SEARCH

SEARCH BY CITATION

Keywords:

  • EZH2;
  • Polycomb;
  • Mammary stem cells;
  • Luminal cell;
  • Differentiation

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Specification of the cellular hierarchy in the mammary gland involves complex signaling that remains poorly defined. Polycomb group proteins are known to contribute to the maintenance of stem cell identity through epigenetic modifications, leading to stable alterations in gene expression. The polycomb protein family member EZH2 is known to be important for stem cell maintenance in multiple tissues, but its role in mammary gland development and differentiation remains unknown. Our analyses show that EZH2 is predominantly expressed in luminal cells of the mouse mammary epithelium. As mammary gland development occurs mostly after birth, the analysis of EZH2 gene function in postnatal development is precluded by embryonic lethality of conventional EZH2 knockout mice. To investigate the role of EZH2 in normal mammary gland epithelium, we have generated novel transgenic mice that express doxycycline-regulatable short hairpin (sh) RNAs directed against Ezh2. Knockdown of EZH2 results in delayed outgrowth of the mammary epithelium during puberty, due to impaired terminal end bud formation and ductal elongation. Furthermore, our results demonstrate that EZH2 is required to maintain the luminal cell pool and may limit differentiation of luminal progenitors into CD61+ differentiated luminal cells, suggesting a role for EZH2 in mammary luminal cell fate determination. Consistent with this, EZH2 knockdown reduced lobuloalveolar expansion during pregnancy, suggesting EZH2 is required for the differentiation of luminal progenitors to alveolar cells.Stem Cells 2013;31:1910-1920


Introduction

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Mammary gland morphogenesis is coordinated by the expression of hormones and growth factors as well as their respective receptors, transcription factors, cell adhesion, and extracellular matrix proteins [1]. The net effect of these signals is the specification of mammary epithelium into a cellular hierarchy composed of multipotent stem cells, progenitor cells, and terminally differentiated cells [2]. The molecular programs involved in determining mammary epithelial cell fate remain poorly understood, but they are likely to involve genetic as well as epigenetic mechanisms.

Polycomb proteins (PcG) are evolutionarily conserved transcriptional regulators that function in establishing and maintaining epigenetic memory during development. Through their regulation of stem cell identity, PcG also provide an important link between stem cells and cancer [3]. Mammalian PcG form two complexes, polycomb repressive complexes 1 and 2 (PRC1 and PRC2), respectively. EZH2, a SET-domain containing methyltransferase, catalyzes the formation of a di- or trimethyl mark on lysine 27 of Histone H3 (H3K27), which is recognized and bound by PRC1, leading to transcriptional repression. Notably, many genes involved in development, stem cell maintenance, and differentiation are targets of H3K27 methylation [4]. In line with this, a role of EZH2 in the maintenance of adult stem cells and progenitor populations has been demonstrated through deletion of EZH2 in diverse cell types including keratinocytes [5], adipocytes [6], and neurons [7].

Little is known about the role of epigenetic regulation during normal mammary gland development [8]. However, evidence for a role of PcG in mammary development has been reported previously [9], and mice with mammary gland specific overexpression of EZH2 developed a multilayered, disorganized ductal epithelium, resulting from expansion of mammary epithelial cells, suggesting a role for EZH2 in mammary gland morphogenesis [10]. Unfortunately, conventional EZH2 knockout mice are embryonic lethal [11], precluding the study of EZH2 function in postnatal mammary gland development and function. Here, we describe the generation and characterization of novel inducible transgenic knockdown mouse strains that allow temporal control over EZH2 knockdown. We show that EZH2 knockdown in newborn mice delays mammary gland development in young virgin mice, due to delayed terminal end bud formation and impaired ductal elongation during puberty. We also demonstrate that EZH2 is required for maintaining the luminal cell pool and that EZH2 knockdown delays lobuloalveolar expansion during pregnancy. Taken together, our results support an important role for EZH2 in mammary luminal cell specification.

Materials and Methods

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Mice

Ezh2-tetKDA and Ezh2-tetKDB mice were generated by targeting constructs for doxycycline-regulatable expression of Ezh2-specific shRNAs into the Rosa26 locus of C57BL/6 embryonic stem cells (ESCs) [12]. Correct targeting was confirmed by Southern blotting for 6/6 ESC clones. To identify shRNA constructs with efficient knockdown, ESC clones were treated with doxycycline (1 μg/ml) for 72 hours. Knockdown of Ezh2 was not seen in ESCs expressing the control constructs “Stop” (containing five thymidines but no shRNA sequence) or “Luc” (encoding an shRNA directed against firefly luciferase). ESC clones expressing the best two Ezh2-specific hairpins (from six tested) showed >90% reduction in Ezh2 mRNA levels following doxycycline administration. These clones were used to generate two independent mouse founder lines, designated Ezh2-tetKDA and Ezh2-tetKDB. The Ezh2-tetKDA and Ezh2-tetKDB hairpins, targeting Ezh2 exon eight or exon 9, respectively, are encoded by the following 5′ to 3′ oligonucleotide sequences: GCAAAGCTTGCATTCATTTCATTCAAGAGATGAAATGAATGCAAGCTTTGC and GCAACACCCAACACATATAAGTTCAAGAGACTTATATGTGTTGGGTGTTGC. Control (StopControl) mice express all elements of the allele but do not express a hairpin. Genotyping was performed by polymerase chain reaction (PCR) with forward primer 5′-CCATGAATTCGAACGCTGACGTC-3′ and one of the following reverse primers: Control: 5′-TATGGGCTATGAACTAATGACCC-3′; Ezh2-tetKDA: 5′-GAATGAAATGAATGCAAGCTTTG-3′; Ezh2-tetKDB: 5′-GAACTTATATGTGTTGGGTGTTGC-3′.

Mice were kept in open cages, and food and water were given ad libitum. All animal experiments were performed in accordance with institutional guidelines and national regulations. Doxycycline hyclate (2 mg/ml, Sigma-Aldrich, Zwijndrecht, Netherlands, http://www.sigmaaldrich.com/nederland; supplemented with 1% sucrose) was administered in the drinking water in darkened bottles and refreshed twice weekly. Initial validation was done on a B6 background; however, fluorochrome-activated cell sorting (FACS) analyses were performed on mice crossed >4 generations to FVB. For some experiments, mice were injected with Bromodeoxyuridine (BrdU) (50 mg/kg) 90 minutes before collection of mammary glands.

Mammary Gland Analysis

Female mice were killed at the indicated ages for analysis. Thoracic (third) glands were used for RNA analysis. One inguinal (fourth) gland from each mouse was paraffin embedded, while the contralateral gland was used for whole-mount analysis as described previously [9]. The extent of ductal outgrowth was measured on inguinal wholemounts as the distance from the center of the lymph node to the leading edge of the ductal mass. Adult female mice were subjected to timed pregnancies, scored by the presence of vaginal plugs and confirmed by examination of embryos at the time of collection of mammary glands.

Tissue Culture

ESCs were cultured on gelatin-coated plates in feeder-free conditions using BRL-conditioned media containing LIF (ES Grow Millipore, Amsterdam, Netherlands, http://www.millipore.com/offices/cp3/nl; 103 units/ml). Primary mouse mammary epithelial cell (MMECs) cultures were isolated from both inguinal and/or thoracic mammary glands of FVB mice and prepared essentially as described [13]. For proliferation assays, FACS sorted cells were counted using trypan blue exclusion, and 100 live cells were seeded in duplicate wells of 24-well plates onto irradiated NIH-3T3 feeder cells. Cells were allowed to form colonies in growth medium containing 20 ng/ml cholera toxin and 5% fetal calf serum (FCS) overnight, which was changed to 1% FCS for an additional 5 days before staining with 0.1% crystal violet.

Quantitative Real Time PCR

For Taqman analyses of Ezh2 RNA in tissues of Ezh2-tetKD mice, total RNA was prepared using Qiagen RNA Plus mini kit. cDNA was synthesized from 1 μg total RNA using the Quantiscript Reverse Transcriptase Kit (Qiagen, Venlo, Netherlands, http://www.bionity.com/en/companies/15228/qiagen-n-v). Primers for murine Ezh2 were from the Assay-On-Demand Kit Mm00468449_m1 (Applied Biosystems, Bleiswijk, Netherlands, http://www.invitrogen.com/site/config/regional/RegionalContactUs/Netherlands). Quantitative real time PCR (qRT-PCR) reactions were performed using TaqMan Universal PCR Mastermix in an ABI PRISM 7900 Sequence Detection System (Applied Biosystems, Bleiswijk, Netherlands, http://www.invitrogen.com/site/config/regional/RegionalContactUs/Netherlands). Input cDNA concentrations were normalized to murine Hp1bp3. Alternatively, RNA from mammary tissue and MMECs was prepared using Trizol reagent (Invitrogen, Bleiswijk, Netherlands, http://www.invitrogen.com/site/config/regional/RegionalContactUs/Netherlands.html), and 1 μg of DNAse-treated (Promega, Leiden, Netherlands, http://nld.promega.com) RNA was reverse-transcribed using Oligo-dT primers (Invitrogen, Bleiswijk, Netherlands, http://www.invitrogen.com/site/config/regional/RegionalContactUs/Netherlands.html), according to manufacturer's instructions, before treatment with RNAse H (Promega, Leiden, Netherlands, http://nld.promega.com). Total RNA from sorted cells was transcribed using VILO (Invitrogen, Bleiswijk, Netherlands, http://www.invitrogen.com/site/config/regional/RegionalContactUs/Netherlands) and the resulting cDNA was analyzed using SYBR green (SYBR green I Master mix, Roche, Almere, Netherlands, http://www.roche.nl) on a Roche LightCycler (Roche Diagnostics, Almere, Netherlands, http://www.roche.nl). Input cDNA concentrations were normalized to HPRT, β-actin, or 18S. Product accumulation was evaluated using the comparative Ct method (2−ΔΔCT). Primer sequences can be found in the Supporting Information.

Immunoblotting

Whole cell extracts were prepared in RIPA buffer (50 mM Tris pH 8.0, 50 mM NaCl, 1.0% NP40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitor cocktail (Complete, Roche, Almere, Netherlands, http://www.roche.nl/portal/eipf/nl/netherlands_portal/roche.nl/xxxcontactxxx). Equal amounts of protein, as determined by Micro BCA Protein Assay Kit (Pierce, Schwerte, Germany, http://www.de.fishersci.com), were resolved on NuPage-Novex 4–12% Bis-Tris gels (Invitrogen, Almere, Netherlands, http://www.roche.nl/portal/eipf/nl/netherlands_portal/roche.nl/xxxcontactxxx) and transferred onto nitrocellulose membranes (0.2 μm, Whatman, AH's-hertogenbosch, Netherlands, http://www.whatman.com). Membranes were blocked in phosphate-buffered saline (PBS) 0.1% Tween-20 (PBST) 5% bovine serum albumin (BSA) for 1 hour, incubated with primary antibodies in PBST 1% BSA overnight at 4°C, and with secondary antibodies coupled to horse-radish peroxide (HRP) for 45 minutes in PBST 1% BSA. Bands were visualized using enhanced chemiluminescence western blotting detection reagent (GE Healthcare, Eindhoven, Netherlands, http://www3.gehealthcare.nl). Primary antibodies are listed in the Supporting Information. Secondary antibodies included HRP-conjugated anti-rabbit and anti-mouse (Invitrogen, Almere, Netherlands, http://www.roche.nl) at 1:10,000.

Immunostaining

For histological analysis, 2 or 5 μm sections of formalin-fixed inguinal mammary glands were stained with hematoxylin–eosin or used for immunolocalization studies, respectively. Immunostaining was performed after microwave antigen retrieval (20 minutes) in 10 mM sodium citrate and blocking in 5% normal goat serum in PBS. Primary antibodies are listed in the Supporting Information. Secondary antibodies included biotinylated goat anti-rabbit IgG (1:500, DAKO, Heverlee, Belgium, http://www.dako.com/nl), goat anti-mouse IgG (1:500, DAKO, Heverlee, Belgium, http://www.dako.com/nl), goat anti-rabbit Alexa 568 (1:250, Invitrogen), and goat anti-mouse Alexa 488 (1:200, Invitrogen, Almere, Netherlands, http://www.roche.nl). Secondary-only antibody controls were included for all immunostains and no nonspecific staining was observed in epithelial cells. Nuclei were visualized using 4″,6-diamidino-2-phenylindole (DAPI) (2 μg/ml, Invitrogen, Almere, Netherlands, http://www.roche.nl). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining was performed using the ApopTag kit (Chemicon, Darmstadt, Germany, http://www.merck.nl). Immunohistochemistry images were captured with an Olympus UC30 camera mounted on a Zeiss Axioskop 40 microscope, using Olympus Cell* Imaging Software. Immunofluorescence images were captured with a Hamamatsu ORCA AG CCD camera on a Zeiss Axio Observer Z1 system, using Axio Vision 4.7 software.

Mammary Cell Preparation

Thoracic and inguinal mammary glands were removed from wild type (Wt) and Ezh2-tetKD females. The lymph node was removed and single cell suspensions were obtained essentially as described by the Visvader/Lindeman group [14] with the following modifications: mammary glands were digested in collagenase/hyaluronidase solution (300 and 100 U/ml, respectively; StemCell Technologies, Vancouver, British Columbia, http://www.stemcell.com/) for up to 90 minutes and treated with dispase (5 mg/ml; StemCell Technologies, Vancouver, British Columbia, http://www.stemcell.com/). Before antibody staining, total cell counts were determined by manual counting using trypan blue exclusion for live cells. Between 3 and 9 mice per genotype were pooled per experiment.

Cell Staining for Flow Cytometry

Blocking was performed in PBS/2% FCS plus 2.4G2 (anti-Fcγ receptor mAb supernatant, prepared in house). Incubations with antibodies listed in the Supporting Information were performed for 30 minutes on ice. DAPI was added to cells at a final concentration of 0.15 μg/ml before analysis. Cell doublets, Lin+ cells (CD45+, TER119+, and CD31+), and dead cells (DAPI+) were discarded. All analyses and sorts were performed using a Cyan analyzer (Beckman Coulter, Woerden, Netherlands, https://www.beckmancoulter.com; with Summit 4.3.01 software) or FACSAria (Becton Dickinson, Breda, Netherlands, http://www.bd.com/nl; nozzle tip diameter 70 μm; 40psi; 30kHz; Diva 6.1.2 software) respectively, and analyzed using FlowJo v.8.8.6 software. To ensure compatibility of the experiments, single color controls and fluorescence minus one (FMO) were carried out and the compensation was calculated by the software.

Statistical Analysis

Statistical analysis was performed using an unpaired t-test (two-tailed, assuming unequal variance) using GraphPad Prism Version 4 software. The p ≤ 0.05 were considered to indicate statistically significant differences.

Results

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Generation and Validation of Mice with Inducible RNAi-Mediated Knockdown of EZH2 Expression

To circumvent the embryonic lethality associated with EZH2 deletion, we generated mice in which EZH2 expression could be downregulated in an inducible manner [12]. The Rosa26 locus of mouse ESCs was targeted with a bipartite construct encoding both an Ezh2-specific shRNA driven by the tet-regulatable H1 promoter and a codon-optimized tet repressor (itetR) [15] driven by the CAGGS promoter (Fig. 1A). Several nonoverlapping shRNAs were designed and knockdown of Ezh2 was assessed in ESCs (Fig. 1B and data not shown). Knockdown of EZH2 in ESCs was achieved within 5 days of doxycycline treatment and could be completely reversed within 4 days of withdrawal of doxycycline from the medium (Fig. 1B). Expression of H3K27me3 was diminished upon EZH2 knockdown, and reexpression coincided with reversal of EZH2 knockdown. Two independent mouse lines were established from the ESC lines encoding the shRNAs that achieved the best knockdown of Ezh2. These lines, which were designated Ezh2-tetKDA and Ezh2-tetKDB, could be distinguished by a genotyping PCR (Supporting information Fig. S1A) and animals bearing the targeted allele were maintained as heterozygotes.

image

Figure 1. Conditional RNAi-mediated knockdown of EZH2 in mice. (A): Strategy of tetR-driven repression of shRNA expression at the Rosa26 locus. H1 promoter driven expression of shRNAs targeting Ezh2 occurs upon administration of doxycycline. The sequences of both nonoverlapping hairpins, expressed by the two independent founder lines Ezh2-tetKDA and tetKDB, are shown. (B): Western blot analysis of EZH2 and H3K27me3 expression in Ezh2-tetKDA embryonic stem cells treated with (+) or without (−) 100 ng/ml doxycycline for the indicated number of days, starting at day 0 (D0). On day 5 (D5) of treatment, doxycycline was maintained (+) or removed (−) from the medium where indicated, to show reversibility of EZH2 and H3K27me3 knockdown. Ponceau S staining is included as a control for protein loading. (C): Taqman analysis of inducible Ezh2 mRNA knockdown in vivo in multiple tissues of 8–12 week old mice Control, Ezh2-tetKDA (TetKDA) or Ezh2-tetKDB (TetKDB) mice (n = 3–4 per group), treated for 10 days with doxycycline plus sucrose (+) or sucrose alone (−) via the drinking water. The average percent remaining Ezh2 transcript levels ± SEM is shown, normalized to Hp1bp3. Each bar represents an average expression value of one mouse cohort normalized to the −Dox control of each tissue. (D): Immunohistochemical analysis of EZH2 expression (strong nuclear staining indicated by arrowheads) in mammary gland sections from Wt and heterozygous (Het) littermate Ezh2-tetKD mice. Representative fields are shown. Mice were given doxycycline for 2 or 5 months for the Ezh2-tetKDA or Ezh2-tetKDB lines, respectively. Scale bars = 50 μm. (E): Knockdown of Ezh2 upon doxycycline treatment of MMECs from founder line A and B. In the three independent experiments shown, MMECs were treated for 5 days in vitro with the indicated concentrations of doxycycline. quantitative real time polymerase chain reaction was performed and Ezh2 mRNA levels are shown relative to untreated cells, normalized to the house-keeping gene Hprt. Abbreviations: Dox, doxycycline; Het, heterozygotes; M. gland, mammary gland; Wt, wild type.

Download figure to PowerPoint

As a proof of principle, we tested whether global knockdown of EZH2 expression was sufficient to recapitulate the embryonic lethality observed in conventional Ezh2 knockout embryos [11]. C57BL6/J or FVB dams were mated to two independent Ezh2-tetKDA or three independent Ezh2-tetKDB males and maintained on doxycycline throughout pregnancy. Pups were collected on the day of birth and genotyped for the presence of the tetKD allele. No transgenic animals were detected in the litters from either line (Supporting Information Fig. S1B), demonstrating that Ezh2-tetKD pups did not survive until birth.

Next, we validated the efficiency of EZH2 knockdown in adult mice following 10 days of doxycycline treatment (Fig. 1C). In control mice, Ezh2 transcript levels were not affected by the presence or absence of doxycycline. However, in doxycycline-treated Ezh2-tetKDA and Ezh2-tetKDB, mice Ezh2 mRNA levels were reduced by 60–90% in all tissues tested, including the mammary gland. The only exception to this was the brain in Ezh2-tetKDA animals, where only 20% knockdown was achieved. Even long-term treatment of Ezh2-tetKD mice with doxycycline (5 months) was well tolerated (Supporting Information Fig. S1C) and resulted in sustained knockdown of EZH2 (Supporting Information Fig. S1D and data not shown). Upon macroscopic and histopathologic analysis, all tissues in these mice appeared normal.

In the adult mammary gland, Ezh2 mRNA levels were reduced by approximately 70% in Ezh2-tetKD mice compared with nontreated transgenic mice following 10 days of doxycycline treatment (Fig. 1C). Moreover, EZH2 expression in mammary glands could easily be detected in Wt littermates (Fig. 1D), but was reduced in doxycycline treated Ezh2-tetKD mice. Finally, in vitro doxycycline treatment induced robust Ezh2 knockdown in MMECs derived from Ezh2-tetKDA and Ezh2-tetKDB mice (Fig. 1E).

Thus, we conclude that we have developed a novel mouse model that allows in vivo manipulation of EZH2 expression in multiple tissues. Our preliminary analysis demonstrates that knockdown of EZH2 results in embryonic lethality but is well tolerated in adult mice, even for longer periods.

Knockdown of EZH2 Results in Defective Outgrowth of the Prepubertal Mammary Gland

Next, we wanted to test the effect of knockdown of EZH2 on mammary gland development. Expression of EZH2 was high in MMECs from adult FVB virgin mice and further increased during pregnancy, corresponding with high levels of H3K27me3 (Fig. 2A). The expression pattern of EZH2 overlapped with expression of the PRC1 member BMI1, which is required for mammary gland development during puberty, but not for alveolar differentiation during pregnancy [9].

image

Figure 2. Knockdown of EZH2 delays outgrowth of the prepubertal mammary gland. (A): Western blot analysis of EZH2 and H3K27me3 expression in MMECs derived from Wt virgin mice or mice at the indicated stages of pregnancy (L = lactation; I = involution). A nonspecific band running below H3K27me3 is indicated with an asterisk. β-ACTIN is shown as a control for loading. Expression of BMI1 is included as a positive control. Samples are representative of two independent sets of MMEC. (B): Whole-mount analysis of Wt and Ezh2-tetKD mammary glands from 4-week-old mice, treated with (+DOX) or without (−DOX) doxycycline from birth. The lymph node is visible as a dark mass. A higher magnification of the terminal end bud structures is shown in the right-hand panels. The direction of growth is indicated. Scale bars = 1 mm and 0.25 mm for left and right panels, respectively. (C): Quantitative real time polymerase chain reaction analysis of Ezh2 knockdown in mammary glands of 4-week-old Wt (white bars) and Ezh2-tetKD (black bars) females of both founder lines, treated with (+DOX) or without doxycycline (−DOX) from birth. Results represent mean Ezh2 transcript levels ± SEM, normalized to Hprt. Results are from at least two independent experiments and represent the mean ± SEM of 4–7 mice for each condition for each group. *p < .02. (D): Ductal elongation in inguinal mammary glands from 4-week-old Wt (white bars) and Ezh2-tetKD (black bars) females of both founder lines, treated with (+DOX) or without doxycycline (−DOX) from birth. Results represent the mean ± SEM of 4–7 glands from 3 to 7 independent mice for each condition for each group. **p < .005. (E): Immunohistochemical analysis of EZH2 expression (brown nuclear staining) in representative TEBs and ducts in mammary gland sections from 4-week-old Wt and Ezh2-tetKD mice treated with doxycycline from birth. Scale bars = 50 μm and 100 μm in TEBs and ducts, respectively. Numbers of ducts with fewer (F) or greater (G) than 20% BrdU-positive cells were counted in a mammary gland section from 4-week-old Wt and Ezh2-tetKD mice (n = 4 per group) treated with doxycycline from birth. Each dot represents an individual duct. All mammary gland sections from Wt mice contained ducts with >20% BrdU-positive cells, while these were only found in two of 4 Ezh2-tetKD sections. Data in (G) represent 11, 11, 2, 3 and 3, 0, 0, 2 ducts observed in four independent Wt and Ezh2-tetKD mice, respectively. Abbreviations: BrdU, bromodeoxyuridine; Dox, doxycycline; wk, week; TEB, terminal end bud; Wt, wild type.

Download figure to PowerPoint

To determine whether EZH2 is also required for mammary gland development during puberty, we studied the effect of Ezh2 knockdown in female mice that were treated with doxycycline continuously from birth. Whole-mount analysis of inguinal mammary glands revealed that, despite the presence of a normal rudimentary mammary gland structure (anlagen) at 2 weeks of age (Supporting Information Fig. S2A), mammary gland outgrowth was delayed in both Ezh2-tetKD founder lines compared with Wt littermates at 4 weeks of age (Fig. 2B, lower two panels). The delay in mammary gland outgrowth was still evident in 6-week-old mice (Supporting Information Fig. S2A and S2B), but appeared to catch up and was more or less comparable with that observed in Wt mice by 9 weeks of age (Supporting Information Fig. S2A and S2B). Despite initially robust Ezh2 knockdown in 4-week-old mice (75% in Ezh2-tetKDA and 67% in Ezh2-tetKDB; Fig. 2C), Ezh2 mRNA levels were reduced by only 20% in 6-week-old mice and comparable with Wt levels by 9 weeks of age (Supporting Information Fig. S2C). This indicates selection against cells with diminished EZH2 expression, suggesting that these cells were either eliminated or that silencing of the tetKD allele was occurring. These results indicate that, whereas EZH2 knockdown could be sustained for more than 5 months in mammary glands of adult mice, it is lost in mammary glands of young mice.

Unlike other tissues in the body, mammary gland development mostly occurs after birth [16, 17], raising the possibility that this tissue is more sensitive to EZH2 loss. Indeed, we found that mammary gland outgrowth was inhibited more severely than other tissues. EZH2 knockdown mice were 19% (Ezh2-tetKDA) to 54% (Ezh2-tetKDB) smaller than their Wt littermates (Supporting Information Fig. S3A). Consistent with an overall smaller body size, inguinal lymph node length in Ezh2-tetKDA and Ezh2-tetKDB mice was also reduced by 14% and 32%, respectively (Supporting Information Fig. S3B). In comparison, ductal elongation showed a 75% reduction in both transgenic lines compared with Wt littermates (Fig. 2D, p < .005), strongly arguing for the specific requirement of EZH2 for mammary gland outgrowth.

Despite the reduced ductal elongation, the mammary epithelium present in 4-week-old Ezh2-tetKD mice retained normal architecture and expressed markers of myoepithelial (smooth muscle actin) and luminal cells (E-cadherin), as well as estrogen receptor (Supporting Information Fig. S4A). However, in addition to failing to invade the surrounding stroma, the vast majority of distal end structures in 4-week-old Ezh2-tetKD mice did not have the characteristic appearance of terminal end buds (TEBs) (Fig. 2B). These structures represent the most proliferative compartment of the pubertal mammary epithelium and form the mammary gland by invading the surrounding stroma [18]. Given the absence of normal TEB structures, we hypothesized that EZH2 is required for TEB formation or maintenance. Indeed, we were able to detect prominent EZH2 expression in both luminal and basal layers as well as in the cap cells of TEBs from 4-week-old Wt mice (Fig. 2E). In contrast, EZH2 expression was low in proximal ducts close to the nipple, where it was found in scattered luminal and basal cells throughout the ducts. As expected, EZH2 expression was markedly decreased in ducts and distal end structures from doxycycline-treated Ezh2-tetKD mice (Fig. 2E).

To determine whether the delay in ductal outgrowth was due to a difference in cell proliferation, we measured BrdU incorporation in 4-week-old Ezh2-tetKD mice. We did not observe a significant difference in overall BrdU incorporation in Ezh2-tetKD mice compared with Wt mice (Fig. 2F). However, ducts in which BrdU incorporation was 20% or higher (representing the highly proliferating TEBs) [19] were completely absent or present in small numbers in sections from mammary glands from Ezh2-tetKD mice (Fig. 2G). The fact that we observed a significant reduction in apoptotic cells in mammary glands from 4-week-old Ezh2-tetKD mice, supports the notion that reduced elongation is the result of impaired TEB cell proliferation rather than TEB collapse (Supporting Information Fig. S4B).

These findings suggest that a reduction in Ezh2 from birth onwards delays mammary gland outgrowth, most likely due to impaired TEB proliferation. This is consistent with a role for EZH2 in mammary gland morphogenesis.

EZH2 Maintains the Luminal Progenitor Pool in the Prepubertal Mammary Gland

To further characterize the mammary gland defects associated with EZH2 knockdown in prepubertal mice, we examined mammary epithelial cell preparations from the mammary glands of 4-week-old Ezh2-tetKD mice and their Wt littermates, using FACS. The cell surface markers CD24 and CD29 can resolve two distinct mammary epithelial populations enriched for basal/myoepithelial (CD24+CD29hi) or luminal (CD24+CD29lo) cells [20]. The luminal cell population can be further subdivided into CD61+ luminal progenitor and CD61 mature luminal cells [21] (Fig. 3A).

image

Figure 3. EZH2 maintains the luminal progenitor pool. (A): Representative fluorochrome-activated cell sorting dot plots showing the percentage of the indicated populations in the Lin fraction (top) and in the LinCD24+ fraction (bottom) of mammary glands from 4-week-old Wt and Ezh2-tetKD mice. (B): Histogram showing the percentage of the indicated Lin cell populations in mammary glands from 4-week-old Wt and Ezh2-tetKD mice. Data represent mean ± SEM. Between 3–7 mice, treated with doxycycline from birth, were pooled per genotype per experiment, in at least five independent experiments. (C): Quantitative real time polymerase chain reaction analysis of Ezh2 in mammary glands from 4-week-old Wt and Ezh2-tetKD mice treated with doxycycline from birth. Gata3, SMA, and PDGFRβ transcripts are included as controls for purity of sorted luminal, basal and stromal populations, respectively. Mean transcript levels ± SEM are shown, normalized to 18S, from at least three independent experiments. Abbreviation: Wt, wild type.

Download figure to PowerPoint

As expected from the reduced ductal outgrowth observed in doxycycline-treated Ezh2-tetKD mice, somewhat fewer lineage-negative (LIN) cells were recovered from Ezh2-tetKD mice than from Wt littermates (Supporting Information Fig. S5A). However, the percentage of LIN cells was not different between genotypes (Fig. 3B), nor was there any difference between genotypes with respect to numbers of CD24CD29lo stromal cells or the CD24+CD29hi basal-enriched fraction within the LIN population (Fig. 3A, 3B; Supporting Information Fig. S5A). In contrast, there was a twofold reduction (p < .01) in the percentage of LINCD24+CD29lo luminal cells recovered from doxycycline-treated Ezh2-tetKD mice compared with Wt littermates with an accompanying increased proportion of CD29loCD24 stromal cells (Fig. 3A, 3B; Supporting Information Fig. S5A). Indeed, EZH2 knockdown had a stronger negative impact on the number of CD24+CD29lo luminal cells than on the other LIN populations (Supporting Information Fig. S5A). Compared with Wt littermates, mammary glands from doxycycline-treated Ezh2-tetKD mice showed a strong reduction in the number of LINCD24+CD29loCD61+ luminal progenitors but only a modest reduction in the number of LINCD24+CD29loCD61 differentiated luminal cells (Supporting Information Fig. S5B). Consequently, the CD61+ luminal progenitor-enriched fraction was reduced approximately 1.5-fold in Ezh2-tetKD mammary glands relative to Wt littermates (p < .01) (Fig. 3B), with a concomitant increase in the percentage of CD61 differentiated luminal cells (p < .05). In the mammary glands of untreated Ezh2-tetKD mice, the total number (Supporting Information Fig. S5C) as well as the percentage of all LIN populations (Supporting Information Fig. S5D) were comparable with those recovered from Wt mammary glands.

These findings suggest that a reduction in Ezh2 from birth onwards leads to a net loss of luminal cells coupled with enhanced differentiation to maintain the pool of differentiated luminal cells. This is consistent with a normal role for EZH2 in restricting differentiation of luminal progenitors to more mature cells. Indeed, Ezh2 transcript levels were highest in sorted luminal progenitor cells compared with stromal cells or other stem/progenitor populations in 4-week-old Wt mice (Fig. 3C). As expected, Ezh2 mRNA was reduced in all populations sorted from doxycycline-treated Ezh2-tetKD mice (Fig. 3C) but, surprisingly, not in the mature luminal cell fraction. Sorted cell populations from untreated Ezh2-tetKD mice expressed levels of Ezh2 that were comparable with Wt (Supporting Information Fig. S6A). As the transcription factor GATA3 is known to be important for mammary gland development and in particular luminal cell differentiation [21, 22], we tested whether the effects of EZH2 knockdown on the luminal cell pool were related to changes in Gata3 expression. However, Gata3 expression in mammary epithelial populations sorted from doxycycline-treated Ezh2-tetKD mice was comparable with Wt littermates (Fig. 3C). Partial compensation of EZH2 function by the methyltransferase EZH1 has been reported in ESCs [23], and in hematopoietic cells [24]. However we did not find evidence of compensation by EZH1 upon EZH2 knockdown in sorted mammary epithelial cell populations (Supporting Information Fig. S6B). Despite the reduced numbers of luminal cells in mammary glands from doxycycline-treated Ezh2-tetKD mice, EZH2 knockdown had no effect on sorted luminal progenitors in in vitro colony formation assays (Supporting Information Fig. S6C). This could suggest that EZH2 does not affect luminal differentiation per se, but it is more likely that sufficient luminal progenitors with incomplete EZH2 knockdown are present to form colonies in vitro.

We conclude from these results that EZH2 is important for maintenance of the luminal progenitor cell pool and may restrict differentiation of luminal progenitors to more mature cells in a manner that this is independent of GATA3.

Knockdown of EZH2 Results in Defective Lobuloalveolar Differentiation During Pregnancy

We next determined whether EZH2 knockdown affects lobuloalveolar differentiation of mammary glands during pregnancy. To this end, female Ezh2-tetKD mice and Wt littermates were maintained on doxycycline from birth until breeding age (9 weeks or older), and mammary glands were analyzed following timed matings.

In line with the high levels of EZH2 in MMECs from mid-pregnant glands (Fig. 2A), our analysis suggests that EZH2 is required for normal lobuloalveolar development. Reduced lobuloalveolar development was observed by mid-pregnancy (14.5dP), and persisted to late pregnancy and parturition (1.5 dL) (Fig. 4A and data not shown). Despite the persistence of Ezh2 transcript levels in doxycycline-treated Ezh2-tetKD mammary tissue during mid-pregnancy (Fig. 4B), Ezh2 expression was significantly reduced at parturition. At this time, glands from Wt littermates contained well-differentiated lobuloalveolar units consisting of fully expanded alveoli filled with milk (Fig. 4C). In contrast, the Ezh2-tetKD mammary tissue contained large areas of adipocytes and a low density of lobuloalveolar units. The lumens of alveoli in Ezh2-tetKD glands appeared less dilated, and some were closed, reminiscent of immature alveoli of Wt mice. Notwithstanding these morphological abnormalities, the majority of alveolar lumina in glands from Ezh2-tetKD mice were filled with milk, as evident from the presence of fat droplets in the histological analysis (Fig. 4C, right panels). Moreover, pups born from doxycycline-treated Ezh2-tetKD females had visible milk spots.

image

Figure 4. EZH2 is required for lobuloalveolar expansion during pregnancy. (A): Whole-mount analysis of mid-pregnant and lactating Wt and Ezh2-tetKD mammary glands from females treated with doxycycline from birth (P = pregnancy; L = lactation). Scale bars = 1 mm. All mice were 9 weeks or older at the time a plug was detected. (B): Quantitative real time polymerase chain reaction analysis of Ezh2 transcript levels in mid-pregnant and lactating mammary glands of Wt (n = 3) and Ezh2-tetKD (n = 3–4) females treated with doxycycline from birth. Values represent the mean ± SEM, normalized to Hprt, from at least two independent experiments. *p < .03. (C): Representative hematoxylin–eosin sections of lactating mammary glands from doxycycline-treated Wt and Ezh2-tetKD mice at low (left panel) and high (right panel) magnification showing normal milk production despite reduced lobuloalveolar expansion in Ezh2-tetKD mice. Scale bars are indicated. (D): Immunofluorescence staining of NKCC1 and Npt2b transporters in mammary glands from virgin and 1.5 dL doxycycline-treated Wt and Ezh2-tetKD mice. Luminal cells of adult virgin ducts show clear membrane staining of the NKCC1 transporter in both Wt and Ezh2-tetKD mice, and only occasional cells express NKCC1 in Wt lactating glands (white arrowheads). Lactating glands from Ezh2-tetKD mice had diffuse NKCC1 expression (white arrowheads). Conversely, the Npt2b transporter is not expressed in virgin glands, but becomes expressed at the apical surface of the lumen of alveoli during lactation. Scale bars = 40 μm. (E): Model of the differentiation hierarchy within mammary epithelium. Our results suggest a model in which EZH2 maintains the luminal progenitor pool during mammary gland development and promotes differentiation to an alveolar cell type during pregnancy. Abbreviations: DAPI, 4″,6-diamidino-2-phenylindole; Dox, doxycycline; Wt, wild type.

Download figure to PowerPoint

To further characterize the morphological phenotype in Ezh2-tetKD mammary tissue, we examined the expression of specific markers for the discrimination of ductal and alveolar epithelial cells. The Na-K-Cl cotransporter one (NKCC1) is present at high levels in the ductal epithelial cells of virgin mammary gland and is diminished during pregnancy and at parturition [25]. NKCC1 expression was detected in adult virgin glands of doxycycline-treated Ezh2-tetKD mice and Wt littermates (Fig. 4D). At parturition however, EZH2 knockdown resulted in more widespread, persistent expression of NKCC1 in luminal cells, suggesting incomplete differentiation of these cells to an alveolar cell type. This was in contrast to Wt tissues, where only occasional cells expressed NKCC1 (Fig. 4D).

In contrast with NKCC1, the Na-Pi type IIb cotransporter (Npt2b) is not detectable in virgin tissue but becomes expressed at the apical membrane from day 15 of pregnancy and remains present during lactation [25]. As expected, the apical membranes of secretory alveoli in Wt mice strongly expressed Npt2b (Fig. 4D). Consistent with the milk production we observed, the majority of alveoli in Ezh2-tetKD glands also expressed Npt2b, indicative of normal secretory function [26]. Nevertheless, the reduced density of lobuloalveolar units in lactating glands from Ezh2-tetKD mice treated with doxycycline suggested the partial failure of ductal epithelial cells to undergo correct differentiation into secretory alveolar cells. This effect appeared to be independent of GATA3 and ELF5, the transcription factors known to be required for alveolar differentiation [21, 22, 27, 28], as transcript levels of both Gata3 and Elf5 in Ezh2-tetKD glands were comparable with those from Wt littermates (Supporting Information Fig. S7A). Consistent with normal expression of Elf5, mRNA levels of milk-specific proteins α- or β-casein, or WAP were unaltered (Supporting Information Fig. S7B). Also transcript levels of the signal transducer and activator of transcription (STAT) factors Stat5a, Stat5b, and Stat3 were unchanged in lactating mammary glands from doxycycline-treated Ezh2-tetKD mice compared with Wt (Supporting Information Fig. S7C). No differences were observed in the percentage of Ki67-positive (Supporting Information Fig. S7D) or TUNEL-positive (Supporting Information Supporting Information Fig. S7E) epithelial cells in lactating glands of doxycycline-treated Ezh2-tetKD mice and Wt littermates, suggesting the reduced density of lobuloalveolar units did not result from reduced proliferation or increased apoptosis.

These results suggest that EZH2 is required specifically during pregnancy-induced differentiation of the mammary epithelium and that this may be due to reduced differentiation of luminal progenitor cells into alveolar cells.

Discussion

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

The analysis of EZH2 gene function in postnatal development was hitherto precluded by embryonic lethality of conventional EZH2 knockout mice. To circumvent this, we generated novel transgenic mouse lines expressing doxycycline-inducible hairpins targeting Ezh2 and used these mice to investigate whether EZH2 is required for normal development of the mammary gland. As the use of RNAi technology can result in off-target effects [29], we used two independent founders lines for our studies, expressing independent hairpins targeting Ezh2. There may of course be threshold effects associated with residual EZH2 expression in Ezh2-tetKD mice, and genetic inactivation of EZH2 in the mammary gland might have a stronger phenotype. Indeed, while this study was under review, Pal et al. reported stronger phenotypes associated with MMTV-cre-mediated deletion of Ezh2 in the mammary gland [30]. However, these stronger phenotypes might also result in part from cre-associated toxicity [31] and/or from lactation defects seen in certain MMTV-cre strains [32].

While we did not observe changes in the mammary gland upon acute knockdown of EZH2 in adult virgin mice, continuous EZH2 knockdown from birth resulted in defective mammary gland outgrowth during puberty. The different requirement of EZH2 in developing versus adult mammary glands in virgin mice implies EZH2 activity is not required in the absence of a growth stimulus. This is in line with the role we observe for EZH2 during pregnancy. Consistent with the hypercellular and supernumerary TEBs in mammary glands with EZH2 overexpression [10], knockdown of EZH2 resulted in fewer TEBs and delayed ductal elongation. This suggests that EZH2 knockdown disrupts the formation or maintenance of the TEBs. In contrast to Li et al. who reported increased proliferation in EZH2 overexpressing epithelium [10], we could not detect an overall decrease in proliferation in Ezh2-tetKD mammary glands. It is nevertheless conceivable that small differences in proliferation rate could result in the net effect of reduced ductal elongation.

In line with reduced ductal outgrowth, Ezh2-tetKD mammary glands contained fewer luminal progenitors. Additionally, EZH2 knockdown had a stronger negative impact on CD61+ luminal progenitors than on CD61 luminal cells, representing differentiated luminal cells [21]. The absence of EZH2 knockdown in this latter population suggests that there is particularly strong selection in these cells against reduced EZH2 expression, indicating that EZH2 is important for their maintenance. However, we cannot rule out trivial factors such poor expression of the hairpin in this cell type.

Although the molecular pathways important for maintenance and differentiation of mammary epithelial cells remain unclear, the transcription factor GATA3 is known to be important for luminal cell differentiation [21, 22] as Gata3 deletion blocks luminal differentiation. We did not observe a difference in Gata3 expression in EZH2 knockdown cells however, suggesting that the maintenance of differentiated luminal cell numbers in Ezh2-tetKD mice is not due to derepression of GATA3. This is in contrast to Li et al. who observed an upregulation of GATA3 in EZH2-overexpressing mammary epithelium [10]. However, as those experiments were not performed in sorted cells, our results would predict that the observed GATA3 upregulation results from increased numbers of luminal cells (which express high levels of GATA3) present in the mammary glands of EZH2 overexpressing mice. Indeed, we observed significantly reduced levels of Gata3 expression in unsorted mammary epithelium from doxycycline-treated Ezh2-tetKD mice (Supporting Information Fig. S8B).

How does knockdown of EZH2 elicit the phenotype we observe in the mammary gland? From birth until puberty, the mammary gland is a rudimentary and largely dormant structure. Around 3 weeks however, the mammary gland develops rapidly in response to changes in circulating hormone levels and TEBs appear. The number of mouse models exhibiting pubertal phenotypes of impaired ductal elongation and TEB formation is extensive and results from perturbation of many genes important for normal morphogenesis [1]. We speculate that the mammary gland phenotypes we observe upon EZH2 knockdown are likely due to a reduction in H3K27me3 marks, resulting in derepression of silenced PRC2 target genes. It should however be noted that recent findings in multiple cell types suggest EZH2 inactivation can be partly compensated for by EZH1 [23, 24, 33], which can catalyze trimethylation of H3K27, albeit weakly [34]. We did not find evidence of compensation by EZH1, at least not at the transcriptional level. In this respect, the analysis of mice deficient for EED, which is required not only for trimethylation but also for mono- and dimethylation of H3K27 [35], will likely solve the issue of redundancy between EZH2 and EZH1.

Despite the initial effect of EZH2 knockdown on mammary gland development, mammary gland outgrowth continued and indeed caught up in older virgin mice. In line with this, the mammary glands from 4-week-old Ezh2-tetKD animals retained sufficient numbers of luminal progenitors to produce colony assays in vitro, an assay used to assess stem/progenitor potential [36]. Presumably also in vivo, these luminal progenitors contribute to the eventual outgrowth of the mammary gland in older Ezh2-tetKD animals. Mammary gland outgrowth coincided with a rebound of Ezh2 mRNA levels despite continuous doxycycline treatment, strongly suggesting selection against knockdown or compensatory increase in transcription of Ezh2 by an unknown mechanism. In support of this notion, selection for retention of a functional Gata3 allele was also observed in conditional Gata3 knockout mice [22].

Our results indicate that EZH2 levels become limiting again during pregnancy, resulting in reduced lobuloalveolar development. Presumably the rapid differentiation of the mammary gland in response to hormonal signals during pregnancy does not allow sufficient time for selection to take place, resulting in EZH2 insufficiency. It seems likely however, that selection against Ezh2 knockdown also occurs during pregnancy-induced differentiation, as we could only achieve a 50% reduction in Ezh2 expression. The reduced lobuloalveolar density we observe may be due to aberrant specification of luminal progenitors to alveolar cells upon reduced EZH2 expression. While we did not observe an obvious reduction in side branching in adult Ezh2-tetKD mice, we cannot rule out the possibility that reduced lobuloalveolar density arises from the presence of fewer ducts with the capacity to form alveolar units. Interestingly, the effects of EZH2 knockdown are the inverse of those seen for deletion of the PRC2 complex member BMI1, which causes premature lobuloalveolar differentiation [9], suggesting that EZH2 and BMI1 have opposing roles during pregnancy-induced differentiation of luminal cells. Knockdown of EZH2 during puberty did not significantly alter numbers of differentiated luminal cells, despite fewer progenitors, suggesting that EZH2 knockdown may promote differentiation toward ductal epithelium. We therefore speculate that EZH2 knockdown may also be required for maintaining the alveolar lineage (Fig. 4E). It is conceivable that EZH2 knockdown results in enhanced differentiation towards ductal epithelial cells at the cost of alveolar cells, causing a block in lobuloalveolar development later on. Notably, the impaired lobuloalveloar differentiation in EZH2 knockdown glands does not appear to be due to decreased expression of the known mediator of luminal cell differentiation Elf5 [27, 28] or Gata3 [21, 22], suggesting that yet other downstream targets are important for this process.

EZH2 is often found overexpressed in human breast cancer, and particularly in poorly differentiated tumors that are thought to arise from luminal progenitors [37]. Consistent with a role for EZH2 in these cells, our results show that EZH2 is required for the maintenance of the luminal cell pool and luminal progenitor cell lineage. It will be interesting to determine whether in human mammary epithelium EZH2 plays a similar role in luminal cells. Together with the observation that overexpression of EZH2 in the mouse mammary gland results in expansion of mammary epithelial cells but not in mammary tumor formation [10], our results suggest that EZH2 is important for maintenance of the luminal cell pool, but that collaboration with additional somatic mutations is required for tumor onset.

Conclusion

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Our data reveal a key role for EZH2 in the maintenance of the luminal cell pool and luminal progenitor cell lineage. Further experiments are required to determine whether these phenotypes are due to derepression of silenced PRC2 target genes.

Acknowledgments

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

We thank Jim Turner and Jürg Biber for their generous gifts of antibodies against NKCC1 and Npt2b, respectively; Anita Pfauth and Frank van Diepen for cell sorting; The NKI animal facility for animal care; the animal pathology department for histology; Lauran Oomen and Lenny Brocks for assistance with digital microscopy; Sandra Büchel and Lisa Antoni for Taqman analysis; and Drs. Wilbert Zwart and Renee van Amerongen for critically reading the manuscript. This work was supported by the European Commission project EuroSyStem (200270, the TI Centre for Translational Molecular Medicine (CTMM) BreastCare project, and the Cancer Genomics Centre Netherlands). E.M.M. was supported by Marie Curie Fellowship MC 237486 and NHMRC Training fellowship 575577. A.P. is currently affiliated with the Laboratory of Mammary Gland Biology, National Cancer Centre, Singapore.

Disclosure of Potential Conflicts of Interest

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

V.B. and J.S. receive income from TaconicArtemis. All other authors indicate no potential conflicts of interest.

References

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
stem1437-sup-0001-suppfig1.tiff10290KSupporting Information Figure 1
stem1437-sup-0002-suppfig2.tiff14357KSupporting Information Figure 2
stem1437-sup-0003-suppfig3.tiff19KSupporting Information Figure 3
stem1437-sup-0004-suppfig4.tiff11913KSupporting Information Figure 4
stem1437-sup-0005-suppfig5.tiff1360KSupporting Information Figure 5
stem1437-sup-0006-suppfig6.tiff1298KSupporting Information Figure 6
stem1437-sup-0007-suppfig7.tiff1478KSupporting Information Figure 7
stem1437-sup-0008-suppfig8.tiff12KSupporting Information Figure 8
stem1437-sup-0009-suppinfo.doc250KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.