Lineage Enforcement by Inductive Mesenchyme on Adult Epithelial Stem Cells across Developmental Germ Layers§


  • Renea A. Taylor,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    Search for more papers by this author
  • Hong Wang,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    Search for more papers by this author
  • Sarah E. Wilkinson,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    Search for more papers by this author
  • Michelle G. Richards,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    Search for more papers by this author
  • Kara L. Britt,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    Search for more papers by this author
  • François Vaillant,

    1. The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria 3050, Australia
    Search for more papers by this author
  • Geoffrey J. Lindeman,

    1. The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria 3050, Australia
    Search for more papers by this author
  • Jane E. Visvader,

    1. The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria 3050, Australia
    Search for more papers by this author
  • Gerald R. Cunha,

    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    2. Department of Urology, The University of California, San Francisco, California 94143, U.S.A
    Search for more papers by this author
  • Justin St. John,

    1. Clinical Sciences Research Institute, University of Warwick, CSB-University Hospital, Coventry, CV2 2DX United Kingdom
    Search for more papers by this author
  • Gail P. Risbridger

    Corresponding author
    1. Centre for Urological Research, Monash Institute of Medical Research, Monash University, Melbourne, Victoria 3168, Australia
    • Centre for Urological Research, Monash Institute of Medical Research, Monash University, 27-31 Wright Street, Clayton, Victoria 3168, Australia
    Search for more papers by this author
    • Telephone: +613 9594 7408; Fax: +613 9594 7420

  • Author contributions: R.A.T. and G.R.C.: conception and design, data analysis and interpretation, manuscript writing; H.W.: collection and/or assembly of data, data analysis and interpretation; S.E.W. and J.S.-J.: collection and/or assembly of data, data analysis and interpretation, manuscript writing; M.G.R. and K.B.: collection and/or assembly of data; F.V., G.J.L, and J.E.V.: provision of study material; G.P.R.: conception and design, financial support, data analysis and interpretation, manuscript writing.

  • First published online in STEM CELLS EXPRESS October 27, 2009.

  • §

    Disclosure of potential conflicts of interest is found at the end of this article.


During development, cell differentiation is accompanied by the progressive loss of pluripotent gene expression and developmental potential, although de-differentiation in specialized cells can be induced by reprogramming strategies, indicating that transdifferentiation potential is retained in adult cells. The stromal niche provides differentiating cues to epithelial stem cells (SCs), but current evidence is restricted to tissue types within the same developmental germ layer lineage. Anticipating the use of adult SCs for tissue regeneration, we examined if stroma can enforce lineage commitment across germ layer boundaries and promote transdifferentiation of adult epithelial SCs. Here, we report tissue-specific mesenchyme instructing epithelial cells from a different germ layer origin to express dual phenotypes. Prostatic stroma induced mammary epithelia (or enriched LinCD29HICD24+/MOD mammary SCs) to generate glandular epithelia expressing both prostatic and mammary markers such as steroid hormone receptors and transcription factors including Foxa1, Nkx3.1, and GATA-3. Array data implicated Hh and Wnt pathways in mediating stromal-epithelial interactions (validated by increased Cyclin D1 expression). Other recombinants of prostatic mesenchyme and skin epithelia, or preputial gland mesenchyme and bladder or esophageal epithelia, showed foci expressing new markers adjacent to the original epithelial differentiation (e.g., sebaceous cells within bladder urothelium), confirming altered lineage specification induced by stroma and evidence of cross-germ layer transdifferentiation. Thus, stromal cell niche is critical in maintaining (or redirecting) differentiation in adult epithelia. In order to use adult epithelial SCs in regenerative medicine, we must additionally regulate their intrinsic properties to prevent (or enable) transdifferentiation in specified SC niches. STEM CELLS 2009;27:3032–3042


The ability of adult, tissue-specific epithelial stem cells (SCs) to transdifferentiate into other cell types is particularly important in regenerative medicine. Transdifferentiation occurs when a cell transforms into a different cell type normally outside its resident differentiation state. In the developing embryo, cell differentiation is accompanied by the progressive loss of pluripotent gene expression and developmental potential [1]. Nevertheless, de-differentiation in committed cells can be induced by somatic cell nuclear transfer and non-oocyte-based approaches such as fusion of somatic cells with embryonic SCs, treatment of somatic cells with extracts of pluripotent cells, and/or retroviral transduction of somatic cells to overexpress pluripotency genes [2]. Therefore, the capability for reprogramming and transdifferentiation is retained into adulthood [3, 4]. The potential for transdifferentiation of differentiated somatic (adult) epithelial SCs opens up novel avenues for regenerative medicine and avoids the ethical issues surrounding the use of embryonic SCs, but it is crucial to ensure the safety of this process [1].

Another approach used to reprogram epithelial differentiation is varying the stromal microenvironment. Previously, we and others have shown that prostatic stroma induces prostatic lineage commitment and differentiation in human embryonic SCs (ESCs) [5], spermatagonial SCs (SSCs) [6], and isolated prostatic SCs [7–9], demonstrating the dominance of differentiating cues from stroma in the SC niche that determine SC fate. The stromal niche also provides transdifferentiating cues to SCs residing within adult epithelia, but stroma-induced transdifferentiation is restricted to tissue types within the same germ layer lineage. Using tissue recombination, where selected epithelia that contain putative SCs can be recombined with foreign or heterotypic mesenchyme [10], Cunha and colleagues showed that normally growth-quiescent adult epithelial tissue could be stimulated by embryonic/neonatal mesenchyme to undergo profound morphogenetic and differentiative changes. For example, within endoderm-derived tissues, mouse prostatic mesenchyme induced prostatic differentiation in adult rat, mouse, and human bladder cells as well as postnatal mouse vaginal epithelial cells [11–14]. Similarly, within mesoderm-derived tissues, seminal vesicle epithelial differentiation could be induced from epithelium of the adult ureter, ductus deferens, or epididymis [15, 16]. Likewise, within ectoderm-derived tissues, embryonic skin epithelial cells could be induced to differentiate into mammary ducts producing casein [17]. In all cases, the newly induced epithelial phenotype corresponded to an alternate phenotype within the repertoire of the original germ layer. Arising from these observations, it remains unknown if stroma associated with one germ layer can redirect or reprogram differentiation of adult SCs from a different germ layer origin. Along these lines, SSCs transdifferentiated in vivo into tissues of all germ layers, including prostatic, uterine, and skin epithelium, when recombined with the appropriate mesenchyme [6]. Similarly, adult testicular cells (including germ cells) and neural SCs gave rise to mammary epithelial ducts in fat pad transplantation studies, further confirming the dominance of the niche over the SCs autonomous phenotype [18]. To test if developmental stroma can reprogram “epithelial” SCs across different developmental germ layers, we generated heterotypic tissue recombinants composed of ectoderm-derived epithelia and endoderm-associated mesenchyme.



Timed pregnant SD/OB rats were obtained from Monash Animal Services. Urogenital sinus mesenchyme (UGM) and urogenital epithelium were obtained from male rat embryos at 16 days gestation (plug day = day 0), and rat seminal vesicle mesenchyme (SVM) were obtained from male pups on the day of birth (day 0) as previously described [5, 19]. Both UGM and SVM function as potent prostatic inducers when isolated at the time of epithelial budding and branching morphogenesis; seminal vesicle development occurs after birth in rats [20]. β-actin-BALB/c-GFP mice were bred at Monash Medical Centre Animal Facility. Mammary glands were harvested from fetal (E14.5), immature (2-3 weeks postnatal), or adult (20-40 weeks) female BALB/c-GFP mice. Preputial glands, bladder, and esophageal tissues were obtained from embryonic BALB/c-GFP mice (E16.5–E18.5) as previously described [21]. Enriched mammary SCs (MaSCs) were isolated from 8 week old adult Rosa26 mice as previously described [22]. All animal handling and procedures were carried out in accordance with National Health and Medical Research Council guidelines for the Care and Use of Laboratory Animal Act and according to the Animal Experimentation and Ethics Committee at Monash Medical Centre, Clayton, Australia.

Tissue Collection

Tissue collection from mouse and rat tissues involved microdissection of urogenital sinus, seminal vesicles, mammary glands, preputial glands, esophagus, and foot pad skin. This was performed in a modified watch glass (Maximov depression slides (Fisher, Pittsburgh, PA, using a dissecting microscope (SZX12, Olympus Corporation, Tokyo, Japan, in the presence of dissecting media (basal medium of Dulbecco's Modified Eagels Media (DMEM) and Hams F-12 (1:1 vol/vol) supplemented with penicillin and streptomycin (100 U/ml) and fungizone (20 μg/ml) at pH 7.3. Mesenchyme and epithelia were obtained by mechanical separation following digestion in 1% trypsin (Difco, Detroit, MI, in Hank's calcium- and magnesium-free Balanced Salt Solution (Gibco, Invitrogen, Vic, Australia, for 60 minutes at 4°C, followed by washing in DMEM plus 10% fetal calf serum (FCS) to neutralize the trypsin. Epithelial ducts and associated mesenchyme were physically dissected apart using an iris scalpel and fine forceps from each tissue type as previously described [23]. Enriched MaSCs were isolated by fluorescence-activated cell sorting for lineage markers (LinCD29HICD24+/MOD) from 8 week old adult Rosa26 mice as previously described [22].

Tissue Recombinants

Heterotypic tissue recombinants (i.e., tissues from both mouse and rat) were generated by combining intact mouse mammary epithelia (∼5,000 cells) or isolated (∼1,500 cells; MaSCs) cells in 10 μl rat tail collagen together with rat prostatic stroma (∼250,000 cells; either UGM or SVM). Sheaths of intact stroma were placed into collagen gels for recombination with intact or isolated epithelial ducts or cells (i.e., prostatic stroma + mammary gland epithelium) for 16-24 hours prior to grafting. The procedures of handling isolated cells, tissue recombination, and subrenal grafting were performed as previously described [5, 19]. Briefly, heterotypic tissue recombinants were grafted under the kidney capsule of adult male immune-deficient SCID mice bearing subcutaneous 10 mg testosterone implants to augment androgen levels. Recombinants were grown in host mice for 4 weeks (sufficient time for maximal terminal differentiation and maturation of epithelial cells [19]). Mesenchyme and epithelia of each tissue type were grafted alone as a method of detecting contamination. At the time of harvest, mice were anesthetized with Avertin (Sigma Cat #T4, 840-2; 0.4 mg/g of body weight), then killed by cervical dislocation. Grafts were imaged and measured, removed from kidneys and either fixed with Bouin's or formalin for histological analysis or placed in DMEM 2% FCS for epithelial cell isolation.

Flow Cytometry and RNA Extraction

For epithelial cell isolation, grafts were sliced into 1 mm2 pieces and centrifuged (5 minutes, 1,500 rpm, 4°C) in DMEM 2% FCS. Tissue was resuspended in DMEM 2% FCS containing collagenase (300 U/ml) and hyaluronidase (100 U/ml) and gently rotated at 37°C for 45 minutes. Enzymatic activity was neutralized with DMEM 2% FCS. Organoids were digested to single cells using 0.1% Trypsin-Versene by gentle rotation at 37°C for ∼20 minutes. Cells were collected by centrifugation (10 minutes, 1,500 rpm, 4°C) and resuspended in DMEM 2% FCS. The total number of cells per graft was estimated, and multiple grafts were pooled into separate biological replicates to obtain sufficient RNA; ∼3-5 grafts were pooled to generate mammary or prostate epithelial tissues, whereas ∼10-12 grafts were pooled to generate prostate-s + mammary-e recombinant epithelial samples.

Propidium iodide (PI; Sigma, St. Louis,; 10 μg/ml) was added to cell suspensions of >10,000 cells/ml per sample. Cells were sorted into GFP+ (i.e., epithelial) and GFP (i.e., stromal) populations, and PI+ populations (discarded) using a MoFlo XDP flow cytometer (Beckman Coulter, Fullerton, CA, in conjunction with Summit V5 software. Sorted cells were centrifuged (5 minutes, 3,000 rpm, 4°C), supernatant was discarded, and cells were resuspended in 350 μl RLT Buffer (Qiagen, Hilden, Germany, with 10 μl β-mercaptoethanol/ml. Cell lysates were homogenized, and samples were stored at −80°C until RNA extraction. RNA was extracted using the Qiagen RNeasy Mini Kit and subsequently treated with TURBO DNA-free (Ambion, Austin, TX,, in order to eliminate DNA, according to manufacturer's instructions. RNA was concentrated and purified using the Qiagen RNeasy MinElute Cleanup Kit, and sample RNA concentrations were measured using a Nanodrop Spectrophotometer ND-1000 in conjunction with ND-1000 V3.3.1 computer software.

RT2 Profiler Polymerase Chain Reaction Arrays

Mouse Signal Transduction PathwayFinder RT2 Profiler PCR Arrays (SABiosciences, Cat#PAMM-014E, Frederick, MD, http://www. were used and performed on an ABI 7900HT Fast Real-Time PCR System (Applied Biosystems (AB), Foster City, CA, in conjunction with SDS V2.3 computer software. Samples included epithelium (e) from tissue recombinants with prostatic stroma (s) including prostate-s + mammary-e (n = 2; Group 1) and prostate-s + prostate-e (n = 2; Group 2) as well as mammary epithelium (n = 3; Group 3) and prostatic epithelium (n = 3) from adult GFP mice (set as control gene expression levels). Arrays were performed according to the manufacturer's instructions using the SA Biosciences RT2 First Strand Kit (Cat #C-03) and RT2 SYBR Green/ROX qPCR Master Mix (Cat #PA-012-8), and plates were loaded using a CAS-1200 robot (Corbett Robotics, and were centrifuged (1 minute, 1,000 rpm, room temperature) after loading and prior to running in Applied Biosystems 7900HT Fast Real-Time PCR System (Applied Biosystems). PCR data was formatted using SDS V2.3 and Microsoft Excel software, then analyzed by importing into the SABiosciences RT2 Profiler PCR Array Analysis online software.

Staining and Immunohistochemistry

All sections were stained for hematoxylin and eosin and heterotypic (rat/mouse) recombinants were stained with DAPI to visualize nuclei. Standard immunohistochemistry was performed using the DAKO Autostainer Universal Staining System (DAKO A/S, Denmark, Antibodies to AR, Foxa1 (HNF-3α, C-20; 0.5 μg/ml), Gata-3 (HG3-31; 2 μg/ml), and Nkx 3.1 (T19; 1 μg/ml) were purchased from Santa Cruz (Santa Cruz, CA,; to PR, ERβ, and CKH were purchased from DAKO; to smooth muscle α-actin from Sigma; to ERα from Novocastra (Newcastle upon Tyne, U.K.,; to Cyclin D1 from Abcam (Cambridge, U.K.,; 2 μg/ml) and used according to company specifications and as previously reported [5]. For (double) immunofluorescence (α-actin/CKH, α-actin/CK18 or GFP), antigen sites were retrieved by microwaving in 0.01 M citrate buffer, pH 6 (for α-actin/CKH or α-actin/CK18) or 0.05 M Tris buffer/0.01% EDTA pH 9 (for GFP). Non-specific immunoreactivity was blocked with MOM mouse Ig blocking reagent (Vector Laboratories, Peterborough, U.K.,; stock MOM Ig blocking reagent in 250 μl of TBS) and 30 minutes in DAKO (DAKO, Ely, Cambridgeshire, U.K., protein block. Sections were incubated with antibodies against α-actin (mouse IgG2a clone 1A4; 1:1000; Sigma), keratin 18 (mouse IgG1 clone Ks18.04; 1:2 dilution; Progen, Heidelberg, Germany,, high molecular weight cytokeratin (mouse IgG1 clone 34βE12; 1:14 dilution, DAKO) or GFP (rabbit IgG, 0.5 μg/ml, Invitrogen) overnight at 4°C. After washing in 0.05% Tween in TBS, sections were stained with goat anti-mouse IgG2a-Alexa647 (1:500 dilution; Invitrogen), anti-mouse IgG1-Alexa 488 (1:500 dilution; Invitrogen), or goat anti-rabbit IgG-Alexa 488 (1:250 dilution; Invitrogen). Sections were washed and mounted with Vectashield-DAPI (Vector Laboratories). All sections were examined using a Nikon A1 Rsi confocal microscope (Nikon, Japan, with lasers exciting at 402, 488, and 638 (emission detection at 425-475, 570-620, 662-737). Multicolor images were collected sequentially in three channels and captured using the Nikon NIS Elements imaging software. For Cyclin D1 immunolocalization, semiquantitation involved systematic counting of >1,000 epithelial cells from n = 3-6 individual tissue recombinants to account for regional variation in morphology and localization.

Analysis of Mitochondrial DNA

Individual cell populations were isolated using laser-capture microdissection. Paraffin sections of 5 μm were mounted on membrane slides (P.A.L.M. Microlaser Technologies, Germany, and stained with GFP antibody after deparaffinization. Slides were dehydrated in graded ethanol solutions and finally cleaned in xylene. The GFP positive glands and the negative stained stroma around the glands were dissected separately using PALM Laser-Capture Microdissection system (P.A.L.M. Microlaser Technologies) and capped in a solution of 10 mM Tris, 20 μg/ml proteinase K, 10 mM EDTA, and 1% Tween 20. Dissected tissues were incubated at 55°C for 12 hours to lyse DNA and stored at −20°C for mtDNA analysis. In order to perform mtDNA analysis and sequencing, 200 ηg of total DNA was amplified in 50 μl reactions containing 1x buffer, 1.5 mM MgCl2, 200 mM dNTPs (all Bioline, London, U.K.,; 0.5 μM CytBF primer (GAGGACAAATATCCATTCTGAGG); and 0.5 μM CytBR primer (AGATGGAGGCTAGTTGGCC), to produce 660 bp of the CytB gene. The reaction was performed in a BioRad Thermocycler (BioRad, Hercules, CA, with an initial denaturation of 95°C for 5 minutes followed by 35 cycles of 94°C for 45 seconds, 56°C for 30 seconds, and 72°C for 90 seconds. All PCR products were resolved on a 2% agarose gel at 100 V for 1 hour and purified using the QIAquick Gel Extraction Kit (Qiagen). Purified PCR products were then sequenced according to the automated direct sequencing protocol [24] using a GeneAmp 9,700 and the ABI PRISM BigDye terminator cycle sequencing ready reaction kit (AB, Foster City, CA) with primers CytBF and CytBR. Electrophoresis of cycle sequencing products was performed on an ABI PRISM 377 sequencer (AB). Sequences were viewed in “4Peaks” Version 1.7.2 and aligned in ClustalW ( Restriction enzyme digest was performed using 40 ηg of each PCR product and 5 U BSe I and 1x buffer (New England Biolabs, MA, USA, at 37°C for 90 minutes followed by enzymatic inactivation at 65°C for 10 minutes. Products were then resolved on 3% agarose electrophoretic gels.

Statistical Analysis

Six or more replicates were obtained for all tissue recombinations transplanted under the renal capsule of at least three male hosts. For histology and immunohistochemistry, at least six tissue recombinants were examined systematically, sampling throughout the entire tissue. For semiquantitation of Cyclin D1 immunolocalization, data were analyzed to determine normality and significant differences were determined by one-way analysis of variance followed by Tukey-Kramer's multiple comparison post hoc tests. Statistical analyses were performed using Graph Pad Prism 5.01 (Graph Pad Software, Inc., San Diego, CA Data are expressed as the mean ± standard error of the mean, and differences between groups were considered significant at p < .05.


Prostatic Stroma Supports Differentiation of Mammary Gland Epithelium

Rat-mouse heterotypic tissue recombinants of prostatic stroma plus GFP+ mouse mammary epithelium (derived from 2-3 week postnatal mammary glands) were compared with homotypic prostatic tissue recombinants (embryonic prostate-s + prostate-e; derived from E17.5 male mice) or mammary grafts (postnatal mammary-s + mammary-e). All three grafts types were grown in male immune-deficient SCID host mice for 4 weeks. Survival (based on identification of implanted epithelial cells indicating successful development of tissues) of heterotypic prostate-s + mammary-e and homotypic prostatic tissue recombinants was observed in 100% of grafts, whereas mammary grafts grown in male host mice showed reduced graft survival (31.25%, n = 16 grafts; Table 1), indicating that the addition of prostatic stroma significantly increased survival of mammary epithelial cells in the male host environment (Table 1).

Table 1. Efficiency of various epithelia to transdifferentiate into stromal-induced phenotype
inline image

In order to determine the identity of epithelial and stromal cell types, immunolocalization of α-actin was performed and showed significant differences in morphology between mammary grafts and heterotypic (prostatic-s + mammary-e) or prostatic tissue recombinants grafted into male host mice. Mammary grafts contained only a few atrophic epithelial ducts (presumably due to an absence of an estrogenic endocrine environment; Fig. 1A) with flattened luminal cells abutting α-actin-positive myoepithelial cells; α-actin-positive smooth muscle cells were not detected in adipose stroma (Fig. 1B). Grafted prostatic-s + mammary-e tissue recombinants exhibited abundant ductal growth and ductal branching morphogenesis where glandular ducts, lined by α-actin-positive myoepithelial and tall columnar luminal epithelial cells, were embedded in stroma containing random bundles of α-actin-positive smooth muscle cells (Fig. 1C--1D). Prostatic tissue recombinants were devoid of α-actin-positive myoepithelial cells but contained organized layers of α-actin smooth muscle cells in a mature prostatic stroma (Fig. 1E--1F). Dual labeling of basal epithelial cells (using cytokeratins) and myoepithelial (using α-actin) showed that mammary epithelium grafted alone or in combination with prostatic stroma co-express myoepithelial and basal cell markers (α-actin and CKH; Fig. 1G--1H). In contrast, prostatic grafts contained a discontinuous layer of CKH-positive, α-actin-negative basal cells along glandular ducts, surrounded by organized layers of α-actin-positive smooth muscle in the stroma (Fig. 1I; Individual panels in supporting information Fig. S1A--1E). Collectively, based upon expression of the above cytoskeletal and contractile proteins, heterotypic prostate-s + mammary-e tissue recombinants maintained mammary gland differentiation (i.e., α-actin+/CKH+ myoepithelial cells) that were surrounded by a fibromuscular stroma containing α-actin-positive smooth muscle cells. These results were observed irrespective of the age or developmental stage of the mammary epithelia, including developing (E14.5; supporting information Fig. S2A), immature (2-3 weeks postnatal; Fig. 1), or adult (20-40 weeks; supporting information Fig. S2B), or the type of mesenchyme (UGM or SVM) utilized. Control grafts of prostate-s (either UGM or SVM) alone gave rise to epithelium-free fibrous connective tissue when placed in collagen gels and grafted under the kidney capsule of male host mice.

Figure 1.

Prostatic stroma supports differentiation of mammary gland epithelium. Heterotypic tissue recombinants of rat prostatic mesenchyme (s; stroma) plus mouse mammary gland epithelia (e; epithelia) were grown in male host mice. Control tissues were intact mammary glands or homotypic prostate recombinants grown in male host mice. (A–F): Immunolocalization of α-actin was used to visualize glandular duct and presence of myoepithelial cells (marked by arrow) or stromal smooth muscle cells (marked by ∗). (F): Inset is representative concentration matched IgG negative control for images (A–F). (G–I): Dual immunofluorescent labeling of α-actin/CKH (merged image with co-localization-yellow; CKH-green; α-actin-red; DAPI-blue) confirmed co-localization of basal and myoepithelial cell markers (α-actin+/CKH+; arrow) in mammary glands and prostate-s + mammary-e tissue recombinants, but not mouse prostate where basal cells do not contain α-actin. (J): Immunolocalization of GFP protein (FITC labeled anti-GFP) confirmed the epithelia in prostate-s + mammary-e tissue recombinants was derived from BALB/c-GFP mouse mammary epithelia. (K): DAPI staining of nuclei confirmed epithelial cells were of mouse (M) origin and stroma was rat-derived (R) nuclei (glandular ducts outlined in dashed line). (L): Heterotypic prostate-s + mammary-e tissue recombinants were analyzed for cell fusion events by assessing contributions of mouse and rat mitochondrial (mt)DNA following amplification of the CytB gene. Pure populations of epithelium (Lanes 1 and 2) or stroma (Lane 3) were isolated using laser capture microdissection. Polymerase chain reaction (PCR) was also performed to generate CytB PCR with mouse (Lane 4), rat (Lane 5) templates, or a combination of both (Lane 6). Equal mixtures of mouse and rat CytB mtDNA, as shown in Lane 6, were used to demonstrate the predicted pattern of mtDNA banding if fusion had taken place (Mouse & Rat). Left lane = 100 bp Ladder.

We used several methods to confirm the origin of epithelial cells and exclude the possibility of epithelial contamination. First, GFP+ mouse epithelium could be visualized in association with non-GFP rat prostatic mesenchyme before and after the grafting (supporting information Fig. S3A). Immunohistochemistry for GFP protein expression revealed specific localization to epithelial cells and not the surrounding mesenchyme (Fig. 1J; individual panels in supporting information Fig. S1F). Visualization of nuclei using DAPI showed that prostate-s + mammary-e recombinants contained mouse epithelial cells (“speckled” nuclear appearance), whereas surrounding stroma was rat-derived (“smooth” nuclear appearance; Fig. 1K). Second, to exclude cell fusion events, we examined the CytB gene of the mouse or rat mitochondrial (mt) genome since mesenchyme/stroma were derived from rat and epithelia from mouse. Using laser capture microdissection of heterotypic tissue recombinants including GFP+ mammary epithelia (derived from mice) and prostatic stroma (derived from rats), we analyzed extracts from each cell source for mtDNA. Restriction enzyme digest of PCR products for the CytB gene revealed that only one population of mtDNA was present in each of the isolated populations of cells from the tissue recombinants specific to the transplanted cell type. This confirmed that mouse--rat cell fusion had not occurred (Fig. 1L; Lanes 1-3), while Figure 1L (Lane 6) and supporting information Figure S4 demonstrate the likely outcomes if equal contributions of mtDNA were present following such fusions.

Prostatic Stroma Induces Dual Mammary- and Prostate-Specific Factors

In addition to unique prostate and mammary gland patterns of cytoskeletal and contractile protein expression, prostatic and mammary epithelial cell differentiation involves the establishment of unique regulatory programs indicative of cell fate, including steroid receptor expression and developmental cell-specific transcription factors. In order to determine whether mammary gland epithelia retained the molecular machinery for estrogenic response even in the presence of prostatic mesenchyme and male systemic hormones (i.e., stromal-directed differentiation and lineage commitment), we examined the expression of nuclear hormone receptors and tissue-specific transcription factors known to be involved in prostatic and mammary fate specification in grafts of heterotypic prostate-s + mammary-e tissue recombinants, which were compared with mammary grafts grown in male host mice (supporting information Fig. S2). Prostate-s + mammary-e tissue recombinants maintained prototypic female (estrogen-regulated) hormone receptors (ERα and PR) in epithelial cells (Fig. 2A) and expressed ERβ in both stromal and epithelial cells (Fig. 2B), while gaining expression of a prototypic male (androgen-regulated) nuclear hormone receptor, AR, in mammary gland derived epithelial cells (Fig. 2C). Immunolocalization of GATA-3, a mammary gland luminal epithelial specification factor [25], remained strongly expressed in luminal cells of mammary epithelia, even when grafted in combination with prostatic stroma and grown in male host mice (Fig. 2D). Uniquely, we also detected immunolocalization of two of the earliest known prostate-specific markers, Nkx3.1 [26] and Foxa1 (HNF-3β, [27, 28]) in epithelial cells of prostate--mammary tissue recombinants (Fig. 2E, 2F). Collectively, we showed co-immunolocalization of androgen- and estrogen-regulated hormone receptors, as well as prostate- (Nkx3.1 and Foxa1) and mammary-specific (GATA-3) transcription factors in epithelial ducts of heterotypic prostate-s + mammary-e tissue recombinants grown in male host mice. Thus, prostatic stroma induced the generation of chimeric epithelium expressing features of both ectodermal and endodermal origin, that is, activation of dual prostatic (endoderm) and mammary gland (ectoderm) transcriptional programs. Importantly, the combination of AR, Nkx3.1, and Foxa1 expression is not usually observed in mammary epithelium, but this unique expression of prostate-associated proteins was induced in mammary epithelium by prostatic stroma. The dual morphology and expression of nuclear hormone receptors and cell fate specification transcription factors was observed uniformly in all glandular areas. In each tissue recombinant, we did not observe regions of prostate-like glands adjacent to mammary-like glands; that is, each duct expressed a uniform “mixed” phenotype.

Figure 2.

Expression of hormone receptors and transcription factors in tissue recombinants. Representative photomicrographs showing immunolocalization of steroid hormone receptors and transcription factors in heterotypic tissue recombinants of rat prostatic mesenchyme (s; stroma) + mouse mammary gland epithelia (e; epithelia) that were grown in male host mice. (A): Estrogen receptor (ER)-α and progesterone receptor (PR; inset). (B): ERβ. (C): Androgen receptor (AR; inset is representative concentration matched IgG negative control for all images). (D): GATA-3. (E): Nkx3.1. (F): Foxa1.

Prostatic Stroma Induces Differentiation of Enriched MaSCs

In order to determine whether stromal-induced changes in epithelial cell fate were induced directly in adult mammary SCs or by transdifferentiation of terminally differentiated cells, a second source of mammary epithelium, SC-enriched populations (enriched-MaSCs), were obtained using an established cell surface marker phenotype (LinCD29HICD24+/MOD) from 8 week old FVBN or adult Rosa26 mice. Single cells within this MaSC population are multipotent and self-renewing, possessing the ability to reconstitute a complete mammary gland in vivo [22, 29]. Heterotypic tissue recombinants of prostate-s + MaSCs were grown in male host mice for 4 weeks; enriched MaSCs grafted alone (or with mammary stroma; n = 4) failed to survive. When grown with prostatic mesenchyme, enriched MaSCs underwent growth and differentiation, producing glandular ducts surrounded by fibromuscular stroma (Fig. 3A). Morphology was similar to intact mammary gland epithelium, that is, α-actin-positive myoepithelial and tall columnar luminal epithelial cells surrounded a stroma containing random bundles of α-actin smooth muscle cells (Fig. 3B). We confirmed the origin of epithelial cells (excluding the possibility of epithelial contamination) by visualization of nuclei using DAPI (i.e., mouse-derived epithelial cells surrounded by rat-derived stromal cells; Fig. 3C) or β-galactosidase expression from Rosa26 mice (supporting information Fig. S3B). Analysis of steroid receptor expression and developmental cell-specific transcription factors revealed that enriched MaSCs expressed epithelial androgen- (AR) and estrogen-regulated hormone (ERα and ERβ) receptors, as well as mammary (GATA-3) and prostate (Nkx3.1 and Foxa1) specific transcription factors in epithelial ducts composed of prostate-s + MaSCs grown in male host mice (Fig. 3D--3I). Expression and localization of the above markers was comparable to tissue recombinants of prostate-s + mammary-e, although Foxa1 was only expressed in ∼30% of glandular epithelial cells (Fig. 3I) compared to >90% in prostate-s + mammary-e grafts (Fig. 2F).

Figure 3.

Prostatic stroma induces prostatic phenotype in enriched mammary stem cells (MaSCs). Representative photomicrographs showing heterotypic tissue recombinants of rat prostatic mesenchyme (s; stroma) + mouse enriched-MaSCs isolated from FVBN or Rosa 26 mice; CD29HICD24+/MOD grown in male host SCID mice. (A): H&E shows glandular ducts (derived from MaSCs) embedded in prostate stroma. (B): Immunolocalization of α-actin shows α-actin-positive myoepithelial cells and α-actin-smooth muscle cells in stroma. (C): DAPI staining of nuclei indicates mouse (M) and rat (R) cells (glandular ducts outlined in dashed line). (D–I): Immunohistochemistry for steroid hormone receptors and transcription factors. (D): Estrogen receptor (ER)-α. (E): ERβ. (F): Androgen receptor (AR; inset is higher power). (G): GATA-3. (H): Nkx3.1. (I): Foxa1.

Stromal-Induced Lineage Commitment Involved Hedgehog and Wnt Signaling

In order to investigate the signaling pathways involved in the stromal-epithelial interactions between endoderm-associated prostatic stroma and ectoderm-derived (GFP+) mammary epithelium [Group 1; prostatic-s + GFP+ mammary-e] and prostate epithelium [Group 2; isolated from prostatic-s + GFP+ prostatic-e] tissue recombinants, mRNA analysis using Mouse Signal Transduction PathwayFinder RT2 Profiler PCR Arrays were performed. Epithelial cells expressing GFP were isolated from Group 1 and Group 2 following 4 weeks grafting in male SCID mice (n = 2 biological replicates of pooled grafts per group). For comparison, non-grafted GFP+ prostate epithelium was isolated from adult male mice (control; n = 3) and mammary gland epithelium (Group 3) was isolated from 6-7 week old female mice (n = 3 biological replicates of pooled samples). Although several genes were up-regulated in multiple pathways, the most consistent and elevated fold changes were observed in genes involved in Wnt and/or Hedgehog (Hh) signaling pathways in mammary epithelium following prostatic stromal induction (Fig. 4A). A subset of genes was uniquely up-regulated compared to prostate epithelium (e.g., En1, Ccnd1 and Birc5), whereas others showed similar expression between prostate and mammary epithelium induced by prostatic stroma (e.g., Bmp2 and Myc). For the full list of genes and relative fold changes, see supporting information Table S1, or for heatmap clustering, see supporting information Figure S5. Using immunohistochemistry, we confirmed that the product of Ccnd1, Cyclin D1, was uniquely increased in mammary epithelium induced by prostatic stroma, and that this increase in protein expression was significant when compared to mammary or prostatic control tissues (Fig. 4B–4D). This was observed in terms of intensity of immunostaining, and the percent of Cyclin D1-positive epithelial cells; using semiquantitation, we showed that Cyclin D1 was expressed in 58.41 ± 2.57% of mammary glands compared to 75.70 ± 4.13% (p < .05) in epithelial cells of prostatic-s + mammary-e tissue recombinants or 66.66 ± 3.04% in prostatic epithelium (Fig. 4E).

Figure 4.

Wnt and Hedgehog genes are induced in mammary epithelia by prostatic stroma. Gene expression changes in tissue recombinants of [Group 1] prostatic stroma (s) plus mammary epithelium (e; n = 2) or [Group 2] tissue recombinants of prostatic-s + prostatic-e (n = 2) compared to intact prostate epithelium derived from adult male mice (n = 3). GFP+ mammary or prostate epithelium was isolated by fluorescence-activated cell sorting. (A): Summary of genes involved in Hedgehog or Wnt signaling represented as ratio of fold change in [Group 1] mammary epithelium (induced by prostatic stroma) or [Group 2] prostate epithelium, compared to prostatic epithelium (relative to Hsp90ab1; positive values indicate up-regulation and negative values indicate down-regulation). Photomicrographs representing localization of Cyclin D1 in (B): 6-7 week old intact mammary ducts or (C): heterotypic tissue recombinants of rat prostate-s + mammary-e and (D): prostate (generated by prostatic-s + prostatic-e; inset is concentration matched IgG control for (B–D)) following 4 weeks growth in male host mice. (E): Semiquantitation of Cyclin D1 localization represented as % positive epithelial cells of intact mammary glands (black bar, n = 3), tissue recombinants of prostatic stroma-s + mammary-e (grey bar, n = 6) or tissue recombinants of prostatic-s + prostatic-e (white bar, n = 4). Data is represented as mean ± SEM; different superscripts indicate groups that are significantly different to control (prostatic epithelium; p < .05).

Stromal-Induced Transdifferentiation of Other Ectoderm-Derived Epithelium

In order to show that transformation of ectoderm-derived epithelia to an endoderm-associated phenotype was not restricted to hormone-responsive tissues such as mammary gland and prostate, we tested ectoderm to endoderm transition using adult epithelial cells from nonreproductive tissues. To do this, we generated heterotypic tissue recombinants of rat prostatic stroma plus mouse skin epithelium (derived from foot skin of adult male BALB/c-GFP mice) that were grown under the renal capsule of male host SCID mice. Visualization of nuclei using DAPI confirmed that epithelial cells were derived from mouse skin and not from epithelial contamination of the rat prostate mesenchyme (supporting information Fig. S3C).

Under the influence of prostatic stroma, skin epithelium differentiated into glandular ducts that contained regions of multilayered squamous CKH-positive, AR-negative, Foxa1-negative epidermis adjacent to foci of prostate-like ducts composed of a polarized columnar epithelia containing basal (marked by CKH) and luminal cells expressing AR and Foxa1 (Fig. 5A--5H). Glandular ducts were surrounded by AR-positive stroma (derived from prostatic mesenchyme) containing thick sheaths of α-actin smooth muscle (Fig. 5E--5G). These data showed lineage enforcement by inductive mesenchyme resulting in chimeric tissue recombinants containing foci of glandular epithelial ducts expressing prostatic markers and a prostatic morphology adjacent to epithelium, which maintained skin morphology and differentiation. All grafts were recovered, and prostate-like regions were observed as isolated foci in 33% (n = 6) of tissue recombinants (Table 1).

Figure 5.

Prostate stroma induces transition of ectoderm-derived skin epithelia. Heterotypic recombinants were generated from endoderm-associated prostatic mesenchyme (stroma; s) and ectoderm-derived skin (foot pad) epithelium (e) and subrenally grafted into male SCID mice. Immunolocalization of (A–B): high molecular weight cytokeratins (CKH), (C–D): Foxa1, (E–F): androgen receptor (AR), and (G–H): α-actin. Recombinants showed areas of epidermal differentiation (marked by arrow) adjacent to induced foci of prostate-like glands (marked by ∗) expressing AR, Foxa1 together with a polarized columnar epithelia adjacent to squamous skin epithelium.

Stromal-Induced Transdifferentiation of Endoderm-Derived Epithelium

Following the demonstration of transdifferentiation of ectoderm-derived epithelia from mammary and skin to prostate by endoderm-associated stroma, we tested if the reverse process could occur whereby endoderm-derived epithelia could be induced to express ectoderm-like phenotypes under the influence of ectoderm-associated mesenchyme. To do this, we generated tissue recombinants composed of rat preputial gland mesenchyme (preputial glands are exocrine sebaceous glands associated with the external genitalia of rats and mice) combined with mouse bladder or esophageal epithelium (derived from adult BALB/c-GFP mice). In preputial gland-s + bladder-e or preputial gland-s + esophageal-e recombinants, we demonstrated the formation of chimeric tissue recombinants containing foci of induced preputial sebaceous glands adjacent to bladder and/or esophageal epithelia that maintained its original cell fate (Fig. 6A--6C). Sebaceous glands that exhibit a unique characteristic histodifferentiation are strictly skin (ectoderm) derivatives and have never been shown to differentiate from endoderm-derived cells.

Figure 6.

Preputial gland stroma induces transition of endoderm-derived epithelia. Heterotypic recombinants were generated from ectoderm-associated preputial gland mesenchyme (stroma; s) and endoderm-derived bladder or esophageal epithelium (e) and subrenally grafted into male SCID mice. (A-B): H&E staining showed foci of sebaceous epithelia cells (marked by ∗) were observed adjacent to transitional or stratified bladder epithelia (marked by arrow). (C-D): Immunostaining of Foxa1, localized to regions with intrinsic bladder differentiation (marked by arrow), but not in foci of sebaceous epithelia responding to preputial gland stroma in bladder epithelia (marked by ∗). (E): H&E staining showed foci of sebaceous epithelia cells (marked by ∗) were also observed in esophageal epithelium (marked by arrow) under the influence of preputial gland mesenchyme (inset shows low power image of esophageal epithelium and adjacent foci of sebaceous acini).

Under the influence of preputial gland mesenchyme, bladder epithelia retained its urothelial phenotype and expressed Foxa1 (previously reported in adult bladder [30]), compared to adjacent foci of sebaceous epithelial cells that did not express Foxa1, confirming the transformation of epithelia in those select regions (Fig. 6C, 6D). Visualization of nuclei using DAPI confirmed that bladder and esophageal epithelium were mouse-derived, whereas preputial gland stroma was rat-derived (supporting information Fig. S3D--S3E). All grafts were recovered, and sebaceous epithelial cells were observed as isolated foci in 66% (n = 6) and 33% (n = 6) of bladder esophageal recombinants, respectively (Table 1).


Here we showed that organ-specific mesenchyma can enforce lineage commitment and alter terminal differentiation of adult epithelial SCs across endoderm and ectoderm germ layer boundaries. Our studies primarily focused on the differentiation of mammary gland epithelia induced by prostatic stroma. We showed that transdifferentiation occurred in the absence of cell fusion and transfer of cellular components or genetic information and was facilitated by enriched MaSCs (LinCD29HICD24+/MOD). We demonstrated the complexity and generality of the principle of stroma as a director of adult SC fate in a variety of other epithelia including skin, bladder, and esophagus. These tissues span different developmental origins (ectoderm and endoderm) and are either hormone-independent (skin, bladder, esophagus) or hormone-dependent (mammary gland).

Although it is known that nonsomatic testicular germ cells are pluripotent and can undergo directed differentiation [31–33], our data show that, in the absence of pluripotency, committed somatic adult epithelial SCs can “move” towards a new differentiated state, across germ layer boundaries, with commitment to a new cell lineage while retaining limited properties of their original phenotype, supporting the dominance of the tissue microenvironment over the intrinsic nature of SC differentiation. Although transdifferentiation of adult epithelial cells by stroma has been extensively demonstrated within tissues of the same embryological origin, our data provide novel evidence that stroma-induced differentiation also occurs between tissues of different embryological origins (i.e., endoderm and ectoderm), specifically mediated via enriched populations of MaSCs. Similarly, transdifferentiation of isolated multipotent SCs was recently demonstrated using fetal and adult neural SCs that produced mammary epithelial cells in the inductive mammary fat pad [5, 34]. Advances in isolation and characterization of tissue-specific SCs will allow further investigation of their differentiation potential and how they are influenced by stromal microenvironments. Our data imply that mammary SCs show a more limited range of differentiation capacity compared with SSCs and ESCs, since under the same experimental conditions, reported by us and others [35], mammary epithelial SCs were unable to fully transform into a prostatic phenotype but instead maintained several of its original differentiation markers, in addition to expressing new ones. In tissue recombinants of other tissue types (including skin, bladder, or esophagus), we showed focal areas of new phenotype (induced by stroma) adjacent to regions of original differentiation markers. Whether these partial or focal changes in differentiation are determined by the distribution of SCs (i.e., areas where epithelial differentiation was maintained were devoid of multipotent SCs) remains unknown.

Lineage commitment requires activation and repression of key transcription factors that act as downstream targets of sequential and reciprocal cross-talk between the epithelial SCs and the surrounding mesenchyme [36]. The transformation of mammary epithelia to gain markers of prostatic differentiation involved the establishment of unique regulatory programs that control cell fate including steroid receptors and developmental cell-specific transcription factors. Here we showed the expression of androgen-regulated prostate-specific proteins including AR, Nkx3.1 [26], and Foxa1 (HNF-3β) [28] in mammary epithelium of heterotypic tissue recombinants. Androgen receptor and Nkx3.1 are two of the earliest markers of prostatic differentiation that are expressed at low/undetectable levels in mammary gland epithelium. Foxa1 interacts with AR and multiple prostatic enhancers to control androgen-induced activation of both human and rodent prostate-specific genes [18]. Foxa1 binding is also essential for ER--chromatin interactions and subsequent expression of estrogen target genes [27]. We showed that expression of mammary-associated estrogen-regulated transcription factors (GATA-3 and ERα) was maintained in the same glandular epithelia. GATA-3 is a critical component of the master cell-type-specific transcriptional network, as well as key determinant of mammary gland morphogenesis and luminal cell specification [25, 37]. Expression of GATA-3, ERα, and PR is also highly correlated in mammary development and cancer [38]. These data provide evidence of activation of both androgen- and estrogen-regulated pathways in heterotypic prostate--mammary tissue recombinants that were grown in male hosts and thus exposed to androgen stimulation. These tissue recombinants were analyzed following 4 weeks growth in host mice, and although we cannot predict whether the cellular localization or expression of these factors would alter after a longer period of time, it is unlikely since this time is widely used to allow sufficient time for maximal differentiation of prostate and mammary epithelial cells [12, 19, 21].

We were unable to address the functional potential of the mammary glands induced by prostatic stroma to produce milk proteins, since tissue recombinants were hosted in male mice. Previously, Cunha and colleagues generated tissue recombinants of UGM plus adult mammary gland epithelium (MGE) grown in female mice [21]. When exposed to female hormonal milieu (lacking systemic androgens), mammary epithelia maintained their original tissue identity (maintenance of α-actin-positive myoepithelial cells) under the influence of prostatic stroma but showed evidence of squamous metaplasia. Squamous metaplasia is a response characteristic of prostate epithelium (but not mammary gland epithelium) to estrogen stimulation [39], indicating that the prostatic stroma influenced mammary epithelial differentiation in the female host estrogenic endocrine environment. Tissue recombinants of UGM+MGE grown in female hosts did not produce milk protein even under the stimulation of lactogenic hormones (generated by pituitary grafts) [21], suggesting the stromal microenvironment had a major influence on the functional differentiation and biological activity of mammary epithelium, similar to the findings reported here in male host mice.

Numerous signaling molecules have been implicated in niche control of cell fate, including Hh, Wnts, BMPs, fibroblast growth factor, transforming growth factor β, and Notch [40, 41]. During prostate organogenesis, systemic androgens stimulate ARs in developing prostatic mesenchyme, and stromally-derived “andromedins” are released and induce prostatic morphogenesis in adjacent epithelium [42]. A critical contribution of stromal TGF-β signaling to prostatic differentiation was demonstrated using TGF-β type II receptor stromal conditional knockout mice (Tgfbr2fspKO), where mouse bladder epithelium underwent prostatic differentiation when recombined with wild-type UGM, but not Tgfbr2fspKO UGM [13]. Additionally, Wnt signaling was identified as a key component of stromal--epithelial signaling. Here, we report up-regulation of genes involved in the Wnt and Hh signaling pathways in mammary epithelial cells isolated from heterotypic tissue recombinants. Activation of Wnt and/or Hh signaling are known to be involved in SC differentiation and cell fate determination of multiple tissues [43, 44]. Our data show that Wnt and Hh downstream targets genes were up-regulated in mammary epithelial cells by a prostatic stroma, particularly Ccnd1 (expression of protein product Cyclin D1 confirmed by immunolocalization), suggesting that prostatic stroma induced proliferation of mammary epithelium, as well as altering differentiation and cell fate, irrespective of their stage of lineage commitment or differentiation. In the prostate gland, stromal activation of Wnt or Hh signaling acts in a paracrine manner to influence epithelial proliferation and differentiation [45–48], and here we show that these stromal--epithelial interactions are maintained in prostate--mammary recombinants.


Overall, we showed that inductive embryonic/neonatal mesenchyme can provide a unique microenvironment (or “niche”) for transdifferentiation of adult somatic epithelial SCs, which clearly retain the ability to be reprogrammed by stroma, even outside their normal developmental germ layer repertoire. Thus, for safe transplantation of multipotent adult epithelial SCs in regenerative medicine, the intrinsic properties of the SCs must be correctly aligned with the cues from the stromal microenvironment, as mismatch of epithelial SC-stromal cell niche can result in mixed tissue phenotypes or onset of disease.


The authors are grateful to Tameeka Hill, Camden Lo, Vivian Vasic, and James Nyugen for technical assistance. The research was supported by grants from the Australian Research Council (ARC #DP0987059) and Peter and Lyndy White Foundation. RAT is the recipient of the Peter Doherty Fellowship (#284395) and Prostate Cancer Foundation of Australia Rotary Research Fellowship. GPR is an NH&MRC Principal Research Fellow (#43796). The research by JV, GL, and FV was supported by grants from the NHMRC and Victorian Breast Cancer Research Consortium. Acknowledgment of funding: Australian Research Council, NH&MRC, Peter and Lyndy White Foundation, Prostate Cancer Foundation of Australia, Victorian Breast Cancer Research Consortium.


There are not conflicts or competing financial interests relating to this work.