Research ArticleStem Cell Niche

Stromal Gli2 activity coordinates a niche signaling program for mammary epithelial stem cells

See allHide authors and affiliations

Science  21 Apr 2017:
Vol. 356, Issue 6335, eaal3485
DOI: 10.1126/science.aal3485

Double duty for mammary stem cell niche

The stem cell niche is a complex local signaling microenvironment that regulates stem cell activity for tissue and organ maintenance and regeneration. As well as responding locally, during puberty, the mammary gland stem cell niche also responds to systemic hormonal signals. Zhao et al. have found that Gli2, a transcriptional effector of Hedgehog signaling, coordinates the niche-signaling program and activates expression of receptors for the mammatrophic hormones estrogen and growth hormone throughout the mammary gland (see the Perspective by Robertson). Disease may result not only from stem cell defects, but also from dysregulation of the microenvironment.

Science, this issue p. eaal3485; see also p. 250

Structured Abstract


The stem cell niche is a complex signaling microenvironment that acts locally to sustain stem cell activity in tissue maintenance and regeneration. Although the cellular constitution and signaling activity of the stem cell niche is coming into focus in a variety of tissues, genetic regulatory factors that specify the niche are less clear. The activity of such factors is particularly intriguing in organs such as the breast, where the niche provides local signals for tissue homeostasis but also must be entrained by circulating hormones that induce the dramatic changes of puberty.


The epithelial stem cells of a variety of organs respond to secreted signals generated in subjacent stromal cells, in a manner often dependent on activity of the Hedgehog (Hh) signaling pathway. Gli2, which encodes the major transcriptional effector of Hh signaling, is expressed in a subset of stromal cells adjacent to the ductal epithelium of the mouse mammary gland and is highly expressed at terminal end buds and end structures of the pubertal and adult mammary gland, respectively, which represent the sites of mammary epithelial stem cells.


Stromal ablation of a conditional Gli2fl allele with Fsp1Cre, a stromally expressed recombinase allele (producing Gli2∆S mice), caused a delay in mammary ductal development, reduced the number of mammary gland stromal cells and volume of extracellular matrix, and caused abnormal mammary duct distension. Stromal Gli2 ablation did not alter development of the ovary or pituitary, nor their production of mammatrophic hormones such as estrogen or growth hormone, but did affect ductal regeneration, as indicated by a fivefold decrease in outgrowth efficiency of mammary stem cells (MaSCs) transplanted into Gli2∆S mammary glands. These findings suggest that Gli2 specifies a stromal niche signaling program that critically regulates MaSC activity. FACS-isolated mammary stromal cells showed Gli2-dependent expression of factors that stimulate epithelial stem cell renewal, ductal outgrowth, and morphogenesis, including specific members of the IGF, WNT, FGF, and HGF families of secreted peptides. Single-cell analysis showed expression of these factors in a subset of stromal cells in a manner dependent on Gli2 function. Receptors for the mammatrophic hormones estrogen and growth hormone were also expressed in a Gli2-dependent manner, and estrogen and growth hormone treatment of cultured stromal cells induced these secreted peptide factors, suggesting that a critical role of Gli2 in mammary stromal niche cells is to render these cells responsive to systemic mammatrophic hormones. Supporting this conclusion, mammary gland implants of polymer fragments releasing IGF1 and WNT2 rescued Gli2∆S ductal growth phenotypes, whereas growth hormone–releasing polymer did not.


We find that the Hedgehog pathway transcriptional effector GLI2 specifies a stromal cell niche signaling program that supports mammary epithelial stem cells in pubertal and virgin adult mice. This program includes expression of factors such as IGF and WNT, and GLI2 acts in part by causing expression of receptors for hormones such as estrogen and growth hormone, thus entraining local stem cell niche activity to systemic mammatrophic hormones. Our work illustrates the central role of the stromal niche in controlling epithelial stem cell activity and suggests that niche failure may underlie pathogenesis of certain diseases, including the deficient breast development and hormonal insensitivity associated with the human disorder, combined pituitary hormone deficiency.

Stromal GLI2 coordinates mammary stem cell niche signaling program.

Transcriptional regulation by Hedgehog pathway effector GLI2 in stromal cells of the mammary gland coordinates a hormone-responsive niche signaling program that directs epithelial stem cell activity during the changes of puberty (schematic, left). Gli2-deficient stromal cells in mouse mammary are unable to support normal ductal outgrowth and morphogenesis of transplanted, green fluorescent protein–labeled stem cells (right).


The stem cell niche is a complex local signaling microenvironment that sustains stem cell activity during organ maintenanceorgan maintenance and regeneration. The mammary gland niche must support its associated stem cells while also responding to systemic hormonal regulation that triggers pubertal changes. We find that Gli2, the major Hedgehog pathway transcriptional effector, acts within mouse mammary stromal cells to direct a hormone-responsive niche signaling program by activating expression of factors that regulate epithelial stem cells as well as receptors for the mammatrophic hormones estrogen and growth hormone. Whereas prior studies implicate stem cell defects in human disease, this work shows that niche dysfunction may also cause disease, with possible relevance for human disorders and in particular the breast growth pathogenesis associated with combined pituitary hormone deficiency.

The Hedgehog (Hh) signaling pathway regulates embryonic patterning and postnatal homeostasis of multiple tissues (13), acting via its major transcriptional effector, GLI2. A specific function for Gli2 in ductal morphogenesis of the mammary gland was suggested by the defects noted in Gli2−/− mutant mouse embryonic mammary gland rudiments engrafted into adult hosts (4). In addition, breast underdevelopment at puberty is associated with the human disorder CPHD (combined pituitary hormone deficiency) (5), a substantial subset of which is caused by loss-of-function or trans-dominant mutations of the GLI2 gene (68). In contrast, Gli2 function appears to be dispensable for embryonic specification of mammary gland rudiments, which instead shows a requirement for expression of the GLI3 repressor of Hh pathway targets in underlying somitic mesoderm (1, 912). Adding further complexity to the role of Hh pathway activity, Gli2 is expressed in the epithelium of embryonic mammary gland rudiments, whereas Gli2 expression in pubertal and virgin adult mice occurs exclusively in stromal cells (12).

Periductal Gli2 expression in developing mammary

To investigate the role of Gli2 in postnatal mammary gland growth, we examined Gli2 expression in the developing gland and manipulated its activity through genetic ablation or hyperactivation. Gli2 expression occurs intermittently along mammary ducts, and is highly enriched at terminal end buds of the pubertal gland (5 weeks; Fig. 1, A and B) and at terminal structures of the adult female gland (10 weeks; Fig. 1, E and F). These regions of the pubertal and adult ducts undergo extensive proliferation and have been suggested as sites for mammary epithelial stem cells (1318). Gli2-expressing cells in cross sections of mammary ducts appear restricted to stromal regions in both pubertal (Fig. 1, C and D) and adult mammary gland (Fig. 1, G and H), as previously reported (12), and in adult male mammary gland as well (fig. S1A). This stromal pattern of Gli2 RNA and protein expression was confirmed by quantitative polymerase chain reaction (qPCR) and Western blot assays of fluorescence-activated cell sorting (FACS)–sorted stromal versus luminal or basal epithelial cells (Fig. 1, I and J, and fig. S1B). The close proximity of Gli2 expression to mammary ducts (Fig. 1, A to H) suggests the possibility of an inductive influence from epithelium in mediating stromal Gli2 expression. Such an inductive effect was confirmed by the absence of β-galactosidase (β-Gal) expression in cleared mammary fat pads of Gli2nLacZ/+;NSG mice (Fig. 1K, left panel) and appearance of stromal β-Gal expression upon transplantation of cleared fat pads with FACS-isolated mammary epithelial cells (Fig. 1K, right panel, blue arrowheads; see Methods). We found that transforming growth factor–β1 (TGFβ1), expressed in mammary epithelial cells (19), markedly induced expression of Gli2 in FACS-isolated stromal cells in vitro in a dose-dependent manner (4.5-fold and 8-fold at 5 and 50 ng/ml, respectively; Fig. 1L). The dependence of Gli2 expression on TGFβ1 signaling was confirmed in vivo by a 2.2-fold reduction of mammary gland stromal Gli2 mRNA upon treatment of mice with the TGFβ receptor (TGFβR) inhibitor, LY2109761 (Fig. 1M). Response to TGFβ does not include expression of Gli2 in the mammary epithelium; this and other differences in epithelial versus stromal responses may help account for the complexity of TGFβ effects on mammary development and in carcinogenesis (19).

Fig. 1 Gli2 expression in mammary is restricted to stroma and induced by TGFβ.

(A and B, E and F) Representative whole-mount carmine alum stains and β-Gal activity stains (X-Gal) of the fourth mammary gland of Gli2nLacZ/+ mice at 5-week-old pubertal (A and B) or 10-week-old adult (E and F) stages (n = 6 animals for each stage). Note heavier X-gal staining at terminal structures (arrowheads). Scale bar, 500 μm. (C and G) X-gal/Eosin counterstain and (D and H) CK14/β-Gal immunofluorescence stain of cross sections showing stromal expression of Gli2 (arrowheads). Scale bar, 50 μm. (I and J) Stromal cells, epithelial basal cells, and epithelial luminal cells from 10-week-old mice (n = 5) were FACS-isolated and analyzed by qRT-PCR for Gli2 mRNA expression. Gli2 levels were 6.6, 14.2, 2.0, and 2.3% for total lineage (Lin), stromal, luminal, and basal epithelial cells, respectively. (K) X-Gal staining showing epithelial cell–induced stromal Gli2nLacZ expression in mammary whole mount (scale bar, 1 mm) and in EosinY counterstained mammary sections (insets; scale bar, 50 μm). Red arrows point to the vasculature, and blue arrows point to the transplanted epithelial ducts. n = 3 mice (6 transplants) per condition. (L) FACS-sorted mammary stromal cells derived from Gli2nLacZ/+ mice were stimulated as indicated, and representative changes of β-Gal activity were measured and normalized to control samples. n = 3 replicates per condition. (M) Gli2 qPCR analysis showing reduction of Gli2 transcripts in FACS-isolated mammary stromal cells of mice treated with TGFβ inhibitor, LY2109761, for 5 days as compared with vehicle treatment (n = 3 mice per treatment).

Stromal Gli2 in pubertal mammary development

To genetically ablate Gli2 function in stromal cells by site-specific recombination, we examined the extent of coincident expression of β-Gal from the Gli2nLacZ reporter with Cre recombinase from the Fsp1Cre allele of the fibroblast specific protein 1 gene. In mice of the genotype Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ, Cre recombinase causes a switch of expression of the membrane-associated red fluorescent protein mT to expression of the membrane-associated green fluorescent protein mG (fig. S1C). In these mice Gli2 promoter activity, indicated by expression of β-Gal, is enriched in mG-expressing cells (fig. S1D), and a quantitative comparison revealed that most cells expressing β-Gal from the Gli2nLacZ allele also expressed mG (73 ± 12%; Fig. 2, A and B). The Fsp1Cre driver thus appears suitable for Cre-mediated ablation of a floxed conditional allele of Gli2 (Gli2fl) in stromal cells of the mammary gland, which is the exclusive site of Gli2 expression in mammary glands of pubertal and virgin adult females. Fsp1Cre combined with Gli2fl/fl or Gli2fl/nLacZ, hereafter referred to as Gli2∆S (genetic ablation of Gli2 is characterized in fig. S2A) produced a significant delay of mammary gland development at pubertal stages (5 weeks; fig. S2B). By adult stages, Gli2∆S mice gradually developed a hypoplastic mammary gland with greatly enlarged ducts and severely disrupted terminal end structures (6 to 10 weeks; Fig. 2C and fig. S3). Histological examination and FACS analysis further revealed reduced levels of type I, type II, and type IV collagen fibers and a fivefold reduction in periductal stromal cells (2.3% versus 10.9%; Fig. 2E and fig. S2C). The combined ductal and stromal abnormality, as measured by the ratio of ductal diameter to stromal thickness (ductal:stromal ratio or DSR), was highly significant in Gli2∆S mice (Fig. 2D).

Fig. 2 Stromal Gli2 activity is essential for normal mammary development.

(A) Representative immunostaining showing coincident stromal cell expression of Gli2nLacZ and Fsp1Cre in mammary sections of Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/+ mice. Gli2-expressing cells were pseudocolored in red. Membrane-localized mGFP encircles nuclear-localized β-Gal. Scale bar, 50 μm. (B) Venn diagram showing that Fsp1Cre-expressing cells include the majority (73 ± 12%) of Gli2-expressing cells as well as other stromal cells. (C) Representative whole-mount carmine alum staining of second and third mammary glands (i, v) and selected regions of interest (ROI; ii, vi), and H&E staining (iii, vii), and trichrome staining (iv, viii) of second and third mammary glands in 10-week-old Gli2WT and Gli2ΔS littermate mice. Enlarged ducts and disrupted terminal structures in Gli2∆S are apparent (arrowheads in v, vi). The ductal-to-stromal ratio (DSR, panels iii, vii) is defined as the diameter of the ductal lumen (blue) divided by the total width of ductal stroma (yellow). Scale bars, 1 mm (i, v), 200 μm (ii, vi), 50 μm (iii, iv, vii, viii). (D) DSR score, as defined in (C), for ducts of the indicated width classes in Gli2WT versus Gli2ΔS mice. A significantly higher DSR of ducts in various width classes is shown associated with Gli2ΔS mice (10 ducts of the indicated width within each class from the second and third mammary glands of littermate mice were analyzed). (E) Representative analysis of FACS-isolated stromal cells from littermate Gli2WT versus Gli2ΔS mice (n = 5) for expression of Fsp1Cre-activated mGFP. A fivefold reduction of Fsp1-labeled stromal cells in the Gli2ΔS mammary gland is shown.

To examine the effect of increased Gli2 activity, we used Fsp1Cre to drive stromal expression of SmoM2, which encodes a constitutively activating form of the seven-transmembrane Smo component of the Hedgehog pathway (20, 21), in mice of the genotype Fsp1Cre;R26SmoM2. These mice had a maximum life span of 12 weeks (fig. S4A) but, in sharp contrast to Gli2∆S, displayed a marked expansion of the mammary gland stromal compartment, as indicated by increased collagen expression, a 20-fold increase in the percentage of stromal cells expressing the proliferative marker Ki67 (1.2 ± 0.5 versus 23 ± 4), and increased expression of Gli1 (fig. S4, B to D). Both the stromal expansion and proliferative effects of stromal SmoM2 expression were reversed by additional ablation of Gli2, which functions downstream of Smo, in mice of the genotype Fsp1Cre;R26SmoM2;Gli2fl/fl. This reversal demonstrates that SmoM2 effects are dependent on Gli2. The contrasting effects of Gli2 genetic ablation or hyperactivation indicate that the overall abundance of stromal elements within stromal cells is dictated by the presence and activity of Gli2.

Stromal Gli2 supports mammary stem cell activity

The developmental delay and terminal defects in mammary ducts of Gli2∆S mice suggested that ductal epithelial cell proliferation might be impaired. We indeed noted a qualitative decrease in epithelial staining for the Ki67 indicator of proliferative activity in adult mammary gland sections from Gli2∆S mice (fig. S5A) and further found by quantitative FACS-based analysis a 2.6-fold reduction of epithelial cells labeled by incorporation of the nucleoside analog EdU (5-ethynyl-2′-deoxyuridine) in Gli2∆S as compared to Gli2WT mice (1.3% versus 0.5%; Fig. 3A). In contrast, markers indicative of epithelial differentiation were not significantly different in their expression in immunofluorescence staining (fig. S5A), and the relative proportions of FACS-isolated luminal versus basal populations in Gli2∆S mice were not notably different from Gli2WT mice (lower right panels in fig. S5A). To further investigate reduced levels of proliferation, we cultured mammary epithelial cells from Gli2∆S and Gli2WT mice ex vivo in a Matrigel-based colony-forming assay. We found that primary epithelial cells from Gli2WT mice formed three times as many colonies as cells from Gli2∆S mice (Fig. 3B). Collectively, these data indicate that removal of Gli2 in Fsp1Cre/+ stromal cells leads to a defect in the proliferation of the mammary epithelium.

Fig. 3 Stromal Gli2 supports mammary epithelial stem cell regeneration.

(A) Percentage of EDU+ mammary epithelial cells as determined by FACS analysis 5 hours after EDU injection from Gli2WT (Gli2flox/+) and Gli2ΔS (Fsp1Cre:Gli2nLacZ/flox) littermate mice (n = 5 per genotype). (B) Colonies formed per 103 input epithelial cells within 10 days of culture in Matrigel were measured in triplicate from five individual Gli2WT or Gli2ΔS mice. (C and D) A LinCD24+CD49fhiGFP+ basal epithelial cell population enriched in MRU was FACS-isolated from EGFP mice and mixed with similarly isolated LinCD24+CD49fhitdTomato+ basal epithelial cells from Gli2WT or Gli2ΔS mice for competitive repopulation assays. Cell mixtures transplanted into bilateral pairs of cleared fat pads were analyzed for outgrowth after 9 weeks. In the representative outgrowths shown, the red epithelial cells from Gli2WT mice generate a higher proportion of the outgrowth (44 ± 8.5%) as compared to red epithelial cells derived from Gli2ΔS mice (7 ± 3.5%). Scale bar, 2 mm. No differences in stromal abundance were noted in trichrome staining of cross sections of repopulated mammary glands. Scale bar, 50 μm. Ten independent transplants into five NSG animals per condition were used to generate the bar graph in Fig. 2D. (E) LinCD24+CD49fhiGFP+ cells were FACS-isolated, serially diluted, and transplanted into cleared fat pads of immunocompromised (NSG) Gli2WT or Gli2ΔS mice. Outgrowths were analyzed by whole-mount immunostaining, with representative images shown for outgrowths from transplantations of 2000 cells (E). Cross sections show no differences of CK14/CK8 expression in ducts grown in NSG;Gli2WT or NSG;Gli2∆S hosts, although ducts in NSG;Gli2∆S hosts were distended. Scale bars, 5 mm (whole mount) and 50 μm (sections). (F) Results from transplantation of a range of cell numbers are shown as total outgrowths and as the proportion of these outgrowths that are greater than 20 mm2 in size. Outgrowth efficiency is calculated on the basis of total outgrowths by L-CalcTM software of StemCell Technologies. Data are presented as mean ± SEM, and significance was calculated by an unpaired Student’s t test (*P < 0.05%, **P < 0.01).

The growth and regeneration of mammary epithelium depend on the activity of mammary epithelial stem cells (MaSCs), functionally characterized by measurement of mammary repopulating units (MRUs) upon transplantation of FACS-isolated cells into cleared mammary fat pads (22, 23). The activity of MaSCs depends critically upon signals provided by the mammary stroma (24). As MaSCs are enriched in myoepithelium (22, 23), the basal layer of the mammary epithelium that directly contacts the stromal environment, we sought to investigate how stromal loss of Gli2 may affect the epithelial stem cell niche. To this end, we first compared MaSCs from Gli2WT or Gli2∆S mice in a competitive mammary gland repopulation assay by FACS isolation of cells from the upper portion (~50%) of the basal epithelial cell population (Lin CD24+ CD49fhigh; Fig. 1I) of Gli2∆S or littermate Gli2WT mice; these upper basal epithelial cells, enriched in MaSCs (22, 23), are fluorescent red, as mT expression is not affected by the stromal Fsp1Cre driver. These red fluorescent cells were mixed with similarly FACS-isolated green fluorescent cells [expressing enhanced GFP (EGFP)] from control mice, and the mixtures were injected into the left and right cleared fat pads, respectively, of recipient NSG animals (Fig. 3C). Within 9 weeks of transplantation, the red fluorescent cells from Gli2WT mice contributed to the regenerated ductal tree to the same extent as green fluorescent control cells (44%:56%), whereas red fluorescent cells from the Gli2∆S mice contributed minimally (7%:93%) (Fig. 3, C and D, and fig. S5B). Cross sections of mammary ducts from these transplanted animals showed no notable differences in periductal stromal cells and no ductal enlargement in the outgrowths derived from epithelial cells of Gli2WT or Gli2∆S mice (Fig. 3C). Thus, MaSCs derived from the upper basal epithelial population from Gli2∆S mice are defective in their regenerative activity, with reduced colony-forming potential in vitro (Fig. 3B) and a substantially impaired ductal outgrowth potential in vivo in competitive transplantation experiments (Fig. 3C).

To further examine the effect of stromal Gli2 function on MaSCs, we measured MRUs of FACS-isolated upper basal epithelial cells from GFP-expressing mice upon transplantation into cleared mammary fat pads of Gli2∆S or Gli2WT littermate mice (Fig. 3E). We found that transplantation into control host animals with at least one Gli2WT allele in their mammary stroma generates ductal outgrowths fivefold more efficiently as compared to Gli2∆S hosts (outgrowth efficiencies: 1/618 versus 1/2920; Fig. 3F). Ductal outgrowth in Gli2∆S mice was limited to a small area (<20 mm2), and these outgrowths were often malformed, with enlarged mammary ducts and a thin layer of stromal cells (Fig. 3E and fig. S5C) with reduced Gli2 expression (fig. S5D). Notably, the difference in MaSC functional activity measured in these transplantation assays is solely a function of the Gli2 status of the host stromal environment, as the donor cells are from mice that are wild-type except for expression of GFP. These findings thus suggest that the Gli2 activity in FSP1+ cells controls a niche-signaling program that regulates mammary stem cell activity and that this program is disrupted in Gli2∆S mice.

Locally acting factors mediate stromal Gli2 activity

In addition to input from local stromal cells, mammary morphogenesis and the MaSC pool are also regulated by other factors, including the systemic hormonal environment (25) as well as immune cell surveillance (26). We therefore investigated the possibility that Gli2 activity in other cells or organs affects mammary development. To this end, we first analyzed coexpression of Fsp1 and Gli2 in ovary, the major organ for female hormone production (e.g., estrogen and progesterone). We observed few double-positive cells in corpus luteum (0.16%, n = 1927), and no coexpression in follicles (0%, n = 2418) (Fig. 4, A to D), suggesting that Gli2 ablation is a rare event in the ovaries of our Gli2∆S mice. In addition, and consistent with the lack of Fsp1 and Gli2 coexpression in ovary, Gli2WT and Gli2∆S mice had statistically indistinguishable estrogen levels (Fig. 4E, P = 0.5) that were comparable to those previously reported (27). We similarly noted little coexpression of Fsp1 and Gli2 in the pituitary and no significant change in serum growth hormone (GH) or prolactin levels (Fig. 4, F to H). These data collectively suggest that the Gli2∆S mice maintain a hormonal environment similar to that of Gli2WT littermates during development.

Fig. 4 Stromal Gli2 ablation does not affect mammatrophic ovarian or pituitary hormones.

(A to D) Staining for β-Gal and FSP1/S100a4 was performed in a Gli2nLacZ/+ mouse ovary. β-Gal (green), FSP1/S100A4 (red), and nuclei (DAPI, blue) are shown in corpus luteum (A and B) and follicle (C and D). Quantitative analysis (B and D) involved counting more than 2000 cells. Scale bar, 100 μm. (E) Serum from Gli2WT (n = 12) and Gli2∆S (n = 8) mice was collected and probed for estrogen level by ELISA. No significant difference was found between wt and Gli2∆S mutant groups (P = 0.5). (F) Representative H&E (left) immunostaining (middle and right) showing that Gli2-expressing cells are distinct from Fsp1Cre-labeled cells in the anterior lobe of pituitary gland in Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/+ mice. β-Gal (Gli2)–expressing cells were pseudocolored in red. The lack of Gli2 expression in the posterior lobe of the pituitary gland is apparent. (G and H) Quantification of growth hormone and prolactin levels in serum by ELISA in Gli2WT or Gli2ΔS mice.

We next investigated potential coexpression of Fsp1 and Gli2 in mammary macrophages, which are reported to play a role in mammary development during puberty (26). Although we did observe Gli2 expression in a minuscule fraction (~1%) of mammary cells expressing the macrophage marker F4/80 (Fig. 5A, n = 1247), we detected no Fsp1 expression in F4/80+ cells (Fig. 5B, n = 1248), indicating that Fsp1Cre-driven ablation of Gli2 in mammary macrophages is highly unlikely. To further exclude a possible defect in mammary macrophages in the Gli2∆S phenotype, we transplanted wild-type stromal cells or macrophages into the mammary fat pads of pubertal Gli2∆S;NSG recipient mice (Fig. 5C). We noted that although stromal cells successfully rescued ductal enlargement with restoration of periductal collagen and extracellular matrix (Fig. 5, C to F, n = 3), macrophage transplantation did not provide any rescuing activity. These results suggest that macrophages contribute little if anything to the epithelial activity lost in Gli2∆S mice. Furthermore, the sufficiency of locally implanted stromal cells to rescue the Gli2∆S mammary phenotype excludes a role for loss of systemic factors such as pituitary or ovarian mammatropic hormones in the Gli2∆S mammary phenotype.

Fig. 5 Effects of Gli2 ablation in Gli2S mice are mediated by mammary stromal cells.

(A) Mammary gland from a 6-week-old Gli2nLacZ/WT mouse was paraffin sectioned and stained for β-Gal (Gli2 expression) and F4/80 (macrophage marker). Sections containing terminal end buds (TEBs) were analyzed and quantified for Gli2 (green) and macrophage (red) costaining. Scale bar, 100 μm. (B) Mammary gland from a 6-week-old wt mouse was paraffin sectioned and stained for FSP1/S100A4 and the macrophage marker F4/80. Sections containing TEBs were analyzed and quantified for FSP1/S100A4 (green) and macrophage (red) costaining. Scale bar, 100 μm. (C) Fsp1-expressing stromal cells and macrophages were sorted from an Fsp1Cre;mTmG;Gli2nLacZ/+ mouse mammary gland and injected into the fat pads of Fsp1Cre;Gli2fl/fl;NSG recipient mice. Two months after injection, recipient mice were euthanized for whole-mount analysis. (D) Fsp1-expressing stromal cells rescue mammary duct enlargement, whereas macrophages do not. N = 3 experimental repetitions. (E and F) Mammary glands from Gli2∆S;NSG recipients with injected donor macrophages and stroma were collected and sectioned for H&E staining, and DSR (ratio of ductal diameter, blue arrow, to total stromal width, yellow arrow) was determined as in Fig. 2. P < 0.01. Error bars indicate SD.

To identify stromal factors induced by Gli2 activity, we compared gene expression of FACS-isolated Fsp1Cre-marked stromal cells from Gli2∆S or Gli2WT littermates by microarray analysis) and identified genes encoding paracrine factors such as Igf1, Wnt2, Hgf, Fgf7, and Bmp7; we also identified down-regulated genes encoding estrogen receptor (Esr1), growth hormone receptor (Ghr), and other genes typically expressed in stromal cells. Of particular interest among the paracrine factors, WNT signals sustain the activity of MaSCs in vivo and in ex vivo culture (28, 29), and functional inactivation of Igf1 or its epithelially expressed receptor, Igf-1r, leads to development of hypoplastic mammary glands with terminal end bud abnormalities (30, 31), similar to the defects we observe in Gli2∆S mice. The down-regulation of these genes was confirmed by qRT-PCR of bulk stromal cells from Gli2WT and Gli2∆S mice (Fig. 6A) and was further quantified by qRT-PCR from single cells (fig. S6; summarized in Fig. 6B). Igf1 and Wnt2 were expressed in 7% and 9% of individual Gli2WT stromal cells, respectively, but in 0% of Gli2∆S stromal cells (Fig. 6B). Our single-cell analysis was validated by the expression of Gli2 in 19% of Fsp1Cre-marked Gli2WT stromal cells (Fig. 6B), consistent with our immunofluorescence analyses (Fig. 2, A and B). The Gli2nLacZ reporter is expressed within the thin stromal layer that remains in Gli2∆S mice in a pattern similar to that of Gli2nLacZ within control mice (Fig. 6C), indicating that these cells may be correctly specified but fail to function normally, owing to loss of GLI2 protein activity. These results collectively show that the Gli2 transcription factor is critical for expression of mammary stem cell niche signals.

Fig. 6 Stromal Gli2 regulates a niche signaling program for mammary stem cell activity.

(A) Fsp1Cre-labeled mammary stromal cells FACS-isolated from Gli2WT (Fsp1Cre;Rosa26mTmG/+) or Gli2ΔS (Fsp1Cre;Rosa26mTmG/+;Gli2flox/nLacZ) littermate mice were analyzed by qRT-PCR for expression of selected genes, shown as a percentage of HPRT expression, and measured in triplicate from each of three individual Gli2WT or Gli2ΔS mice. Three major classes of stromally expressed genes are down-regulated in stromal cells of Gli2ΔS mice (see text). (B) Gene expression in single, FACS-isolated Fsp1Cre-labeled stromal cells from Gli2WT and Gli2ΔS mice was analyzed by qRT-PCR. The results (fig. S4) are summarized here as the percentage of single cells analyzed that express the indicated genes. (C) Sections of Gli2WT (Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/+) and Gli2ΔS (Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/flox) mammary glands immunostained for collagen and β-Gal. Selected areas (dotted lines) are shown at higher magnification in the green channel only (β-Gal–expressing cells) in the right panels. The frequency of β-Gal–labeled cells per 100-μm length of ductal epithelium is shown in the lower left of the right panels. Scale bar, 50 μm. (D and E) Fsp1Cre-labeled mammary stromal cells FACS-isolated and cultured from Gli2WT or Gli2ΔS mice were stimulated with murine growth hormone (mGH) (D) or 17β-estradiol (E) for 24 hours and assayed for expression of Igf1 or Hgf, respectively. Expression of Igf1 and Hgf in Gli2ΔS is reduced as compared to Gli2WT. Measurements in panels (D) and (E) were made in triplicate from each of three individual Gli2WT or Gli2ΔS mice. Error bars indicate SEM.

Mammatrophic hormone response requires stromal Gli2

Previous studies have shown that MaSCs respond to the mammatrophic hormones estrogen and GH (32, 33). MaSCs do not, however, express the receptors for these hormones (34), indicating that the mammatrophic hormone effect is indirect. Mammary stroma is an important mediator of hormone effects on MaSC activity (24), and we find that expression of estrogen receptor and GH receptor is markedly reduced, from 7 and 14%, respectively, to 2% of individual stromal cells in Gli2∆S mice (Fig. 6B). We therefore examined the Gli2 dependence of hormone response in FACS-isolated Fsp1Cre-marked mammary stromal cells and indeed found that stimulation with GH or with estrogen revealed significantly impaired induction of their target genes Igf1 or Hgf, respectively, in cells from Gli2∆S as compared to Gli2WT mice (2.88-fold reduction for Igf1 and 3.16-fold reduction for Hgf; Fig. 6, D and E), reflecting the significant reduction in expression of receptors for these hormones (fig. S7).

These defects in hormonal response of mammary stroma are reminiscent of the incomplete response of patients suffering from the disorder CPHD to hormonal supplementation. This supplementation is provided because multiple peptidic pituitary hormones and steroidal hormones induced by pituitary hormones are deficient as a result of the congenital hypopituitarism associated with CPHD. Supplementation with GH rescues height defects in these patients (35, 36), but abnormal breast development often is not improved, even with estrogen supplementation (5). Gli2∆S mice produce normal levels of pituitary hormones such as GH and prolactin (Fig. 4, G and H), as Gli2 expression in the pituitary is not affected (Fsp1Cre and Gli2nLacZ expression in the pituitary do not overlap; Fig. 4F). Despite intact pituitary hormone production, however, Gli2∆S mice nevertheless show abnormal mammary gland development, as described above. Gli2∆S mice thus do not share the pituitary hormonal deficiencies of CPHD patients but do share the resistance to hormone treatments, which likely contributes to the pathogenesis of mammary abnormalities in CPHD patients. Consistent with this interpretation, we find that hGLI2, like mGli2, is expressed predominantly in stromal breast cells (fig. S8).

If, as our findings suggest, stromal Gli2 mediates hormonal response and the expression of niche signals that support epithelial stem cell activity, it should be possible to rescue mammary defects seen in Gli2∆S mice by providing missing niche signals from an exogenous source. We therefore tested the effect of slow-release polymer implants containing secreted protein signals in the intact mammary of Gli2∆S mice, focusing in particular on insulin-like growth factor 1 (IGF1) and a canonical WNT family protein because of the well-established roles of these signals in mammary development (2831), and using GH as a control. Ex vivo assays showed that GH, IGF1, and WNT2b proteins embedded in the slow-release polymer, poly(ethylene-co-vinyl acetate) (Elvax40), were gradually released over 2 to 8 days and were able to stimulate pathway-specific target genes (fig. S9, A to G). For in vivo experiments, the effects of control or factor-containing polymer implants surgically placed in the mammary glands of Gli2∆S mice were assessed after 7 days by whole-mount analysis and qPCR (Fig. 7A).

Fig. 7 Slow-release IGF1, WNT2B polymer implants rescue epithelial growth defects in Gli2∆S mice.

(A) Schematic of slow-release polymer-based implantation experiments to test activity of GH, IGF1, and combined IGF1/WNT2B. (B) Control and GH-containing polymers (dashed lines) were implanted bilaterally into immunodeficient Gli2ΔS (NSG: Gli2ΔS) mice, and ductal morphology was examined 7 days after implantation (n = 3 implantations; one representative image of three is shown). Scale bar, 1 mm. (C) Numbers of terminal structures within 2 mm of the implantation site are shown as fold increase relative to control implants. (D) Induction of Igf1 induction in control and GH-containing implants. GH failed to induce terminal growth or Igf1 expression in NSG;Gli2ΔS mice. (E) Control and IGF1-containing polymers (dashed lines) were implanted bilaterally into NSG;Gli2ΔS mice at pubertal (p40) or adult (p70) stages. Ductal morphology was examined after 7 days (n = 3 implantations; one representative image of three is shown). Scale bar, 1 mm. (F and G) Numbers of terminal structures (F) within 2 mm of implants and IGF1-induced expression of Egr1 and M-Csf (G) were calculated and normalized to control implants. (H) Control and IGF1/WNT2b-containing polymers (dashed lines) were implanted bilaterally into NSG;Gli2ΔS mice, followed by morphological analyses after 7 days (n = 3 implantations; one representative image of three is shown). Scale bar, 1 mm. (I and J) Numbers of terminal (I) structures and induction of IGF1 or WNT2b target genes (J) were quantified as fold increase relative to control implants. IGF1/WNT2b implants induce more terminal end growth as compared to IGF1 alone. Data are normalized to the mean values in the control and are presented as mean ± SEM; statistical significance was calculated by an unpaired Student’s t test (*P < 0.05%, **P < 0.01).

We indeed confirmed, that, although GH-containing Elvax40 implants increased the number of terminal structures fivefold in Gli2WT;NSG host mice (5.625 ± 1.21; fig. S10, B and C), in Gli2∆S;NSG mice these implants failed to stimulate Igf1 expression and did not change terminal structure growth at the implantation site (Fig. 7, B to D). In contrast to GH, IGF1-containing implants increased terminal structure formation in the vicinity of the implants sevenfold in pubertal (psd40) and fivefold in adult (psd70) Gli2∆S mice (Fig. 7, E and F, and fig. S10A). IGF1 target genes Egr1 and M-Csf also were reliably induced near the site of the implants (Fig. 7G), confirming the efficacy of IGF1 delivery. As compared to IGF1 alone, implants containing both IGF1 and WNT2b showed an even greater effect in inducing a 13-fold increase in terminal structure growth at the adult stage (Fig. 7, H and I) in Gli2∆S mice and consistently induced the expression of M-Csf and Axin2 (Fig. 7J); Axin2 expression indicates Wnt response, which promotes the survival and self-renewal ability of mammary epithelia (29). Together these results suggest that, although the mammary abnormalities in Gli2∆S mice fail to respond to GH stimulation, the terminal growth defects of the mammary ductal tree can be rescued in part by local supplementation with the GLI2-dependent paracrine factors IGF1 and WNT2b.


Although the cellular constitution and signaling activity of the stem cell niche is coming into focus in a variety of tissues, the genetic regulatory factors that specify the niche and thus create a favorable environment for stem cell function have not been identified. The activity of such factors is particularly intriguing in organs such as the breast, where the niche provides local signals for tissue homeostasis but also must be entrained by circulating hormones that induce the dramatic changes of puberty. Using a Gli2nLacZ reporter and a stromal-specific Cre driver, we find that Gli2 is predominantly expressed in periductal stroma of the mammary gland of pubertal and virgin adult mice and encodes a functionally critical regulator that specifies niche activity in support of mammary epithelial stem cells (Fig. 8). One aspect of this regulatory activity is the induction, direct or indirect, of stem cell support factors, including IGF and WNT signaling proteins. The GLI2 transcription factor thus appears to control the niche signaling program for MaSCs. Gli2∆S mice lacking stromal Gli2 function consequently suffer from a severe defect in regenerative activity of epithelial stem cells. A second aspect of Gli2 activity is the induction of receptors for mammatrophic hormones such as growth hormone and estrogen in stromal cells such that they become competent to respond to circulating hormones, thereby placing niche signaling activity under hormonal control. This hormonal response is also lost in Gli2∆S mice (Fig. 8). Thus, our analysis provides insight into the mechanism that coordinates systemic mammatrophic hormone action and activity of the local epithelial stem cell niche, namely, that a single transcription factor, GLI2, controls both the expression of hormone receptors and the stromal niche signaling program.

Fig. 8 Mammary stem cell niche dysfunction in Gli2∆S mice and CPHD patients results from stromal loss of Gli2.

Normal Gli2 function in mammary stroma (left) coordinates mammary stem cell niche signaling, response to estrogen and growth hormone, and stromal expansion, leading to normal mammary stem cell maintenance and hormone–induced growth and morphogenesis at puberty. Stromal loss of Gli2 function in Gli2∆S mice or CPHD (right) leads to disruption of the niche signaling program and loss of mammatrophic hormone response at puberty.

Our findings in Gli2∆S mice imply that a defective stem cell niche environment may underlie the pathogenesis of breast defects seen in the human disorder CPHD. Although several degenerative and aging-related diseases have been tentatively ascribed to defects in stem cells themselves, dysfunction of a supporting niche has not previously been linked to human disease (37). By indicating that niche failure can cause disease, our study asserts the central role of the niche in controlling stem cell activity and raises the possibility of niche dysfunction in other human diseases. Furthermore, by showing that Gli2-dependent mammary niche signaling factors can rescue aspects of the Gli2∆S mammary in mice, our work suggests a potential therapeutic intervention for patients with hormone-refractory CPHD.



Female mice at pubertal (5 weeks) or adult (10 weeks) stages were used for all experiments. Gli2nLacZ mice (38) were first bred with Gli2flox mice (39), then with Rosa26mTmG/+ mice (40), to generate Gli2flox/nLacZ or Gli2nLacZ/+;Rosa26mTmG/+ mice. Fsp1Cre mice (41) were crossed to Gli2flox/flox to generate Fsp1Cre;Gli2flox/+ or with Rosa26mTmG/mTmG mice to generate Fsp1Cre;Rosa26mTmG/+ strain. Fsp1Cre;Gli2flox/+ mice were then crossed with Gli2flox/nLacZ mice to obtain littermate Gli2flox/+ and Fsp1Cre;Gli2flox/nLacZ mice. The Fsp1Cre;Rosa26mTmG/+ strain was crossed to Gli2flox/flox strain to generate Fsp1Cre;Rosa26mTmG/+;Gli2flox/+ mice which were subsequently bred to Gli2nLacZ/+;Rosa26mTmG/+ to obtain littermate Fsp1Cre;Rosa26mTmG/+, Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/+ and Fsp1Cre;Rosa26mTmG/+;Gli2nLacZ/flox mice. NSG mice were extensively bred and back-crossed with Fsp1Cre;Gli2nLacZ/flox through six generations to obtain Gli2flox/+;NSG, Gli2nLacZ/+;NSG, Fsp1Cre;Gli2flox/flox;NSG, and Fsp1Cre;Gli2nLacZ/flox;NSG mice. All mouse strains were obtained from Jackson Laboratories, except as otherwise indicated. For each experiment, mice of the same genotypes were randomly selected for experiments and investigators were blinded for outcome assessment. All procedures were performed under an assigned protocol (14586) approved by the Administrative Panel on Laboratory Animal Care (APLAC) at Stanford University.

Antibodies and reagents

The following antibodies were purchased: chicken polyclonal anti-β-Gal antibody (Abcam, CAT:ab9361), rabbit polyclonal anti-CK14 antibody (Covance, CAT:PRB-155P), rabbit polyclonal anti-Ki67 antibody (Abcam, CAT: ab15580), Cy3-conjugated mouse monoclonal anti-smooth muscle actin antibody (Sigma, CAT: C6198), rat anti-CK8 antibody (DSHB, TROMA-I), rabbit polyclonal anti-collagen I antibody (Abcam, CAT: ab21286), goat polyclonal anti-NKCC1 antibody (Santa Cruz, CAT:sc-21545), guinea pig anti-Gli2 antibody was obtained from Dr. Eggenschwiler at University of Georgia, APC (allophycocyanin)–conjugated rat anti-mouse F4/80 (Biolegend), rabbit polyclonal anti-S100a4 (Biolegend). Dylight 488-labeled goat anti-chicken IgY, Alexa 488-labeled donkey anti-rabbit immunoglobulin G (IgG) and Alexa 647-labeled donkey anti-goat IgG were purchased from Jackson ImmunoResearch. Alexa 488-labeled goat anti-rabbit IgG, Alexa 594-labeled goat anti-rabbit IgG, and Alexa 647-labeled goat anti-rat IgG were purchased from Invitrogen. Trichrome straining kit (25088-100) was purchased from Polysciences Inc. Rat/mouse growth hormone enzyme-linked immunosorbent assay (ELISA) kit was purchased from Millipore (EZRMGH-45K). Mouse prolactin ELISA kit was purchased from Abcam (ab100736). Mouse Igf1 ELISA kit was purchased from Raybiotech (ELM-IGF1). Mouse Wnt2B ELISA kit was purchased from MYBioSource (MBS946031). Click-IT®Plus EdU Alexa Fluor®647 flow cytometry assay kit (C10634) was purchased from Invitrogen. Poly(ethylene-co-vinyl acetate) (Elvax40) beads (340502) were purchased from Sigma. Recombinant mouse Igf1 (791-MG-050) and recombinant mouse Wnt2b (customized quantity) were purchased from R&D Systems. Mouse growth hormone (CYT-540) was purchased from Prospecbio. 48.48 Dynamic Array Chips (BMK-M.48.48) for single-cell analysis were purchased from Fluidigm. Ortho-nitrophenyl-β-galactoside (ONPG) was purchased from Thermo-Scientific (catalog no. 34055). TGFβRI/II inhibitor, LY2109761, was purchased from XCESSBIO in a customized quantity.

Mammary cell preparation and flow cytometry

Mammary glands from adult mice (10-week-old) were dissected and minced. Single-cell suspensions were made by digestion of minced tissue with culture medium [Dulbecco’s modified Eagle’s medium (DMEM)/F12, 5% fetal bovine serum, 1% penicillin-stretomycin-glutamine] containing collagenase/hyaluronidase cocktail solution (StemCell Technolgies, 07912) for 3 hours at 37°C, followed by sequential-5 min treatments with red blood cell lysing buffer (Sigma, R7767), 0.25% trypsin-EDTA, and deoxyribonuclease (0.1 mg/ml) in dispase (0.1 U/ml) with gentle pipetting. Cell suspension was then filtered through 40-μm cell strainers and applied to subsequent FACS isolation. Phycoerythrin (PE)/Cy7-conjugated anti-mouse CD24 (rat, clone M1/69), APC–conjugated anti-mouse CD49f (Rat, clone GoH3), PerCP/Cy5.5 conjugated anti-mouse CD31 (rat, clone 390), PerCP/Cy5.5 conjugated anti-mouse CD45 (rat, clone 30-F11), and PerCP/Cys5.5 conjugated anti-mouseTer119 (rat, TER-119) were purchased from BioLegend. Antibodies were then incubated with strainer-filtered mammary cells in Hanks’ balanced salt solution (HBSS) with 2% fetal bovine solution for 20 min on ice. All FACS analysis and cell sortings were performed using FACS aria II (Becton Dickinson).

EdU flow cytometry

In vivo EdU labeling was accomplished by intraperitoneal injection of EdU (150 μg per 10 g of body weight). Mammary glands were collected after 3 hours and prepared as single-cell suspensions, followed by click-it chemistry for EdU staining and surface markers–based staining (CD24-PE/Cy7, CD49f-APC, CD31-PerCP/Cy5.5, CD45-PerCP/Cy5.5, Ter119-PerCP/Cy5.5) to separate stromal, luminal, and basal populations. EdU-stained cells in the stromal-negative population were then presented as a percentage of parental cells.

Liquid β-Gal assay

FACS-isolated mammary stromal cells were cultured ex vivo and treated with the indicated concentrations of TGFβ with or without TGFβRI/II inhibitor, LY2109761. Cells were lysed and incubated with colorimetric β-Gal substrate, ortho-nitrophenyl-β-galactoside (ONPG). β-Gal activity as indicated by OD420 was measured at the indicated conditions.

In vitro colony-formation assay

To compare ex vivo colony-forming activity of mammary epithelial cells derived from different backgrounds, we adopted an established 3D On-Top assay (42). Briefly, Matrigel (BD Bioscience) was thawed and applied to the bottom of 24-well plate at a volume of 120 μl per well, followed by polymerization at 37°C for 15 min. FACS-isolated mammary epithelial cells were resuspended in stock buffer (DMEM/F12, 10% FBS) at a density of 2 × 105 cells per ml, serial diluted in Matrigel containing buffer (DMEM/F12, 10% FBS, 50% Matrigel), and plated on top of the polymerized Matrigel. Culture medium was changed every 48 hours. Primary colony numbers were scored after 14 days in culture, followed by sequential phosphate-buffered saline (PBS)–EDTA treatment and extraction of 3D-cultured mammary colonies into single-cell suspensions.

Histology analysis

Gli2WT and Gli2ΔS mice were euthanized at the indicated ages. The 2nd/3rd and/or the 4th mammary glands were dissected. For whole-mount carmine alum analysis, mammary glands were fixed in Carnoy’s fixative for 3 hours, followed by serial rehydration, carmine alum staining, and serial dehydration. Mammary glands were then cleared with xylenes and mounted with permount. For whole-mount X-Gal analysis, mammary glands were fixed in X-Gal fixative (2% paraformaldehyde, 0.2% glutaraldehyde) for 30 min at ambient temperature, followed by serial rehydration, overnight carmine alum staining, dehydration, and clearance with xylenes. For X-Gal staining, hematoxylin and eosin (H&E) staining, and trichrome staining, mammary glands were fixed and cut as paraffin sections for H&E or trichrome staining or as cryosections for X-Gal staining. X-Gal, H&E, and trichrome staining on mammary sections was performed as described previously (43).

Immunofluorescence and confocal microscopy

Dissected mammary glands were fixed in 4% paraformaldehyde, serially dehydrated, and processed as 4-μm paraffin sections. Paraffin sections were cleared with xylenes, rehydrated, and subjected to antigen-retrieval in citric acid–based buffer by the pressure cooker method (44). For immunostaining, tissue sections were rinsed in PBS, then blocked in blocking solution containing 3% bovine serum albumin (BSA), 0.25% Triton X-100, and serum of the host species used to generate the secondary antibody (10% goat serum or 10% donkey serum). An hour after blocking, diluted primary antibodies in blocking buffer were incubated with sections overnight at 4°C in a humidified chamber. Sections were then washed three times with PBS containing 0.25% Triton X-100 and incubated with appropriate secondary antibodies diluted 1:1000 in blocking solution for 1 hour at ambient temperature, followed by three washes and mounting with Prolong Gold mounting reagent with 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen). Immunofluorescence images were obtained using a Zeiss LSM 510 inverted confocal microscope, processed for publication with Zeiss LSM5 Image Browser software, and are representative images from one of three experiments.

Mammary fat pad transplantation

FACS-isolated cells were resuspended in a cocktail buffer with 50% DMEM/F12 (20%FBS), 50% Matrigel, and 0.04% Trypan Blue (Sigma). Cells were injected into cleared mammary gland fat pads of 3-week-old female mice. Repopulated mammary glands were harvested 9 weeks after transplantation. Outgrowths of repopulated mammary ductal structures were imaged under a fluorescence dissection microscope (Leica). For macrophage and stromal cell rescue assay, macrophages (identified by F4/80) and Fsp1+ stromal cells (identified by GFP) from the mammary glands of Fsp-cre Gli2nLacZ/wt mTmG mice were injected into the inguinal fat pads of 3-week-old Gli2ΔS NSG recipient mice at 10,000 cells per gland with 50% Matrigel + DMEM F12 medium. Two months after the injection, mammary glands were dissected and analyzed by whole-mount and immunohistochemistry (IHC).

Mammary in vivo competitive repopulation assay

Green color–labeled LinCD24+CD49fhighGFP+ basal cells isolated from Gli2WT mice were mixed in equal number with red color–labeled LinCD24+CD49fhiGFP+ basal cells isolated from Gli2WT mice (green Gli2WT: red Gli2WT) or from Gli2ΔS (green Gli2WT: red Gli2ΔS) mice. These mixtures were transplanted bilaterally into cleared mammary gland fat pads of immune-deficient NSG mice (non-obese diabetic, severe combined immunodeficiency, IL-2 gamma chain-deficient). Outgrowths in this competitive mammary repopulation assay were examined 9 weeks after transplantation.

Mammary regeneration assay

Lin-CD24+CD49fhiGFP+ basal cells derived from EGFP+ mice were serially diluted as indicated and transplanted into cleared mammary gland fat pads of Gli2WT;NSG mice with at least one functional allele of Gli2 (Gli2flox/+;NSG or Gli2nLacZ/+;NSG) or into Gli2ΔS;NSG littermates (Fsp1Cre;Gli2flox/nLacZ;NSG or Fsp1Cre;Gli2flox/flox;NSG). Mammary outgrowth frequency and size of outgrowths were measured. Outgrowth frequency (repopulation frequency) was calculated by L-Calc software of StemCell Technologies. Owing to limited numbers of matching Gli2WT;NSG and Gli2 ΔS;NSG within a single litter, data presented in the table summarize transplantation results in mice from several different litters.

RNA extraction and microarray analysis

Total RNAs of FACS-isolated stromal population from Gli2WT and Gli2ΔS mice were extracted by Trizol and RNeasy Plus min kit (Qiagen). Concentration and quality of extracted RNAs were determined by Agilent 2100 Bioanalyzer (Agilent Technologies). RNA samples were hybridized to the Mouse Gene 2.0 ST microarray chips and scanned in accordance with the manufacturer’s protocol (data available at Expression values normalized and differentially expressed genes were identified by Transcriptome Analysis Console (TAC). Genes showing a fold change of 1.5 (log2) were highlighted in functional annotation analyses. Expression changes of genes of interests by microarray analysis were validated via quantitative reverse transcription–polymerase chain reaction (RT-PCR) analysis.

Single-cell FACS isolation and data analysis

GFP-labeled mammary stromal cells of Gli2WT and Gli2ΔS mice were FACS-isolated (5000 cells per genotype), followed by single-cell FACS isolation into 96-well plates. Sorted single cells on a 96-plate format were lysed by hypotonic pressure, and mRNA expression levels (Ct, threshold cycle) of 48 genes in 48 cells were measured by microfluidic single-cell RT-qPCR reaction in two 48 by 48 format Fluidigm chips (45, 46) (Fluidigm, CA). Data analysis was performed with Matlab (Mathworks, MA).

Slow-release polymer implantation assay

GH, IGF1, and IGF1/WNT2B were prepared as slow-release proteins in Elvax40-based polymers (4749). Briefly, lyophilized BSA (20 mg) alone or in a mixture with the indicated amount of GH, IGF1, or IGF1/WNT2B were dispersed in 0.125 ml of Elvax40 that had been dissolved in dichloromethane (20% wt/vol). The mixture was quick-frozen and dried in a rotary evaporator, then the polymer matrix with entrapped proteins was cut to size (0.5 to 1 mg) and surgically implanted. Kinetics and activities of released proteins were validated in ELISA analyses and in qPCR analyses of their target gene expression, respectively. Control Elvax40 polymer or Elvax40 polymers containing the above-mentioned proteins were implanted bilaterally into pubertal or adult Gli2ΔS mice. Mammary ductal morphology was examined 7 days after surgery. In parallel experiments, mammary ducts in the vicinity (1 to 2 mm) of implants were dissected and prepared as single-cell suspensions. Cells derived from various implants were plated in a 24-well Petri dish for 2 hours to deplete fibroblasts, followed by collection of supernatant enriched in epithelial cells. Total RNAs were extracted from epithelial cells and converted to cDNAs, which were then applied to gene-specific (Bio-Rad) preamplification and subsequent qPCR analyses.

GH, estrogen, and prolactin ELISA analyses in serum

Peripheral blood was drawn by retro-orbital bleeding in the amount of 200 μl from Gli2WT and Gli2ΔS mice. Mice were allowed to recover for 5 days until the next collection. Blood samples were kept at ambient temperature for 30 min, followed by centrifugation at 3000 rpm for 10 min. Serum in the supernatant was either used immediately for ELISA analyses or aliquoted and stored at –80°C until next use. Serum samples were diluted and subjected to standard ELISA assays to determine GH or prolactin levels in accordance with the manufacturer’s protocol. Differences between genotypes are considered significant if P < 0.05. Estrogen levels of Gli2WT and Gli2ΔS mice were determined by 17β-Estradiol ELISA kit (from Enzolifesciences) according to the manufacturer’s instructions.

Statistical analysis

Statistic analysis was performed with GraphPad Prism v5. Data are presented as mean ± SEM, and group differences were examined with a two-tailed Student’s t test. P values were calculated on the basis of three independent experiments unless otherwise specified. A value of P < 0.05 is considered statistically significant.

Primers used in real-time RT-PCR analysis

Primers used for preamplification of Igf1, Egr1, mCsf, and Axin2 in cDNAs of dissected implants were purchased from Bio-Rad. Primers used for qPCR analyses of Col1a1, Col1a2, Col2a1, Pdgfrα, Pdgfβ, and Esr1 were previously reported (5052). Others were purchased as QuantiTect Primer Assays from Qiagen. QuantiTect primer assays used for real-time RT-PCR include mGli2 (QT00291711), Ghr (QT00109900), mBmp7 (QT00096026), mIgf1 (QT00154469), mWnt2 (QT00118503), mWnt2b (QT00115451), mFgf7 (QT00172004), mHgf (QT00158046), mEgr1 (QT00265846), mCsf1 (QT01164324), mIl-3Rα (QT00251293), mAxin2 (QT00126539), and mLef1 (QT00148834).

Supplementary Materials

References and Notes

Acknowledgments: This work was supported by a Susan G. Komen fellowship to C.Z. and funding from the U.S. Department of Defense, the Breast Cancer Research Foundation, and the Ludwig Institute for Cancer Research. The cell imaging core of Stanford Neuroscience Microscopy Service and the FACS core of Stanford Institute for Stem Cell Biology and Regenerative Medicine were supported by NIH core grants. We are grateful to K. M. Loh and J. Sage for critical reading and helpful discussion of this manuscript. P.A.B. is an investigator of the Howard Hughes Medical Institute.

Stay Connected to Science

Navigate This Article