Stem Cell-Like Differentiation Potentials of Endometrial Side Population Cells as Revealed by a Newly Developed In Vivo Endometrial Stem Cell Assay

Background Endometrial stem/progenitor cells contribute to the cyclical regeneration of human endometrium throughout a woman's reproductive life. Although the candidate cell populations have been extensively studied, no consensus exists regarding which endometrial population represents the stem/progenitor cell fraction in terms of in vivo stem cell activity. We have previously reported that human endometrial side population cells (ESP), but not endometrial main population cells (EMP), exhibit stem cell-like properties, including in vivo reconstitution of endometrium-like tissues when xenotransplanted into immunodeficient mice. The reconstitution efficiency, however, was low presumably because ESP cells alone could not provide a sufficient microenvironment (niche) to support their stem cell activity. The objective of this study was to establish a novel in vivo endometrial stem cell assay employing cell tracking and tissue reconstitution systems and to examine the stem cell properties of ESP through use of this assay. Methodology/Principal Findings ESP and EMP cells isolated from whole endometrial cells were infected with lentivirus to express tandem Tomato (TdTom), a red fluorescent protein. They were mixed with unlabeled whole endometrial cells and then transplanted under the kidney capsule of ovariectomized immunodeficient mice. These mice were treated with estradiol and progesterone for eight weeks and nephrectomized. All of the grafts reconstituted endometrium-like tissues under the kidney capsules. Immunofluorescence revealed that TdTom-positive cells were significantly more abundant in the glandular, stromal, and endothelial cells of the reconstituted endometrium in mice transplanted with TdTom-labeled ESP cells than those with TdTom-labeled EMP cells. Conclusions/Significance We have established a novel in vivo endometrial stem cell assay in which multi-potential differentiation can be identified through cell tracking during in vivo endometrial tissue reconstitution. Using this assay, we demonstrated that ESP cells differentiated into multiple endometrial lineages in the niche provided by whole endometrial cells, indicating that ESP cells are genuine endometrial stem/progenitor cells.


Introduction
Human endometrium lines the uterine cavity and regenerates, differentiates and regresses with each menstrual cycle under hormonal control throughout the course of a woman's reproductive life [1]. These morphological and functional features of human endometrium can be reproduced in an in vivo endometrial regeneration model in which severely immunodeficient NOD/SCID/ccnull (NOG) mice are xenotransplanted with dispersed human endometrial cells under the kidney capsule followed by hormonal treatment [2]. It is likely that the cyclical regeneration of human endometrium is achieved through tissuespecific stem cell system(s) in which human endometrium contains a pool of multipotent stem cells capable of cyclically producing progenitor cells that further differentiate into each endometrial cell component [3,4]. Indeed, several investigators (including our group) have provided evidence suggesting the existence of endometrial stem/progenitor cell cells and their possible roles in humans [5,6]. The candidate endometrial stem/progenitor cells include clonogenic endometrial cells [7], endometrial side population (SP) cells that possess a Hoechst 33342 low-fluorescence profile [8][9][10][11], CD146 + CD140b + stromal cells [12], CD29 + CD73 + CD90 + stromal cells [13], and W5C5 + stromal cells [14]. There exists, however, no consensus regarding which endometrial population represents the stem/progenitor cell fraction [6,15].
Tissue-specific, candidate stem cells have been identified in many types of tissues based on the SP phenotype [16]. This characteristic is due to the unique ability of the primitive cells to pump out the DNA-binding dye Hoechst 33342 via the ATPbinding cassette transporter G2 (ABCG2) [16]. Primitive hematopoietic precursors from bone marrow were the first SP cells identified with this technique [17]. We previously demonstrated that SP cells, but not main population (MP) cells, both isolated from the human endometrium, regenerate human endometriumlike tissues in vivo when xenotransplanted under the kidney capsule of NOG mice [10]. However, a low efficiency of reconstitution was observed, possibly because endometrial SP (ESP) cells may require a specific ''niche'' provided by other endometrial cell components to reconstitute the entire endometrium in vivo as well as in in vitro culture. Indeed, successful proliferation of ESP cells in the presence of endometrial MP (EMP) cells in conventional media [10] suggests that EMP may provide a ''niche'' appropriate for activation of ESP.
In the present study, we modified and improved our previous in vivo endometrial regeneration model [2] to characterize the stem cell properties of human ESP cells. Using our newly developed in vivo endometrial stem cell assay, we have demonstrated that ESP cells are significantly more capable of differentiating into multiple endometrial lineages than are EMP cells in assistance with the niche provided by whole endometrial cells.

Isolation of ESP and EMP Cells from Human Endometrium
As shown in Figure 1A, we first dissociated human endometria mechanically and enzymatically and purified epithelia-enriched and stroma-enriched fractions as previously described [1]. These two fractions were mixed, and the mixture (designated as singly dispersed endometrial cells (SDECs)) was then stained with Hoechst dye followed by flow cytometric analysis and cell sorting for isolation of ESP and EMP cells ( Figure 1A, right upper panel). The ESP fraction disappeared upon treatment with 50 mM reserpine, an ABCG2 blocker ( Figure 1A, right lower panel). The percentage of ESP cells among SDECs was 3.1360.61% (mean6SEM, n = 8).

Surface Marker Characterization of ESP and EMP Cells
We next characterized the expression patterns of several cell surface markers in ESP cells in comparison with EMP cells. As shown in Figure 1B, there were no significant differences in the percentages of cells positive for CD326 (epithelial) [18] and CD10 (stromal) [19] Figure 1B). There was no difference in the percentages of W5C5-positive cells, purportedly putative human endometrial mesenchymal stem-like cells [14], between ESP and EMP fractions (11.261.3% vs. 13.663.7%; n = 6, each group) ( Figure 1B). CD140b + CD146 + cells, reportedly putative human endometrial mesenchymal stem-like cells [12], were significantly more abundant in the ESP fraction than in EMP cells (4.160.9% vs. 1.560.5%, P,0.05; n = 6, each group) ( Figure 1C). Thus, flow cytometric analysis demonstrated that ESP and EMP fractions contained similar proportions of epithelial and stromal markerpositive cells; however, endothelial cells and mesenchymal stemlike cells were significantly more abundant in ESP than in EMP cells.

In vivo Stem Cells Assay Using ESP and EMP Fractions
ESP cells alone can give rise to endometrium-like tissues with delineated glandular structures in vivo when xenotransplanted under the kidney capsule of NOG mice [10]. However, the reconstitution efficiency has been very low (#10%). We suggest that this is because ESP cells require a ''niche'' provided by other endometrial cell components to reconstitute the entire endometrium in vivo and also to proliferate and differentiate into various endometrial lineages in vitro. Indeed, co-culture with EMP cells improves the poor proliferative potential of ESP cells [10] suggesting that the EMP fraction may provide a ''niche'' essential for elicitation of ESP stem cell activity.
These ideas and data prompted us to develop an in vivo endometrial stem cell assay in which multilineage differentiation of endometrial stem cells could be identified through cell tracking during endometrial tissue reconstitution. Figure 2A illustrates the procedures for the in vivo stem cell assay. Stem or non-stem cells -ESP or EMP cells in this study -were infected with lentivirus immediately after cell sorting without cell culture to express bioluminescent/fluorescent marker(s). Specifically, we used redemitting firefly luciferase (RedFluc) and tandem Tomato (TdTom) that can be detected by bioluminescence and fluorescence imaging (BLI and FLI, respectively). The infected cells were mixed with SDECs, resuspended in collagen gel, and then implanted under the kidney capsule of ovariectomized NOG mice. These mice were treated with estrogen and progesterone for about eight weeks, and then subjected to BLI followed by nephrectomy. The graft-bearing kidneys were also subjected to BLI and FLI, histological and immunohistochemical analyses. The NOG mice (and their kidneys) transplanted with TdTom-expressing ESP cells and TdTom-expressing EMP cells were designated TdTom-ESP and TdTom-EMP, respectively.
We performed BLI of the ventrally positioned TdTom-ESP and TdTom-EMP mice one week after xenotransplantation and detected bioluminescent signals in locations corresponding to the right kidney ( Figure 2B). Quantification of the measured light revealed that the signal intensities reflecting the number of RedFluc-expressing cells were not significantly different between the ESP-TdTom mice (4.2862.33610 5 photon count, n = 5) and the EMP-TdTom mice (3.4361.26610 5 photon count, n = 6) ( Figure 2C). BLI and FLI analysis of NOG mouse kidneys excised eight weeks after xenotransplantation clearly showed that intense focal bioluminescent and fluorescent spots were confined to the transplanted sites ( Figure 2D) indicating the existence of TdTomlabeled ESP and EMP cells or their descendants in these kidneys.

Histological and Immunohistochemical Analysis of in vivo Reconstituted Tissues
Detectable bioluminescent and fluorescent signals indicate not only the presence of the labeled cells but also the existence of reconstructed tissue containing those cells in the kidney. Indeed, an apparent cystic mass was macroscopically observed at the site of the transplant used for the in vivo stem cell assay ( Figure 3A, left and middle panels). H&E staining revealed that the cystic mass possessed well-delineated glands and stroma like endometrium ( Figure 3A, right panel). The mass was positive for human vimentin (Vm), as demonstrated by immunofluorescence studies using anti-human vimentin (Vm) antibody that recognizes only human Vm, a representative marker for endometrial stromal cells.
In this stem cell assay, ESP or EMP cells (1610 4 each) were infected with lentivirus without cell culture to introduce TdTom and then mixed with propidium iodide (PI)-negative SEDCs (4.9610 5 cells each). In contrast to ESP alone as previously reported [10], the efficiency of endometrial tissue reconstitution was 100% (n = 11) when the cell mixture was used for transplantation. Since the majority (approximately 98%) of the initial mixture was SDECs, it is quite reasonable that the high reconstitution efficiency was the same as that obtained in the previous study in which only SEDCs were used for in vivo reconstitution assay. In other words, lentivirally TdTom-marked ESP and EMP cells did not affect the reconstitution potentials of SEDCs. Rather, TdTom-positive ESP were focally present in the reconstituted tissue as shown by a dotted line in Figure 3B (middle and right panels), collectively suggesting that ESP cells focally contributed to the endometrial tissue reconstitution driven by nonlabeled SDECs.
We next investigated which fraction, ESP or EMP, was more involved in the endometrial regeneration and which endometrial lineage they differentiated into. Thus, to determine how ESP and EMP contribute to the genesis of each endometrial linage, we performed fluorescent co-staining of TdTom in combination with each of the differentiation markers including Vm (stroma), cytokeratin (Ck, epithelium), a-smooth muscle actin (a-SMA, smooth muscle), and CD31 (endothelium) to identify the behavior of TdTom-labeled cells.
Immunofluorescence clearly showed that TdTom + Vm + , TdTom + Ck + and TdTom + CD31 + cells were more abundant in the reconstituted tissues derived from TdTom + ESP cells than in those from TdTom + EMP cells ( Figure 4, A, D, and J). There was no difference in the number of TdTom + a-SMA + cells between the TdTom + ESP fraction and the TdTom + EMP fraction ( Figure 4G). Endometrial epithelial cells exhibited evenly distributed TdTom signals, whereas the other types of endometrial cells, particularly stromal cells, displayed tiny dot signals ( Figure 4 and Figure S1).
The percentage of cells doubly positive for an endometrial lineage marker and TdTom among lineage marker + cells reflects the potential of TdTom + ESP and EMP fractions for differentiation into the corresponding endometrial lineage. To quantify the percentage of endometrial lineage-positive TdTom + cells in the immunofluorescent slides, we performed an image analysis of 36 visual fields (three visual fields per section, three sections per sample, four samples) from each differentiation marker by TissueQuest software (Tissue Gnostic) and show the representative scattergrams for each endometrial linage (Figure 4, B, E, H, and K). In agreement with results of the qualitative analysis of immunofluorescence signals, the quantitative ''TissueQuest'' analysis revealed that the percentages of TdTom + Vm + , TdTom + Ck + and TdTom + CD31 + cells among each lineage marker + cells were significantly higher in the reconstituted tissues originated from TdTom + ESP cells than in those from TdTom + EMP cells. Specifically, the data showed the following: TdTom + Vm + , 26.263.4% vs. 11.062.6%, P,0.0005; TdTom + Ck + , 58.765.6% vs. 36.965.3%, P,0.01; TdTom + CD31 + , 31.966.0% vs. 6.864.3%, P,0.0005, Figure 4, C, F, and L. In contrast, there was no difference in the percentages of TdTom + a-SMA + cells between TdTom + ESP and TdTom + EMP fractions (27.865.5% vs. 19.364.5%; Figure 4H). Thus, these results collectively suggest that ESP cells have in vivo multilineage differentiation potentials characteristic of somatic stem cells and that they are capable of differentiating into endometrial stromal, epithelial, smooth muscle, and endothelial cells.
ESP cells, but not EMP cells, are known to preferentially and exclusively express ABCG2, a stem cell marker [19]. To trace the stem cell-like behavior of transplanted TdTom + cells, we investigated co-expression of ABCG2 and TdTom in the reconstituted tissue. Immunofluorescence studies using antibodies against ABCG2 and TdTom revealed that very few but significant numbers of double positive cells could be detected. Importantly, they were observed only in TdTom-ESP mice but never in TdTom-EMP mice ( Figure 5). The intensity of the ABCG2 fluorescence was not strong enough to be subjected to quantitative ''Tissue-Quest'' analysis. Taken together these results suggest that the majority of ESP cells may give rise to several endometrial lineages whereas a remnant minor fraction of ESP cells may remain undifferentiated and possibly retain a stem cell phenotype even eight weeks after transplantation.

Discussion
Many groups (including ours) have identified, isolated, and characterized putative endometrial stem/progenitor cells and their relevant cells through a variety of methods [6,15]. It seems, however, that each method produces cells that are distinct from the cells isolated by another method [6,15]. There are several in vitro assays to identify and characterize stem cell activity, including clonogenic assays and multi-lineage differentiation assays. In particular, clonal activity is routinely used for the identification of putative somatic stem cells in many adult tissues including the endometrium [20][21][22]. If a particular cell does not have significant clonogenic activity, it may not be considered a stem/progenitor cells. It remains possible, however, that in vitro culture lacks the proper conditions for the maintenance of viability, growth, and differentiation. In vitro multi-lineage differentiation assays may have similar drawbacks. Thus, an in vivo stem cell assay is absolutely needed to identify bona fide stem cells. Therefore, in vivo repopulation assays are considered to be stringent approaches to the identification and characterization of hematopoietic stem cells [23]. Indeed, it has been reported that most hematopoietic cells capable of long-term repopulation in vivo may not be detected as clonogenic cells in standard culture systems [24,25].
We previously demonstrated that human ESP cells are capable of regenerating human endometrium-like tissues in vivo when xenotransplanted under the kidney capsule of NOG mice [10]. Although the previous study showed that ESP clonal efficiency was low compared to EMP cells [10], the in vivo regeneration and multi-lineage differentiation potentials of ESP cells strongly suggest that ESP cells are likely endometrial stem/progenitor cells [10,26]. The regeneration of well-structured endometrium, however, was inefficient likely because the niche provided by other endometrial cells was lacking. This drawback makes it difficult to compare the differentiation capacity of ESP and EMP cells [10]. We hypothesized that ESP might behave as progenitor/stem cells when surrounded an appropriate microenvironmental niche. Therefore, we developed a novel in vivo stem cell assay in which a mixture of ESP cells and whole endometrial cells function together as a transplanted ''niche'' [2]. For this purpose, EMP cells as the control or ESP cells must be distinguished from the ''niche'' cells and, therefore, they must be labeled with bioluminescent and fluorescent proteins to track their contribution to each endometrial lineage during endometrial tissue reconstitution. In this assay, we have mixed dispersed cells with a collagen gel and placed the suspension under the kidney capsule using a soft flexible thin tube. This new cell transplantation technique was much easier and much less invasive than our previous procedure in which a simple liquid suspension of dispersed cells was injected through the mouse kidney parenchyma using a micropipette [2].
Using this assay, we have successfully shown a high reconstitution rate, approximately 100%, and demonstrated that ESP cells could differentiate into several endometrial lineages, including epithelial, stromal, and endothelial cells. These results suggest that the ESP fraction may contain a higher proportion of stem/progenitor cells capable of differentiating into endometrial components than does the EMP fraction. Indeed, flow cytometric analysis showed that the proportion of CD140b + CD146 + cells, reportedly endometrial mesenchymal stem-like cells [12], was significantly higher in ESP cells than in EMP cells. However, it is conceivable that the ESP fraction contains several types of partially differentiated endometrial lineages at higher frequencies than the EMP fraction and that those differentiated cells, in turn, may survive and expand during tissue reconstitution. In this context, one can argue that the magnitude of the contribution may reflect  the proportion of pre-existing differentiated cells, but not that of stem/progenitor cells. To address this criticism, we compared the expression of several differentiation markers in ESP and EMP fraction. We found that there were no significant differences at least in CD326 + (epithelial) and CD10 + (stromal) cells ( Figure 1B), suggesting that the higher contribution of ESP cells to epithelium and stroma in recapitulated tissues may be due to an abundance of stem/progenitor cells rather than that of well-differentiated stromal and epithelial cells.
In contrast to epithelial and stromal cells, CD31 + (endothelial) and CD34 + (endothelial, stem) cells were more abundant in the ESP fraction than the EMP fraction ( Figure 1B), which agrees with our previous results [10]. In this context, the abundance of differentiated endothelial cells in ESP cells may result in a higher contribution of ESP to endothelial cells than EMP. CD31 + and CD34 + cells, however, are not necessarily well-differentiated endothelial cells. Indeed, endothelial progenitor cells are also positive for CD31 and CD34 [27]. Furthermore, we have previously shown that ESP cells exhibit an endothelial progenitor cell-like phenotype [10]. Thus, it is more likely that the stem cell activity of ESP may also contribute to the genesis of endothelium during tissue reconstitution.
ESP cells represent the most likely candidates for endometrial stem/progenitor cells; however, there are two possibilities with regard to the composition of the ESP fraction. First, it could contain ''master'' uterine stem cell(s) capable of giving rise to almost every type of endometrial lineage. Secondly, there may be progenitors for every type of endometrial lineage present in the ESP fraction and they could generate each type of fully differentiated cell component. The latter possibility is more likely based on the present results that the ESP fraction contains several types of cells positive for epithelial, stromal, and endothelial markers. Our new in vivo stem cell assay will allow us to further identify which fraction of the ESPs is responsible for the genesis of each endometrial lineage. Based on the present results showing high stem cell activity of the ESP fraction compared to that of the EMP fraction, we will compare in vivo stem cell activity among CD326 + SP, CD326 2 SP and whole SP. The number of CD326 + SP is so small that they alone may not be able to give rise to identifiable tissues. However, with assistance from whole endometrial cells as ''niche'' cells, we expect that CD326 + SP may be able to contribute to the formation of one or more endometrial lineages.
As shown in Figure 4E, some glands mainly consist of TdTom + cells but other glands consist of both TdTom + and TdTom 2 cells in TdTom-ESP mice. It has been reported that each gland of human eutopic endometrium is monoclonal and therefore may have originated from a single or multiple stem/progenitor cells with uniform clonality at the bottom of each endometrial gland [28]. However, this interpretation seems inconsistent with our present data. Since the disorganized and unpolarized mixtures of non-labeled whole endometrial cells and TdTom-labeled ESP cells are the starting material for endometrial tissue reconstitution in this assay, it is possible that both non-labeled ESP cells present in whole endometrial cell populations and TdTom-labeled ESP may together contribute to the genesis of glands in a disorganized manner therefore resulting in ''hybridized'' or ''mosaic'' gland formation.
Endometrial epithelial cells exhibited evenly distributed TdTom signals, whereas the other types of endometrial cells, particularly stromal cells, displayed tiny dot signals (Figure 4 and Figure S1). The difference in the pattern of TdTom signals was not due to an artifact of immunostaining and imaging procedures, because the difference was still observed in the same section ( Figure S1). It is possible that the pattern of intracellular distribution of TdTom may depend on the cell type. Indeed, the localization of TdTom varies according to the cell type [29], and DsRed2, another type of RFP, shows dotted pattern in cells derived from rhabdomyosarcoma [30].
In summary, we have established a novel in vivo endometrial stem cell assay that enabled us to track the differentiation of the ESP fraction into each endometrial lineage. We observed a high efficiency of endometrial tissue reconstitution. The results of this study further consolidate characterization of the ESP fraction as previously reported [8,9,10,26]. Many groups (including ours) have identified putative endometrial stem/progenitor cells and their relevant cells through a variety of methods [6,15]. Nevertheless, it remains to be determined how many types of stem/ progenitor cells are present in the human endometrium, how they differ in phenotype and function, and the nature of the hierarchical relationships that extend across the various types of precursors. To gain consensus regarding which endometrial population represents bona fide stem/progenitor cells in terms of in vivo stem cell activity, this newly developed in vivo endometrial stem cell assay will be useful for identification and/or validation of human endometrial stem/progenitor cells reported thus far and in the future.

Ethics Statement
This study was approved by the Keio University Ethics Committee, and informed written consent was obtained from each patient prior to tissue collection. Procedures performed on animals were also approved by the Keio University Ethics Committee and conducted in accord with the Guide for the Care and Use of Laboratory Animals of the Keio University School of Medicine.

Tissue Collection
Secretory phase endometrial specimens without any abnormalities or malignancies were collected from consenting women (aged 38-49 years) with normal menstrual cycles undergoing total abdominal hysterectomy for benign gynecological diseases. We used two endometrial specimens for grafting and six for flow cytometric analysis. The use of these human specimens was approved by the Keio University Ethics Committee.

Isolation and Flow Cytometric Analysis of ESP and EMP Cells
Endometrial specimens were separated and dissociated into endometrial stromal and glandular epithelial single cell fractions as described previously [2]. The mixture of two fractions (SDECs) was washed in calcium-and magnesium-free Hanks' Balanced Salt Solutions (HBSS) (Sigma-Aldrich) supplemented with 2% FBS, 10 mM Hepes Buffer Solution (Sigma-Aldrich) (HBSS+) and suspended at 2610 6 cells/mL in HBSS+ and stained with 5.0 mg/ mL Hoechst 33342 (Sigma-Aldrich) for 90 min at 37uC, as described previously [17,31]. For flow cytometry, fluorescein isothiocyanate (FITC)-conjugated or phycoerythrin (PE)-conjugated antibodies and propidium iodide (PI) (Sigma-Aldrich) were simultaneously added to Hoechst-stained cells suspended in HBSS+. Cells were incubated on ice for 30 min, pelleted, and washed with HBSS+. The antibodies are listed in Table S1. Flow cytometric analysis and cell sorting were performed as described in Methods S1. After collecting 1610 5 events, the SP population was defined as previously reported [17]. Samples were analyzed using Kaluza software (Beckman Coulter).

Lentiviral Infection of Transduced ESP and EMP Cells and Preparation of Endometrial Grafts
ESP and EMP cells were infected with lentivirus by centrifugation without cell culture as described in Methods S1. Infected ESP or EMP cells (1610 4 cells each) were mixed with 4.9610 5 PInegative SDECs and resuspended in rat-tail collagen (BD Biosciences) neutralized according to the manufacturer's instructions, and the mixtures were dispensed in 15 mL aliquots and incubated for 30 min at 37uC as described previously [32]. They were designated TdTomato-ESP and TdTomato-EMP, respectively. The collagen suspensions were overlaid with DMEM+ and incubated overnight. The following day, TdTomato-ESP or TdTomato-EMP was implanted under the kidney capsule of oophorectomized NOD/SCID/ccnull (NOG) mice. We used two human specimens and the procedure was performed in duplicate or triplicate for each specimen. Thus, a total of five TdTomato-ESPs and six TdTomato-EMPs were transplanted.

Xenotransplantation and Hormonal Treatment
NOG mice [33] were used for xenotransplantation experiments. Either TdTomato-ESP or TdTomato-EMP was transplanted under the kidney capsule (for details, see Methods S1). At transplantation, both recipients' ovaries were removed and the recipient was implanted subcutaneously with two E 2 pellets (1.5 mg of E 2 per pellet; Innovative Research of America, Sarasota, FL, USA). A P 4 pellet (15 mg of P 4 per pellet; Innovative Research of America) was subcutaneously implanted six weeks after the transplantation. These xenotransplanted mice were nephrectomized according to the experimental protocol (for details, see Methods S1).

Histology and Immunohistochemistry
H&E-staining and immunofluorescence analyses were performed on cryosections derived from kidneys transplanted with TdTomato-ESP or TdTomato-EMP that were air-dried, washed, and fixed. After permeabilization and blocking, tissue sections were incubated with the pre-titrated primary antibodies listed in Table S2. For indirect fluorescence staining, the first antibodies were visualized by incubation with secondary antibodies conjugated with Alexa Fluor 488 (green) (Life Technologies, Carlsbad, CA, USA) and Alexa Fluor 555 (red) (Cell Signaling Technology, Danvers, MA, USA). Images were collected as described in Methods S1. For image cytometry, images were analyzed using the TissueQuest software (TissueGnostics, Vienna, Austria) as described in Methods S1.

Bioluminescence Imaging (BLI) and Fluorescence Imaging (FLI)
We used a Xenogen-IVIS 100 cooled CCD optical macroscopic imaging system (SC BioScience Corporation, Tokyo, Japan) for BLI and FLI. For in vivo BLI, OVX-NOG mice xenotransplanted with lentivirally engineered SDECs were anesthetized with 3% isoflurane and given a retro-orbital injection of D-luciferin (SC BioScience Corporation) (150 mg/kg body weight). All images were analyzed as described in Methods S1. To quantify the measured light, regions of interest (ROI) were defined over the transplanted area and all values were examined from an equal ROI. NOG mouse kidneys excised eight weeks after xenotransplantation were placed on culture dishes and subjected to FLI followed by BLI in the presence of 150 mg/ml D-luciferin.

Statistics
Statistical analysis was carried out using IBM-compatible SPSS for Windows version 19.0.0 (SPSS Inc., Chicago, IL, USA). Results are expressed as means 6 SEM. Distribution of each sample was assessed with the Shapiro-Wilk test. When the distributions of both samples were normal their dispersions were assessed with the Levene test. We compared homoscedastic samples using a two-sample t-test or for heteroscedastic samples, we used the Welch test. When at least either of the samples was not normally-distributed, they were compared using the Mann-Whitney test. P values less than 0.05 were considered statistically significant.