Differentiation Patterns of Uterine Carcinomas and Precursor Lesions Induced by Neonatal Estrogen Exposure in Mice.

Exposure to estrogenic chemicals during early development can result in adverse health effects later in life ( Masse et al. 2009 ; Walker and Ho 2012 ; Ho et al. 2017 ; Herceg et al. 2018 ). The most infamous example occurred when the potent synthetic estrogen diethylstilbestrol (DES) was administered to an estimated 2-10 million pregnant women in the United States from 1938-1971 in a misguided effort to prevent miscarriages and premature delivery ( Giusti, Iwamoto, and Hatch 1995 ; IARC 2012 ; Reed and Fenton 2013 ). The daughters of women treated with DES while pregnant were later found to have increased lifetime risk for a broad spectrum of reproductive health outcomes, including infertility, vaginal adenosis, high-grade cervical intraepithelial neoplasia, and vaginal and cervical adenocarcinoma ( Hoover et al. 2011 ; Verloop et al. 2010 ; Hatch et al. 2001 ). Although DES has been contraindicated for use in pregnancy since 1971, this case study continues to highlight the importance of exposures that occur during sensitive windows of development and to serve as a classical model for developmental reprogramming, in which a shift in epigenetic processes early in life permanently alters physiological responses and disease risk later in life ( FDA 1972 ; Jefferson et al. 2011 ; Jefferson et al. 2013 ; Laronda et al. 2012 ). In the most widely used experimental model of early-life estrogen effects, female CD-1 mice are treated with an exogenous estrogen on postnatal day (PND) 1-5 ( Newbold, Bullock, and McLachlan 1990 ; Newbold et al. 2001 ; Suen et al. 2016 ). This short-term exposure can result in a high incidence of uterine carcinoma by 18 months of age, as well as other changes including adenomyosis, basal and squamous cell metaplasia, and atypical hyperplasia of the endometrial glands ( Newbold and McLachlan 1982 ; Newbold, Bullock, and McLachlan 1990 ; Suen et al. 2016 ). This spectrum of effects has been attributed to the timing of estrogen exposure, which corresponds with key periods of cellular differentiation and gland formation in the mouse reproductive tract ( Kurita 2011 ; Cooke et al. 2013 ). The neonatal exposure mouse model has been used for over four decades, and carcinogenic effects have been reported across dozens of studies, but to date there has not been a detailed characterization of uterine epithelial phenotypes in relation to the development of neoplastic lesions. Previous work has identified several different factors that contribute to the uterine effects of early-life estrogen exposure. In rodent models, for example, endogenous estrogens play a critical role in promoting DES-induced lesions, which are prevented by ovariectomy prior to puberty or conditional deletion of estrogen receptor alpha (ERα) ( Ostrander, Mills, and Bern 1985 ; Newbold, Bullock, and McLachlan 1990 ; Hendry et al. 1999 ; Couse et al. 2001 ). Other evidence implicates disruption of signaling pathways related to differentiation, which leads to permanent shifts in epithelial cell fate patterns throughout the reproductive tract ( Kurita and Cunha 2001 ; Kurita, Mills, and Cunha 2004 ; Kurita 2011 ; Jefferson et al. 2011 ). More recent findings point to early epigenetic modifications in key developmental proteins such as sine oculis-related homeobox 1 (SIX1), which may mediate persistent genomic and morphologic changes in the uterus ( Jefferson et al. 2013 ; Suen et al. 2016 ; Ho et al. 2017 ). It is still unclear, however, how these different processes interact over time to drive carcinogenesis. The goal of this study was to describe the histopathologic and immunohistochemical features of uterine epithelial changes following neonatal exposure to either DES or the soy phytoestrogen genistein (GEN) in female CD-1 mice up to 18 months of age. We show that both estrogens induce abnormal cell populations within endometrial glands by early adulthood. These cell types exhibit distinct differentiation profiles and predominate within atypical hyperplastic and neoplastic lesions that occur later in life.

The caudal cervix was lined by stratified squamous epithelium ( Fig. 1A and 1B ) that varied based on normal features of estrous cycle stage ( Dixon et al. 2014 ) in all age groups. These features included a superficial mucification layer extending from the vaginal epithelium along the squamous cervix in a subset of mice ( Fig. 1B ). Squamous epithelium was replaced by columnar glandular cells at or near the squamocolumnar junction (SCJ) ( Fig. 1A ). One to two layers of cuboidal basal cells (similar to “reserve cells” in the human cervix) were often present subjacent to the luminal columnar epithelium in this transition zone from cervix to uterine body ( Fig. 1C and 1D ). These basal cells varied in the degree to which they extended anteriorly past the SCJ. In some CON mice, they extended to the middle portion of the uterine body, either continuously from the squamous cervix or as scattered discontinuous aggregates. The latter cell populations were often indistinct on H&E staining and confirmed only with corresponding basal cell IHC markers. In the uterine body and horns, the central lumen and simple tubular glands of CON mice were lined by a single layer of cuboidal to columnar glandular epithelium ( Fig. 1E and 1F ). Glands lacked a peripheral layer of myoepithelial or basal cells, which is present in prostate, mammary, and other contractile glands. The normal endometrial stroma was clearly demarcated from the myometrium. Physiologic variations in endometrial morphology such as mitotic and apoptotic epithelial cells and stromal edema were evident based on estrous cycle stage, as described elsewhere ( Dixon et al. 2014 ). Glandular cystic dilation and adenomyosis of endometrial glands were observed as age-related background findings ( Suen et al. 2016 ). Exposure to DES and GEN resulted in a spectrum of uterine abnormalities, consistent with those described previously in this model (e.g., Newbold, Bullock, and McLachlan 1990 ; Newbold et al. 2001 ; Suen et al. 2016 ). Effects on the squamous cervix included hyperkeratosis, reduced mucification, and a low incidence of adenosis at the fornices but no atypical hyperplastic or neoplastic lesions. Uteri from DES and GEN mice had reduced overall glandular complexity and loss of demarcation between the endometrial stroma and surrounding myometrium. Endometrial epithelial lesions induced by neonatal DES and GEN exposure included glandular cystic dilation, adenomyosis, basal cell metaplasia, squamous cell metaplasia, atypical hyperplasia, and carcinoma ( Table 2 ). The overall pattern of effects was similar between the two types of estrogens. The following minor differences were observed in GEN compared to DES groups: lower incidence of basal cell metaplasia and atypical hyperplasia at 6 mo; higher incidence of glandular cystic dilation at 12 mo; and higher incidence of squamous metaplasia at 18 mo. The earliest and most prevalent uterine effect of neonatal estrogen treatment was basal cell metaplasia ( Fig. 2A and 2B , lower image). This change consisted of 1-2 layers of epithelial cells underlying morphologically normal columnar epithelium lining the central lumen and endometrial glands ( Fig. 2A ). Basal cells were cuboidal with round to oval nuclei, often with perinuclear clearing ( Fig. 2B ). This bilaminar appearance was often accompanied by rudimentary or multilocular luminal morphology, giving some glands an immature appearance. Basal cell metaplasia was most commonly identified in glands of the uterine body but was also found variably along the length of the uterine horns. At 2-3 mo of age, 70-89% of DES and GEN mice exhibited basal cell metaplasia, whereas basal cells were rarely observed in CON mice (beyond the transition zone of the cervix and uterine body). Previous reports have included basal cell metaplasia under the diagnosis of squamous cell metaplasia (e.g., Newbold, Bullock, and McLachlan 1990 ; Newbold et al. 2001 ). Here, we split these terms based on stratification and maturation to keratinocytes. In squamous cell metaplasia, a lineage of stratified squamous epithelial cells typically extended from the basement membrane to the lumen, replacing the glandular epithelium and often including fully differentiated superficial keratinocytes ( Fig. 2C and 2D ). Basal cell metaplasia was characterized by compartmentalization of basal and luminal layers and did not show evidence of keratinocyte maturation, replace glandular epithelium, or border luminal spaces. Although basal cell metaplasia was highly associated with (and considered to be a likely precursor lesion for) squamous cell metaplasia, the vast majority of these lesions did not show clear squamous maturation. Atypical focal glandular hyperplasia was defined by irregular endometrial glands with cytological and architectural pleomorphism ( Fig. 3A ). Features included increased epithelial cell numbers; abnormal growth patterns with mixed luminal, basal, squamous, and in rare cases mucous cells; small and irregular glandular lumens, often intraepithelial; and excessive crowding of glands. Cellular features included anisokaryosis, anisocytosis, and loss of polarity for luminal cells but no evidence of local or vascular invasion. Carcinomas induced by DES and GEN were characterized by the presence of atypical glandular structures with local invasion into the surrounding stroma or myometrium ( Fig. 3B – 3E ). The most common tumor location was the uterine body near the bifurcation of uterine horns, although carcinomas also occurred in the uterine horns and were often multicentric ( Fig. 3B ). The growth pattern was highly infiltrative and did not typically result in mass lesions that could be observed grossly. The most common morphologic subtype consisted of small glandular structures with round lumens <50 μm in diameter ( Fig. 3C ), often containing secretory material and large vacuolated cells ( Fig. 3D ). Neoplastic cells along glandular margins were often difficult to distinguish from adjacent stromal cells ( Fig. 3D ), and vascular invasion was noted in rare cases. Carcinomas typically occurred on a background of atypical glandular hyperplasia (which was recorded as a distinct lesion). Fibrosis and inflammation were variably present along the margins of neoplastic glands. All carcinomas contained acini with lumen formation, consistent with glandular differentiation. In rare cases, neoplastic glands contained luminal cells with abundant finely granular and basophilic cytoplasm, resembling mucous cells. However, none of the carcinomas had a pure luminal cell phenotype with expansile growth. All tumors showed some degree of basaloid differentiation, which in some cases progressed to squamous metaplasia with or without full keratinization ( Fig. 3E ). In these cases, it was difficult to distinguish adenocarcinoma with squamous differentiation from squamous cell carcinoma or adenosquamous carcinoma. According to INHAND terminology ( Dixon et al. 2014 ), this latter term should be used for tumors with “at least 10% or more squamous differentiation.” Although a number of carcinomas in DES and GEN mice met this criterion, a diagnosis of adenosquamous carcinoma was not considered appropriate in many other cases given the distinction between basal and squamous cell differentiation described above. Based on these considerations, and the predominance of basal and luminal cell differentiation as the central feature, we propose calling these tumors adenobasal carcinomas with or without squamous differentiation. To support morphologic observations and better understand the molecular phenotype of these metaplastic and neoplastic lesions, we applied a panel of IHC markers related to cellular differentiation, receptor expression, and apoptosis. Nuclear expression of SIX1 was present in the stratified squamous epithelium of the ectocervix and transition zone in all mice ( Fig. 4A , Table 3A ). No differences were observed in SIX1 expression between CON and DES or GEN mice at these sites. SIX1 staining was most intense in basal/suprabasal layers but often extended full thickness to the superficial squamous layers (unlike the basal markers P63 and CK14, as described below). When present, basal cells in the transition zone of the cervix were also positive for SIX1, whereas overlying glandular cells showed variable labeling ( Fig. 4A ). In the uterine body and horns, SIX1 labeling was consistently absent throughout luminal/glandular epithelium of CON mice ( Fig. 4B ). In DES and GEN mice, nuclear SIX1 expression occurred in basaloid and squamous cells of all metaplastic glands ( Fig. 4C ). Within the luminal compartment of these glands, SIX1 was typically present only in cells that had aberrant SIX1+ basal cells directly subjacent to them ( Fig. 4D ). SIX1 expression was also present in low numbers of scattered luminal cells within morphologically normal uterine glands (later identified as CK14+/CK18+ cells, as described below) ( Fig. 4E ). In hyperplastic and neoplastic glands, there was moderate to strong nuclear expression of SIX1 in basal and luminal cells ( Fig. 4F ), in contrast to other basal cell markers (P63 and CK14), which were only present in the basal compartment. Strong nuclear P63 labeling was present in basal/suprabasal cells of the squamous cervix and basal (reserve) epithelial cells near the SCJ in all mice ( Fig. 5A , Table 3A ). Luminal/glandular endometrial epithelial cells in CON, GEN, and DES mice diffusely lacked P63 staining. Areas of basal cell metaplasia showed strong expression of P63 specifically within basal cells (but not overlying luminal cells) ( Fig. 5B ). Foci of squamous cell metaplasia also had P63 staining within basal/suprabasal cells, similar to normal squamous cervix ( Fig. 5C ). Endometrial epithelium with atypical hyperplasia and all uterine carcinomas showed moderate to strong expression of P63 in basal but not luminal cell populations ( Fig. 5D ). Strong cytoplasmic CK14 labeling was present diffusely within all layers of the cervical squamous epithelium and in basal/reserve cells near the SCJ of all mice ( Fig. 6A , Table 3B ). Endometrial expression of CK14 was most prominent within metaplastic glands (specifically within basal cells, as for P63) ( Fig. 6B ). Luminal epithelial cells of the endometrium diffusely lacked CK14 staining ( Fig. 6C ) with the exception of small populations of scattered cells not directly associated with basal or squamous cell metaplasia ( Fig. 6D ). These CK14+ luminal cells were present in very low numbers (<1% of total cells) in a subset of CON mice and appeared to co-express CK18 but not P63 based on comparison of serial sections; SIX1 labeling was either absent or equivocal. In DES and GEN mice, this type of CK14+ luminal cell was more prevalent ( Table 3B ) and consistently co-expressed both CK18 and SIX1. Hyperplastic and neoplastic lesions showed moderate to strong expression of CK14 in basaloid cells lining acinar structures, areas of squamous differentiation, and a small subset of luminal-type epithelial cells ( Fig. 6E and 6F ). Expression of CK18 was not present in normal squamous epithelium of the cervix ( Fig. 7A ) but was observed in the superficial mucification layer when present ( Table 3B ). The lower incidence of cervical CK18 expression in DES and GEN mice was due primarily to the lack of mucification. CK18 was also observed in small foci of adenosis at the fornices in low numbers of DES and GEN mice. In the endometrium, CK18 was expressed in all surface luminal/glandular epithelial cells of all mice. Cellular localization was predominantly membranous, to a lesser extent cytoplasmic, and most prominent along apical/luminal margins. In foci of basal metaplasia, CK18 staining was observed specifically within the luminal compartment ( Fig. 7B ). In hyperplastic and neoplastic glands, CK18 expression was present within luminal cells of all lesions evaluated ( Fig. 7C and 7D ). CK18 expression thus complemented CK14 and P63 expression and confirmed the two-compartment phenotype of these tumors. Double-staining for CK14 and CK18 was performed to confirm co-expression of these markers in a distinct subset of uterine epithelial cells. Normal squamous cervical epithelium did not contain CK14+/CK18+ cells, whereas cells of the mucification layer, when present, were variably positive. Reduced mucification in DES and GEN mice accounted for the lower incidence of CK14/CK18 labeling in the cervix ( Table 3B ). Low numbers of these cells were also present within glandular epithelium near the SCJ in a subset of CON, DES, and GEN mice ( Fig. 8A ). In the endometrium, CK14+/CK18+ cells were either completely absent or limited to low numbers of scattered columnar cells lining the central lumen in a subset of CON mice ( Fig. 8B ). In DES and GEN mice, CK14+/CK18+ cells were observed within morphologically normal (non-neoplastic), hyperplastic, and neoplastic glands ( Table 3B ). In non-neoplastic epithelium, CK14+/CK18+ cells were luminal and did not include P63+ basal or squamous cells ( Fig. 8C and 8D ). In DES and GEN mice, dual-labeled cells were present in the epithelium of both the central lumen and smaller deeper glands, and some of these cells appeared to have processes expanding into the stroma ( Fig. 8D and 8E ). Dual labeling was more prevalent in neoplastic compared to non-neoplastic epithelium ( Table 3B ). In neoplastic glands, CK14+/CK18+ cells were most common along the deep margins of glands and often associated with an invasive cellular morphology in the absence of basal metaplasia. The majority of neoplastic glands had distinct CK14+/CK18− and CK14−/CK18+ compartments, while a subset of glands, often with only a single epithelial layer, showed CK14+/CK18+ staining ( Fig. 8F ). Carcinomas thus consisted of a mosaic with luminal (SIX1+/P63−/CK14−/CK18+), basal (SIX1+/P63+/CK14+/CK18−), and mixed (SIX1+/P63−/CK14+/CK18+) cell types ( Fig. 9A – 9E ). While small subsets of neoplastic glands (or portions of individual glands) showed basal-only, luminal-only, or mixed-only cells, the vast majority of glands within each uterine carcinoma showed a combination of cells with basal and luminal differentiation. Speckled membranous CD44 labeling was present in basal/suprabasal layers of the squamous cervical epithelium in all mice ( Fig. 10A , Table 3C ). In the endometrium, epithelial CD44 labeling was observed in some CON mice but not others, and no clear pattern could be discerned based on cycle stage. When present, staining of endometrial glandular epithelium in CON mice was sporadic, often scant, and limited to deeper glands ( Fig. 10B ). Strong CD44 labeling was also noted in immune cells. Cellular localization was membranous in cervical epithelium, cytoplasmic in stromal/immune cells, and membranous +/− cytoplasmic in glandular epithelium ( Fig. 10C ). In DES and GEN mice, positive membranous staining was often (but not always) present in metaplastic basal cells. Weak to moderate membranous CD44 staining was also observed within basal layers of neoplastic glands in a subset of tumors but was not a prominent feature ( Fig. 10D ). In cervical squamous epithelium, nuclear ± cytoplasmic TUNEL was present mainly in scattered cells within the superficial cornification layer and in cells that had already sloughed into the central lumen ( Fig. 11A , Table 3C ). In the endometrium, TUNEL+ cells were scattered throughout the luminal epithelium and tended to be more prevalent along the central lumen compared to deeper glands ( Fig. 11B ). No differences in cervical or endometrial epithelial TUNEL were noted between CON and DES or GEN mice at 2-3 or 18 mo of age. In metaplastic basal cells, TUNEL was completely absent or present in only a few scattered cells (estimated at <1% of total) ( Fig. 11C ). In areas of squamous metaplasia (non-neoplastic) or squamous differentiation (neoplastic), TUNEL was generally limited to the superficial layer and sloughed intraluminal cells, similar to normal squamous cervix ( Fig. 11D ). Neoplastic glands without full squamous differentiation (and not including sloughed cells) had either no TUNEL labeling or very few positive cells (<1%). Nuclear expression of both ERα and PGR was present throughout normal cervical squamous and endometrial glandular epithelium of all mice ( Table 3D ). In the normal endometrium, ERα labeling was often present diffusely within glandular epithelium ( Fig. 12A ), whereas PGR expression varied from sporadic patchy labeling of basal and luminal cells to more diffuse labeling of all epithelial cells ( Fig. 12B ). No clear differences in epithelial ERα or PGR expression were observed between CON and DES or GEN mice ( Table 3D ). ERα and PGR were also expressed within all metaplastic, hyperplastic, and neoplastic uterine glands, including luminal and basal cells ( Fig. 12B–F ). Previous studies have documented uterine changes in CD-1 mice induced by neonatal DES exposure up to PND 22 ( Yoshida, Newbold, and Dixon 1999 ; Yoshida, Newbold, and Dixon 2000 ). Early effects included decreased glandular cell proliferation, abnormal expression of estrogen-related differentiation markers, and the development of fingerlike cytoplasmic processes in the intercellular basal regions of epithelial cells. Our goal here was to track DES and GEN effects from early to late adulthood. At 2-3 mo of age, all control mice showed evidence of estrous cyclicity based on uterine and vaginal histology. The only morphologic lesion in the uterine epithelium among DES and GEN mice at this time was basal cell metaplasia, which was represented by small foci scattered along the periphery of the primary endometrial lumen and endometrial glands within both the uterine body and horns ( Table 2 ). All other glandular lesions induced by DES and GEN were present by 6 mo of age and increased in incidence with age. The incidence of uterine carcinoma in DES and GEN treated mice ranged from 0% at 2-3 mo to 7-16% at 6 mo, 29-35% at 12 mo, and 33-40% at 18 mo. No cases of uterine carcinoma were observed in CON mice at any time point assessed. Morphologic and molecular characteristics of abnormal cell types and neoplastic lesions induced by DES and GEN are summarized in Fig. 13 . At a cellular level, the SIX1+/P63+/CK14+/CK18− basal cells and associated SIX1+/P63−/CK14−/CK18+ luminal cells increased with age from small sparse foci at 2-3 mo of age to become widespread within endometrial glands in older DES and GEN mice. Accordingly, the IHC scores for basal cell markers SIX1, P63, and CK14 increased at least 2-fold between 2-3 mo and 18 mo of age ( Table 3 ). These SIX1+ basal and luminal cells appeared at the same time, and there was no morphological evidence indicating a direct lineage between them ( Fig. 13 ). A smaller population of mixed/bipotential cells co-expressing CK14 and CK18 was also present at both 2-3 and 18 mo of age. These cells were more prevalent in neoplastic compared to non-neoplastic glands but otherwise did not appear to increase with age in non-neoplastic glands. There were no clear age-related effects in the expression of CD44, TUNEL, ERα, or PGR in non-neoplastic endometrial glands in any of the groups ( Table 3C and 3D ). No differences in IHC marker expression were observed between DES and GEN groups at any of the observed time points ( Table 3 ).

Exposure to estrogenic chemicals during key windows of development can increase risk of reproductive tract cancer. This carcinogenic effect has been associated with altered patterns of cellular differentiation (e.g., reviewed in Kurita 2011 ). The goal of this study was to characterize uterine epithelial phenotypes resulting from neonatal exposure to two different reference estrogens, DES and GEN. Both chemicals induced abnormal populations of luminal, basal, and mixed endometrial epithelial cells. These three distinct cell populations were established by early adulthood and later comprised the vast majority of cells within all uterine carcinomas. These findings support the idea that cancer risk later in life can be influenced by altered patterns of cellular differentiation and programming established during early development. The most prominent early change induced by neonatal estrogen exposure was an increase in uterine basal cell metaplasia. This lesion was characterized by 1-2 layers of cuboidal cells distributed between the basal lamina and the glandular epithelium of the central endometrial lumen and along the perimeter of deeper uterine glands. Basal cells were rarely present in control mice beyond the junctional zone of the squamous cervix and uterine body. While this change may be seen sporadically as a background finding in other mouse strains, it is generally restricted to the uterine body and not associated with cancer (e.g., Couse et al. 1997 ). Endometrial basal cells in DES and GEN mice showed positive labeling for SIX1, P63, CK14, ERα, and PGR, low TUNEL expression, and sporadic CD44 expression, consistent with a mature basal cell phenotype. These cells increased with age so that by 18 mo they were a predominant feature of the endometrial epithelium (typically >50% of glands affected). The second abnormal cell population induced by neonatal estrogen exposure consisted of columnar luminal cells overlying metaplastic basal cells. These cells were positive for CK18 and negative for P63 and CK14, similar to normal glandular epithelium. The distinguishing feature was nuclear labeling for SIX1, which was generally absent from normal endometrial luminal epithelium (beyond the junctional zone with cervix). SIX1 is a transcription factor widely expressed during development and involved in tissue differentiation ( Christensen et al. 2008 ; Yajima et al. 2014 ). It is also overexpressed in various human cancers and has thus been considered an oncofetal protein because of its potential dual roles in development and cancer ( Wu et al. 2015 ). Previously, we identified luminal SIX1+ “founder” cells at the time of DES and GEN exposure (PND5) and showed increasing SIX1 expression with age ( Suen et al. 2016 ). Here, we found that SIX1 is expressed in both P63+/CK14+/CK18− basal cells and P63−/CK14−/CK18+ luminal cells when directly adjacent to basal cells. This distinctive localization suggests that SIX1 may mediate some type of cross-talk between abnormal basal and luminal cell populations. The third cell population induced by neonatal estrogen exposure showed a mixed phenotype with both basal and luminal elements. These cells co-expressed CK14, CK18, and SIX1 but not P63 and could be identified only through IHC labeling. Morphologically, they had a columnar luminal phenotype and did not associate with P63+ basal cells. Low numbers of CK14+/CK18+ cells were present sporadically in the cervical and endometrial epithelium of the central lumen in control mice, but these cells (in the endometrium at least) had no or equivocal labeling for SIX1. In older DES and GEN mice, CK14+/CK18+ cells were present along the basal margin of endometrial glands. When present in neoplastic lesions, they often showed features of microinvasion including disruption of the basement membrane and extension of cytoplasmic processes into the adjacent stroma. To our knowledge, no prior studies have specifically identified CK14+/CK18+ cells in the endometrial epithelium. In other tissues, such as prostate and mammary gland (which normally have a layer of CK14+ basal/myoepithelial cells, unlike endometrium), several reports have described small fractions of CK14+/CK18+ cells and proposed that these represent a progenitor or stem cell population (e.g., Wang et al. 2001 ; Smith, Mehrel, and Roop 1990 ; Buono et al. 2006 ). Additional studies are needed to determine whether SIX1+/P63−/CK14+/CK18+ cells are specific to this mouse model or also present in human uterine cancers. All uterine carcinomas and associated hyperplastic lesions in this study had a combination of basal and glandular components. We consider this to be a dual or mixed phenotype rather than a primary adenocarcinoma with squamous differentiation (or squamous cell carcinoma with glandular differentiation) based on the presence of all three abnormal cell types described above in all hyperplastic lesions and tumors evaluated. Still, it is not clear which of these cell types actually progresses to malignancy. Based on morphologic features of atypia and local invasion and the general lack of a differentiated state, we hypothesize that the population of mixed P63−/CK14+/CK18+/SIX1+ cells is the cell of origin for both P63+/CK14+/CK18−/SIX1+ basal cells and P63−/CK14−/CK18+/SIX1+ luminal cells and the most likely tumor progenitor candidate ( Fig. 13 ). This idea is further supported by the observations that (1) P63+/CK14+/CK18− basal cells in other sites such as mammary gland or vagina are not promoted to cancer by neonatal estrogen exposure and (2) SIX1+/P63−/CK14−/CK18+ luminal cells do not show clear morphologic evidence of neoplastic progression. Glandular structures within carcinomas consisting of only SIX1+/P63−/CK14−/CK18+ luminal cells were rare (and when present, difficult to identify as neoplastic based on morphology), filling of glandular lumens to form solid tubules (as sometimes seen with other types of luminal cell adenocarcinomas) was not observed, and local invasion typically occurred along the basal margin of neoplastic glands by CK14+ cells. Future work in genetically-modified mouse models could be used to delineate the role of differentiation factors like SIX1 in regulating these distinct cell types and promoting uterine carcinogenesis. A growing body of evidence supports the idea that chemical exposure during development can increase cancer risk later in life through epigenetic processes. However, there remains a critical need for mechanistic information and biomarkers that will enable identification of susceptible populations. Here we examined epithelial differentiation patterns over time following neonatal estrogen exposure and identified cell populations associated with endometrial cancer development. These results inform the relationship between early developmental exposures, aberrant cellular programming, and reproductive outcomes later in life. Identification of the molecular drivers of these differentiation patterns may provide insights into appropriate biomarkers and/or therapeutics for human endometrial cancer.

All animal studies were conducted following the recommendations of the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. All procedures involving animals were performed at the National Institute of Environmental Health Sciences (NIEHS) according to an approved Institutional Animal Care and Use Committee protocol. Female CD-1 mouse pups were obtained from the NIEHS breeding colony and given daily subcutaneous injections (0.02 ml) of 1 mg/kg DES or 50 mg/kg GEN in corn oil or corn oil vehicle as negative control (CON) during PND 1-5, as described previously ( Newbold, Bullock, and McLachlan 1990 ; Newbold et al. 2001 ; Jefferson et al. 2011 ; Suen et al. 2016 ). All mice were weaned at PND22 and housed 5 females/cage in static cages on a designated aging rack that was separated from (but in the same room as) male mice and breeding lines. Reproductive tracts were collected at 2-3, 6, 12, or 18 months (mo) of age. The age range of 2-3 mo old mice resulted from staggered availability of female pups from the NIEHS colony. Mice for the 6, 12, and 18 mo time points originated from a previous study by Suen et al. (2016) . The total number of mice evaluated per exposure group was n=30-33, n=26-30, and n=30 for 6, 12, and 18 mo time points, respectively. Mice for the 2-3 mo time point were obtained from the same breeding colony and exposed to vehicle CON, DES, or GEN (total n=9-13 per exposure group) using the same dosing protocol as for the 6, 12, and 18 mo time points. Mice were euthanized by CO 2 asphyxiation, and uteri were collected and processed as previously described ( Suen et al. 2016 ). The cranial half of each right uterine horn was frozen for molecular analysis and thus not available for histopathology. The remaining tract, including cervix and cranial vagina, was fixed flat in 10% neutral buffered formalin for 18 hours at 4°C and then transferred to 70% ethanol and stored at 4°C until processing. Tissues were processed using standard histologic procedures, paraffin-embedded, sectioned longitudinally at 6 μm, and either stained with hematoxylin and eosin (H&E) or left unstained for immunohistochemistry (IHC). IHC analysis was performed on the same mice used for histopathology. Reproductive tracts were sectioned until the central uterine lumen could be observed in plane with the cervicovaginal epithelium in order to have a continuous view of the epithelium from the uterus to the cranial vagina. Histology was performed by the Histology Core Laboratory of the NIEHS/National Toxicology Program (NTP). A single H&E-stained section of uterus and cervix was examined for each mouse. Where applicable, histopathologic diagnoses were based on standard criteria and nomenclature for neoplastic and non-neoplastic lesions presented by the International Harmonization of Nomenclature and Diagnostic Criteria for Lesions in Rats and Mice (INHAND) Project ( www.toxpath.org/inhand.asp ) ( Dixon et al. 2014 ). Any discrepancies with INHAND terminology are described in the results section. Severity of non-neoplastic lesions was qualitatively scored using a generic 0-4 scale (0=absent, 1=minimal, 2=mild, 3=moderate, 4=severe) based on lesion extent and complexity. All pathology data, tabulations, and observations were recorded by a board-certified veterinary pathologist (CEW). All mice 2-3 mo of age (n=9-13/exposure group) and a subset of 18 mo mice (n=9-12/exposure group) were evaluated using a panel of IHC stains. The 18 mo subset included arbitrarily-selected CON mice (n=9) and all GEN (n=10) and DES (n=12) mice with uterine carcinomas, as determined previously ( Suen et al. 2016 ). For the time-course evaluation, we also included descriptions of lesions and IHC labeling in mice at 6 and 12 mo of age. For each mouse, serial sections were stained for each of the following markers: SIX1 as a developmental differentiation factor and putative cancer progenitor cell marker; tumor protein 63 (P63) and cytokeratin 14 (CK14) as markers of basal cell differentiation; cytokeratin 18 (CK18) as a marker of luminal (glandular) cell differentiation; CD44 as a putative endometrial stem cell marker; terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) as a marker of apoptosis; and the hormone receptors estrogen receptor alpha (ERα) and progesterone receptor (PGR). Immunohistochemical staining was performed at the NIEHS/NTP IHC core facility using standard protocols ( Painter, Clayton, and Herbert 2010 ) ( https://www.niehs.nih.gov/research/resources/protocols/protocols-immuno/index.cfm ). Reagent preparations are listed in Table 1 . Heat-induced epitope retrieval was performed using 1X Antigen Decloaker, pH 6.0 (Cat # CB910M, Biocare Medical, Concord, CA; SIX1, P63, CD44, ERα, PGR) or 1X Nuclear Decloaker, pH 9.5 (Cat # CB911M, Biocare Medical; CK18, Dual CK14/CK18) in the Decloaker pressure chamber for 5 min at 120°C. Labeling for CK14 alone did not require antigen retrieval. Endogenous peroxidase activity was quenched using 3% H 2 O 2 for 15 min. For single antigen immunolabeling, non-specific sites were blocked using either 10% animal serum (Jackson ImmunoResearch, West Grove, PA) followed by the avidin/biotin blocking kit (Vector Laboratories, Burlingame, CA) or Rodent Block M (BioCare Medical). Sections were incubated with the appropriate primary antibody or negative control antibody for 1 hour at room temperature, followed by the secondary antibody for 30 min at room temperature. Sections were treated with the R.T.U. VECTASTAIN Universal ABC Kit (Vector Laboratories; SIX1, P63, CK14, CK18, CD44, and PGR) or the Streptavidin SS Label (Biogenex Laboratories, San Ramon, CA; ERα) and visualized using 3,3′-diaminobenzidine (DAB) chromogen. TUNEL staining was performed using the ApopTag Plus Peroxidase In Situ Apoptosis Detection Kit (Cat # S7101, Millipore, Billerica, MA) using the manufacturer’s instructions. Dual immunolabeling was also performed for CK14 and CK18 to identify cells co-expressing these markers. For this staining, non-specific sites were blocked using the Antibody Blocker/Diluent (Enzo Life Sciences, Farmingdale, NY) for 10 min at room temperature. Sections were then incubated with a mixture of primary anti-CK14 and anti-CK18 antibodies or negative control antibodies for 1 hour at room temperature, followed by incubation with a 1:1 dual mix of MULTIVIEW PLUS mouse HRP/rabbit AP (Enzo Life Sciences, Farmingdale, NY). The HIGHDEF green AP and HIGHDEF DAB chromogens (Enzo Life Sciences, Farmingdale, NY) were used to visualize CK14 (rabbit antigen-antibody complex) and CK18 (mouse antigen- antibody complex), respectively. All immunolabeled sections were counterstained with hematoxylin, dehydrated through graded ethanol, cleared in xylene, and cover-slipped. Appropriate positive control tissues (mouse skin and intestine) were stained with each experiment. Stratified squamous cervical and vaginal epithelium present on each slide was also used as an internal positive control for specific markers (e.g., SIX1, P63, and CK14). Immunolabeling within endometrial and cervical epithelium was evaluated by a pathologist and assigned a qualitative labeling score from 0 to 4 based on the estimated percentage of labeled target cells and overall staining intensity. Corresponding H&E-stained sections were used as needed to confirm lesion- or cell-specific labeling. Dual labeling for CK14/18 was represented by a discrete forest green color and evaluated as for a single label; it was not possible to visually distinguish intensities of the component markers. Co-localization of IHC labels other than CK14 and CK18 was evaluated by comparing adjacent IHC-stained sections. All slides were scanned using the Aperio AT2 Scanner (Leica Biosystems Inc., Buffalo Grove, IL). Representative images were captured either from digital slides using Aperio ImageScope v. 12.3.0.5056 (Leica Biosystems Inc., Buffalo Grove, IL) or from glass slides using a Nikon Eclipse E600 light microscope (Nikon Instruments Inc., Melville, NY) and Lumenera Infinity 2-3C digital camera (Ottawa, ON). Statistical analyses were performed using GraphPad Prism, version 7.0 ( https://www.graphpad.com/quickcalcs/contingency2/ ). Incidence data between age-matched treatment groups were compared using a two-tailed Fisher exact test.