Noninvasive and real-time assessment of reconstructed functional human endometrium in NOD/SCID/γ_c^null immunodeficient mice

Abstract

Human uterine endometrium exhibits unique properties of cyclical regeneration and remodeling throughout reproductive life and also is subject to endometriosis through ectopic implantation of retrogradely shed endometrial fragments during menstruation. Here we show that functional endometrium can be regenerated from singly dispersed human endometrial cells transplanted beneath the kidney capsule of NOD/SCID/γ_c^null immunodeficient mice. In addition to the endometrium-like structure, hormone-dependent changes, including proliferation, differentiation, and tissue breakdown and shedding (menstruation), can be reproduced in the reconstructed endometrium, the blood to which is supplied predominantly by human vessels invading into the mouse kidney parenchyma. Furthermore, the hormone-dependent behavior of the endometrium regenerated from lentivirally engineered endometrial cells expressing a variant luciferase can be assessed noninvasively and quantitatively by in vivo bioluminescence imaging. These results indicate that singly dispersed endometrial cells have potential applications for tissue reconstitution, angiogenesis, and human–mouse chimeric vessel formation, providing implications for mechanisms underlying the physiological endometrial regeneration during the menstrual cycle and the establishment of endometriotic lesions. This animal system can be applied as the unique model of endometriosis or for other various types of neoplastic diseases with the capacity of noninvasive and real-time evaluation of the effect of therapeutic agents and gene targeting when the relevant cells are transplanted beneath the kidney capsule.

Human endometrium lines the uterus and comprises luminal and glandular epithelial cells, stromal fibroblasts, vascular smooth muscle cells, endothelial cells, and immune competent cells. These cell components coordinately participate in the cyclical changes of human endometrium, including proliferation, differentiation, and tissue breakdown and shedding under the influence of estrogen and progesterone during the menstrual cycle. This unique system of cyclic tissue regeneration also depends on the cyclical growth and regression of the blood vessels that supply the endometrium (1). In addition, angiogenesis is deeply involved in the pathogenesis of endometrium-derived disorders such as endometriosis (2). Endometriosis, one of the most common gynecological diseases, is characterized by the presence of functional endometrial-like tissue outside the uterine cavity. It is an estrogen-dependent disorder associated with substantial morbidity; however, the etiology and pathophysiology are not well elucidated (3). To study the physiology of normal endometrium and the pathogenesis of endometriosis, a variety of in vivo models using small animals has been developed by using the transplantation of autologous or heterologous endometrial cells/tissues or endometriotic tissues (4).

In the present study, taking advantage of newly developed severe immunodeficient mice, NOD/SCID/γ_c^null (NOG) mice (5), we have developed a mouse model that satisfies the following requirements: (i) the transplant of human origin is quantitatively and characteristically uniform in each animal, (ii) functional and morphological changes characteristic of human eutopic and/or ectopic endometrium are reproduced, and (iii) the transplant is assessable for a long term in a noninvasive, real-time, and quantitative manner. By using this model, evidence has been obtained suggesting a previously uncharacterized mechanism underlying the regeneration and remodeling of human endometrium and the pathogenesis of endometriosis.

Results

Reconstruction of Human Endometrial Tissues in NOG Mice.

We first isolated and dissociated endometrial cells mechanically and enzymatically from human cycling endometrium. We transplanted 5 × 10⁵ of these singly dispersed endometrial cells (SDECs) beneath each kidney capsule of ovariectomized (OVX)-NOG mice. Multiple immunological dysfunctions, including cytokine production incapacity and functional incompetence of T, B, and natural killer (NK) cells, may explain the high success rates of xenografts in NOG mice (5). Xenotransplanted NOG mice were subjected to different hormonal treatments (Fig. 1A).

Fig. 1.

Reconstruction of human endometrium-like tissue from SDECs. (A) Experimental design and hormonal treatment. NOG mice were OVX, xenotransplanted with SDECs, and then treated without (no-Tx) or with s.c. implantation of usually two E₂ pellets alone (E₂) or with a daily s.c. injection of P₄ (red arrows) for the last 2 weeks (P₄) or in combination with E₂ (E₂ + P₄); they were finally nephrectomized 10 weeks after transplantation. (B) Macroscopic and microscopic findings of the transplanted site (arrowhead) of NOG mice 10 weeks after xenotransplantation. H&E staining was performed on the transplanted lesion of NOG mice treated with E₂, P₄, or E₂ + P₄ as indicated. The borders between the reconstituted tissue and the mouse kidney (K) are indicated by the dotted lines. (C–G) Immunofluorescence staining of the endometrial reconstructs in the E₂ + P₄-treated NOG mice by using antibodies against cytokeratin (Ck) and Vm (C), CD13 and CD9 (D), CD56 or CD14 in combination with CD45 (E), Vm and α-SMA (F), or human CD31 (hCD31) and α-SMA (G), followed by Hoechst (Ho) staining. (F) Arrows point to the regions prominently immunoreactive for α-SMA. A small box marks a region shown at higher magnification in the adjacent panel as indicated. (G) Arrows and arrowheads indicate hCD31⁺/α-SMA⁺ cells and hCD31⁻/α-SMA⁺ cells, respectively. (Scale bars: 100 μm.)

Surprisingly, endometrium-like tissues were found in all of the transplanted kidneys of NOG mice (n = 30) hormonally treated for 10 weeks (Fig. 1B), whereas a very tiny but macroscopically identifiable tissue was reconstituted in only 1 of 12 nonhormonally treated mice (data not shown). Treatment with estradiol (E₂) and progesterone (P₄), or E₂ + P₄ treatment, generated a large cystic mass (Fig. 1B, arrowhead) with a fine surface vasculature (Fig. 1B Inset). Hematoxylin and eosin (H&E) staining demonstrated that well delineated glands and stroma were present in the transplanted lesions dissected from E₂- or E₂ + P₄-treated mice (Fig. 1B). Furthermore, the growth of the endometrial transplants together with the enlargement of the uterus were E₂ dose-dependent [ supporting information (SI) Fig. 6].

Immunofluorescence studies revealed that the reconstructed tissues, but not the mouse kidney, were stained exclusively with anti-human vimentin (Vm) antibody (clone V9) (Fig. 1C). As V9 can recognize only human Vm, the reconstructed tissue was clearly of human origin. In addition, the stroma was positive for CD10 (data not shown) and CD13, both endometrial stromal cell markers (6), whereas the glandular structure was positive for cytokeratin and CD9, both epithelial markers (6) (Fig. 1 C and D). Human endometrial tissue and decidualized endometrium in pregnancy contain large numbers of CD45-positive leukocytes, the vast majority of which are CD56⁺ NK cells with the other populations being CD14⁺ macrophages and T cells (7). In agreement with this profile, CD45⁺ cells were abundant in the reconstructed tissue, and the proportions of CD56⁺ and CD14⁺ cells, both of human origin, were similar to those in the eutopic endometrium (Fig. 1E).

Expression of α-smooth muscle actin (αSMA) is ubiquitously prominent in the uterine myometrium but is mainly confined to vascular smooth muscle cells in the endometrium (8). In the Vm-positive regenerated tissue, αSMA antibody preferentially and potently reacted with two distinct regions: one adjacent to the mouse kidney parenchyma and the other beneath the serous membrane (Fig. 1F, arrowheads). These two regions mainly consist of αSMA-positive spindle-shaped fibroblastic cells morphologically similar to myometrial cells (Fig. 1G). The vast majority of these densely populated αSMA⁺ cells did not colocalize with CD31⁺ endothelial cells (Fig. 1G, arrowheads). In the same area, however, some αSMA⁺ cells were present along with cells positive for CD31, an endothelial cell marker (Fig. 1G, arrows). These results collectively suggest that some, but not most, αSMA⁺ cells contribute to the formation of the vessels, presumably as pericytes. Thus, the reconstituted tissue possessed hierarchical architecture distinctly composed of endometrium- and myometrium-like layers (Fig. 1G), closely resembling the normal uterine structure at the interface between endometrium and myometrium. Absence of αSMA⁺ cells in SDECs (data not shown) indicated that they did not contain differentiated myometrial cells, which may account for the lack of spiral arterioles in the reconstituted endometrial tissues, because spiral arteries arise from the myometrium. Furthermore, given the potential of endometrial stroma for differentiation into smooth muscle (9), it is likely that the generation of myometrial-like components may be caused by the transdifferentiation of SDECs into smooth muscle cells. Alternatively, it also remains possible that SEDCs include a population of undifferentiated smooth muscle precursor cells. Taken together, SDECs have the potential to regenerate the endometrium ectopically with tissue polarity, glandular structures, and endometrial cell components.

Neovascularization of the Reconstructed Endometrium by Human–Mouse Chimeric Vessels.

We further investigated the vascular components of the E₂ + P₄-exposed endometrial reconstruct. As shown in Fig. 2A, human CD31⁺ (arrows) or mouse CD31⁺ (arrowheads) endothelial cells coexisted abundantly in the reconstructed tissue in E₂ + P₄-treated NOG mice; however, they did not appear to anastomose with each other (Fig. 2A Right). Intriguingly, cells positive for TER-119, a mouse erythrocyte marker, were found not only in the mouse CD31⁺ vessels (SI Fig. 7) but also in the vessels consisting of human CD31⁺ endothelial cells (Fig. 2B, arrowheads), indicating that the human-originated vasculature functioned as a circulation system in the reconstructed endometrium.

Fig. 2.

Vascularization of the reconstructed endometrium predominantly composed of human–mouse chimeric vessels. Immunofluorescence staining of the reconstituted endometrial tissue (A) and the mouse kidney parenchyma adjacent to the reconstituted tissue (B–E) in the E₂ + P₄-treated NOG mice by using antibodies against human CD31 (hCD31) and murine CD31 (mCD31) (A), hCD31 and TER-119 (B), hCD31 and Vm (C), Vm and αSMA (D), or mCD31 and Vm (E), followed by Hoechst (Ho) staining. (A) Arrows and arrowheads indicate hCD31⁺ cells and mCD31⁺ cells, respectively. (B) Arrowheads show hCD31⁺ vessels containing mouse erythrocytes positive for TER-119. (D) A small box marks a region shown at higher magnification in the adjacent panel as indicated. (E) Arrowheads point to mCD31⁺ vessels that were present together with human Vm⁺ cells. (Scale bars: 100 μm.)

To address the presence of chimeric vessels composed of human and mouse endothelial cells, we performed histological analyses focusing on the mouse kidney adjacent to the reconstituted endometrium. We found that human-originated (human Vm-positive) cells migrated into the mouse kidney parenchyma and most were positive for human CD31 (Fig. 2C). The anti-αSMA antibody reacts with both mouse and human smooth muscle cells. Therefore, as most Vm⁺ cells also are positive for human CD31, αSMA⁻/Vm⁺ cells (Fig. 2D, green) corresponded to human CD31⁺ cells in the mouse kidney parenchyma, whereas αSMA⁺/Vm⁻ cells (Fig. 2D, red) were mouse smooth muscle cells. Thus, there existed chimeric vessels of two distinct species in the mouse kidney parenchyma (Fig. 2D Center and Right), which was confirmed by the colocalization of murine CD31⁺ endothelial cells and human Vm-positive cells (Fig. 2E, arrowheads).

Hormone-Dependent Changes of the Reconstructed Endometrium.

It is well known that the endometrial P₄ receptor (PR) is up-regulated by E₂ exposure. In agreement, PR was expressed prominently in the E₂-exposed reconstruct (Fig. 3A). In addition, glandular cells became pseudostratified (Fig. 3A, arrowheads), reflecting a high level of E₂-dependent proliferation activity.

Fig. 3.

Hormone-dependent morphological and functional changes of the reconstructed endometrium. Immunofluorescence staining of the reconstituted endometrial tissues in the E₂-treated (A) or E₂ + P₄-treated (B) NOG mice by using antibodies against PR and Vm (A) and Vm and PRL (B), followed by Hoechst (Ho) staining. (A) Arrowheads show pseudostratified glandular cells. (C) Hormonal treatment protocol (cyclic E₂ + P₄) for induction of menstruation-like tissue breakdown and shedding. (D) Macroscopic findings of the transplanted site (arrows) of a NOG mouse treated with cyclic E₂ + P₄. (E) H&E and immunofluorescence staining of the transplanted lesion in the cyclic E₂ + P₄-treated mice. The glandular structure was partially disrupted (arrowheads), and hemorrhage occurred in the stroma (arrow). (B and E) A small box marks a region shown at higher magnification in the adjacent panel as indicated. (Scale bars: 100 μm.)

Decidualization is the P₄-induced differentiation of E₂-primed endometrial stromal cells, which occurs during the P₄-dominated secretory phase and pregnancy. Decidualized stromal cells become enlarged and round, and they produce many bioactive substances, including prolactin (PRL) (10). Accordingly, PRL was expressed prominently in the stromal cells, but not glandular cells, of E₂ + P₄-exposed endometrial reconstructs (Fig. 3B), indicating the successful spatiotemporal induction of PRL in this animal model. In addition to the strong staining of PRL caused by relatively high levels of P₄, H&E staining revealed edematous stroma with predecidual changes, accumulation of mononuclear leukocytes beneath the glandular epithelium, and tortuous glands (SI Fig. 8), all of which are distinctive morphological changes of secretory-phase endometrium.

One of the unique events of human cycling endometrium is the breakdown and shedding of tissue, i.e., menstruation, which is induced by the withdrawal of P₄. To provoke menstrual changes in the reconstructed endometrium, OVX-NOG mice were subjected to cyclic treatment with E₂ + P₄ (cyclic E₂ + P₄) after xenotransplantation and thereafter to P₄ withdrawal (Fig. 3C). Grafts collected a week after the second P₄ withdrawal contained large blood-filled cysts that were similar to red spot lesions (Fig. 3D, arrowheads) typical of aggressive endometriosis (11). H&E and immunofluorescence staining of the cystic lesion revealed that the glandular structures partly were destroyed (Fig. 3E, arrowheads) and that hemorrhage had occurred in the degenerative stroma (Fig. 3E, arrow).

Lentiviral Introduction of Reporter Genes into Primary Endometrial Cells.

Bioluminescence imaging (BLI) recently has emerged as a useful tool for tumor, hematopoietic, and neural cell tracking studies in living animals (12–15). To apply this animal model to the BLI system, we have developed a recombinant lentivirus capable of introducing and stably expressing both the Venus [a yellow fluorescent protein (YFP) mutant] (16) gene and the click beetle red-emitting luciferase (CBR luc, a luciferase variant) (17) gene in the primary culture endometrial cells. These two reporter markers are useful for cell sorting and the detection of living cells from outside the body.

Fluorescence microscopy revealed YFP signals were detected in ≈20–60% of the primary culture SDECs infected with the recombinant lentivirus (Fig. 4A). To quantitatively assess YFP intensity, these SDECs were harvested and stained with propidium iodide (PI). Then, PI-negative SDECs were sorted by flow cytometry and further divided into three groups based on YFP intensity: high (I), low (II), and negative (III) (Fig. 4B). These three subpopulations and noninfected SDECs (IV) were plated onto the corresponding wells of a six-well dish, cultured, and subjected to BLI (Fig. 4C). BLI revealed the luminescence intensity to correlate well with the YFP intensity (Fig. 4C).

Fig. 4.

Expression of both fluorescence and luminescence markers in the lentivirally transduced SDECs. (A) Phase-contrast and fluorescence microscopy of the primary cultures of lentivirally transduced endometrial stromal (Upper) and glandular (Lower) cells. (B) A representative flow cytometric profile of the PI-negative fraction (red box at Left) of the lentivirally transduced SDECs consisting of three subpopulations (Right) based on the fluorescence intensity: high (I), low (II), and negative (III) subpopulations. (C) Macroscopic luminescence image of a six-well dish where each subpopulation, as sorted in B, was cultured in the corresponding well. Noninfected SDECs were cultured in the “IV” well.

Noninvasive and Quantitative Assessment of the Reconstituted Endometrium by BLI.

The OVX-NOG mice were transplanted with the lentivirally engineered SDECs expressing CBR luc beneath the kidney capsule, simultaneously implanted with two E₂ pellets, and subjected to BLI analysis 8 weeks after xenotransplantation. BLI of the ventrally positioned NOG mouse revealed bioluminescence signals in locations corresponding to the bilateral kidney (Fig. 5A Upper). To confirm that these signals did indeed originate from the transplanted kidney, a dorsally positioned laparotomized mouse (Fig. 5A Lower Left) and its excised kidneys (Fig. 5A Lower Right) were subjected to BLI, which clearly indicated that the intense focal bioluminescent spots were confined to the transplanted sites.

Fig. 5.

Optical bioluminescence images and noninvasive quantitative assessment of the endometrial tissues reconstructed from lentivirally transduced SDECs in living NOG mice. (A) A bioluminescence image of the endometrial reconstructs expressing CBR luc in a ventrally positioned NOG mouse treated with E₂ alone. Bioluminescence images of the laparotomized mouse at the dorsal position (Lower Left) and its excised kidneys (Lower Right). (B–D) Representative BLI (Upper) and serial photon count measurements (Lower) of xenotransplanted OVX-NOG mice treated for different durations with the various indicated doses of E₂ pellets (B), with E₂ in combination with daily injections of ICI 182,780 (C), or with cyclic E₂ + P₄ treatment to induce artificial menstrual cycle-related changes (D) in accordance with a similar protocol as shown in Fig. 3C.

To determine whether the growth behavior of the reconstructed tissue can be assessed quantitatively and sequentially, we treated the xenotransplanted OVX-NOG mice without or with one or two E₂ pellets. Sequential BLI of the mice at 3–9 weeks after xenotransplantation revealed that the signal intensities reflecting the volume of the reconstructed tissue were enhanced in an E₂ dose- and time-dependent manner (Fig. 5B), in agreement with the macroscopic findings (SI Fig. 6). In contrast, the signal intensity was kept low and constant or rather decreased in nonhormonally treated mice (Fig. 5B).

To validate our BLI system as a tool for drug testing, we measured the bioluminescence signals in xenotransplanted OVX-NOG mice treated with E₂ in combination with ICI 182,780 (Tocris Cookson Inc., Ellisville, MO), a pure estrogen antagonist. In contrast to the mice treated with E₂ alone (Fig. 5B), the signal intensity was decreased 2–3 months after the cotreatment with E₂ and ICI 182,780 (Fig. 5C), indicating that the antagonistic effect of ICI 182,780 can be assessed noninvasively.

Finally, we monitored the dynamic changes of the endometrial reconstructs during an artificial menstrual cycle induced by cyclic E₂ + P₄ treatment (Fig. 5D). Sequential BLI revealed that the signal intensities fluctuated dramatically in accord with the addition and withdrawal of P₄ (Fig. 5D), indicating that the decrease and reincrease in signal intensity faithfully reflected tissue breakdown and regression after P₄ withdrawal as shown in Fig. 3 C and D and, presumably, subsequent tissue regeneration, respectively (Fig. 5D). Thus, the repeated menstrual cycle-related changes of the transplants can be noninvasively monitored in living NOG mice.

Discussion

Human endometrium is believed to regenerate from the lower basalis layer, a germinal compartment that persists after menstruation, to give rise to a new functionalis layer (18). The potential of singly dissociated endometrial cells to reconstitute the endometrial tissue raises a possibility that endometrial stem/progenitor cells present in the functionalis layer also may participate in the physiological regeneration of the endometrium through their recycle and reuse. Our hypothesis seems to be supported by the recent report that label-retaining cells, which are thought to be tissue stem/progenitor cells, are present not only at the endometrial–myometrial junction but also in luminal epithelium and stroma adjacent to luminal epithelium and near blood vessels in mice (19).

Several experiments in which human endometrial tissues were transplanted into immunodeficient mice such as SCID and nude mice have shown that endometriotic lesions derive their blood supply from the surrounding vascular network (20–22). Furthermore, native human-originated graft vessels disappear gradually, but host vessels instead invade during revascularization of endometrial explants (21–24). As NK cells play a critical role in the IL-12-mediated inhibition of angiogenesis (25), the lack of human endometrial graft-originated vessels in SCID and nude mice in previous studies (20–24) may be attributable to the presence of functioning NK cells, which would attack various immature precursor cells that show low MHC expression level (26). In contrast, we demonstrate here that human-derived vessels are abundant in the endometrial reconstructs, and that they invade into the mouse kidney parenchyma and become connected with mouse vessels functioning as a circulation system. Because NK cells are functionally incompetent in the NOG mouse (5), the human-originated neovascularization is allowed to take place. Given that impaired NK cell cytotoxicity contributes to inadequate removal of ectopic endometrial cells from the peritoneal cavity (27), this endometriosis model using NOG mice may be suitable for the study of the pathogenesis of endometriosis.

Our findings raise the possibility that endothelial cells or their progenitors derived from human endometrium may have a unique angiogenic potential to migrate, invade, and form the vasculature in host tissue, even when originating from a different species. Bussolati et al. (28) have reported that tumor-derived endothelial cells, but not normal endothelial cells, possess a unique neoangiogenic ability to grow in immunodeficient mice and to form vascular structures linked with the mouse circulation. It therefore is conceivable that, in addition to the peritoneal environment (29), the angiogenic potential of the endometrial endothelial cells per se may be one of the critical determinants for the establishment and development of the endometriotic lesion. Antiangiogenic therapy recently has been proposed as an alternative treatment for endometriosis (23, 30). This therapeutic strategy might be strengthened further by targeting endometrium- or endometriosis-specific angiogenesis, as postulated here.

The animal model described here has several advantages over the endometriosis models in current use. First, a single-cell suspension is adequate for several experimental procedures such as cell selection, genetic engineering, and quantitative assessment, as compared with the dissected sections of endometrial or endometriotic tissues used for the transplants in many endometriosis models (4). The use of SDECs also allows relatively easy production of several or even dozens of homogeneous mice, because only a small number of endometrial cells is required for the reconstruction of the endometrium. Second, the regenerated tissues exhibit abundant vascularization and the preservation of endometrial cell components and tissue organization, which enable the long-term maintenance of the reconstructed structure and the hormone-dependent changes characteristic of human cycling endometrium and/or endometriotic explants. Third, taking advantage of the lentivirus and CBR luc, whose emitted light has the capacity to pass through the thickened tissue, the transplants can be assessed for a prolonged period in a noninvasive, real-time, and quantitative manner.

Thus, it is demonstrated here that SDECs have the capacity for tissue regeneration and reconstruction. Combining this unique potential together with NOG mice and lentivirus-mediated cell engineering, we report the unique animal model suitable for the study of endometrial physiology/pathophysiology and the pathogenesis of endometriosis through noninvasive, real-time, and quantitative assessment of ectopically reconstituted endometrium-like tissues. Furthermore, this animal model system can be potentially applicable for drug testing and gene-target validation not only in endometriosis but also in other various types of neoplastic disease when the relevant primary culture cells or cell lines are transplanted beneath the kidney capsule.

Materials and Methods

Detailed protocols can be found in SI Methods.

Tissue Collection.

Proliferative phase endometrial specimens (n = 20) without any abnormalities or malignancies were collected from consenting women (aged 39–51 years) with normal menstrual cycles undergoing total abdominal hysterectomy for benign gynecological diseases. The use of these human specimens was approved by the Keio University Ethics Committee.

Preparation of SDECs.

Human endometrial specimens were dissociated mechanically and enzymatically, filtrated, isolated by Ficoll gradient, and finally separated into endometrial stromal and glandular epithelial cell fractions as described previously (31, 32), with certain modifications. In brief, endometrial tissue samples were cut into small pieces and digested with 0.2% (wt/vol) collagenase (Wako, Osaka, Japan) and 0.05% DNase I (GIBCO, Carlsbad, CA) in DMEM for 1.5 h at 37°C on a shaking rotor. The tissue digest was filtrated through a sterile 407-μm mesh filter, followed by a 40-μm cell strainer (BD Biosciences, Bedford, MA). The glandular fragments retained in the strainer were recovered by back-flushing.

Stromal cell suspension was layered over Ficoll-Paque PLUS (Amersham Biosciences, Piscataway, NJ) and centrifuged to remove red blood cells. The media/Ficoll interface layer mainly containing stromal and immune cells was aspirated and washed. Glandular fragments were digested with 0.05% trypsin-EDTA solution (Sigma–Aldrich, St. Louis, MO) and 0.05% DNase I by pipetting, washed, and filtrated through a 40-μm cell strainer to yield a single-cell suspension. Mixtures of these two fractions were designated SDECs and were used as transplants or subjected to cell culture and lentiviral infection. Flow cytometric analysis revealed that the percentages (mean ± SD) of CD9, CD13, CD31, and CD45 were 12.0 ± 7.2%, 40.2 ± 9.5%, 2.4 ± 1.0%, and 46 ± 13.0% in SDECs, respectively. Procedures for preparing SDECs and the representative flow cytometric data are summarized schematically in SI Fig. 9.

Xenotransplantation and Hormonal Treatment.

NOG mice (5) were used for xenotransplantation experiments. The single-cell suspension of SDECs (5 × 10⁵) in 5–10 μl of DMEM was injected underneath the kidney capsules (for details see SI Methods). At transplantation, NOG mice were OVX and implanted s.c. without (Fig. 1A, “no-Tx”) or with one or usually two E₂ pellets (Fig. 1A, “E₂”) (1.5 mg of E₂ per pellet; Innovative Research of America, Sarasota, FL). For P₄ or E₂ + P₄ treatment, 1 mg of P₄ was injected s.c. every day for the last 2 weeks (Fig. 1A). For cyclic E₂ + P₄ treatment, xenotransplanted OVX-NOG mice implanted with two E₂ pellets were subjected to two cycles of daily P₄ injections for 2 weeks with a 3-week interval (Figs. 3D and 5D). In some mice implanted with an E₂ pellet, 100 μg per body weight per day of ICI 182,780 was injected s.c. every day from the first day of transplantation (Fig. 5C). These xenotransplanted mice were nephrectomized according to the experimental protocol (for details see SI Methods).

Histological and Immunofluorescence Analysis.

Serial cryosections of the kidneys excised from the xenotransplanted NOG mice were air-dried, washed, and fixed. After permeabilization and blocking, tissue sections were incubated with the pretitrated primary antibodies listed in SI Table 1. For indirect fluorescence staining, the first antibodies were visualized by incubation with secondary antibodies conjugated with Alexa Fluor 488 (green) or 568 (red) (Molecular Probes, Eugene, OR). Nuclei were stained with Hoechst 33258. Images were collected as described in SI Methods.

Infection of Lentivirus Expressing CBR luc and Preparation of Transduced SDECs.

The gene fragment of CBR luc was excised from the pCBR-Basic vector (Promega, Madison, WI) and cloned into pCSII-EF-MCS-IRES2-Venus (33). Viral stock was produced as described in ref. 15. SDECs were cultured, infected with lentivirus, continuously propagated, harvested, and then dissociated into single cells. The PI-negative/Venus-positive SDECs were sorted by a MoFlo (Cytomation, Fort Collins, CO) and plated onto a culture dish or transplanted beneath the kidney capsule on the dorsal side in OVX-NOG mice.

BLI.

We used a Xenogen-IVIS 100 cooled CCD optical macroscopic imaging system (SC BioScience Corporation, Tokyo, Japan) for BLI. For in vitro imaging, SDECs were plated at equal density and imaged in the presence of 150 μg/ml d-luciferin (SC BioScience Corporation). For in vivo imaging, OVX-NOG mice xenotransplanted with lentivirally engineered SDECs were anesthetized and given an i.p. injection of d-luciferin (150 mg/kg body weight). All images were analyzed as described in SI Methods. To quantify the measured light, regions of interest (ROI) were defined over the transplanted area and all values were examined from an equal ROI.

Abbreviations

NOG

NOD/SCID/γ_c^null

SDECs

singly dispersed endometrial cells

OVX

ovariectomized

NK

natural killer

E₂

estradiol

P₄

progesterone

E₂ + P₄

treatment with E₂ and P₄

Vm

vimentin

αSMA

α-smooth muscle actin

PR

P₄ receptor

PRL

prolactin

BLI

bioluminescence imaging

YFP

yellow fluorescent protein

CBR luc

click beetle red-emitting luciferase

PI

propidium iodide.