A Quiescent Bcl11b High Epithelial Cell is Required for Mammary Gland Regeneration

Summary

Stem cells in many tissues sustain themselves by entering a quiescent state to avoid genomic insults and to prevent exhaustion caused by excessive proliferation. In the mammary gland, the identity and characteristics of quiescent epithelial stem cells are not clear. Here, we identify a quiescent mammary epithelial cell population expressing high levels of Bcl11b and located at the interface between luminal and basal cells. Bcl11b^high cells are enriched for cells that can regenerate mammary glands in secondary transplants. Loss of Bcl11b leads to a Cdkn2a-dpendent exhaustion of ductal epithelium and loss of epithelial cell regenerative capacity. Gain and loss of function studies show that Bcl11b induces cells to enter the G0 phase of the cell cycle and become quiescent. Taken together, these results suggest that Bcl11b acts as a central intrinsic regulator of mammary epithelial stem cell quiescence and exhaustion, and is necessary for long-term maintenance of the mammary gland.

eTOC summary

Cai et al. (2016) describe a quiescent mammary stem cell population labeled by Bcl11b and located at the luminal-basal interface that supports mammary gland regeneration. Bcl11b sustains this population by inducing cell cycle regulators that promote the dormant state.

Introduction

The mammary gland contains a ductal system consisting of basal and luminal cells that generates a milk-producing organ during pregnancy. Following weaning, the mammary gland ductal system undergoes involution and the proper long-term maintenance of the proliferation capacity of the mammary epithelia is required for multiple rounds of female reproductive cycles.

There are advantages provided by quiescence in tissue-specific stem cells; they may avoid proliferation associated genome damage that can cause the accumulation of deleterious mutations and/or the initiation of apoptosis (Codega et al., 2014; Foudi et al., 2009; Wilson et al., 2008). Despite extensive studies using a number of techniques including flow cytometry (Shackleton et al., 2006; Stingl et al., 2006), lineage tracing (Plaks et al., 2013; Rios et al., 2014; van Amerongen et al., 2012; Van Keymeulen et al., 2011; Wang et al., 2015), and in vitro culture (Prater et al., 2014; Zeng and Nusse, 2010), the cellular hierarchy of the mammary gland is still controversial. Some have suggested that the mammary gland is maintained by separate basal and luminal progenitors, while others have suggested a bipotent basal cell progenitor that can generate both basal and luminal cells (Rios et al., 2014; Van Keymeulen et al., 2011). Regardless of the hierarchy of the mammary epithelium, it is important to understand the molecular regulation of the long-lived epithelial cells, which have the greatest proliferation capacity.

For most of their life, quiescent stem cells (i.e. hematopoietic stem cells or skin stem cells) remain at minimal cycling rate and metabolic activities to preserve their long-term self-renewal ability under physiological condition. Upon stress or injury, they can be activated by growth signals and give rise to multiple cell types to orchestrate a homeostatic architecture of the organs for regeneration (Wilson et al., 2008). In the mammary gland, reminiscent of quiescent hematopoietic stem cells (HSCs), label retention assays suggest the existence of a quiescent long-lived cell population with extensive self-renewal ability (dos Santos et al., 2013; Pece et al., 2010; Smith, 2005). Understanding the intrinsic molecular network that specifies the quiescence program of long lived mammary epithelial cells could shed light on the regulation of long-term tissue homeostasis, organ regeneration, cancer relapse, aging and many other pathological degenerative diseases.

In this study, through single cell gene expression analysis of mammary epithelia, we identified a quiescent population within mammary basal cells that expressed high level of Bcl11b. We demonstrated that Bcl11b is a major cell intrinsic factor that is functionally required for maintaining a minority of epithelial cells that express the basal cytokeratin Krt17 in a dormant state. Bcl11b knockout mice had impaired mammary gland development and mammary epithelial cells were unable to regenerate mammary glands after transplantation. Thus, Bcl11b is required to preserve epithelial cell long-term proliferation capacity and to maintain normal mammary gland homeostasis.

Results

Bcl11b is a Nuclear Protein Highly Expressed in CD49f^highCD24^medLineage⁻ Cells and Specifically Localized to the Mammary Duct Basal Layer

The composition of mammary epithelium at the single cell level was investigated. We first focused on CD49f^highCD24^medLin⁻ cells, which are greatly enriched for mammary repopulating units (MRUs) as measured by transplantation assays (Shackleton et al., 2006; Stingl et al., 2006). To better understand the molecular regulation of long-term proliferation capacity as measured by transplantation assays, we isolated the various populations of mammary epithelial cells based on the expression of CD49f and CD24: Basal1 (CD49f^highCD24^medLin⁻) cells, which are enriched for cells with the greatest proliferation capacity, Basal2 (CD49f^highCD24^lowLin⁻) and Lum1 (defined as CD49f^lowCD24^highLin⁻ cells), Lum2 (defined as CD49f^lowCD24^medLin⁻ cells) (Stingl et al., 2006) (Fig. 1A, S1A–C) were analyzed using sensitive single-cell multiplexed gene PCR (Dalerba et al., 2011). The expression of a number of transcription factors (which were selectively expressed by a subset of mammary epithelial cells) was screened in the CD49f^highCD24^medLin⁻ population. In this screen, Bcl11b, a C2H2 zinc finger transcription factor and a member of various chromatin-remodeling complexes (e.g. SWI/SNF, NURD)(Cismasiu et al., 2005; Krasteva et al., 2012), was found to be highly expressed in a subset of Basal1 compared to Basal2 cells or luminal epithelial cells (Fig. 1A,B, S1C–F). Intriguingly, superimposed clustering analysis demonstrated that Bcl11b is not uniformly expressed at high levels in individual CD49f^highCD24^medLineage⁻ cells (Fig. 1B). Instead, Bcl11b mRNA positive cells comprised only 4.8% of the Basal1 cells that expressed the basal cytokeratin Krt17 (12 out of 248 cells) with both decreased frequency and levels of expression observed in Basal2 and luminal populations (Fig. 1C and S1A–D). This expression pattern was further confirmed by real time PCR analysis of the corresponding mammary populations (Fig. 1D). We couldn’t detect Bcl11b expression in the overall luminal population, Cd14+ luminal progenitors, or Cd14- mature luminal cells (Fig. S1G).

Fig 1. Bcl11b expression is restricted to CD49f^highCD24^medLineage⁻ cells and shows sporadic localization in the basal layer.

(A) FACS plot of a mammary gland from an 8–12 week old adult mouse with surface markers CD24 and CD49f. Red circles show gates for Basal1(CD49f^highCD24^medLineage⁻), Basal2(CD49f^highCD24^lowLineage⁻), LUM1(CD49f^lowCD24^highLineage⁻) and LUM2(CD49f^lowCD24^lowLineage⁻) populations. (B,C) Single cell gene expression analysis of Bcl11b in Basal1, Basal2 and Luminal cells. Cell and gene clustering are superimposed with the expression panel in (B). Red color indicates high expression, green indicates low expression and gray indicates no expression. The Ct value of Bcl11b in the original gates is shown in (C). In this scale bar, blue color indicates low Ct value (high expression), while red color indicates high Ct value (low expression). (D) Real time PCR quantification of Bcl11b mRNA levels in Basal1, Basal2 and Luminal cells with ActB as an internal control. Data were presented as Mean±SEM. N=3 (E) Immunofluorescence of the mammary gland from an 8 week old mouse. Red: Krt14; Green: Bcl11b; Blue: DAPI. Scale bar, 20μm (F) Schematic diagram of Bcl11b tdTomato reporter mouse. (G) A representative FACS plot of a mammary gland from a Bcl11b^tdtomato/wt reporter mouse. The TdTomato expression level is shown for EpCAM+ mammary cells. (H) Tdtomato expression level is shown for CD49f^highCD24^medLineage⁻ population. (I) Real time PCR confirming Bcl11b^high population isolated from Bcl11b tdtomato reporter mice enriched for Bcl11b mRNA expression. N=3, p<0.01. (J) Mammary bifurcating mammary duct of a 12-week-old Bcl11b^tdtomato/wt mouse. Stacked images were shown on the left panel. Optical sections were shown on the right panel. Yellow arrowheads point to individual tdTomato positive cells. Scale bar, 100μm. See also Figure S1

Next, we examined the mammary tree for the location of the Bcl11b positive cells. In adult mice, immunofluorescence analysis showed that Bcl11b positive cells were sparsely scattered within the interface of the basal and luminal cell layers of the duct in the mammary gland (Fig. 1E). Interestingly, Bcl11b positive cells were rarely seen in the basal cells of some ducts, which only had one layer of luminal cells and appeared to be mature (Fig. 1E). The immunochemistry corroborates the single cell gene expression data showing that Bcl11b is exclusively expressed by less than 1% of cells located at the interface of the basal and luminal cells.

To isolate Bcl11b^high cells from the mammary gland, we used a Bcl11b tdTomato reporter mouse (Li et al., 2010), which has an inserted IRES-tdTomato cassette at the 3’UTR of the Bcl11b locus (Fig. 1F). Consistent with the quantitative RT-PCR and immunochemistry results, a distinct population of tdTomato-high cells was exclusively restricted to cells adjacent to the basal population (Fig. 1G). Only a subset of the MRU-enriched CD49f^highCD24^medLineage⁻ cells (5.7±3.5%, mean±S.D.) expressed high levels of Bcl11b as measured by tdTomato fluorescence intensity (n=24, p<0.0001) (Fig. 1H, I, S1H). These data suggested that CD49f^highCD24^medLineage⁻ cells contain a minority population of cells that express high levels of Bcl11b RNA and protein. To track Bcl11b expression in intact mammary glands, Bcl11b^tdTomato/wt reporter mice were analyzed by multi-photon microscopy. In these mice, individual cells expressing high levels of Bcl11b were sporadically distributed at the rim of the mammary gland, in the branching mammary tips, bifurcating ends, and scattered in the mature ducts (Fig. 1J). Thus, single cell gene expression, immunofluorescence, flow cytometry and live cell imaging all corroborated the rarity of Bcl11b^high cells are located adjacent to both the basal and luminal cells in the mammary ducts.

Conditional Knockout of Bcl11b Impairs Mammary Gland Development and Regeneration Capacity

The fact that Bcl11b is an epigenetic regulator of transcription and a potential fate determinant prompted us to further investigate whether Bcl11b played a functional role in the mammary epithelium and mammary gland development. We deleted Bcl11b expression in the mammary gland by crossing Krt14-cre mice with Bcl11b^flox/flox mice (Golonzhka et al., 2009). Embryonically activated Krt14-cre can trace the majority of the mammary epithelia including both luminal and basal cells, allowing us to investigate Bcl11b’s global function within the mammary gland. Analysis of the Bcl11b mutant mammary gland showed substantially retarded postnatal development with significant reductions (~50%) in mammary tree size (p<0.001) (Fig. 2A, B). This suggests that Bcl11b is required for proper mammary morphogenesis.

Fig. 2. Deletion of Bcl11b causes mammary gland hypoplasia.

(A) Whole mount image of Bcl11b^flox/flox and Krt14-cre Bcl11b^flox/flox inguinal mammary gland at week 9. (B) Quantification of mammary gland size from littermates (control and Bcl11b mutant) at adulthood (10–14 weeks). Data are plotted as Mean±S.D. (p<0.0001, n=16). (C) Image of collagenase digested mammary gland fragments from Krt14-cre mTmG Bcl11b^wt/wt and Krt14-cre mTmG Bcl11b^flox/flox mice showing GFP and tdTomato. (D) FACS analysis of Krt14-cre mTmG Bcl11b^wt/wt and Krt14-cre mTmG Bcl11b^flox/flox mice with CD49f and EpCAM. EpCAM positive mammary epithelia were gated and plotted using GFP. (E) Quantification of GFP positive cells of whole mammary epithelia, basal cells and luminal cells. Data were presented as mean±SD. (n=5) (F) Total MRU number quantified by limiting dilution transplantation of lineage negative mammary cells from control and Krt14-cre Bcl11b^flox/flox mice. Data were normalized to the control MRU number. n=3, p<0.01 (G) Images and pie chart of the secondary transplant of the mammary outgrowths from control and Krt14-cre Bcl11b^flox/flox mice. (H) Whole mount images of control (adeno-cre Bcl11b^wt/wt) and mutant (adeno-cre Bcl11b^flox/flox) transplants. (I) Pie chart of limiting dilution transplant of adenovirus-cre mediated Bcl11b knockout mammary epithelia. (J) ELDA plot of limiting dilution transplant of adenovirus-cre mediated Bcl11b knockout mammary epithelia. N=3, p<0.01. See also figure S2.

To specifically interrogate Bcl11b null cells, we crossed Krt14-cre Bcl11b^flox/flox mice with an mTmG reporter mouse to label the cells that had expressed cre. Consistent with previously reported analyses of Krt14-cre mice, our control mammary gland (Krt14-cre Bcl11b^wt/wt mTmG) showed predominantly green ducts (cre+) and a small amount of red epithelia (cre-) (Fig. 2C, top panel). Interestingly, in the mutant (Krt14-cre Bcl11b^flox/flox mTmG) mice, the mammary glands showed a dramatic reduction in green cells with a striking increase in observable red epithelia, suggesting that these mammary ducts were made up of a significant number of cells that had escaped Bcl11b deletion by cre (Fig. 2C, bottom panel). When quantified by flow cytometry, the percentage of green cells in the mammary epithelia dropped from 84.4±5.0% (mean±SD; n=4) in the control mouse to 34.3±15.77% (mean±SD; n=5) in the mutant mouse (p<0.001; Fig. 2D, E).

In addition, the green cell reduction trend also held true in both basal and luminal mammary subpopulation (Fig. 2E). This data suggests that the Bcl11b null cells had a reduced ability to maintain the mammary gland compared to their wild type counterparts. Interestingly, Bcl11b knockout affected the luminal cells (which do not express detectable levels of Bcl11b) more dramatically than the basal cells. This suggests that the Krt14 positive cells had impaired ability to give rise to luminal cells after Bcl11b deletion, reminiscent of a defect in lineage commitment and multi-potency at some time during the mammary gland development.

We speculated that the impaired mammary gland of Bcl11b knockout mice resulted from a proliferation defect. Hence, we tested the mammary tree reconstitution capacity of Bcl11b mutant cells by limiting dilution transplant. Indeed, Bcl11b embryonic deletion caused a 75% reduction of the total MRUs in Bcl11b null mammary cells compared to wild-type mammary cells (Fig. 2F). When we passaged the primary outgrowths to secondary recipient mice, the Bcl11b mutant cells were unable to regenerate new mammary trees (Fig. 2G, S2A). Together, these data suggest embryonic deletion of Bcl11b impairs the long-term regeneration capacity, and results in a gradual exhaustion of the mammary epithelium.

To ask whether Bcl11b plays a role in maintaining the regeneration capacity of the adult mammary epithelia, adenovirus-cre was used to abruptly and efficiently knock-out Bcl11b due to high transient expression of cre (Russell et al., 2003). Since adenovirus will not integrate into the genome, the temporary cre activity will not permanently injure the cells. Taking advantage of this property of adenoviruses, mammary cells isolated from Bcl11b^wt/wt and Bcl11b^flox/flox mice were infected with adeno-cre-GFP. GFP+ mammary epithelial cells were isolated by sorting and transplanted into recipient mice (Fig. S2B). Analysis of the recipients’ mammary fat pads revealed a prominent reduction (~20 fold) of mammary repopulating activity in Bcl11b knockout versus wild-type cells (N=3, p<0.01; Fig. 2H–J and Table S1). This indicates that when Bcl11b is efficiently deleted in adult mammary gland, the regeneration capacity of adult MRUs is significantly impaired. These data reinforce the hypothesis that Bcl11b is necessary for the maintenance of adult mammary gland epithelium.

Bcl11b^high Cells are Highly Competent in Mammary Gland Reconstitution

Since Bcl11b is expressed in a subset of mammary cells and is required for mammary gland reconstitution, we were intrigued by the biology of Bcl11b^high cells. The proliferation capacity of Bcl11b^high CD49f^highCD24^medLin⁻ cells was determined using in vitro and in vivo assays. First, Bcl11b^high and Bcl11b^low CD49f^highCD24^medLineage⁻ cells were isolated from the Bcl11b^tdTomato/wt reporter mouse and subsequently cultured on matrigel using culture conditions enabling the growth of murine MRUs (Zeng and Nusse, 2010). Bcl11b^high cells formed ~8 fold more colonies than their Bcl11b^low counterparts (Fig. 3A), which indicates that cells with the highest expression of Bcl11b are greatly enriched for cells with the ability to form colonies of mammary epithelial cells. Furthermore, the Bcl11b^highCD49f^highCD24^medLin⁻ population has in vitro developmental potential, as colonies formed from a single Bcl11b^high cell clearly contain both basal and luminal epithelia in the single cell gene expression analysis (Fig. S3A, B). These data suggest that Bcl11b^high cells maintain their proliferation potential in the in vitro organoid cultures.

Fig. 3. Cells expressing high levels of Bcl11b are enriched for CD49f^highCD24^medLineage⁻ cells.

(A) Colony formation assay for CD49f^highCD24^medLin⁻ expressing high versus low levels of Bcl11b. 1K cells/well were seeded for each condition and cultured for a week. Number of colonies were counted and plotted in the bar graph. The data are presented as mean ±S.E.M (N=3, p<0.01) (B) ELDA plot of Limiting dilution transplant of the Bcl11b^high and Bcl11b^low CD49f^highCD24^medLineage⁻ populations (C,D,E) Single cell transplants of Bcl11b^high CD49f^highCD24^medLineage⁻ cells. A schematic diagram of single cell sorting and a representative Bcl11b^high single cell image in a Terasaki plate is shown in (C,D). A whole mount image of a representative singe cell transplant outgrowth is shown in (E) with the table showing the numbers of total injections and outgrowths (E). (F) The mammary outgrowths from Bcl11b^highCD49f ^highCD24^medlin⁻ cells were dissected and dissociated to single cells. Single cell suspensions from two distinct mammary outgrowths were passaged to 6 recipient mice for secondary transplant. Pie chart shows the tree coverage of the mammary fat pad. See also Figure S3.

Next, we investigated whether Bcl11b^high CD49f^highCD24^medLineage⁻ cells have reconstitution activity. We transplanted Bcl11b^high and Bcl11b^low CD49f^highCD24^medLineage⁻ cells into syngeneic mice. Consistent with the colony formation assay, Bcl11b^highCD49f^highCD24^medLineage⁻ cells displayed a 6-fold enrichment in their capacity to engraft mice over Bcl11b^low cells (p<0.05, N=3) (Fig. 3B, and Table S2). Single cell transplants of Bcl11b^high CD49f^highCD24^medLineage⁻ into syngeneic mice revealed an engraftment frequency of 1/9.5 (4 outgrowths of 38 single cell injections; Fig. 3C–E). In addition, the outgrowths generated from Bcl11b^high CD29^highCD24^medLineage⁻ cells can be efficiently passaged and reconstitute new mammary trees in a secondary transplant (Fig. 3F). These data suggest that Bcl11b directly regulates the long-term regeneration ability of mammary cells rather than indirectly affecting mammary gland development through a cell extrinsic mechanism.

Bcl11b^high Cells are Quiescent

For many somatic tissues, stem cells utilize a strategy of switching between an active proliferating state and a resting state to preserve long-term regeneration potential (Adorno et al., 2013; Fuchs, 2009; Wilson et al., 2008). Actively proliferating mammary progenitors were recently identified by the cell surface marker Procr. We were therefore intrigued by the relationship between Bcl11b^high and Procr^high mammary cells. Interestingly, FACS analysis showed that Bcl11b and Procr marked two distinct populations of CD49f^highCD24^medLin⁻ that do not overlap but are instead mutually exclusive (Fig. 4A), which is further supported by our single cell gene expression analysis (Fig 4B). This led us to dissect the cellular status of these two distinct populations. DNA content analysis of these two populations showed that 11% of Procr^high CD49f^highCD24^medLin⁻ were actively proliferating with cells in 4N fraction while less than 1% of the Bcl11b^high CD49f^highCD24^medLin⁻ were in the 4N fraction (Fig. 4A). This was further corroborated by an EdU incorporation assay (only labels replicating cells), which showed that Bcl11b^high cells did not incorporate any EdU (Fig. 4C). To determine whether Bcl11b^high cells are in G₁ or G₀, we co-stained a 12-week old mouse mammary glands with Bcl11b and the proliferation antigen, Ki67; Ki67 is expressed by actively proliferating cells throughout the cell cycle, but not by cells in G₀. Analysis of Bcl11b and Ki67 co-localization revealed that the two genes were rarely expressed in the same cells in both ducts and terminal end buds (Fig. 4D, E).

Fig. 4. Bcl11b^high cells in mammary gland are quiescent.

(A) Flow plot showing Bcl11b^high cells are distinct from Procr+ cells within CD49f^highCD24^medLin⁻gate. DNA content analysis of Bcl11b^high and procr+ CD49f^highCD24^medLin⁻. Procr+ CD49f^highCD24^medLin⁻contained more proliferating cells (11%) than Bcl11b^high CD49f^highCD24^medLin⁻ (0.67%). (B) Single cell gene expression analysis of Procr+ and Procr- basal cells. (C) EdU incorporation assay for Bcl11b^high cells. Adult mice (8–12weeks) were injected via IP with 1.25mg/20g EdU for one day and analyzed for Edu incorporation (red), Bcl11b (green), Krt14 (blue) 24 hrs later. (D,E) Immunofluorescence co-staining of Bcl11b and Ki67 in mammary gland frozen sections from a pubescent mouse showing mammary ducts and terminal end bud. Green: Bcl11b. Red: Ki67. Blue: DAPI (F) Single cell gene expression analysis showing that the mRNA expressions of Bcl11b and Ki67 are mutually exclusive. (G) Left panel: CD49f^highCD24^medLineage⁻ cells were sorted and stained with Pyronin Y and Hoechst 33342. The G₀, G1 and SG₂M cells were sorted and subjected to real time PCR. Right panel: real time PCR analysis of Bcl11b expression level in various cell populations. N=3 (H) Images showing day1 and day11 culture of PKH26 stained CD49f^highCD24^medLineage⁻ cells, and sorting gates of PKH26 retaining and non-retaining cells. (I) Real time quantification of Bcl11b expression in the PKH26 retaining (PKH26^high) and PKH26 low (PKH26⁻) populations. Data are presented as mean ±S.E.M (N=4, p<0.05). See also Figure S4

This was further bolstered by an unbiased single cell gene expression analysis on Bcl11b and Ki67 mRNA expression in adult mouse mammary gland (Fig. 4F and S4A). As a second hallmark of G₀ cells, quiescent stem cells have a low RNA content and can be identified as a Pyronin Y^low-Hoechst^low population (Gothot et al., 1997). To confirm Bcl11b^high cells are slow cycling, we sorted CD49f^highCD24^medLin⁻ that were Pyronin Y^high Hoechst^high(G₂), Pyronine^highHoeschst^low (G₁) or Pyronin Y^low-Hoeschst^low (G₀) and subjected them to real-time PCR analysis for Bcl11b. As predicted, Bcl11b was highly expressed in the Pyronin Y^low-Hoechst^low population, which are the putative G₀ cells (Fig. 4G). To ask whether Bcl11b expression is closely related to phenotypic slow cycling cells, we performed a long label-retaining assay in breast organoid cultures using the membrane dye PKH26 (Pece et al., 2010). Our result revealed Bcl11b is consistently enriched by ~3 fold in the PKH26^high population (Fig. 4H, I). Recently, in vivo long label retaining assay using an induced H2B-GFP identified a cell surface protein Cd1 as a marker for long label retaining cells (LRCs) (dos Santos et al., 2013); hence, we isolated Cd1+ and Cd1- basal cells. We found that Bcl11b was consistently enriched about 5 fold in the Cd1+ population compared with the Cd1- population (Fig. S4B). These data collectively demonstrate that the vast majority of Bcl11b^high cells are G₀ label-retaining cells with a reduced rate of proliferation, both in vivo and in vitro.

Bcl11b Functionally Regulates Quiescence of Mammary Epithelial Cells

As transcription factors often determine a specific expression profile, gain and loss of function assays were done to address whether Bcl11b itself regulates the cycling rate of CD49f^highCD24^medLin⁻ cells. We found that enforced expression of Bcl11b significantly repressed colony growth of CD49f^highCD24^medLineage⁻ cells (Fig. S4C, D). To eliminate the possibility that this was an overexpression artifact, we constructed an inducible Bcl11b vector using the Tet-on system to fine tune induced Bcl11b expression levels. In the murine mammary epithelial Comma D β cell line (Deugnier et al., 2006), which has a heterogeneous expression of Bcl11b (Fig. S5), induced Bcl11b expression dramatically repressed proliferation (Fig. S4E). Next, we investigated the effects of Bcl11b overexpression in CD49f^highCD24^medLineage⁻cells using a competition growth assay. We mixed pInducer-Bcl11b transduced CD49f^highCD24^medLineage⁻ cells (GFP+) with untransduced CD49f^highCD24^medLineage⁻ cells (GFP-) in a 1:1 ratio and cultured them in the presence or absence of Doxycycline (Fig. 5A–C). Doxycycline-induced Bcl11b expression strikingly reduced the percentage of GFP positive cells from 43% in the control cells to 11% in the Bcl11b-induced cells (p<0.01; Fig. 5A–C). The inhibition of CD49f^highCD24^medLineage⁻ cell growth is not an artifact of the construct or Doxycycline, as the overexpression of a control transcription factor (Nr2f1) caused an opposite phenotype of increased proliferation after induced expression (Fig. 5A–C). These data suggest that Bcl11b functionally suppresses mammary cell proliferation.

Fig. 5. Bcl11b functionally regulates mammary epithelial cells.

(A) Schematic diagram showing the strategy of the competition growth assay. Sorted CD49f^highCD24^medLineage⁻ cells were infected with a Bcl11b inducible construct. Infected cells (GFP+) and uninfected cells (GFP−) were mixed in a 1:1 ratio and cultured in the absence (Dox−) or presence (Dox+) of Doxycycline. Colonies were digested to single cells and analyzed by flow cytometry. (B) Competition growth assay with induced Bcl11b or Nr2f1 in the presence or absence of 100ng/ml Doxycycline for 3 days (C) Quantification of GFP percentage in Bcl11b or Nr2f1 competition growth assay. Data are presented as mean ±S.D. (N=3). (D) Images of the colonies arising from CD49f^highCD24^medLineage⁻ cells in a 3D culture from Krt14-cre Bcl11b^wt/wt and Krt14-cre Bcl11b^flox/flox mice. Images on day3 and day5 are shown. (E) The colony size of each individual colony was quantified and plotted over time. (F,G) Quantification of the sizes of colonies in a single cell culture assay from Krt14-cre Bcl11b^wt/wt and Krt14-cre Bcl11b^flox/flox mice. Single CD49f^highCD24^medLineage⁻ cells were sorted to 96-well plates (1cell/well) and cultured for colony formation. Data are presented as mean ±S.D. (p<0.05) (H) Induced deletion of Bcl11b with a Rosa26-creERT2 Bcl11b^flox/flox mTmG mouse. 8 week-old Rosa26-creERT2 Bcl11b^flox/flox mTmG mouse was administered with one dose 1mg/10g tamoxifen and left in husbandry for 3 weeks. Mammary glands were dissected and analyzed by flow cytometry. (I) Gene expression analysis of mouse basal cells at different stages of mouse pregnancy. Mammary glands from mice at different pregnancy stages p0, p6.5, p12.5, p18.5, lactation L1, were harvested and dissociated into single cells. Sorted basal cells were subjected to real time PCR analysis. (J) Hormone regulation of Bcl11b expression in mammary gland. Adult mice were ovariectomized and treated with vehicle (corn oil) or Estrogen (10ug)+Progesterone (1mg). Bcl11b expression was quantified by real time PCR, compared with pregnant mammary gland at stage p12.5. N=3 See also Figure S5,6.

To elucidate how Bcl11b is involved in the cell cycle, we performed microarray gene expression analysis of control and Bcl11b-induced cells. Gene enrichment analysis showed that cells expressing low levels of Bcl11b (DOX- uninduced) were significantly enriched for G₂ pathway factors and the pre-replicative complex subunits, which are required for DNA duplication (Fig. S6A). Moreover, genes required for G₁ progression, including Ccnd1, Ccne1, Ccne2, Cdk2, Cdk6, E2f1, were also strikingly enriched in Bcl11b-uninduced cells (Fig. S6B). Notably, mKi67, which distinguishes G₁ and G₀, was significantly down regulated by Bcl11b (Fig S6B). To ensure these results were not unique to the Comma D β cell line, this gene down-regulation pattern was confirmed by real time PCR of cells from FACS sorted primary CD49f^highCD24^medLin⁻ cells (Fig. S6E, F). This gene expression analysis suggests that Bcl11b is targeting genes that regulate the early G₁ phase of the cell cycle to prevent cell cycle progression. Consistent with this idea, we also observed up-regulation of cell cycle repressors such as Foxo1, Foxo3 and Foxo4, in the Bcl11b-induced cells (Fig. S6C). The Foxo factors repress cell cycle in part via repression of expression of Cyclin D (Schmidt et al., 2002). Notably, we found that Sall2, a gene required for serum-starvation induced proliferation arrest (Liu et al., 2007), was up-regulated upon Bcl11b expression (Fig. S6C). These data suggest that Bcl11b systemically and consistently inhibits G₁–G₂ cell cycle progression, which induces mammary cells into a quiescent state (Fig. S6D).

To ask whether Bcl11b plays a role in the maintenance of the quiescent mammary cells, we measured the proliferation of Bcl11b-loss of function CD49f^highCD24^medLineage⁻ cells (Krt14-cre Bcl11b^fl/fl) in an organoid assay. The Bcl11b loss of function cells produced colonies that were 50% larger compared to their wild type counterparts (Fig. 5D–G). When we induced the depletion of Bcl11b in vivo with a Rosa26-creERT2 Bcl11b^fl/fl mTmG mouse, the basal cells of the Bcl11b null population (GFP+) exhibited a significant expansion over that of the Bcl11b wild type population (GFP−) (Fig. 5H). These gain and loss of function experiments strongly suggest that Bcl11b is required for maintaining CD49f^highCD24^medLin⁻ cells in a quiescent state.

Since it has been recently reported that steroid hormones tightly regulate the reconstitution activity of mammary cells during estrus cycle and pregnancy (Asselin-Labat et al., 2010; Joshi et al., 2010), we asked whether hormones regulates Bcl11b. We first isolated mammary basal cells at various stages of mouse pregnancy, and subjected them to quantitative PCR analysis. Interestingly, we found a prompt increase of Ki67 expression at early pregnancy followed by a deep decline in the later pregnancy. In contrast, Bcl11b expression gradually decreased throughout the full term of pregnancy, suggesting that epithelial cells downregulated Bcl11b to allow drastic proliferation followed by differentiation (Fig 5I). To ask whether the decrease of Bcl11b during pregnancy is due to elevation of hormones, we treated ovariectomized mice with estrogen and progesterone. Indeed, Bcl11b’s level was consistently downregulated to a similar level in pregnancy (Fig. 5J). These data suggest that steroid hormones act to downregulate Bcl11b in the Bcl11b high cells to promote the expansion of mammary epithelial cells.

Bcl11b Regulates CD49f^highCD24^medLin⁻ cells Homeostasis via Cdkn1a and Cdkn2a

The G₁ phase of the cell cycle is tightly regulated by cyclin dependent kinase inhibitors (CDKIs) through inhibition of kinase activity of the cyclinD/CDK4/6 and cyclinE/CDK2 complexes. Certain CDKIs have been reported to reduce proliferation of stem cells in certain tissues (Cheng et al., 2000; Kippin et al., 2005; Matsumoto et al., 2011; Zou et al., 2011). We sought to investigate the possibility that Bcl11b inhibits proliferation by regulating CDKIs. We performed real time PCR for all the CDKIs using samples from the competitive growth assay. We found that upon Bcl11b induction, Cdkn1a/p21 mRNA was up-regulated, while other CDKIs showed only minor or insignificant changes (Fig. 6A, B). p21 has been shown to retard both hematopoietic stem cell and neural stem cell proliferation (Cheng et al., 2000),(Kippin et al., 2005; Marques-Torrejon et al., 2013). Recently, using a Cdk2 sensor, p21 was shown to be responsible for reducing proliferation in multiple epithelial cell lines (Spencer et al., 2013). We reasoned that if Bcl11b exerted cell cycle repression through p21, knockout of p21 should impair Bcl11b’s ability to repress CD49f^highCD24^medLin⁻ proliferation. To test this, we performed a competition growth assay with CD49f^highCD24^medLin⁻ using a p21 knockout mouse. Induction of Bcl11b did not repress cell growth as efficiently in the p21 mutant cells as in wild-type cells (Fig. 6C). The reduced proliferation of the GFP positive cells was partially rescued (11.74±1.1% GFP positive cells in Dox+ WT vs 19.0±2.0% GFP positive cells in Dox+ p21 mutant, p<0.01). This data suggests that p21 is one of the pathways that Bcl11b regulates to control the cell cycle.

Fig. 6. Bcl11b regulation of CD49f^highCD24^medLineage⁻ proliferation and engraftment via Cdkn1a(p21) and Cdkn2a(p16).

(A) Schematic Diagram showing the process of real time PCR analysis of Bcl11b induced and uninduced populations in colonies formed by CD49f^highCD24^medLin⁻. (B) mRNA expression levels of selected CDKIs (p21, p16, p19, p27, p57) after induced Bcl11b expression in CD49f^highCD24^medLineage⁻ cells in 3D culture. GFP+ cells from Dox-(0ng/ml) and Dox+(50ng/ml) samples were sorted for real time gene expression analysis. (C) Competition growth assay with Bcl11b induction in p21 knockout mice. Note: p21 knockout partially rescues Bcl11b-mediated repression of proliferation. N=3 (D) Gene expression plot of Cdkn2a(p16) versus Bcl11b from the currently available 10,000 mouse microarrays. Each dot represents one microarray dataset. (E) Gene expression plot of p53 versus Bcl11b from the currently available 10,000 mouse microarrays. Each dot represents one microarray dataset. (F) Schematic diagram showing the design of the p16 shRNA and p53 shRNA rescue assays. Basal1 cells from Bcl11b^flox/flox mice were sorted and transduced with p16 or p53 shRNA lentivirus and cultured for a week on 3D matrigel. Mammary colonies were then dissociated and transduced with adeno-cre-RFP and cultured overnight. The GFP and RFP double positive cells were then sorted and transplanted into NSG recipient mice. (G) Limiting dilution transplants of adeno-cre mediated Bcl11b knockout basal cells infected with p16 shRNA or p53 shRNA vectors. In contrast to p53 shRNA, which did not show significant difference in repopulation frequency, p16 shRNA did rescue the engraftment by ~3 fold. (N=3, P_pSIH vs p16 shRNA<0.05, P_{pSIH vs p53shRNA}=0.63).

In certain somatic tissues it has been shown that a significant proportion of the tissue stem cells are slow cycling. For example, HSCs need to remain in the slow cycling state to maintain self-renewal ability (Essers et al., 2009; Wilson et al., 2008). Over-proliferation after various genetic perturbations or tissue damage by insults such as toxins, often leads to stem cell exhaustion in vivo and cellular senescence in vitro, mediated by either the p53 or INK4a (Cdkn2a) pathway (Beausejour et al., 2003; Lin et al., 1998; Schmitt et al., 2002). We asked whether Bcl11b knockout-mediated mammary exhaustion is dependent on the p53 or Cdkn2a pathways. To test whether Bcl11b has any correlation with Cdkn2a and p53 expression, we applied a boolean analysis (Sahoo et al., 2008) to a pool of >10,000 published mouse microarray datasets from GEO. Interestingly, the expression of Bcl11b was mutually exclusive to that of both p53 and Cdkn2a (Fig. 6D–E; S_{Bcl11b-Cdkn2a}=9.45>3, ER_{Bcl11b-Cdkn2a}=0.05<0.1; S: Boolean Statistics, ER: Error Rate), suggesting that there might be a direct or indirect molecular link between Bcl11b and the Cdkn2a or p53 pathways.

To test whether Cdkn2a and/or p53 are biologically important for Bcl11b function, we constructed lentiviral p53 shRNA and Cdkn2a shRNA vector (Adorno et al., 2013), and knocked down these two genes in the Bcl11b knockout (adenovirus-cre mediated) mammary epithelia. We subsequently transplanted these cells into recipient mice. Limiting dilution transplant analysis showed that p53 knockdown did not change the MRU frequency compared with Bcl11b knockout alone. However, Cdkn2a knockdown significantly increased MRU frequency by ~3 fold (p<0.05; Fig. 6F, G and Table S3). This data suggests that the MRU defect in the Bcl11b mutant mice is mediated at least in part by the Cdkn2a pathway. Since Cdkn2a has been linked to impaired stem cell function and senescence add Park nature Bmi1 paper reference (Park et al., 2003; Sato et al., 2015; Sharpless, 2004), our data suggest loss of Bcl11b might result in MRU and mammary gland exhaustion via the Cdkn2a senescence pathway.

Discussion

In this study, we have identified a dormant population of mammary epithelial cells that are important for normal mammary gland homeostasis. These cells rely on Bcl11b to maintain quiescence. Bcl11b is necessary for the long-term maintenance of the mammary gland at least in part via prevention of a Cdkn2a-mediated mammary epithelial cell exhaustion. Most other genes that have been linked to maintenance of the mammary epithelium such as Bmi1 and Bcl11a are expressed by a larger percentage of both the luminal and basal cells in the mammary epithelium (Khaled et al., 2015; Pietersen et al., 2008). By contrast, only a small number of mammary epithelial cells that express the basal cytokeratin Krt17 also express detectable levels of Bcl11b mRNA or protein as measured by immunochemistry (Fig.1E) or quantitative single cell RT-PCR (Fig.1B). These results suggest that basal epithelial cells must cycle into a quiescent, Bcl11b^high state in order to maintain a normal mammary gland.

In many somatic organ systems, stem cells have to fulfill seemingly opposing demands of rapid proliferation to produce the mature, functional cells of the organ/tissue, while also cycling slowly in order to preserve long-term regenerative potential(Harmes and DiRenzo, 2009). The mechanism of maintaining a slow cycling quiescent state is the key to understand how cells coordinate these two demands. In the hematopoietic system, HSCs rely on p21 (Cheng et al., 2000) and p57 (Matsumoto et al., 2011) to be maintained at the quiescent state and can be activated by IFNa (Essers et al., 2009). In the mammary gland, slow cycling mammary cells with the ability to gain extensive proliferative capacity have been identified by many approaches including BrdU-labeling (Bussard et al., 2010; Smith, 2005), PKH26 labeling in vitro (Pece et al., 2010), and H2B-GFP pulse chasing (dos Santos et al., 2013). However, neither the intrinsic molecular network that regulates the proliferation state of mammary epithelial cells or the biological consequences of the slow cycling state is fully understood. In this study, we have identified the nuclear factor Bcl11b as a central regulator of the proliferation state of CD49f^highCD24^medLineage⁻ cells. Loss of function studies show that Bcl11b plays an essential role in maintaining CD49f^highCD24^medLineage⁻ cells’ regenerative capacity. At the same time, Bcl11b induces basal epithelial cells to enter a ki67 negative, G₀ state. We further show that loss of Bcl11b results in exhaustion of the mammary epithelium at least in part dependent on Cdkn2a, which is known to regulate senescence programs in many tissues including stem cells and contribute to tissue degeneration in aging in many tissues including the mammary gland (Adorno et al., 2013; Krishnamurthy et al., 2006; Pietersen et al., 2008; Pustavoitau et al., 2016). In contrast, induced Bcl11b expression activates the G1 cell cycle inhibitor p21. Thus, our data identify Bcl11b as a crucial component of the molecular mechanism by which mammary epithelia are maintained long term, in part by evading the Cdkn2a senescence program and entering the Cdkn1a dependent quiescence program.

There are several models of the adult mammary epithelial cell hierarchy (Rios et al., 2014; Van Keymeulen et al., 2011), and our results could be explained by any of the postulated mammary gland hierarchy models. For example, it is possible that the Bcl11b high cells are the true stem cell. A second model is that induction of Bcl11b induces a Krt17⁺ cell to enter a stem cell state. A third model is that induction of Bcl11b induces any mammary epithelial cell to enter a “stem cell state’. There are many other possible models. Regardless of which mammary gland hierarchical model is correct, our data clearly show that Bcl11b is essential for long term maintenance of the ductal epithelium and maintenance of long-term proliferation capacity.

Normal stem cells are thought to be targets for oncogenic transformation and stem cell programs are often used by cancer cells (Behjati et al., 2014; Jamieson et al., 2004; Jan et al., 2012; Shimono et al., 2009). As quiescent mammary cells are a long-lived population, they could be a prime targets for successive and accumulative genomic alterations, ultimately causing malignant transformation. Therefore it will be interesting to dissect the dynamics of this quiescent population during carcinogenesis to determine whether these cells play a role in tumorigenesis. For example, dormant cancer cells are often resistant to cytotoxic agents that mainly target proliferating cells, resulting in increased risk of cancer relapse and metastasis; thus it is imperative to determine whether BCL11B regulates quiescence in breast cancer and contributes to therapy resistance. If so, the elucidation of how BCL11B regulates quiescence could reveal novel therapeutic targets for breast cancer.

STAR Methods

CONTACT FOR REAGENT AND RESCOURCE SHARING

Further information and requests for reagents may be directed to, and will be fulfilled by the corresponding author Michael F. Clarke at Institute for Stem Cell Biology and Regenerative Medicine (ISCBRM), Stanford University. Email: mfclarke@stanford.edu

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animal studies

The Bcl11b^flox/flox mice (C57/BL6 background) were generously provided by Mark Leid’s lab, and were described previously (Golonzhka et al., 2009). The Bcl11b^tdtomato knock-in mice were generously provided by Pentao Liu’s lab, and were described previously (Li et al., 2010). The mTmG (stock number 007676), Krt14-cre mice (stock number 018964) Cdkn1a KO mice (stock number 016565) were purchased from Jackson Laboratory. For analysis of pubertal mammary gland, 4–6weeks female mice were used. For experiments of Bcl11b KO whole mount, limiting dilution transplant, MRU sorting, competition growth assay and immunofluorescence staining, adult female mice (8–16weeks) were used. For each experiment, the age of the female mice used was indicated in the text. All animal procedures were conducted in accordance with a protocol approved by the Stanford University APLAC committee. Mice were maintained in house under aseptic sterile conditions. Mice were administered with autoclaved food and water.

Cell lines

Comma D beta cell line was kindly provided by Dr. Medina, and was described previously (Campbell et al., 1988). Comma D beta cell line was cultured in DMEM-F12 (Invitrogen) supplemented with 2% of Fetal Bovine Serum (Hyclone), 1% PSA (Invitrogen), 10ng/ml EGF(BD) and 5 μg/ml Insulin (Sigma), at 37 degree with 5% CO2.

METHOD DETAILS

Tissue Processing and Flow Cytometry

The 2^nd,3^rd,4^th pair of mammary glands from 8–16 week old virgin C57/BL6 mice were dissected and processed according to the published protocol (Prater et al., 2013) with minor revision. Mammary gland were manually and mechanically minced into 1mm size and digested with 0.5mg/ml Collagenase and 50U/ml hyaluronidase (Stem Cell Technology) for 2 hrs with gentle pipetting every 30mins. Digested mammary homogenate was spun down at 1500 rpm in Allegra^TM 6KR Centrifuge (Beckman Coulter), followed by 5ml ACK lysing buffer (Lonza) treatment for 5min on ice to remove erythrocytes. Then, mammary cells were digested by 5ml 0.25% Trypsin-EDTA (Gibco) for 2–5min, followed by brief DNase I (Worthington) plus Dispase (Stem Cell Technologies) digestion. Dissociated mammary cells were filtered by 40 μm strainer to obtain single cell suspension. For FACS analysis, mammary single cells were stained with CD45 (Biolegend), CD31 (Biolegend), Ter119 (Biolegend), CD49f (Biolegend), EpCAM (Biolegend), CD24 (Biolegend), CD14(Biolegend), CD1d(Biolegend), Procr (Ebioscience) with appropriate conjugated fluorophores for 15min on ice. Then cells were washed and resuspended in HBSS+2%FBS+PSA+DAPI (1ug/ml) at a density of 5 million/ml. Stained samples were analyzed and sorted on FACS Aria II (BD Bioscience) with 100um or 130um nozzle.

Mammary Colony Formation Assay

L1-wnt3A feeder cells (generous gift from Roel Nusse lab) were administered a 40Gy dose of X-ray irradiation, and plated in 96 well plate at a density of 10k/well for 4hrs to allow stable attachment. 30ul of growth factor reduced Matrigel (BD Bioscience) was then overlaid on top of the feeder cells and solidified at 37degree for 10min. 1K/well CD49f^highCD24^medLin⁻ cells were resuspended in 200ul culture media (DMEM F12+2% FBS+PSA+B27+10mM HEPES) supplemented with EGF (10ng/ml, BD Bioscience), Rspo1(250ng/ml, R&D), ROCK inhibitor Y27632 (10uM, Sigma), and Noggin (100ng/ml, R&D), and were overlaid on top of the matrigel. Plate was maintained in 37 degree incubator at 5% CO2 for 1–2 weeks. Single cell colony formation analysis was performed in the same condition.

Real Time PCR

500–2000 primary mammary cells or Comma D beta cells of various mammary populations were directly sorted into Eppendorf tube containing 400ul Trizol (Life Technologies). RNA was extracted according to the manufacturer’s instruction with addition of ultrapure glycogen (Life Technologies) as carrier. RNA was reverse transcribed to cDNA using SuperScript III First Strand Synthesis kit (Life Technologies) according to the manufacturer’s instructions. cDNA was preamplified 15–20 cycles according to the cell number using SybrGreen mastermix (Applied Biosystems) and target gene primer pool designed by IDT (Integrated DNA Technologies). Preamplified cDNA was then subjected to the real time PCR for specific gene target according to manufacturer’s instructions using 7900HT Real Time PCR system (Applied Biosystems). Data were analyzed by SDS2.4 software and Excel. Sybrgreen primers are listed below: Bcl11b: Forward AGGAGAGTATCTGAGCCAGTG, Reverse GTTGTGCAAATGTAGCTGGAAG; p21: Forward CTTGCACTCTGGTGTCTGAG Reverse GCACTTCAGGGTTTTCTCTTG; Ki67: Forward TGCCCGACCCTACAAAATG Reverse GAGCCTGTATCACTCATCTGC; p16: Forward GTGCGATATTTGCGTTCCG Reverse TCTGCTCTTGGGATTGGC; p57: Forward CGCAAACGTCTGAGATGAGTTAG Reverse TCCTGCTACATGAACGAAAGG; p19: Forward CTTCATCGGGAGCTGGTG Reverse AGGCATCTTGGACATTGGG; Cdc25a: Forward TCCATCCCAGTCTCTATCCC Reverse TCAAATCCTGATGCTTCCCAG; Cdk1: Forward TGCAGGACTACAAGAACACC Reverse GCCATTTTGCCAGAGATTCG; Cdk2: Forward GCATTCCTCTTCCCCTCATC Reverse GGACCCCTCTGCATTGATAAG; Cdc25c: Forward GCAAACCTAAGCATTCTGTCG Reverse CAGAGGTCCAGATGAATCCAAG; Wee1: Forward GATCTCCTTTTGCAAGTTGGC Reverse CCAGTCATCTTCATCTCCTTCC; Foxo1: Forward CTACGAGTGGATGGTGAAGAG Reverse TGTGAAGGGACAGATTGTGG; Foxo4: Forward CTACTTCAAGGACAAGGGTGAC Reverse TGCAAGGACAGGTTGTGAC; Top2a: Forward AGTCAGACGTGAGCAGTAATG Reverse CTTCATCCTCATCCTTCTCATCC

Single Cell Gene Expression Analysis

Single-cell gene-expression experiments were performed using Fluidigm's M48 quantitative PCR (qPCR) DynamicArray microfluidic chips (Fluidigm) as described (Dalerba et al., 2011). Single cells were sorted by FACS into individual wells of 96-well PCR plates. Each 96-well plate was preloaded with 5 μl/well of CellsDirect PCR mix (Life Technologies) and 0.1 μl/well (2 U) of Superase In RNase-inhibitor. Following single-cell sorting, each well was supplemented with 1 μl of SuperScript-III RT/Platinum Taq (Life Technologies), 1.5 μl of Tris-EDTA (TE) buffer and 2.5 μl of a mixture of 48 pooled TaqMan assays (Applied Biosystems) containing each assay at 1:100 dilution. Single-cell mRNA was directly reverse transcribed into cDNA (50 °C for 15 min, 95 °C for 2 min), pre-amplified for 20 cycles (each cycle: 95 °C for 15 s, 60 °C for 4 min) and diluted 1:3 with TE buffer. A 2.25 μl aliquot of amplified cDNA was then mixed with 2.5 μl of TaqMan Universal PCR Master Mix (Applied Biosystems) and 0.25 μl of Fluidigm's “sample loading agent,” then inserted into one of the chip “sample” inlets. Individual TaqMan assays were diluted at 1:1 ratios with TE. A 2.5 μl aliquot of each diluted TaqMan assay was mixed with 2.5 μl of Fluidigm's “assay loading agent” and individually inserted into one of the chip “assay” inlets. Samples and probes were loaded into M48 chips using an IFC Controller HX (Fluidigm), then transferred to a BioMark real-time PCR reader (Fluidigm) following manufacturer's instructions. The Taqman probes used in this paper were purchased from Applied Biosystems (Life Technologies) and are listed below: Krt 17: Mm01306857_mH, ActB: Mm00607939_s1, Bcl11b: Mm00480516_m1, Cldn3: Mm01196233_s1, Gapdh: Mm03302249_g1, Krt8: Mm00835759_m1, Krt18: Mm01601702_g1, EpCAM: Mm01227384_m1, Krt19: Mm00492980_m1, Itga6: Mm01333831_m1, Itgb1: Mm00690415_g1, Cd44: Mm01277161_m1, Acta2: Mm00725412_s1, Snai2: Mm00441531_m1, Gata3: Mm00484683_m1, Foxo1: Mm00490671_m1, Axin2: Mm01265783_m1, Ezh2: Mm00468449_m1, Bmi1: Mm03053308_g1, Dll1: Mm00432841_m1, EGFR: Mm01187857_m1, Krt5: Mm00503549_m1, Mme: Mm01285048_m1.

Western Blot

100k Comma D beta cells were lysed by 2X Laemmli SDS sample buffer (100mM Tris pH6.8, 10% glycerol, 4% SDS, 0.01% Bromophenol Blue), and boiled on heat block at 100 degree for 15min. Samples were loaded to 4%–20% precast gradient gel (Bio-Rad) and electrophoresed at 200v for 45min, and transferred to Odyssey® nitrocellulose membrane (LI-COR). Membrane was blocked by PBS+0.1% tween 20+5% Non-fat dry milk for 1hrs RT, and then subjected to primary antibody staining beta Actin (Santa Cruz) 1:500, Rat anti Bcl11b 25B6 (Abcam) 1:1000 1hrs RT or 4 degree overnight. Membrane was then washed by PBST (PBS+0.1% Tween 20) 4X5min, and stained with secondary antibodies HRP-Donkey anti mouse, rat, or rabbit (GE Healthcare) 1:10,000 RT 1hr. Membrane was subsequently washed 4X10min by PBST and developed using SuperSignal® West Dura Extended Duration Substrate (Thermo Scientific) and KODAK X-OMAT LS film (KODAK).

Immunofluorescence

For frozen section, mammary gland was dissected and immediately fixed by 10% formalin for 1hrs followed by PBS washing and 30% sucrose infiltration overnight. The fixed mammary tissue was then embedded in O.C.T. compound (Tissue-Tek) and frozen in −80 degree. Frozen tissue block was sectioned to 14 um at −35degree using Cryostat Leica CM3050 S (LEICA). For immunofluorescence assay, frozen sections were hydrated with PBS for 10min RT. Antigen was retrieved in citrate buffer (10mM Sodium Citrate, 0.05% Tween 20, pH6.0) for 20min at 100 degree in microwave. Sections were blocked with TBS+2% BSA+5% Donkey serum+0.1%Triton X100 for 1 hr RT, and then stained with primary antibody Rat anti-Bcl11b (Abcam), Ki67(Abcam), Krt14 (Covance) overnight, followed by 3X washing by TBST and secondary antibody staining Donkey anti-Rat, rabbit 1:200 (Jackson ImmunoResearch) RT 1hr. After 3X TBST washing and brief DAPI staining (1ug/ml), sections were mounted with Fluoromount® Aqueous Mounting Medium (Sigma). For paraffin section, mammary tissue were fixed by 10% formalin for 2hrs RT and dehydrated by gradient ethanol (70%, 95%,100%) solution. Dehydrated tissue was infiltrated by Xylene solution and embedded with paraffin. Tissue block was sectioned to 10um using Rotary Microtome Leica RM2255 (LEICA). To do the immunofluorescence, paraffin section was de-paraffinized using Xylene and rehydrated with gradient (100%, 95%,70%,0%) ethanol solution and subjected to immunofluorescence staining described above.

Whole Mount

Mammary gland was fixed by Carnoy’s fix (60% Ethanol, 30% CHCl3,10% Acetic Acid) for 4hrs. Fixed tissue was washed with a gradient of ethanol (100%, 95%, 70%, 0%), and then stained with carmin-alum staining solution overnight. The next day, tissue was washed with PBS and a gradient of Ethanol (70%, 95%, 100%). Fat tissue was removed by Xylene treatment, and tissue was mounted using Permount® (Fisher Scientific) for long-term preservation.

Transplantation

For Krt14-cre mediated Bcl11b knockout transplant assay, isolated mammary cells from 12-week-old Bcl11b^flox/wt and Krt14-cre Bcl11b^flox/flox mice were lineage depleted using mammary epithelium enrichment kit (Stem Cells Technologies) according to the manufacturer’s instructions. The depletion efficiency and epithelium percentage were determined by flow cytometry. Suspended cells were pelleted and resuspended with injection media (DMEM F12+50% Matrigel+1%PSA) to 50k/5ul, and serially diluted to 25k/5ul, 12.5k/5ul and 2.5k/5ul. Cells were injected as published. 3-week-old NSG recipient mice (Jackson Laboratory) were anesthetized with Ketamine/Xylazine at a dose of 100mg/kg for Ketamine and 10mg/kg for Xylazine. The inguinal rudimentary tree was removed and 5ul of cell suspension was injected onto the residual fat pad using 25ul Hamilton Syringe. Mice were maintained in aseptic sterile condition for 8–12 weeks before whole mount analysis. For secondary transplant, mammary fat pad at 100K dilution, which exclusively gave rise to full tree, were collected and processed to single cell suspension. Mammary epithelia were sorted based on CD24 and CD49f, and 5k epithelia/pad was injected to 3-week-old NSG mice. The MRU frequency and confidence interval were determined by L-Calc (Stem Cell Technologies) or ELDA (http://bioinf.wehi.edu.au/software/elda/).

For Adenovirus-cre mediated Bcl11b knockout mammary cells transplant, 8–12 week old Bcl11b^wt/wt and Bcl11b^flox/flox mice were lineage depleted using Mammary Epithelium Enrichment Kit (Stem Cell Technologies) and resuspended in culture media (DMEM F12+2%FBS+PSA+10mM HEPES+EGF 10ng/ml+Rspo1 250ng/ml+Noggin 100ng/ml+Rock inhibitor Y27632 10μM) to 10 million cells/ml/well in 24-well non-adherent culture plate (Corning). Cells were transduced with Adenovirus-cre-GFP (Vector Biolabs) at MOI 10 overnight. The next day, cells were pelleted and treated with 400ul TrypLE^TM Select (Gibco)/eppendorf tube at 37 degree for 20mins. Dissociated cells were neutralized with HBSS+2% FBS+PSA and stained with CD45 (Biolegend), CD31 (Biolegend), Ter119 (Biolegend), CD49f (Biolegend), EpCAM (Biolegend). GFP+ epithelia were sorted and then subjected to limiting dilution transplant to NSG recipient mice. Recipient mice were maintained for 10 weeks before analysis. For p53 shRNA and p16 shRNA rescue assay, 50k basal cells/condition from Bcl11b^flox/flox mouse mammary gland were sorted and transduced with pSIH-GFP control virus, pSIH-GFP-p53shRNA virus and pSIH-GFP-p16shRNA virus respectively at MOI 20, and cultured in 96 well plate at a density 2k/well using the colony formation assay protocol above. Colonies were cultured 7–10 days and dissociated by dispase 1mg/ml (BD Bioscience) for 30min followed by TrypLE^TM Select (Gibco) 37degree for 20min to obtain single cells. Single cells were resuspended in 1ml culture media (DMEM F12+2%FBS+PSA+10mM HEPES+EGF 10ng/ml+Rspo1 250ng/ml+Noggin 100ng/ml+Rock inhibitor Y27632 10uM) and transduced by adenovirus-cre-RFP at MOI 10 and cultured in 24-well non-adherent culture plate (Corning) overnight. The next day, cells were pelleted and dissociated with TrypLE^TM (Gibson), and GFP and RFP double positive cells were sorted and transplanted to NSG recipient mice with limiting dilution. Mice were maintained for 6–8 weeks before whole mount analysis.

For Bcl11b^high CD49f^highCD24^medLin⁻ single cell transplant assay, mammary gland from 8–12 week old Bcl11b^td/wt mouse was dissected and processed to single cell suspension, followed by staining and flow cytometry analysis according to the protocol above. Bcl11b^high CD49f^highCD24^medLin⁻ was gated according the tdtomato level of the mature myoepithelial cells. Bcl11b^high CD49f^highCD24^medLin⁻ cells were double sorted to 48-well Terazaki plate (Sigma), containing 5ul culture media/well. Sorted cells were evaluated under the microscope and wells containing single cells were recorded. 5ul matrigel (BD Bioscience) was then added to each well before transplanting to 3-week old C57/BL6 recipient mouse. Recipient mice were maintained for 10 week before whole mount analysis

shRNA knockdown

shRNA against p53 “GTACATGTGTAATAGCTCC” and shRNA against p16 “CATCAAGACATCGTGCGATAT” were cloned into pSIH-GFP Lentiviral vector (Systembio) according to manufacturer’s instructions. Viruses were packaged as described (Adorno et al., 2013). Knockdown efficiency was determined by western blot with Comma D β cell line.

EdU Incorporation Assay

C57/Bl6 mice of 12 week old were injected via IP with 1.25mg/10g EdU. 24hrs after injection, mammary glands were dissected and subjected to EdU label analysis. Click-it reaction and costaining with Bcl11b antibody were performed using Click-iT® Assay Kit(Life Technologies) according to manufacturer’s instructions.

PKH26 Long Label Retaining Assay

CD49f^highCD24^medLin⁻was sorted and stained with PKH26 according to manufacturer’s instructions. Specifically, 10k sorted cells were resuspended in 2μM PKH26 and stained for 5mins with periodic mixing before quenching with equal volume of serum. Then stained cells were cultured on growth factor reduced matrigel with L1 feeder cells for 11days. Colonies were dissociated with dispase (1mg/ml) and trypLE. PKH26 retaining cells and PKH26- cells were sorted by flow cytometer, and then subjected to quantitative real time analysis.

Microarray

pInducer Bcl11b Comma D β cell line was treated with 0ng/ml (control) and 100ng/ml Doxycycline overnight. 200k cells from Sca1+ and Sca1- populations of both conditions were sorted. RNA was extracted by RNeasy plus micro kit (Qiagen) according to manufacturer’s instructions and quantified by Agilent 2100 Bioanalyzer. Library preparation, hybridization and scanning were all performed by Stanford protein and nucleic acid facility (PAN facility). Affymetrix 3'IVT Express protocol was used to generate biotinylated cRNA from 50–500ngs of total RNA. The fragmented and labeled cRNA was hybridized to mouse genome array MOE430 2.0 with 16hr (overnight) hybridization at 45degC at 60rpm in an Affymetrix GeneChip Hybridization Oven 645. The arrays were then washed and stained in an Affymetrix GeneChip Fluidics Station 450. The arrays were scanned using the Affymetrix GeneChip Scanner 3000 7G. Scanned microarray images were imported into Gene Chip Operating Software (GCOS, Affymetrix) to generate signal values, followed by analysis with BRB-ArrayTools (Biometric Research Branch) using default settings. Heatmap of a collection of genes was generated by Cluster and TreeView (Eisen et al., 1998).

QUANTIFICATION AND STATISTICAL ANALYSIS

The variances between groups were first tested by F-test. Then differences between groups were analyzed using unpaired student’s t test with equal variance or unequal variance according to the F-test. Error bars represent standard deviations (± SD) or standard error (±SEM) as indicated in each specific experiment. Each experiment in the paper has been replicated for at least 3 time. In each experiment, N represents the exact times of independent experiments, n represents the sample volume (mouse number, cell number). Statistical parameters and tests are reported in the Figures and corresponding Figure Legends. For limiting dilution analyses, the frequency of mammary repopulating unit was calculated using ELDA software(Hu and Smyth, 2009). Expected frequencies are reported, as well as the 95% confidence intervals (lower and upper values are indicated). No statistical method was used to predetermine sample size, experiments were not randomized and investigators were not blinded to allocation during experiments. Investigators were blinded when assessing outcome of animal experiments (only has ear tag# for identification) during analysis of Bcl11b knockout phenotypes. No data or subjects were excluded from the statistical analysis.

DATA AND SOFTWARE AVAILABILITY

Microarray data for induced Bcl11b expression in Comma D beta cells was deposited to GEO with accession code: GSE62059

Highlights.

• A subset of mammary basal epithelial cells express high levels of Bcl11b

• Loss of Bcl11b impairs mammary gland development and regenerative capacity

• Bcl11b^high cells are quiescent but have potent regenerative activity in transplants

• Bcl11b interacts with cell cycle regulators to induce a quiescent state