infants receive breast milk during the first 6 months (AAP 1997) and, further, that they are fed breast milk exclusively during this time (AAP 2005), because of the numerous demon strated benefits of breastfeeding. Although data are limited, reports estimate that 3-6 million mothers are unable to produce milk or have difficulty breastfeeding each year (Lew et al. 2009). The reasons for this remain unclear, especially given that lactation insufficiency can be the result of psycho social as well as biologi cal factors. However, environ mental chemicals are one candidate explanation for inability to initiate and/or sustain breastfeeding (Neville and Walsh 1995). Impaired lactation may be associated with altered MG develop ment (decreased or unre sponsive breast tissue) and/or endocrine disrup tion (improper hormonal support for lactation). Critical windows include pregnancy and lac tation as well as puberty and the prenatal/ perinatal period. As such, exposure to an EDC during pregnancy has the potential to disrupt lactation in the mother and the daughter. In human studies, strong early findings of asso ciations between serum DDE and shortened lactation in two populations have been only partly replicated, and few other agents have been studied (CupulUicab et al. 2008;Gladen and Rogan 1995;Rogan et al. 1987). In rodents, impaired lactation has been observed in conjunction with altered MG develop ment in one or more genera tions after gestational exposure to dioxin (Vorderstrasse et al. 2004), PFOA (White et al. 2007(White et al. , 2009, atrazine (Rayner et al. 2005), BPA (California Office of Environmental Health Hazard Assessment 2009; Matsumoto et al. 2004), genistein [National Toxicology Program (NTP) 2008], and the candidate pharma ceutical Ziracin (Poulet et al. 2005). For example, atrazine fed to rats during gestation induced MG develop mental changes in offspring, characterized by stunted develop ment, and when these rats were bred, their offspring (secondgeneration) had significantly reduced weight gain, suggesting insufficient milk production (Rayner et al. 2005). In an example of effects on lactation in the dam, exposure of pregnant mice to PFOA decreased pup weight and survival, dimin ished differentiation/growth of dam MG, and induced some alterations in gene expression for milk proteins, which taken together suggest effects on lactation in the exposed dams (White et al. 2007(White et al. , 2009.
Treatment during pregnancy has the potential to affect lactation in both the dam and offspring. Impaired lactation in the dams is typically identified because of decreased pup weight or survival, and impaired lactation in the offspring can be determined only in multi generational studies where offspring are followed through successful reproduction and lactation (Makris 2011). The rodent models and assessment methods used in guideline studies are not adequate for identifying effects on lactation because the surrogate markers of pup weight and post natal survival are not sensitive or specific indicators of impaired lactation (Makris 2011

MG assessment in chemical test guidelines.
MG develop ment can be affected after early exposure to EDCs in rodents. However, few guideline studies for testing environ mental chemicals include pre natal or early life dosing, and MG end points are limited primarily to indirect or surrogate observations during lacta tion and to clinical and pathological evaluation of adult mammary tissue (Makris 2011). For example, the standard 2year rodent cancer bio assay, initiating treatment in young adult animals, is likely to be less sensitive to car cinogens than if develop mental exposures were used, and it cannot provide information on altered susceptibility to carcinogens induced by early life exposures affecting MG develop ment (Hovey et al. 2002;Medina 2007;Rudel et al. 2007;Russo and Russo 2004b;Singh et al. 2000;Thayer and Foster 2007).
To strengthen MG assessment and chemi cal testing, it is a priority to enhance histo pathological evaluation of MG develop ment (e.g., using longitudinal rather than trans verse sectioning so that a larger tissue plane is evaluated), increase attention to evaluation of male MG tissue, and incorporate early life exposures in rodent subchronic and chronic/ carcinogenicity studies. Consistent with these recommendations, the NTP has begun including gestational and lactational dosing in rats assigned to subchronic and carcinoge nicity studies and is taking steps to include early life male and female MG wholemount preparations and longitudinal MG sectioning in reproductive assessment and cancer studies (NTP 2010;Thayer and Foster 2007). Use of these expanded protocols will facilitate link ing altered MG develop ment with later life outcomes.
As a potential addition to some toxicity test guidelines, MG wholemount assessment can demon strate morphological changes in develop ment and differentiation and define the temporal and spatial progression of epithe lial develop ment. Another important reason to include MG assessments in screeninglevel toxicology studies is to ensure that MG effects are identified and can be evaluated in more comprehensive studies. Specifically, data gen erated using whole mounts may be used to trigger further assessment, such as: a) section ing tissue blocks, b) evaluating subsequent (e.g., F 1 ) generations for lactational impair ment, c) maintaining a population longer on study for spontaneous neoplasia evaluation, or d) evaluating altered tumor susceptibility using a carcinogenchallenge protocol. Furthermore, a whole mount may be the only indication of abnormal develop ment in the male MG, which is sensitive to very low doses in some studies (Delclos et al. 2001;Latendresse et al. 2009). However, currently there are no standardized wholemount procedures, and consideration of these data in chemical risk assessment has been limited. The OECD test guideline for an extended onegeneration reproductive toxicity study (OECD 2010) could be revised to include assessment of MG develop ment using whole mounts and/or more thorough histopathology (Hvid et al. 2010). In addition, MG assessment of males and females could be added to the U.S. EPA Endocrine Disruptor Screening Program (EDSP) puber tal develop ment protocols (U.S. EPA 2009aEPA , 2009b. MG develop mental assessment could also be extended to include females in the OECD Test Guideline 407 pubertal protocol (OECD 2008).
Adding MG wholemount procedures to EDSP or OECD test guidelines has raised concerns that a) these assays could be redun dant to endocrinesensitive end points assessed [e.g., ano genital distance, timing of vaginal opening (VO), circulating hormones, and estrous cyclicity], and b) that the procedure is too difficult to be consistently executed across laboratories. However, in some cases MG effects have been observed at lower doses than other EDC outcomes (Table 2), and there is concern that EDSP assays, which identify chemicals affecting estrogen, androgen, or thy roid activities, may not be sensitive to the many mechanisms that can affect breast develop ment. It is reasonable, therefore, to include the MG whole mount in screening studies, at least on a provisional basis, to see if the information gathered is redundant or unique.
Human relevance of rodent models. Rodent models have been widely used to charac terize the influence of susceptibility factors (e.g., ovarian, pituitary, and placental hormones; lifestage and reproductive events) on malignant transformation of the MG, and parallels between rodent and human MG structures and pathologies have been enumer ated (Medina 2007;Russo and Russo 2004b;Singh et al. 2000). Although a few findings in the context of chemicals testing have gener ated concern about human relevance (reviewed by Rudel et al. 2007), an extensive body of breast cancer research demon strates similarities between rodent and human MG develop ment and carcino genesis. These studies indicate that rodent mammary tumors mimic the diver sity of human breast cancers with respect to important initiation processes, histo pathology, hormone dependence, and host-target cell intera ctions (Boylan and Calhoon 1983;Imaoka et al. 2009;Medina 2007;Rudland et al. 1998;Russo et al. 2000;Russo 1993, 2004a;Singh et al. 2000). In general, research indicates greater crossspecies similarities for MG develop ment and cancer than for human menstrual and rodent estrous cyclicity or for human puberty and rodent VO-end points currently included in many EDC test protocols (U.S. EPA 2011). An expert panel on MG tumors con cluded that existing rodent models are use ful as screening tools for identifying potential breast carcinogens (Thayer and Foster 2007). Further, the majority of chemicals that are positive for mammary tumors in the rodent cancer bio assay have some evidence of geno toxicity and many are multi site carcinogens, supporting relevance to humans . Although there are many similarities in the hormonal control of lactation across species, less is known about the utility of the rodent as a model for predicting chemical effects on human lactation. In any case, many risk assessment guidelines operate on the prin ciple that animal effects are considered relevant to humans in the absence of data to indicate otherwise (U.S. EPA 1991(U.S. EPA , 1996(U.S. EPA , 2005. A related issue is the consideration of carcinogenchallenge models as indicators of altered carcinogen susceptibility. DMBA and NMU are primary breastspecific carcino gens that have been widely used in experi ments designed to assess the alteration of the tumor response by hormones or other factors (Kamiya et al. 1995;Medina 2007;Singh et al. 2000). Despite the longstanding use of such carcinogen challenge experiments to assess effects of hormonal or develop mental alterations on tumor susceptibility, the pro tocols are not common in chemical toxicity assessment. Risk assessors have not considered data from carcinogen challenge experiments because of concerns about the protocol rep resenting a chemical mixture study and about the presumed lack of relevance of DMBA or NMU exposure to humans. However, a number of consistent findings of increased susceptibility have been observed in human and rodent studies across multiple MG end points for endogenous hormonal factors, DES, genistein, and dioxin, among others ( Table 1). Models that consider the inter active effects of endogenous hormones and carcinogenic factors across multiple life stages are likely to be more relevant to human health than those with simpler design, because they better reflect the human experience.
Relative sensitivity of MG effects. A lim ited set of studies provide evidence that MG alterations may be more sensitive to some EDCs than are other hormonally responsive end points (Table 2). To precisely determine the relative sensitivity of EDC effects requires studies that include MG as part of a larger set of endocrinesensitive end points. Of the stud ies that simultaneously evaluated MG mor phology and at least one other EDCsensitive end point after develop mental dosing, a subset has detected effects on the MG at dose levels or during exposure periods that did not elicit observable changes in other end points.
Adversity of MG develop mental changes. Hormonal factors either increase or decrease MG tumor susceptibility, and both transient and permanent effects have been observed on MG develop ment. This raises the question of what types of alterations to MG develop ment should be considered adverse. In the context of regulatory evalua tion of chemicals, one point of view is that MG develop mental changes reflect altered growth and develop ment, effects considered adverse by the U.S. EPA Developmental Toxicity Risk Assessment Guidelines (U.S. EPA 1991). For com parison, there is also controversy in the risk assessment community about whether other common markers of altered pubertal timing (e.g., VO, pre putial separation) have human relevance. These end points have never the less been considered adverse, as they are respon sive to endogenous sex steroids, which are important regulators of sexual develop ment conserved across mammalian species. By this reasoning, altered MG growth and develop ment, which is known to have human rele vance, should be considered adverse as well. The question of adversity was discussed by experts gathered at the workshop in the con text of risk assessment. In spite of the out standing questions, the majority perspectives among experts advance the view that MG develop ment and subsequent effects represent a public health outcome of concern and are a priority for future research and assessment. Priority questions, current views, and out standing issues for risk assessment are sum marized in Table 3.
An important question is whether MG develop mental alterations are plausibly related to increased tumor susceptibility by a) epi genetic imprinting of tissue, b) alteration of stem cell populations, or c) increased num ber or ontological duration of TEBs or other structures known to be more vulnerable to carcinogens. Some experts suggest that such agents should themselves be considered carcino gens. Indeed, the International Agency for Research on Cancer (IARC) deems an agent carcinogenic if it is "capable of increas ing the incidence of malignant neoplasms, reducing their latency, or increasing their severity or multi plicity" (IARC 2006). The U.S. EPA defines an effect as adverse if it "reduces the organism's ability to respond to an additional environmental challenge" (U.S. EPA 2010). Applying these defini tions, compounds that cause cancer, either alone or in combination with other factors at a variety of points in a biological chain of events leading to tumor formation, may reasonably be considered carcinogens, includ ing chemicals that increase susceptibility to cancer. Even if such agents are not designated as carcinogens, their profound impacts should encourage the risk assessment community to consider the increase in cancer susceptibility as an adverse effect and therefore to charac terize doses required to elicit the effect. In any case, applying this approach to risk assess ment requires a better under standing of the relation ship between altered MG develop ment and carcino gen susceptibility.

Conclusions and Research Recommendations
Research demonstrates many similari ties between humans and rodents in nor mal and perturbed MG develop ment and carcino genesis. In both humans and rodents, develop mental exposure to hormones affects MG develop ment and carcinogen suscep tibility, and these findings are the basis for on going research to identify chemo preventative agents in humans and to deter mine how EDCs may alter breast cancer risk, pubertal timing, or lactation. EDCs with diverse mechanisms of action, including many not considered primarily estrogenic, alter MG develop ment in rodents. In some Table 3. Priority questions, current views, and issues for improving risk assessment for MG effects. a Priority question for risk assessment application Current views Outstanding issues Are the rat and mouse adequate models for human MG development?
Current knowledge suggests that the rat and mouse are reasonable surrogates.
Lack of information about human pubertal development; mechanisms may differ among species. What is the sensitivity of MG developmental effects? In utero exposure in some studies leads to developmental effects at doses similar to or lower than other developmental and reproductive end points.
Few EDC studies assess both MG development and another sensitive end point of ED; there is a lack of human data to address dose response and a lack of standardized MG development protocol and assessment criteria. Are MG developmental changes adverse?
These changes in MG are considered adverse because they represent alterations in growth and development b and may be a risk factor for lactation and/or cancer outcomes.
Varied definitions of "adversity," depending on scientific discipline and context. volume 119 | number 8 | August 2011 • Environmental Health Perspectives cases, altered MG develop ment can be the most sensitive endocrine end point. The lack of consistent methods for eval uating and reporting MG changes makes it difficult to compare findings across studies, hindering consideration of MG develop mental effects in risk assessment. Continued prog ress will require consistent approaches across laboratories, along with a discussion of unique findings and unanticipated effects. In addition, the relation ships between altered develop ment and effects on lactation or carcino genesis are still being defined. Addressing these research needs [detailed in Supplemental Material (doi:10.1289/ehp.1002864)] is a priority, and enhanced chemical testing and risk assessment are needed to characterize these effects.
Major research initiatives under way include The National Children's Study, which has a number of EDC hypotheses proposed for testing in its longitudinal study (Lewin Group 2003), and the NIEHS Breast Cancer and the Environment Research Centers (NIEHS 2010), which have on going human studies focusing on the relationship between environmental exposures and age of breast develop ment onset and which also support experi mental animal research in this area. Research priorities identified at the workshop [provided in detail in Supplemental Material (doi:10.1289/ehp.1002864)] include further develop ment and validation of the MG whole mount protocol, research to establish the relation ship between effects on MG develop ment and later life outcomes, and issues rele vant to use of these data in risk assessment.