Cells expressing PAX8 are the main source of homeostatic regeneration of adult mouse endometrial epithelium and give rise to serous endometrial carcinoma

During reproductive age, in humans and mice, the endometrial epithelia and stroma undergo cell loss and regrowth in a cyclical manner in response to steroid hormones. During ovulation, there is an increase in progesterone levels that results in differentiation of the endometrium, with subsequent turnover in the absence of embryo implantation. In humans, this turnover involves endometrial shedding manifested by menstruation. In mice, the turnover is accomplished by cell apoptosis and reabsorption. In both species, cyclical homeostatic regeneration of the endometrial epithelium is expected to be maintained by stem cells. However, the presence and location of such cells has been actively debated for the past several decades (reviewed by Gargett et al., 2016; Wang et al., 2020).

A number of recent studies have suggested the location of stem cells in either the glandular (Kaitu'u-Lino et al., 2010; Syed et al., 2020) or luminal (Chan and Gargett, 2006; Huang et al., 2012; Jin, 2019) compartments of the mouse endometrial epithelium. In humans, such cells are commonly thought to be located in the basalis segment of the endometrial glands (Nguyen et al., 2012; Patterson and Pru, 2013; Valentijn et al., 2013). However, according to a recent single-cell transcriptome study of secretory-phase human endometrium, cells showing characteristics of stem/progenitor cells are located in the upper region of the functionalis (Wu et al., 2018 preprint). At the same time, several studies have suggested that endometrial epithelium can be regenerated by stem cells of either stromal/mesenchymal (Huang et al., 2012; Patterson et al., 2013) or bone marrow cell origin (Bratincsak et al., 2007; Du and Taylor, 2007; Ikoma et al., 2009; Mints et al., 2008; Taylor, 2004).

By analogy with stem cells in other organs and tissues (Fu et al., 2018), aberrations in the mechanisms governing endometrial epithelial stem cells can lead to a number of pathological conditions, including cancer. Uterine cancer is the fourth most frequent malignancy and the sixth cause of cancer-related deaths in women in the USA (Siegel et al., 2020). Although the incidence and mortality rates of some cancers, such as lung and colorectal cancers, are declining, both are increasing for endometrial carcinoma (Felix et al., 2017; Siegel et al., 2020). Endometrioid and serous subtypes are the most common subtypes of endometrial carcinoma and are characterized by distinct genetic alterations, pathological phenotypes and clinical behavior. Serous endometrial carcinoma (SEC) is the second most common type of endometrial carcinoma, accounting for ∼10-15% of all cases. SEC often presents at a late stage, recurs even after aggressive, adjuvant therapy and is responsible for the majority of deaths associated with endometrial carcinoma.

Recent extensive integrated genomic analyses of endometrial carcinomas have provided important insights into the repertoire of molecular aberrations characteristic of this malignancy (Berger et al., 2018; Kandoth et al., 2013). According to The Cancer Genome Atlas (TCGA) data, 77% of endometrioid carcinomas contain mutations in PTEN, but few alterations in TP53 (1.1%) and the RB pathway, such as upregulation of CCNE1 (none), CDKN2A (3%), E2F1 (2.2%), CDK2 (none), CDK4 (1.1%) and CDK6 (1.1%), and deletion of RB1 (2.2%). SECs have TP53 mutations in 95% of cases (Berger et al., 2018; Kandoth et al., 2013). Additionally, as the second most common alteration, >70% of SECs have aberrations in the RB pathway, involving upregulation of CCNE1 (35%), CDKN2A (18%), E2F1 (18%), CDK2 (17%), CDK4 (7%) and CDK6 (12%), and deletion of RB1 (7%). Unfortunately, utilization of this information is compromised because the originating cell types have not been determined.

In the present report, we describe a single-cell transcriptomic atlas of the main cell types of the mouse uterus and an epithelial specific subset transcriptome and identify PAX8⁺ cells as the main contributor to homeostatic regeneration of the endometrial epithelium. We also show that conditional inactivation of Trp53 and Rb1 in PAX8⁺ endometrial epithelial cells of adult mice leads to neoplasms that recapitulate human SEC.

To identify cells with expression limited to endometrial epithelium, we have analyzed single-cell mRNA sequencing data published in the Mouse Cell Atlas (MCA; Han et al., 2018). The MCA contains >400,000 single-cell transcriptomic profiles from 51 mouse tissues, organs and cell cultures, including 3761 cells from the uterus. For the uterus, in-depth analysis of the data was not performed. However, two epithelial clusters were identified and annotated. One cluster was characterized by high Ltf expression, whereas the other had high expression of Sprr2f. To improve the resolution of these data, we have re-analyzed the MCA uterus dataset using the Seurat pipeline and sctransform data normalization method to account better for technical noise while preserving biological heterogeneity (Hafemeister and Satija, 2019 preprint). We then prepared a uniform manifold approximation and projection (UMAP) analysis of the main cell types of the mouse uterus (Fig. 1A) and an epithelial subset (Fig. 1B). Furthermore, we performed unsupervised shared nearest neighbor (SNN) clustering and revealed three subpopulations (Fig. 1C; Table S1). Two groups (0 and 1) had preferential expression of Sprr2f and Tacstd2 (Fig. 1D). Consistent with a previous report of luminal epithelium-specific Tacstd2 expression (Filant and Spencer, 2013), TROP2, encoded by Tacstd2, has been detected exclusively in the luminal epithelium (Fig. 1D). Thus, groups 0 and 1 have been annotated as the luminal epithelium. Group 2 was characterized by preferential expression of Ltf and Foxa2. Based on glandular-specific expression of FOXA2 (Jin, 2019; Fig. 1D), this group was annotated as the glandular epithelium. Some markers associated with cell stemness and fate (Itga6, Klf5, Met and Tacstd2) were detected in the luminal subpopulations, whereas others (Aldh1a1, Axin2, Lgr5 and Foxa2) were present in the glandular group (Fig. 1D,E). This suggests a possibility that the endometrium might contain not one but two stem cell pools.

According to single-cell transcriptome analysis, expression of the transcription factor PAX8 was largely present in the endometrial epithelium (Fig. 2A), being detected in both luminal and glandular groups (Fig. 2B). Immunostaining showed the presence of PAX8 in epithelial but not stromal cells of the endometrium (Fig. 2C). To support this observation, we used Tg(Pax8-rtTA2S*M2)1Koes/J (Pax8-rtTA) Tg(tetO-Cre)1Jaw/J (Tre-Cre) mice (Perets et al., 2013) crossed to Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J (Ai9) reporter mice (JAX stock no. 007909). In these mice, the Pax8 promoter drives the expression of reverse transactivating protein (rtTA), which, in the presence of doxycycline, binds to the tetracycline response element (TRE), thereby leading to Cre-loxP-mediated activation of the reporter gene in Ai9 mice (Fig. 2D). In Ai9 mice, expression of red fluorescent protein variant (tdTomato) under the control of CAG promoter at the Rosa26 locus is possible only after Cre-mediated deletion of the stop codon flanked by loxP sites (Madisen et al., 2010). Lineage tracing, using the Pax8-rtTA Tre-Cre Ai9 mice, revealed that >90% of endometrial epithelial cells were labeled 2 days after a single intraperitoneal administration of doxycycline (Fig. 2E). The majority of the luminal and glandular epithelium continued to express tdTomato for ≥300 days after the doxycycline pulse. These results support the notion that epithelial cells are responsible for long-term maintenance of the epithelium throughout the estrous cycle.

Using the Pax8-rtTA Tre-Cre model, but with Confetti mice [Gt(ROSA)26Sor^{tm1(CAG-Brainbow2.1)Cle}/J; JAX stock no. 013731] instead of Ai9 reporter mice (Fig. 3A), we observed random labeling of individual cells with either green fluorescent protein (GFP) or red fluorescent protein (RFP) in both luminal and glandular epithelial compartments 3 days after a single doxycycline pulse (Fig. 3B-D). However, at later stages, both compartments became composed increasingly of monochromatic groups of cells (Fig. 3B-D). These findings show clonal expansion of PAX8-expressing endometrial epithelium.

Given that TP53 mutations are present in 95% of SEC cases and alterations in the RB pathway in >70% of SEC cases, we conditionally inactivated Trp53 and Rb1 in the endometrial epithelium in adult (6- to 8-week-old) virgin Pax8-rtTA TRE-Cre Trp53^loxP/loxP Rb1^loxP/loxP Ai9 mice. We observed formation of endometrial neoplasms in 81% (17 of 21) mice between 109 and 400 days after doxycycline administration (Fig. 4A). Histological characterization of uteri revealed invasion of malignant glands into the myometrium and serosa, marked cytological atypia and morphological similarity to human SEC (Fig. 4B-D). Supporting efficient Cre-loxP-mediated gene recombination, neoplastic cells expressed tdTomato (Fig. 4E). Beginning at 60 days post-injection, early dysplastic lesions characterized by loss of polarity, cellular atypia and cell proliferation were observed in both luminal (Fig. 4F) and glandular (Fig. 4G) endometrial epithelium.

By 300 days after doxycycline administration, 45% of mice developed dysplastic lesions in the epithelium of the uterine tube (also known as the oviduct or Fallopian tube; A.F.-N. and A.Y.N., unpublished observations.) Eighteen percent of the same mice developed more advanced but localized neoplasms. These lesions were similar to serous tubal intraepithelial carcinomas (STICs) and early high-grade serous carcinomas previously described in mouse models with combined inactivation of Trp53 and Brca1, Brca2 or Pten (Perets et al., 2013), or an amino-terminal truncated version of SV40 large T antigen (T121), which inactivates all members of the RB family (Rb1, p107 and p130) (Zhang et al., 2019). However, tubal lesions developed later than endometrial neoplasms, with no STICs being observed <154 days after doxycycline induction.

Human SECs are characterized by diffuse p53 staining, cytoplasmic p16 (also known as CDKN2A), reduced expression of estrogen receptor (ER; also known as ESR1) and progesterone receptor (PR; also known as NR3C3) and an increased number of Ki67⁺ (also known as MKI67⁺) cells (Fig. 5A). A similar pattern of p16, ER, PR and Ki67 expression has been observed in mouse endometrial neoplasms (Fig. 5B). Taken together with the histological features, this mouse model mimics many features of the human disease and is well suited for studying SEC.

To test whether PAX8⁺ cells were preferentially susceptible to Trp53 and Rb1 mutations, we inactivated these genes in FOXJ1⁺ ciliated endometrial epithelial cells using Foxj1^{tm1.1(cre/ERT2/GFP)Htg}/J (FoxJ1^CreERT2::GFP) mice. According to both mRNA levels and tdTomato expression in FoxJ1^CreERT2::GFP Ai9 mice, the frequency of FOXJ1⁺ cells was about one-third of that of PAX8⁺ glandular epithelial cells (Fig. 6A,B; M.B., A.F.-N. and A.Y.N, unpublished observations). However, none of the FoxJ1^CreERT2::GFP Trp53^loxP/loxP Rb1^loxP/loxP Ai9 mice (n=14) developed any pathological lesions by 400 days after a single pulse of tamoxifen. In addition, neoplastic lesions were not observed in the epithelium of the uterine tube despite efficient labeling of tubal ciliated cells in our model (Fig. 6C).

A mouse endometrial epithelium RNA-sequencing cell atlas during developmental stages has been reported previously (Wu et al., 2017). Of note, that study was focused only on CD326⁺ epithelial cells isolated by fluorescence-activated cell sorting and therefore did not allow identification of other epithelium-specific markers and non-epithelial cells. Our present analysis addresses these problems and provides a description of all main cell types of the mouse uterus.

Recently, two studies based on cell lineage tracing with tamoxifen-controlled Cre expression have been published (Jin, 2019; Syed et al., 2020). The first study used limited labeling and mathematical analysis in Krt19-Cre/ERT2 and FoxA2-Cre/ERT2 mice crossed with reporter mice (Jin, 2019) and concluded that the stem cell population resides in the intersection zone between luminal and glandular epithelium. By contrast, the second study reported that Axin2⁺ glandular cells can act as drivers of epithelial homeostasis and regeneration (Syed et al., 2020). Cell fate tracing was initiated at either diestrus (Jin, 2019) or metestrus (Syed et al., 2020), which complicates comparative interpretation of the results. Following of AXIN2⁺ cells was for periods ≤70 days for virgin mice. It has also been noted that AXIN2⁺ cells contribute to luminal epithelium only over long periods (Syed et al., 2020). Thus, the existence of alternative pools of luminal and/or glandular stem/progenitor cells cannot be excluded. This possibility is supported by our single-cell transcriptome findings suggesting that the endometrium might contain more than one stem cell pool.

Our studies involve the evaluation of the fate of PAX8-expressing cells. PAX8 is a member of the paired box (PAX) family of transcription factors. In adult mice, it is expressed in a number of tissues, including the epithelium of the uterine tube and the renal excretory system. In the endometrial epithelium, it is expressed in all stages of the estrous cycle (Syed et al., 2020). According to a single-cell transcriptome analysis, higher levels of PAX8 mRNA are observed in the glandular epithelium than in the luminal epithelium. However, immunohistochemical analysis did not show any discernable differences in PAX8 expression. The mechanisms and functional significance of this observation remain to be studied.

Based on a single color tracing of PAX8⁺ cells (Syed et al., 2020) and analysis of cancer-associated mutations (Suda et al., 2018), recent studies have suggested that normal endometrial epithelial cells might expand clonally. Our findings based on two-color cell labeling provide direct support to this notion. It remains to be determined whether clonal expansion of the endometrial epithelium is a result of a neutral competition between endometrial epithelial stem cells, as has been shown for Lgr5⁺ cells in the small intestine (Snippert et al., 2010).

Our studies are consistent with the existence of endometrial epithelial stem cells. However, they cannot completely exclude non-epithelial contributions in non-homeostatic conditions, such as postpartum endometrial regeneration and artificial decidualization (Huang et al., 2012; Patterson et al., 2013). None of the studies reporting contributions of bone marrow-derived cells (Bratincsak et al., 2007; Du and Taylor, 2007; Ikoma et al., 2009; Mints et al., 2008; Taylor, 2004) included experiments testing for cell fusion, which is a common reason for the detection of bone-marrow-specific markers in non-hematopoietic tissues. Thus, these results await further confirmation.

The main strengths of autochthonous genetically modified mouse models of cancer include precise control of the initiation of genetic alterations and the ability to study cancer initiation and progression in an immunocompetent environment (Day et al., 2015; HogenEsch and Nikitin, 2012). Previously, it has been reported that endometrial epithelium-specific knockout of Trp53 in Ksp1.3-Cre-expressing cells results in the development of endometrial intraepithelial carcinoma (the precursor of serous carcinoma) and serous carcinoma at 14-16 months of age (Wild et al., 2012). The use of this model for studies of adult endometrial carcinoma pathogenesis is complicated by constitutive expression of Ksp1.3-Cre in the embryonic precursor structures that give rise to most of the epithelia of the adult urinogenital system.

Our new model of SEC addresses the problem of constitutive expression and allows introduction of the initiating events in the adult mouse. It also confirms the crucial role of  if noe mutations and RB1 pathway alterations in the pathogenesis of SEC. Thus, this model should be a valuable tool for future fundamental and translational studies. At the same time, our SEC model has two limitations and might require further refinements. First, we have noticed that in addition to development of uterine tube malignancies, ≤45% of mice form STICs and high-grade serous carcinomas in the uterine tube. These lesions develop later than the endometrial carcinomas, thereby reducing potential interference with the interpretation of our results. However, salpingectomy might allow establishment of a more accurate model of SEC. Given that most mice with serous uterine tube lesions also have SEC, it is possible that at least some of these lesions might represent metastasis from the neoplastic endometrial epithelium.

The second potential limitation is doxycycline-independent TRE-Cre activity in some cell types, such as thymic epithelial cells (Zhang et al., 2019) and blood cells (A.F.-N. and A.Y.N., unpublished observations). Depending on a combination of genetic alterations and mouse genetic backgrounds, TRE-Cre expression can lead to fast progression of background malignancies. As indicated in the Materials and Methods, ∼10% of TRE-Cre Trp53^loxP/loxP Rb1^loxP/loxP mice developed histiocytic sarcomas and were not included in further analyses. Development of Pax8-Cre/ERT2 mice might be a better approach in future for modeling neoplastic lesions of the endometrial epithelium.

Our findings suggest that transformation of FOXJ1 ciliated cells might require other initiating genetic alterations. Indeed, it has been proposed that clear cell endometrial carcinomas can arise from ciliated endometrial cells (Cochrane et al., 2017). Compared with SECs, clear cell endometrial carcinomas have a lower frequency of TP53 mutations (30-40% versus 95%) but contain other alterations, such as microsatellite instability (15% versus 5%) and PTEN mutations (30% versus 10%) (An et al., 2004; Lax et al., 1998).

The Tg(Pax8-rtTA2S*M2)1Koes/J (Pax8-rtTA) Tg(tetO-Cre)1Jaw/J (Tre-Cre) mice (Perets et al., 2013), Foxj1^{tm1.1(cre/ERT2/GFP)Htg}/J (FoxJ1^CreERT2::GFP; JAX stock no. 027012), Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} (Ai9) mice (JAX stock no. 007909) and Confetti mice [Gt(ROSA)26Sor^{tm1(CAG-Brainbow2.1)Cle}/J; JAX stock no. 013731] were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). The Trp53^loxP/loxP and Rb1^loxP/loxP mice, which have Trp53 and Rb1 genes, respectively, flanked by loxP alleles, were a gift from Dr Anton Berns (The Netherlands Cancer Institute, Amsterdam, The Netherlands). All the experiments and maintenance of the mice followed the recommendations of the Guide for the Institutional Laboratory Animal Use and Care Committee.

For lineage tracing experiments, Pax8-rtTA Tre-Cre Ai9 and Pax8-rtTA Tre-Cre Confetti 6-week-old mice received a single dose (12 μl/g body weight) of doxycycline (6.7 mg/ml in sterile PBS) by intraperitoneal injection. For tumor induction experiments, doxycycline was administered to 6- to 10-week-old Pax8-rtTA Tre-Cre Trp53^loxP/loxP Rb1^loxP/loxP Ai9 mice and control mice. For tamoxifen induction of Cre expression in FoxJ1^CreERT2::GFP mice, 6-week-old mice received a single dose (8 μl/g body weight) of tamoxifen (25 mg/ml in corn oil; Sigma-Aldrich, St Louis, MO, USA; cat# T5648) by intraperitoneal injection. In our cohorts, ∼10% of TRE-Cre Trp53^loxP/loxP Rb1^loxP/loxP mice developed histiocytic sarcomas and were not included in further analyses. All mice were euthanized by CO₂ inhalation, and further analyses were carried out.

All tissues were fixed in buffered 4% paraformaldehyde overnight at 4°C, followed by standard tissue processing and paraffin embedding. Histological and immunohistochemical staining was carried out on 4-μm-thick tissue sections. For immunohistochemistry, antigen retrieval was performed by incubation of deparaffinized and rehydrated tissue sections in boiling 10 mM sodium citrate buffer (pH 6.0) for 10 min. The primary antibodies against ER, FOXA2, Ki67, P16, PAX8, PR, tdTomato/RFP, TP53 and TROP2 were incubated at 4°C overnight, followed by incubation with secondary biotinylated antibodies [30 min, at room temperature (RT)]. A modified Elite avidin-biotin peroxidase (ABC) technique (Vector Laboratories, Burlingame, CA, USA; pk-6100) was performed at RT for 30 min. Hematoxylin was used as the counterstain. All primary antibodies used for immunostaining are listed in Table S2.

The MCA (Han et al., 2018) expression matrices for uterus (Uterus1 GEO: GSM2906478; and Uterus2 GEO: GSM2906479) and related cell type metadata were downloaded (GEO: GSE108097) and re-analyzed using a custom version of the Seurat v.3.1.0 R package (Stuart et al., 2019). Seurat was used for data integration and normalization, principal component reduction, SNN clustering, UMAP visualization and differential gene expression analysis. We combined the Uterus1 and Uterus2 datasets and normalized the data using a regularized negative binomial regression model (sctransform R package) in order to remove technical effects while preserving biological variability (Hafemeister and Satija, 2019 preprint). On the normalized data, we performed SSN clustering with a resolution of 0.4, and we visualized the data using UMAP (Becht et al., 2018). Differential expression analysis was performed using the ‘FindAllMarkers’ function of Seurat, using Wilcoxon's signed-rank test and considering only genes with >log2(0.25) fold-change and expressed in ≥25% of cells in the cluster.

All human specimens were de-identified. They were not collected specifically for the purpose of this research.

Statistical comparisons were performed using Student's two-tailed unpaired t-test and a χ² test with InStat 3 and Prism 8 software (GraphPad Software Inc., La Jolla, CA, USA). Survival curves were computed using the Kaplan–Meier method, and the survival comparisons were analyzed by log-rank tests. Significance was determined as P<0.05.

We thank Blaine A. Harlan for her technical contributions at the early stages of this project, and Minseok (Joseph) Kim for his TROP2 immunostaining.