SIX1 Regulates Aberrant Endometrial Epithelial Cell Differentiation and Cancer Latency Following Developmental Estrogenic Chemical Exposure.

SIX1 is robustly expressed in the vagina and ectocervix, but the protein is not normally detectable in the neonatal or adult uterus, despite low levels of transcript expression, unless the mice are exposed neonatally to estrogenic chemicals ( 9 , 11 , 13 ). The neonatal lethal phenotype of the global knockout mice precludes determination of whether SIX1 has a role in perinatal uterine development. To determine if SIX1 is involved in postnatal uterine differentiation or function, Six1 was conditionally deleted in the female reproductive tract during neonatal development ( Fig. S1A ) ( 23 , 28 ). Treatment of Six1 +/+ and Six1 d/d mice with corn oil (CO) on PND1–5 served as a vehicle control for DES treatment. Successful Six1 deletion in CO and DES Six1 d/d mice was documented at both PND5 and 6 mos by real time PCR, immunoblotting, and immunohistochemistry; tissues from Six1 +/+ mice exposed neonatally to DES served as positive controls ( Supplementary Fig. S1B - S1D ). There were no overt morphologic differences observed in uteri or vagina between CO Six1 +/+ and Six1 d/d mice and, although a formal breeding study was not done, CO Six1 d/d female mice were fertile. Together, these data indicated that the Six1 knockout model could be used to achieve selective Six1 ablation in the female reproductive tract during neonatal development without causing gross alterations in tissue morphology or function. Despite the overall normal uterine tissue morphology, adult CO Six1 d/d mice had unexpected changes in cell morphology within some endometrial glands. At both 6 and 12 mos, glandular dysplasia was observed in ~60% of Six1 d/d mice but never in Six1 +/+ mice ( Fig. 1A , Table 1 ). The predominant location of dysplastic glands was at the tip of the uterine horn near the endometrial-oviductal junction and to a lesser extent scattered throughout the horn or near the squamocolumnar junction. When compared with CO Six1 +/+ mice ( Fig. 1B and C ), glands in CO Six1 d/d mice, particularly near the tip of the uterine horn, showed variable degrees of lumen development in that some clusters of dysplastic cells did not have clear lumens while others had microlumens. Abnormal cells within dysplastic glands were characterized by cytologic atypia ( Fig. 1D and E ), including abnormal cell shape and pseudopodia-like appendages that extended from the basal margin and disrupted the adjacent basement membrane ( Fig. 1F and G ). These microinvasive features were consistent with carcinoma but showed no evidence of progression or expansion with age in 6- vs . 12-month-old groups. The appearance of these migrating cells was consistent with the definition of “carcinoma in situ ” or “stage 0 cancer” as described in the National Cancer Institute (NCI) Dictionary of Cancer Terms ( 39 ). The dysplastic glands in the Six1 d/d mice were further characterized using dual IHC and transmission electron microscopy (TEM). Dual IHC for cytokeratin markers of basal (CK14) and glandular (CK18) cells indicated that most cells comprising dysplastic glands and all migrating cells co-expressed both markers (designated CK14+/18+ cells; Fig. 1E - G ) as compared with normal glandular and primary luminal epithelial cells (brown CK14−/18+; Fig. 1C ). CK14+/18+ cells appeared similar to an abnormal cell population previously observed in neonatally DES-exposed mice ( 12 ). To qualitatively assess the apparent absence of a basement membrane adjacent to CK14+/18+ cells within dysplastic glands, TEM was performed on two CO Six1 d/d mice. Within the tissue regions processed and assessed by TEM, morphologically normal glands (H) and two dysplastic glands (I and J) were identified. TEM indicated that glandular epithelial cells in morphologically normal glands had a distinct basement membrane along the basal margin ( Fig. 1H ). In contrast, a basement membrane adjacent to morphologically abnormal cells within dysplastic glands was not observed ( Fig. 1I and J ). Abnormal appearing epithelial cells within dysplastic glands had variably shaped nuclei often containing a single prominent nucleolus, and scant heterochromatin ( Fig. 1I and J ). In areas where the basement membrane was disrupted, adjacent cells and cellular processes extending between the glandular and stromal compartments showed junctional complexes and intracytoplasmic structures that likely represented cytokeratin microfilaments, suggesting epithelial cell origin and epithelial-to-mesenchymal transition. Together, these cellular changes indicate that deletion of Six1 disrupts epithelial cell morphology within some endometrial glands. Furthermore, these data suggest that transient expression of Six1 during postnatal development is required for normal glandular differentiation. These phenotypic abnormalities led us to hypothesize that genes involved in normal endometrial differentiation would be altered in the CO Six1 d/d mice as compared to CO Six1 +/+ mice. Microarray analysis of uteri from CO Six1 d/d vs. Six1 +/+ mice identified 2,856 differentially expressed genes (DEGs) ( Fig. 1K and Supplementary Table S2A ). The most highly significant Gene Ontology (GO) categories of the DEGs were cell division (99 genes; GO Score: 119, p=2.7E-52), nuclear division (80 genes; GO Score: 116, p=6.5E-51), and mitosis (80 genes; GO Score: 116, p=6.5E-51); 71 of these genes were found in all three categories ( Supplementary Table S2B ). In addition, >90% of the DEGs in these three categories were down regulated in Six1 d/d vs. Six1 +/+ . Nine of the top ten down regulated genes from these GO categories encoded either spindle associated proteins ( Spag5 , Kif18b , Ska3 , Mis18bp1 , Ska1 , Aurbk , Ercc6l) or cell cycle regulators ( Ccnb1 and Ccna2) . These findings are consistent with the well-established role of SIX1 in regulating cell proliferation ( 26 ). These DEGs were also analyzed using the NextBio correlation engine to identify correlations with published data sets ( 37 ). NextBio recognized 2,530 of the 2,856 DEGs initially identified in the CO Six1 d/d model; the list of highly overlapped biosets can be found in Supplementary Table S2C . Using the knockout atlas, the top gene perturbation was indian hedgehog ( Ihh ). Of the 2,856 DEGs, 443 genes overlapped with genes altered in uterine tissue from a Ihh conditional knockout (cKO) model ( Fig. 1L ; Supplementary Table S2D ) ( 40 ). Categorizing the overlapped altered genes by up or down-regulation identified 216 similarly down-regulated genes ( p =2.4E-65), including cell cycle genes Ccnb1 , Ccnd1 , Cdk1 and Mcm5 . These findings indicate a positive correlation in gene expression changes when either Six1 or Ihh are deleted in the uterus ( Fig. 1L ). A decrease in Ihh gene expression was confirmed in CO Six1 d/d vs. Six1 +/+ mice ( Fig. 1M ). Conditional uterine deletion of Ihh and Sox17 , an upstream regulator of Ihh , alters endometrial gland development, epithelial differentiation, and epithelial-stromal crosstalk ( 40 , 41 ). Taken together, these findings suggest that deletion of Six1 alters normal uterine differentiation and response, and may function through Ihh regulated pathways. To investigate if conditional Six1 deletion impacts characteristic DES-induced cellular phenotypes, histopathological changes were evaluated in the reproductive tracts of CO and DES-exposed Six1 +/+ and Six1 d/d mice. Regardless of genotype, DES exposure resulted in a similar spectrum of morphological alterations relative to CO mice. These changes included an overall reduction in the number of endometrial glands, as well as basal and squamous metaplasia, adenomyosis, atypical hyperplasia, and carcinoma ( Table 1 and Supplementary Fig. S2 ) ( 6 , 12 , 42 ). However, there were striking differences in the degree of region-specific epithelial metaplasia between DES-exposed Six1 +/+ and Six1 d/d mice. The incidence of basal cell metaplasia was not different between DES-exposed Six1 +/+ and Six1 d/d mice when the whole uterus was evaluated by histopathologic analysis (combined horn and body regions; Table 1 ). However, the severity of basal cell metaplasia at 6 and 12 mos and the incidence and severity of squamous metaplasia at 12 mos was decreased, prompting further investigation of region-specific changes using quantitative image analysis of CK14 and CK18 labeling ( Fig. 2A and S2A ). Within the uterine body, DES-exposed Six1 +/+ and Six1 d/d mice exhibited a similar percentage of metaplastic change at both 6 and 12 mos (as indicated by the percentage of CK14+ epithelium) ( Fig. 2B ). Within the uterine horns, on average 30% of the epithelium in DES-exposed Six1 +/+ mice was CK14+, and this percentage increased with age ( Fig. 2C ). However, less than 10% of the uterine horn epithelium was CK14+ in DES-exposed Six1 d/d mice, a percentage comparable to that in the CO mice, and there was no change with age. Similar alterations in region-specific metaplastic change were observed when data were expressed as total CK14+ tissue area for either region ( Fig. S2B ). These data indicate that DES-induced SIX1 acts as a differentiation factor leading to basal cell and squamous metaplasia specifically in the uterine horns. Similar to our previous results, endometrial carcinoma was observed in ~40% and ~70% of DES-exposed Six1 +/+ mice at 6 and 12 mos, respectively, but not in CO mice of either genotype ( Table 1 , Fig. 3A and B ) ( 13 ). Interestingly, there was a higher incidence of endometrial carcinoma in DES Six1 d/d compared to DES Six1 +/+ mice at 6 mos ( Fig. 3A ). By 12 mos, both genotypes had a similarly high incidence ( Fig. 3B ) and distribution of neoplastic lesions ( Table 1 ). These findings indicate that SIX1 is not required for DES-induced carcinogenesis in the uterus, and instead suggests that it may provide some degree of protection by promoting metaplastic change. Along with adenobasal morphologic features, all carcinomas in DES-exposed Six1 +/+ mice exhibited uniform epithelial SIX1+ labeling, CK14+ and P63+ basal cells, CK18+ luminal cells, and CK14+/18+ epithelial cells within neoplastic lesions in both the uterine body and horns ( Fig. 3C ). Within the uterine body, no clear differences in lesion morphology and staining patterns for CK14, CK18, CK14/18, and P63 were observed between DES-exposed Six1 +/+ and Six1 d/d mice. However, in DES-exposed Six1 d/d mice, uterine horn lesions showed distinct morphologic patterns and IHC markers, including most notably a lack of P63+ and CK14+ epithelial cells with discrete basal morphology ( Fig. 3C ). Rather, abnormal glands in Six1 d/d mice were comprised of a single luminal cell layer that in ~90% of cases contained CK14+/18+ labeled cells. Carcinoma lesions comprised of CK14+/18+ cells in DES Six1 d/d exhibited significant expansion of atypical glands as well as stromal and myometrial invasion that distinguished them from focal glandular dysplasia/carcinoma in situ in CO Six1 d/d described above. Although the degree of basal cell metaplasia was strikingly different between DES Six1 +/+ and Six1 d/d groups, both exhibited neoplastic lesions containing CK14+/18+ cells ( Table 1 ). It is important to note that neoplastic cells within DES-exposed Six1 d/d mice were not confined to a particular uterine area (ie. uterine horn or uterine body; Table 1 ). This finding indicates that conditional Six1 deletion does not shift the location of carcinoma development in a way that is similar to the shift in the appearance of basal cell metaplasia (ie. tumors were not confined to the uterine body similar to the basal cell metaplasia). Taken together, these findings suggest that within the uterine horn, SIX1 is required upstream of P63 for basal cell differentiation and acts as a driver for distinct adenobasal carcinoma morphology following DES exposure. Gene expression analysis of uteri from DES Six1 d/d vs. Six1 +/+ mice identified 993 DEGs ( Fig. 3D and Supplementary Table S2E ). The most significantly altered GO categories were immune response (31 genes; GO Score: 24, p =5.10E-11), epithelial cell differentiation (19 genes; GO Score: 22, p =3.7E-10) and epidermis development (19 genes; GO Score: 20, p =1.3E-9) ( Supplementary Table S2F ). The majority of genes in all of these categories were downregulated. Downregulation of Krt14 and Trp63 was confirmed in DES Six1 d/d vs. Six1 +/+ by real time RT-PCR ( Supplementary Fig. S3A ). Alterations in immune response pathways may arise from the role of SIX1 in uterine epithelial differentiation, which if disrupted, may impact the epithelium’s function as a mucosal immune barrier. The NextBio correlation engine recognized 879 of the 993 DEGs initially identified in the DES-exposed Six1 cKO model; the list of highly overlapped biosets is in Supplementary Table S2G . The phosphatase and tensin homolog (PTEN) knockout bioset highly overlapped with the DES-exposed Six1 cKO model with a gene perturbation score of 100 ( Supplementary Table S2G ). Of the 879 DEGs identified in the DES-exposed Six1 cKO model, 407 genes overlapped with genes altered in prostate tumor tissue from a Pten cKO model ( Fig. 3E , Venn diagram; overlapped gene list can be found in Supplementary Table S2H ) ( 43 ). Categorizing altered genes by up or down-regulation identified 214 genes downregulated in the Six1 cKO model that are upregulated in the Pten cKO model ( p =4.1E-28), indicating a negative correlation in gene expression changes when comparing Six1 with Pten deletion in glandular tumors ( Fig. 3E ). Pten loss leads to ~10-fold increase in Six1 and 1.8-fold increase in Trp63 , similar to increases of these two genes following DES exposure in Six1 +/+ mice ( Supplementary Fig. S2H ), suggesting PTEN pathways regulate Six1 , even though Pten itself is not differentially expressed in DES Six1 d/d uteri. Trp63 appears to be downstream of Six1 because Trp63 is significantly reduced in DES Six1 d/d uteri compared to the DES Six1 +/+ mice ( Supplementary Fig. S3A ). Pten loss is a widely recognized driver of prostate and endometrial carcinogenesis ( 44 , 45 ). Conditional Pten deletion in mouse prostate and endometrial cancer models results in rapid tumor formation, metastasis, and death ( 43 , 46 ). Interestingly, Pten loss leads to highly proliferative squamobasal metaplastic changes, indicating its role in proliferation and cell fate. This disease contrasts with that of DES-induced endometrial cancer, which progresses slowly, is locally highly invasive but rarely metastatic, and appears against a background of abnormally differentiated CK14+ basal and poorly differentiated CK14+/18+ cells but not florid epithelial hyperplasia. Taken together, these findings suggest that DES-induced endometrial cancer functions through a Pten -independent pathway. To determine how many genes exhibited a pattern of “protection” from DES induced gene expression changes, we performed EPIG analysis ( 38 ). The resulting patterns are shown in Supplementary Fig. S3B . Only one pattern exhibited the protection pattern of interest: genes that were up-regulated in DES Six1 +/+ compared to CO Six1 +/+ with DES Six1 d/d having gene expression more similar to CO Six1 +/+ (pattern outlined in red; Supplementary Fig. S3B ). A GO analysis of the 585 genes in this pattern revealed that the 20 most highly significant categories were all related to mitosis/cell division ( Supplementary Tables S2I and J ). These findings suggest that SIX1 mediates some of the impact of DES on cell proliferation in the uterus. The Six1 knockout mouse model highlighted the potential role of poorly differentiated CK14+/18+ cells as a progenitor population in endometrial cancer development. To determine if similar cells were present in human endometrial cancer, dual CK14/18 IHC was performed on human endometrial tissue microarrays ( Table 2 , Fig. 4 ). CK14+/18+ cell labeling was assessed both manually and by digital image analysis. Features of the human CK14+/18+ cells were similar to those identified in mice ( Fig. 4A ). CK14+/18+ labeled cells lacked clear luminal or basal morphology and were often embedded within areas of CK14−/18+ luminal epithelium or migrating into adjacent stroma as atypical individual cells or cell clusters. Manual and digital analyses were generally consistent in the identification of CK14+/18+ labeling, and neither analysis identified any CK14+/18+ labeling in patients with normal or nonneoplastic endometrial tissue. It should be noted that this analysis, using small tissue microarray cores, cannot rule out the possibility of rare CK14+/18+ cells in normal human endometrial glands, as was observed in much larger sections from CO Six1 +/+ mice. Both manual and digital techniques identified CK14+/18+ neoplastic cells as a feature in ~30% of endometrial cancers, although the manual analysis identified 17 more patients with CK14+/18+ cells (Fig. B and C). CK14+/18+ cells comprised only a small portion of cells present within the core; the manual average severity score was 1.7 (~6–10 cells based on scoring criteria) and the average percent area of CK14+/18+ labeled core tissue was 0.12% (~20 cells based on average CK14+/18+ cells size and core area). The manual assessment was likely more accurate because visual inspection of biopsies from the 17 patients considered negative for CK14+/18+ labeling by digital image analysis clearly indicated CK14+/18+ labeling in at least one discrete cell. Further analysis of CK14+/18+ cores identified by manual assessment showed that the presence of CK14+/18+ labeled cells positively correlated with both increasing cancer stage and grade ( Fig. 4D and E ). These findings are consistent with the idea that poorly differentiated CK14+/18+ cells could serve both as progenitor cells that may become transformed in the human endometrium and as a biomarker of more aggressive disease.

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. All genetic mouse founders were bred to FVB/NJ mice for five generations to obtain the lines on an FVB/NJ background. Mice containing loxP sites flanking Six1 ( Six1 fl/fl ) mice were provided by Dr. Pascal Maire ( 23 ). The fifth generation FVB/NJ backcross Six1 fl/+ animals were inbred to generate Six1 +/+ and Six1 fl/fl founder lines, which were then used to generate experimental groups. FVB/NJ mice containing Cre -recombinase under the control of the endogenous progesterone receptor promoter ( Pgr-cre ) were provided by Drs. Franco DeMayo and John Lydon ( 28 ). The Pgr-cre line induces CRE expression as early as postnatal day 3 (PND3) in Pgr -expressing cell types including the uterine luminal epithelium, stroma, and myometrium ( 29 ). Male Six1 fl/fl , Pgr-cre mice were bred to female Six1 fl/fl mice to generate the experimental Six1 conditional knockout line ( Six1 d/d ). Female Six1 +/+ and Six1 d/d pups were given daily subcutaneous injections (0.02 ml) of 1 mg/kg diethylstilbestrol (DES) in corn oil or corn oil (CO) alone as a vehicle control on the day of birth (postnatal day 1, PND1) through PND5, consistent with the previously described DES exposure model ( 6 , 13 , 30 ). 1 mg/kg/day of DES on PND1–5 is an established dose with high efficacy (~90% uterine cancer), which we used to reduce variability in our endpoints in order to understand the molecular mechanisms underlying endometrial carcinogenesis. Sample sizes were calculated based on previously published studies showing that 45% of FVB/N mice exposed neonatally to DES develop cancer by 8 months (mos) of age ( 31 , 32 ). We estimated that using 12 mice per exposure group and genotype would allow us to detect a 45% difference in cancer incidence (one-sided 0.05 level of significance with 80% power) beginning at 6 mos. Mice were euthanized by decapitation on PND5 or CO 2 asphyxiation at 6 and 12 mos and uteri were collected and processed for molecular and histopathologic analysis as previously described ( 12 , 13 ). In 6- and 12-month groups, the cranial half of each right uterine horn was frozen for molecular analyses (microarray and RT-PCR) and thus not available for histopathology. The remaining tract, including left uterine horn, uterine body, cervix, and cranial vagina was used for histopathology and immunohistochemistry (IHC). Human endometrial tissue microarrays (TMA) containing normal, nonneoplastic, normal cancer-adjacent, and malignant biopsies were purchased from a commercial vendor (U.S. Biomax, Inc.; TMA# UT1501, UT803, UT721, EMC961, EMC962). All tissue specimens were obtained with written informed consent according to U.S. federal law. TMAs were prepared from formalin-fixed, paraffin-embedded tissue specimens and freshly cut for IHC staining. Tissue cores from 223 patients across 5 TMAs were assessed for CK14+/18+ cell labeling (normal: 29, normal cancer adjacent: 8, hyperplasia: 5, malignant: 181). Two to three tissue cores were assessed for 211 patients. Four to five cores were assessed for the 12 patients who had redundant tissue cores present on more than one TMA (six patients overlapped on UT803 and UT1501, six patients overlapped on UT803 and UT721). Pathologic diagnoses and descriptions, including FIGO or TNM stage and grade, were provided for most patients. Cancer stage information was provided for 180 of 181 endometrial cancer patients (63/63 patients with CK14+/18+ cells and 117/118 patients lacking CK14+/18+ cells). Cancer patients ranged from stage I-III. There were no stage IV patients. For consistency, TNM scores were converted to FIGO stage based on previously described guidelines ( 33 ). Cancer grade, ranging from grade 1–3 (G1–3), was provided for 159 of 181 endometrial cancer patients (55/63 patients with CK14+/18+ cells and 104/118 patients lacking CK14+/18+ cells). Female mouse reproductive tract 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 IHC. IHC analysis was performed on serial sections from the same mice used for histopathology. Reproductive tracts were sectioned until the central uterine lumen could be observed in plane with the cervicovaginal epithelium. This procedure provided 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). IHC and immunofluorescence (IF) labeling were performed at the NIEHS/NTP IHC core facility using standard protocols ( 34 ). Reagent preparations for both mouse and human tissue are listed in Supplementary Table S1 . Detailed staining methods are provided elsewhere ( 12 ). Brightfield IHC was performed on serial sections of uterus from each mouse for the following markers: SIX1 as a developmental differentiation factor and cancer cell marker; CK14 and tumor protein 63 (P63) as a marker of basal cells; CK18 as a marker of glandular cells; and dual CK14/18 as a marker of poorly differentiated mixed basal/glandular cells. CK14/18 IF was also performed to highlight co-localization of CK14 and 18 labeling. Dual CK14/18 IHC was performed on human endometrial TMAs using the same protocol as described for mice. Appropriate positive and negative control tissues (mouse skin and intestine) were stained with each experiment. For mouse histopathologic assessment, a certified study pathologist examined three H&E-stained sections spaced approximately 30 μm apart for each mouse. Where applicable, histopathologic diagnoses were based on standard criteria and nomenclature for neoplastic and nonneoplastic lesions presented by the International Harmonization of Nomenclature and Diagnostic Criteria for Lesions in Rats and Mice (INHAND) Project ( 35 ). Any discrepancies with INHAND terminology are described in the results section. Severity of nonneoplastic 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. Immunolabeling for dual CK14 and 18 within mouse and human epithelial cells was also evaluated by a pathologist and assigned a qualitative labeling score. For mice, the score ranged from 0 to 4 based on the estimated percentage of CK14+/18+ labeled cells (0=absent, 1=minimal, 2=mild, 3=moderate, 4=severe); corresponding H&E-stained sections were used as needed to confirm lesion- or cell-specific labeling. For humans, the score ranged from 0 to 4 based on the approximate number of CK14+/18+ cells present within the core biopsy (0= absent, 1= 1–5 cells, 2= 6–10 cells, 3= 11–15 cells, 4= >15 cells). CK14+/18+ cell labeling was represented by a discrete forest green color (IHC; mouse and human) and evaluated as for a single label; it was not possible to visually distinguish intensities of the component markers in dual-stained areas. To avoid potential bias, the uterus was read as a single organ for manual H&E and IHC evaluation and was not divided into uterine horn and body regions as in the digital image analysis described below. All pathology data, tabulations, and observations were recorded by a board-certified veterinary pathologist (CEW). Brightfield and fluorescent slides were scanned using the Aperio AT2 Scanner and the Aperio Scanscope FL Scanner, respectively (Leica Biosystems Inc., Buffalo Grove, IL). All fluorescent slides were scanned using the same exposure. Representative brightfield or fluorescent images were captured from digital slides using Aperio ImageScope v. 12.4.0.5043 (Leica Biosystems Inc., Buffalo Grove, IL). IF staining included CK14 (Alexa 568), CK18 (Alexa 488), and DAPI, but CK14 and CK18 colors were artificially inverted during imaging. The Aperio Colocalization Algorithm (Leica Biosystems Inc., Buffalo Grove, IL) was used to quantitate CK14+ or CK18+ labeled area within different reproductive tract regions of interest (ROIs) in slides stained individually for CK14 or CK18. The uterine body ROI was defined as the area from the squamocolumnar junction (but above visible stratified squamous epithelium) up to the horn bifurcation (outlined in Fig. 2 ). In most CO mice, the uterine body region included a small CK14+ area indicating basal/reserve cells typically present in endometrial glands within the transition zone. The uterine horn ROI was defined as the area above the left horn bifurcation up to the utero-tubal junction. Region-specific quantification of CK14+ or CK18+ labeled area was based on the total area containing strong positive pixels per ROI. To account for variability in overall organ size due to epithelial, stromal, and myometrial development, all of which are impacted by neonatal DES exposure, the CK14+ or CK18+ labeled area was represented as a percent of total epithelial area (as determined by adding the CK14+ and CK18+ tissue area on serial sections) rather than a raw positively stained area or as a percentage of the total tissue area. This method determines the proportion of epithelium that is “normal” glandular/simple columnar (CK14−/18+) or squamous/basaloid (CK14+/18−) and how this ratio is altered by treatment and genotype. Section variability between animals could also contribute to variability; therefore, serial sections were used for CK14 and CK18 IHC in this analysis. Despite several attempts to digitally quantitate independent and co-localized CK14 and 18 staining within dual CK14/18 labeled mouse slides, an accurate estimate of CK14+/18+ labeled area could not be generated (CK14+/18+ labeled area was grossly overestimated, whereas CK14+/18− area was underestimated; parameter editing led to significant data loss of either group). For this reason, CK14+/18+ cells were assessed only by manual analysis. The Definiens Tissue Studio TMA Colocalization Algorithm (Definiens, Inc., Cambridge, MA) was used to quantitate CK14+/18−, CK14−/18+, or CK14+/CK18+ labeled area within dual CK14/CK18 stained slides. The percent area of tissue that contained positive labeling was calculated for each core biopsy from a single patient and averaged to create a score for each stain per patient. Note that multiple core biopsies assessed for a single patient showed similar tissue pathologies (e.g., multiple biopsies taken from a cancer lesion or lesions) and that staining scores from biopsies of malignant and normal adjacent tissue from a single patient were not averaged. The average percent area of tissue that defined a single cell for each label (CK14+/18−, CK14−/18+, and CK14+/18+) was calculated to establish minimum labeling cut-offs. TMA cores with positively labeled areas that were less than the average size of one cell for that respective label were considered negative. It was necessary to define an average cell size for each respective label to account for differences across labeling groups (e.g., CK14+/18− basal cells were often larger than CK14−/18+ luminal or poorly differentiated CK14+/18+ cells). Average cell sizes were 87 μm 2 , 49 μm 2 , and 41 μm 2 and average percentages of core area per single cell were 0.0080%, 0.0045%, and 0.0037% for CK14+/18−, CK14−/18+, and CK14+/18+ labeling, respectively. The right anterior uterine horn (near the oviductal-endometrial junction) was collected from two CO Six1 d/d mice at 6 mos. Tissues were fixed in McDowell and Trump 4F:1G fixative overnight and processed using the automated Leica Electron Microscopy Tissue Processor (Leica Biosystems Inc., Buffalo Grove, IL). Samples were rinsed with phosphate buffer, post-fixed in 1% osmium tetroxide, rinsed in water, and dehydrated in an ethanolic series culminating in acetone. The samples were then infiltrated with Poly/Bed 812 epoxide resin. Following polymerization, blocks were trimmed and semithin sections (~0.5 μm thick) were cut, mounted on glass slides, and stained with 1% Toluidine Blue O in 1% sodium borate. Slides were examined with a light microscope to select a ROI and trimmed. Ultrathin sections (80–90 nm thick) were cut, placed onto 200 mesh copper grids, and stained with uranyl acetate and lead citrate per standard protocol. Digital images were captured with a Gatan Orius SC1000/SC600 camera (Gantan Inc., Pleasanton, CA) attached to a FEI Tecnai T12 Transmission Electron Microscope (FEI Company, Hillsboro, OR). TEM evaluation was performed by the NIEHS Electron Microscopy Support Laboratory. RNA was isolated from frozen uterine horn tissue and real time RT-PCR was performed as described previously using Six1 primers (F: CTGCCGTCGTTTGGTTTTAC; R: TTGTAGAGCTCGCGGAAGTT) and normalized to cyclophilin A ( Ppia ) ( 13 ). Transcript expression levels were calculated using the delta Ct method ( 9 , 36 ). Gene expression analysis was conducted on uterine horn tissue from four independent biological replicates for each exposure group and genotype at 6 months-of-age using Agilent Whole Mouse Genome 4 × 44 multiplex format oligo arrays as described previously (Agilent Technologies, Santa Clara, CA) ( 9 ). Data were obtained using Agilent Feature Extraction software (version 12.0), which performed error modeling adjusting for additive and multiplicative noise. The resulting data were processed using Partek Genomics Suite (Version 7.0) software (Partek Inc., St. Louis, MO). To identify differentially expressed probes, a raw data cutoff <10 and log2 transformed analysis of variance (ANOVA) with unadjusted p <0.05 was applied to determine if there was a statistical difference between the means of groups. Principal Component analysis (PCA) was performed to assess variability present in the data. Samples separated into distinct groups based on exposure and genotype except for a single CO Six1 +/+ sample that did not cluster with other groups and was excluded from further analysis. The resulting significantly altered genes were subjected to a ± 1.5-fold change cutoff that were used to generate heat maps in Partek and then uploaded into NextBio Correlation Engine (Illumina Inc., San Diego, CA) for pathway enrichment and correlation to published studies ( 37 ). Microarray data were also evaluated for patterns using the Extracting Patterns and Identifying co-expressed Genes (EPIG) analysis method ( 38 ). Microarray data have been deposited in the Gene Expression Omnibus (GEO) database (accession number GSE138501 ). As described previously, nuclear protein was extracted from uterine horn tissue and SIX1 immunoblotting was performed on PND5 and 6 mos tissue (Suen et al., 2016). Blots were scanned using the HP Scanjet 7650 (Hewlett-Packard, Palo Alto, CA). Images were desaturated in Adobe Photoshop Elements (Adobe, San Jose, CA) to remove color without altering the brightness value of the pixels. Statistical analyses were performed using GraphPad Prism, version 7.0 (La Jolla, CA). Data were analyzed using two-way ANOVA, two-tailed Fisher’s exact test or Chi-square test, Mann Whitney test, and appropriate post hoc tests for multiple comparisons. Tests used are indicated in Figure Legends. Error bars show the standard error of the mean (SEM) for all graphs.

We previously identified basal CK14+/18− and poorly differentiated CK14+/18+ epithelial cell populations associated with neoplastic lesions developing after neonatal DES exposure and showed that SIX1 overexpression precedes and localizes to these abnormal cell populations. These findings led us to hypothesize that SIX1 is a driver of these aberrant differentiation pathways ( 12 ). Here, we show that SIX1 mediates development of CK14+/18− basal cells specifically in the uterine horns, but it is not required for the development of CK14+/18+ cells or endometrial cancer. In the absence of SIX1, epithelial cells co-expressing both CK14 and CK18 become a prominent feature of atypical and neoplastic lesions in the uterine horns, and DES-induced endometrial carcinomas develop more rapidly. These results suggest that SIX1 may act as a cancer-protective factor that facilitates basal differentiation of poorly differentiated CK14+/18+ progenitor cells, preventing or decreasing their transformation. Importantly, these results also provide a mechanistic link between developmental exposure to an estrogenic chemical, altered transcription factor expression, and disease phenotype. Although SIX1 has been studied primarily as a developmental driver and oncoprotein, our findings clearly demonstrate that SIX1 also serves as an epithelial differentiation factor in the endometrium. This finding is consistent with previous studies investigating SIX1 as a regulator of specific stem cell populations. For example, increased expression of SIX1 promotes differentiation in muscle stem cells, whereas ablation of Six1 during muscle regeneration results in an increased number of muscle stem cells and diminished capacity of muscle stem cells to regenerate functional muscle ( 23 , 47 ). SIX1 is also one of several transcription factors that promote differentiation of mesenchymal stem cells into brown adipocytes ( 48 ). Human embryonic stem cells can be induced to differentiate into lacrimal gland epithelial-like cells by simultaneous overexpression of three transcription factors, SIX1, PAX6, and FOXC1 ( 49 ). Notably, lacrimal gland cells have certain features similar to that of the endometrial epithelium, including their glandular morphology and expression of characteristic genes, including CK15, aquaporin 5, and lactoferrin ( 49 ). Together with our findings, these data demonstrate the dynamic and tissue-specific role of SIX1 in regulating progenitor cell populations and promoting differentiation. We propose the following model for the role of SIX1 in carcinogenesis following neonatal DES exposure ( Fig. 5 ). During normal uterine development, a population of poorly differentiated CK14+/18+ epithelial cells arise in the endometrial glands. Under normal conditions (no DES exposure) transient SIX1 expression mediates differentiation of these cells into mature luminal CK14−/18+ cells in the presence of endogenous estrogen. Neonatal exposure to DES results in epigenetic alterations that lead to initiation and promotion of the CK14+/18+ population either directly (e.g. via ER activation) or indirectly (e.g. via increased response to endogenous estrogen levels after puberty) ( 6 , 50 ). The initiated CK14+/18+ cells that may ultimately become transformed serve as a pool of cancer cells-of-origin ( 51 , 52 ). Persistent upregulation of SIX1 enables many of the CK14+/18+ cells to differentiate into mature CK14+/18−/SIX1+ basal cells or CK14−/18+/SIX1+ luminal cells. These basal cells surround luminal cells and may in some cases progress to squamous metaplasia ( 12 ). SIX1+ cell types that exhibit more mature basal (CK14+/18−) or luminal (CK14−/18+) differentiation patterns may be inherently less susceptible to transformation because of their more advanced differentiation state ( 53 , 54 ). In our model, the absence of SIX1 results in a differentiation blockade of CK14+/18+ epithelial cells, leading to dysplastic endometrial glands and a larger pool of progenitor cells that may be later promoted to neoplasia following DES exposure. Treatment strategies for human endometrial cancer are currently driven by surgical staging and tumor histology ( 55 , 56 ). Endometrial cancers are commonly separated into two subtypes characterized by aggressiveness, hormone-responsiveness, and histology. Type I tumors are the most common (~85% prevalence) and are typically less aggressive, dependent upon estrogen, and endometrioid by histopathology. In contrast, Type II tumors are less common (~15%), generally more aggressive, and estrogen independent with serous, clear cell, or undifferentiated morphology ( 57 , 58 ). Early stage endometrioid disease has a favorable (95%) 5-year survival rate, but the survival rate decreases rapidly for later-stage disease (III:67% and IV: 16%) ( 59 ). High stage and/or grade tumors with endometrioid histology have a less favorable outcome and demonstrate molecular characteristics that overlap between both type I and II ( 60 ). Morphologic and molecular sub-classification to identify key tumor signaling pathways will provide information to help develop and guide tissue-agnostic therapeutic strategies in combination with surgery, chemotherapy, and radiation. Here, we show that the CK14+/18+ cell labeling is a feature of 32% of endometrioid tumors and correlates with increasing cancer stage and grade, suggesting its use as a biomarker for less differentiated, high stage or grade Type I tumors. We propose that CK14+/18+ labeling is a novel combination biomarker for endometrial cancer that may support current classification protocols of poorly differentiated and/or more aggressive disease. Identification of clinically actionable biomarkers will continue to improve treatment strategies and disease outcomes and may provide insights into how environmental exposures can impact health and disease.

Abnormal differentiation is a key feature for many types of human cancer. However, the events leading to cellular reprogramming and aberrant cell fate often occur long before malignancy, making it difficult to identify important early drivers. Exposing female mice on days 1–5 of life to estrogenic chemicals provides an established model for early-life alterations in differentiation and development that lead to carcinogenesis later in life ( 1 ). This animal model was originally developed using the potent synthetic estrogen, diethylstilbestrol (DES), a drug widely prescribed during pregnancies in the 1940s-1970s for the prevention of miscarriage (2.5–150 mg/day; 0.033–2 mg/kg/day) ( 2 , 3 ). Human prenatal exposure resulted in reproductive tract abnormalities and increased cancer risk in exposed offspring ( 4 , 5 ). Female mice neonatally exposed to 1 mg/kg/day DES develop abnormalities of the reproductive tract and a high incidence of endometrial cancer later in life ( 6 ). Cancer incidence is dependent on estrogenic activity of the dose and occurs even at very low doses (≤0.0001 mg/kg/day), suggesting that weaker estrogenic chemicals could also cause cancer ( 7 ). In fact, many environmental estrogens such as genistein (phytoestrogen), bisphenol A (plasticizer), and nonylphenol (surfactant) have been assessed in this model and shown to cause endometrial cancer ( 7 , 8 ). Although this exposure model has been widely studied, the mechanisms linking early-life exposure and carcinogenesis later in life remain unknown. Previous work indicates that neonatal exposure to estrogenic chemicals fundamentally alters developmental patterning of the mouse female reproductive tract ( 1 , 9 - 11 ). One of the most highly altered proteins is sine oculis-related homeobox 1 homolog (SIX1), a transcription factor that regulates the development of many tissues and becomes reactivated or overexpressed in multiple types of human cancer. In the mouse female reproductive tract, Six1 transcript and protein expression is normally present in the cervical and vaginal stratified squamous epithelium but is absent from the endometrial glandular epithelium ( 9 , 11 , 12 ). Following neonatal exposure to DES, SIX1 is stably upregulated in the mouse endometrium, further induced by endogenous estrogen exposure, and localized within dysplastic and neoplastic glandular lesions that develop in adulthood ( 9 , 11 - 13 ). These findings suggest that aberrant uterine expression of SIX1 following neonatal estrogenic chemical exposure serves as a functional link between early disruption of cellular differentiation and later development of endometrial carcinomas in this model. SIX1 is a homeodomain-containing transcription factor that plays essential roles in organogenesis and tissue maintenance by regulating cell proliferation, differentiation, survival, migration, and invasion ( 14 - 16 ). SIX1 can function as a transcriptional activator or repressor ( 14 , 17 ) and regulates a diverse network of downstream pathways through interaction with cofactors such as dachshund (DACH) and eyes absent (EYA) ( 14 , 18 , 19 ). Six1 -null mice and pigs display numerous developmental abnormalities, collectively resulting in perinatal lethality ( 14 , 20 , 21 ). In adult mice, SIX1 mediates muscle and kidney tissue regeneration through stem cell maintenance and differentiation ( 22 , 23 ). In humans, loss-of-function mutations in SIX1 disrupt kidney and inner ear development resulting in progressive renal failure and hearing loss characterized as branchio-oto-renal syndrome ( 24 ). In addition to normal developmental and maintenance functions, SIX1 dysregulation and inappropriate reactivation is implicated in several cancer hallmarks and enabling characteristics ( 19 , 25 , 26 ). In animal models and in vitro studies, SIX1 overexpression induces genomic instability, malignant transformation, and metastasis ( 15 , 16 , 19 , 26 ). Upregulation of SIX1 expression has also been described in multiple human cancers, including breast, cervical, and ovarian cancers, and is associated with tumor resistance and poor survival ( 15 ). We and others previously showed that aberrant SIX1 expression is present in a subset of human endometrial cancers ( 13 , 27 ). However, the role of SIX1 in endometrial cancer remains unclear. Collectively, these findings indicate that aberrant uterine SIX1 may have dual roles in both cellular differentiation and carcinogenesis. Here we used a conditional knockout mouse model to investigate the role of SIX1 in normal endometrial epithelial differentiation and carcinogenesis following neonatal DES exposure. We also tested human endometrial tissue biopsies for the presence of molecular changes similar to those observed in our mouse model. The findings reported here may inform cancer biomarker development, therapeutic strategies, and risk assessment for estrogenic chemicals.