Molecular mechanisms of asymmetric divisions in mammary stem cells

Par3 regulates morphogenesis.

Par3 is a scaffolding protein deputed for the spatial localization of many signaling effectors and their recruitment to the apical side. In 2009, Macara et al used limiting dilution transplantation (see Box 1 for a definition) to analyze the relevance of Par3 for mammary gland development [75]. By these means, they proved that Par3 is essential for the complete reconstitution of the mammary gland by a population of mammary cells in a virgin host. In particular, the mammary outgrowths derived from cells interfered for Par3 exhibited defects in the process of TEB elongation [75]. Par3 depletion also leads to the accumulation of cells co‐expressing both luminal (K8) and basal (K14) markers throughout the whole ducts; these cells might represent a population of immature bipotent progenitors [61], [65], [76]. This occurrence could be an indication of increased number of symmetric rather than asymmetric divisions, with the net result of an unbalanced differentiated progeny. Indeed, Par3 silencing reduces the number of myoepithelial cells and, at the histological level, leads to the formation of highly disorganized ducts at the basal layer [75].

Disruption of the Par3/aPKC complex leads to breast cancer.

Par3 binding to aPKC seems to be crucial for organ morphogenesis, as the Par3 splice variant lacking the aPKC‐binding domain is not able to rescue the above‐described phenotype. Indeed, deregulation or mislocalization of aPKC is responsible for epithelial mis‐organization [77], [78], as well as increased invasion of breast cancer cells and poorer prognosis [79]. The mechanism downstream of the Par3–aPKC axis that leads to tumor development seems to progress toward the activation of Stat3 and induction of ECM disaggregation [80]. Furthermore, in an in vitro system of cultured MCF10a acini, constitutive activation of the ErbB2 receptor, which mimics an event that occurs in 15–20% of human breast cancers [81], was shown to lead to severe disruption of apical polarity, as a consequence of the association of ErbB2 with aPKC and Par6 and decreased Par3–aPKC interaction [82].

Par3L loss impairs MaSCs' self‐renewal via regulation of Lkb1 kinase.

Another member of the Par proteins that has been studied for its role in the self‐renewal of mammary stem and progenitor cells is Par3‐like (Par3L). In vivo, primary cells harboring Par3L‐shRNA possess lower reconstitution abilities than WT cells in limiting dilution transplantation assays and, in vitro, form lower number of colonies [83]. Importantly, putative MaSCs interfered for Par3L and marked by GFP positivity, isolated on the basis of the expression of the SHIP promoter [84], [85], were capable of forming colonies that retained a significantly lower number of GFP⁺ cells compared to WT MaSCs, suggesting that Par3L depletion causes loss of mammary stem cells. As also reported for Par3 RNAi [75], this failure of self‐renewal was especially evident in the myoepithelial lineage, with a strong decrease in the generation of K14⁺ differentiated cells. In the case of Par3L‐shRNA however, the authors do not report any accumulation of intermediate progenitors [83], indicating that loss of Par3L might not favor symmetric self‐renewal. Interestingly, the mechanism proposed by the authors for the maintenance of MaSC functions involves the suppression of Lkb1 kinase (the human homolog of C. elegans Par4) activity. Lkb1 is the only member of the family of Par proteins with a recognized tumor suppressor role [86], [87]. It appears that in securing the mechanisms preventing tumor development, many genes with a role in tumor suppression also preserve stem cell state. High levels of Lkb1, for instance, were shown to be deleterious for MaSCs which became unable to form colonies in vitro, and Par3L seems to be the factor that regulates the abundance of Lkb1 inside a cell [83].

The Scribble complex antagonizes polarization.

As regards the basolateral side of the cell, the Scribble complex (composed by Scrib, DLG, and Lgl) is in charge of antagonizing the apical complexes and restricting the apical domain. Of note, Scrib, DLG, and Lgl are lost or mislocalized in many types of cancer, and downregulation of Scribble has been demonstrated to be an initiating event in the transformation of the mammary epithelium [88]. Furthermore, silencing of Scrib in murine mammary cells appeared to cooperate with c‐Myc in fostering proliferation but had no effect on tumor latency, and did not induce metastasis [89]. However, its precise effect on MaSCs and on the regulation of ACD and polarity has not been addressed yet.

Altogether, these findings highlight that polarity proteins affect MaSCs' ability to correctly differentiate and give rise to the mature mammary epithelium after puberty (or upon transplantation in the cleared fat pad). Nonetheless, none of the evidence described here provides direct and mechanistic insights in the perturbation of the ACD program execution. The exhibition of peculiar differentiation markers, or impairment in the generation of the progeny, suggests, though, that the ACD machinery is profoundly involved in the morphogenesis of the mammary gland.

Guided by the knowledge available from model systems like Drosophila, many research groups have been trying to unravel the role of these polarity determinants in the regulation of ACD and tissue apico‐basal polarity also in mammals, but the task has turned out to be extremely challenging, suggesting that the translation of these mechanisms in the mammalian system is not straightforward.

mInsc scaffolding role is essential for proper ACD.

The Par3/Par6/aPKC complex interacts with mInsc, which in turn binds to the microtubule‐associated complex NuMA/LGN/Gαi whose role is to contact and tether the spindle (see Table 1). Molecularly, in asymmetrically dividing stem cells, mInsc is believed to bring LGN to the apical cortex, this way favoring ACD [90]. A recent paper showed that deregulation of mInsc in murine models leads to the skewing of the mode of division of mammary stem cells from asymmetric to symmetric [72]. In this work, the execution of ACD is demonstrated by the localization of LGN and NuMA in a crescent structure above one mitotic spindle pole and is visualized exclusively in the end buds (both cap and body cells), although it is not possible to appreciate whether these dividing cells are MaSCs or dividing progenitors. Intriguingly, in a murine model in which mInsc is overexpressed, the frequency of asymmetric divisions is significantly reduced in favor of the symmetric modality [72]. This was confirmed also in vitro through a label‐retaining assay based on a membrane dye, the PKH26 (henceforth PKH), which allows tracking the proliferative fate of labeled cells (see Box 1). A PKH⁺ stem cell dividing asymmetrically generates one cell that remains quiescent, and therefore retains the label, and another cell that continues to divide. By these means, Ballard et al [72] were able to show that overexpression of mInsc leads to a prevalence of proliferative and likely symmetric divisions, and to larger mammary glands, as observed in the transplantation assay. Whether the mInsc‐induced increase in SCDs is a consequence of its high overexpression that in turn regulates other LGN functions still remains an open issue. The balance between symmetric and asymmetric divisions is also regulated by extrinsic factors, as earlier mentioned in this review, and the mammary gland tissue does not represent an exception in this respect. Indeed, mInsc abundance seems to be dictated by the extent of SLIT2 ligand signaling (expressed by cap and body cells in the end buds), on its receptor ROBO1 [72], and on the downstream activation of Snai1.

Huntingtin (HTT) controls localization of polarity proteins.

Similar to mInsc, HTT also functions as an adaptor between the spindle pole and the microtubule, being able to bind Dynein/Dynactin/NuMA/LGN and determining their apical localization [91]. Cre‐mediated depletion of HTT in the basal compartment of the mammary epithelium (K5‐expressing cells) led to proliferation defects in terms of colony formation and to an unbalanced ratio of basal to luminal cell fate determination [91], [92]. This was accompanied by in vivo expansion of the K8‐K14 double‐positive compartment, a phenomenon already described for Par3 inhibition [75], which indicates the accumulation of potential bipotent progenitors generated by symmetric division. HTT depletion disrupts the apico‐basal polarity of the mammary tissue by randomizing the orientation of the mitotic spindle in the basal cells. Interestingly, HTT loss is a frequent event in mammary carcinogenesis [93].

Aurora‐A kinase (AurkA) associates with the mitotic spindle and orchestrates the localization of fate determinants.

In fly neuroblasts, AurkA contributes to ACD by phosphorylating the Par complex and determining asymmetric localization of Numb [94]. Its impairment causes mis‐oriented symmetric divisions and results in aberrant expansion of the NB compartment [10], [11]. Likewise, in vertebrates, AurkA has been reported to play a role in fate control mechanisms in stem cells. One study confirmed the pivotal role of AurkA in the correct execution of ACD in the normal mammary gland. When the mammary cells were transduced with a mutated form of AurkA, unable to associate with the mitotic spindle, their ability to proliferate and differentiate was impaired, with an unbalanced proportion of basal versus luminal cells generated. Furthermore, the self‐renewal potential of the MaSCs was severely diminished, since the cells expressing AurkA‐mut were not able to survive serial transplantation [95] (see Box 1). As mentioned, AurkA is an important regulator of the spindle positioning in NBs. Regan and colleagues demonstrated that enhanced expression of WT AurkA favors mitotic divisions in the cap cells of the TEBs, with a perpendicular orientation with respect to the basement membrane resulting in an increased proportion of luminal cells [95]. Although also in this case we cannot determine the identity of the dividing cells (stem cells or progenitors), these findings suggest that the perpendicular divisions might be of the asymmetric type, with the daughter cells destined to different localizations in the multilayered epithelium and possibly diverse cell fates, even if this remains to be experimentally assessed. Concordantly, cells expressing a mutated AurkA mainly divide parallel to the basement membrane and only give rise to basal cells. The observed phenomena are clearly compatible with both unilineage and multilineage asymmetric divisions.

Altogether, these observations highlight the importance of regulators of spindle assembly and spindle orientation to determine cell fate in the mammary gland upon mitotic division. In a confined space such as the one in which the mammary epithelial cells need to co‐exist, the localization of the daughter cell after mitosis is crucial, as the paracrine signaling from the surrounding niche has huge influence on cell fate decisions.

β1‐integrin bridges ECM stimuli and intrinsic cell polarity.

Integrins are the immediate sensors of the ECM signals. In the mammary cells, they are involved in a number of signaling networks, transducing the microenvironmental stimuli into intracellular responses [96]. Notably, β1‐integrin (CD29) and α6‐integrin (CD49f) are among the widely accepted surface markers utilized to isolate MaSCs by FACS [61], while the expression of β3‐integrin (CD61) labels luminal early progenitors [97]. Nonetheless, for most of them, a precise role in stem cell biology is not yet known.

β1‐integrin has been widely studied in the context of cytoskeleton regulation and polarity determination. Its specific deletion in the luminal compartment (WAP‐expressing cells) impairs the capacity of mammary fragments to form outgrowths in a recipient's cleared fat pad [98], hinting at defects in the luminal progenitors' ability to self‐renew. Conversely, genetic removal of β1‐integrin in basal cells, which are enriched in MaSCs, profoundly changes the orientation of TEB divisions, leading to severe perturbation of the balance between cell lineages [99]. More specifically, β1‐integrin‐ablated glands fail to reconstitute a mammary tissue upon serial transplantation, exhibiting high impairment in both proliferation and self‐renewal abilities. This phenotype is accompanied by an aberrant TEB‐like structure formation upon induction of pregnancy in β1‐integrin floxed/floxed (β1fl/fl) mice. Collectively, these findings suggest the presence of potential defects in the execution of ACD. Taddei et al addressed, mechanistically, the involvement of β1‐integrin in spindle orientation by measuring the angle of the spindle axis compared to the basement membrane in the TEB basal cells of a pregnant gland, and found that WT cells mainly divide parallel to the basement membrane, while the absence of β1‐integrin allows more perpendicular divisions [99]. These data indicate that, upon pregnancy and consequent expansion of the mammary epithelium, basal cells only rarely undergo perpendicular divisions [99]. This was confirmed also during normal maintenance of mature estrus‐staged glands [100]. In contrast, in the β1‐integrin‐depleted epithelium, basal cells significantly contribute to the luminal compartment as a result of the increased frequency of perpendicular divisions [99]. Thus, the expansion of WT mammary cells at pregnancy might mainly rely on the symmetric self‐renewal of basal cells, which is consistent with the notion that only unipotent stem cells are responsible for gland rearrangements in adult age [67]. Similar findings were obtained by double depletion of the Timp1 and Timp3 metalloproteinases in the ECM of adult virgin glands, further underscoring the importance of the microenvironmental signals in influencing the symmetry of mitotic division [100].

In the context of microenvironmental stimuli, it is worth spending a few words on the hormone‐dependent signals that primarily influence mammary cells' fate decisions. Indeed, ovarian hormones, and progesterone in particular, profoundly impact on MaSC homeostasis since both the basal and the luminal stem cell‐enriched compartments are expanded at diestrus [101]. Excessive stimulation by progesterone seems to actively induce symmetric self‐renewal and increase the size of the MaSC pool (as assessed by the use of CD49f, CD24, CD61 surface markers), likely via paracrine stimulation of Wnt4 and RankL signaling pathways [101]. Nevertheless, direct demonstration of an actual role of hormone stimuli in the regulation of the modality of mitotic division still remains to be provided.

Collectively, these data show how critical polarity is, and how the positioning of SCs and progenitors upon mitotic division has an impact on cell fate specification. At the intracellular level, the choice between ACD and SCD reflects the activation of signaling pathways that either maintain the mother cell identity or contribute to differentiation. In this respect, the Notch pathway appears to be crucial, as discussed more in detail in the next section. Notably, nearly all the fate determinants and niche signals described in this review have been demonstrated to directly or indirectly regulate the Notch pathway. For example, the Par3/aPKC/Par6 complex is known to interact with Numb, the best‐characterized antagonist of Notch activation, determining its asymmetric enrichment at the basal site [102] and its function by phosphorylation in one of the daughter cells in Drosophila neuroblasts [103]. Furthermore, deregulation of the AurkA and HTT proteins promotes Notch activation in displaced suprabasal daughters, which acquire luminal cell fates [92], [95]. Remarkably, different cell fate decisions can be induced in a context‐dependent manner by the Notch ligand, Jagged1 [104], highlighting how much this pathway is sensitive to microenvironmental cues [99], [100], [101].

The effects of Notch functions are highly context‐dependent.

The Notch pathway has a central role in the regulation of both self‐renewal and differentiation of MaSCs in the normal mammary gland during development and morphogenesis [105]. Constitutive expression of active Notch4 has been shown to inhibit differentiation of mammary epithelial cells in vitro and normal gland morphogenesis in vivo [106], [107]. Dontu and colleagues used the mammosphere assay, established by the same group [108], and 3D organoid cultures in Matrigel, to study the role of Notch specifically in stem cell self‐renewal and the regulation of cell fate decisions in normal mammary epithelial cells. By implementing complementary approaches to activate or inhibit Notch signaling, they observed diverse effects on stem and more differentiated populations. Treatment with Notch agonists (DSL or recombinant Dll1) resulted in elevated MaSC self‐renewal through symmetric divisions, as shown by the increasing number of mammospheres upon serial re‐plating and their ability to subsequently give rise to multilineage progenitors in differentiation assays (see Box 1). Treatment of the cultures with Notch antagonists [Notch4 antibodies (N4Ab) or γ‐secretase inhibitors (GSI)], instead, led to rapid mammosphere exhaustion. Notch activation had a proliferative effect also in the more differentiated progenitor populations, favoring, at the same time, commitment to the myoepithelial lineage [109].

Later studies, however, attributed to Notch signaling an anti‐proliferative effect on MaSCs, and commitment to the luminal lineage [110], [111], [112]. Bouras et al reported that Notch homologs 1–3 are mainly expressed in the luminal cells, while Notch4 levels are overall lower and comparable among the different subsets. Knock‐down of Cbf‐1 or γ‐secretase inhibition in the stem population (defined this time as CD29^hiCD24⁺, see Fig 3C) resulted in an expansion of the MaSC pool, which was documented by enhanced in vivo repopulation ability in limiting dilution transplantation assays, and increased clonogenic capacity in Matrigel cultures. On the contrary, upon exogenous constitutive expression of the Notch intracellular domain NICD1, the same cells appeared to lose their capacity to reconstitute a normal mammary gland, as they produced hyperplastic nodules with luminal cell characteristics (luminal marker expression—K8, K18, Stat5a, and E‐cadherin) in vivo, albeit failing to reach terminal differentiation during pregnancy [110].

The outcome of Notch activation or inhibition is, overall, very much dependent on the cell context and developmental stage; in addition, discrepancies between in vivo and in vitro systems in different studies hamper the interpretation of seemingly contradicting results [113]. For instance, enforced expression of NICD1 in luminal progenitors (CD29^lowCD24⁺CD61⁺) conferred mammosphere initiating capacity on this population and, conversely, knock‐down of Cbf‐1 or GSI had a restrictive effect on their proliferation in Matrigel assays [110]. Therefore, the initial observations by Dontu and colleagues in bulk mammosphere cultures could be attributed to the effect of Notch activation on luminal progenitors rather than on MaSCs. Nevertheless, as discussed in previous sections of this review, Notch signaling has been shown to respond to cell–matrix interactions and other microenvironmental cues that are involved in the regulation of ACD in mammary epithelial cells [95], [99], [100].

Numb is asymmetrically partitioned during ACDs and retained by the cell with SC identity.

Upstream of Notch is Numb, which prevents nuclear translocation of NICD and, thus, inhibits its transcriptional activity. In fly NBs, asymmetric partitioning of Numb leads to the generation of two daughter cells with distinct fates. Numb asymmetric localization during mitosis has also been used as a marker in mammary epithelial cells in order to define the modality of division [66], [70], although the direct connection between the asymmetric distribution of Numb and the fate choice of daughter cells has not been yet addressed experimentally in MaSCs. Recently, Tosoni and colleagues reported that in PKH^high isolated MaSCs, Numb is inherited by the daughter cell that retains the stem identity [114]. This is in line with the above‐described expression pattern of Notch homologs in mammary stem and progenitor cells, and the observations of luminal lineage commitment upon enforced activation of Notch signaling in CD29^highCD24⁺ cells [110]. Numb also seems to be involved in the regulation of MaSC ACDs through the modulation of p53, which will be discussed later [114], [115].

Musashi regulates Numb and Notch expression.

Musashi1 (Msi1) is an RNA‐binding protein that is known to regulate Numb gene expression at the translational level. It functions as a positive regulator of Notch, and its specific role in asymmetric cell division was first studied in Drosophila SOPs and the mammalian nervous system [116]. In breast, Msi1 was suggested as a putative marker of the MaSC compartment [117] and was shown to promote luminal progenitor cell expansion in the COMMA‐1D cell line, through activation of the Notch and Wnt signaling pathways [118]. Notwithstanding the lack of a direct link to the regulation of asymmetric division in the context of normal mammary epithelial SCs, the decreased sphere‐forming efficiency of the human breast cancer SUM159T cell line upon RNAi of Msi1 might be suggestive of a switch to the asymmetric mode of division [119]. In the latter study, the authors attributed the loss of self‐renewal capacity of the breast CSCs to the upregulation of NF‐YA and the consequent increase in 26S proteasome activity, altogether promoting NICD degradation.

Wnt‐1 constitutive expression perturbs MaSC functions.

Constitutive Wnt signaling in vivo has been generally associated with the formation of hyperplastic glands and/or the appearance of invasive carcinomas, due to the expansion of basal‐type undifferentiated cells [61], [120], [121]. In particular, substantially higher numbers of “MRUs” (mammary repopulating units—defined as Lin⁻CD29^hiCD24⁺ cells) were scored in the premalignant glands of a Wnt‐1‐driven tumor model as compared to wild‐type glands. This premalignant CD29^hiCD24⁺ population further showed enhanced in vivo regenerative ability and increased tumorigenic potential (the outgrowths were profoundly hyperplastic) upon transplantation, implying a role for Wnt signaling in the self‐renewal of MaSCs [61]. However, other studies have suggested that the mammary epithelial hierarchy could be perturbed in the mouse mammary tumor virus (MMTV) Wnt‐1 model and that excessive Wnt‐1 signaling might also elicit dedifferentiation of committed progenitors to a more stem cell‐like state [122].

Axin2 is a target of the Wnt/β‐catenin pathway and affects MaSC self‐renewal.

In contrast to the above‐mentioned models, Axin2^lacZ/lacZ transgenic mice develop a milder phenotype of hypermorphic, but not hyperplastic, mammary glands. The Axin2 gene is a known target of Wnt/β‐catenin and, at the same time, a negative feedback regulator of the pathway. Thus, disruption of the gene by lacZ provides the possibility to elevate the Wnt signal in a ligand‐dependent and cell‐autonomous manner. Competitive transplantation assays of Axin2^lacZ/lacZ and wild‐type MRUs revealed a correlation between responsiveness to Wnt signaling and repopulation ability, which was mirrored by the relative clonogenic capacity of the two cell types in Matrigel cultures [123]. Similarly, an increase in the absolute number of MaSCs was scored upon continuous exposure of wild‐type MRUs to purified Wnt3a in vitro. An expanding number of colonies were formed at each successive passage, which, with the exception of GATA‐3 downregulation, did not significantly differ from the untreated control in terms of size, cell composition, other differentiation markers, and proliferation and apoptosis indexes. Importantly, the Wnt‐treated colonies retained their ability to generate a fully functional mammary gland upon transplantation even at late passages [123]. Despite the evidence pointing to the involvement of the Wnt pathway in the regulation of MaSC self‐renewal, the possibility that the documented in vivo and in vitro growth potential might derive from a block in differentiation of the dividing cells cannot be formally excluded.

Lipoprotein receptor‐related proteins (LRPs) are implicated in the canonical Wnt pathway as co‐receptors of Frizzleds.

Dkk1, a Wnt inhibitor which functions by interfering with the Lrp5/6 receptor, was shown to limit MaSC (CD29^hiCD24⁺) self‐renewal upon serial passaging in Matrigel [123]. Also, loss of Lrp5 was shown to confer resistance to Wnt‐1‐induced tumorigenesis in mice. In fact, in response to dysregulated Wnt signaling, Lrp5^−/− mice grow hypomorphic glands that are depleted of regenerative potential, as almost no positive outgrowths were scored even at high cell doses in limiting dilution transplantation assays. The authors of the study argue that, apart from the minimal growth of the mammary tree that limits, per se, stem cell numbers, these observations could be potentially attributed to a decrease in the number of symmetric divisions; however, another valid explanation could be that Wnt signaling is essential for MaSC and mammary progenitor survival [124]. Following up this study, Badders and colleagues further elucidated the role of Lrp5‐mediated response to Wnt signaling. In brief, in mammary gland reconstitution experiments, they found that Lrp5‐expressing cells possessed increased repopulation ability compared to bulk mammary epithelial suspensions, whereas Lrp5‐negative cells were depleted of stem cell properties upon transplantation. However, Lrp5^−/−glands were able to reach terminal differentiation and go through multiple rounds of lactation and involution without any apparent functional deficits. This observation, along with the elevated expression of p16^Ink4a and TA‐p63 in Lrp5^−/− cells in vitro, led to the conclusion that the reported loss of regenerative potential could be associated, instead, with increased senescence [125]. In accordance with the findings based on the expression of Lrp5 [125], Axin2⁺ MRUs showed increased in vivo repopulation ability upon transplantation, maintaining a normal histology and the ability to undergo multiple rounds of pregnancy and involution [68], [123]. Nonetheless, parallel lineage tracing experiments, using an Axin2^CreERT2 model, pointed toward a unipotent basal stem cell identity for the Axin2⁺ cells, highlighting the discrepancy between regenerative ability and normal in situ developmental potential of Wnt‐responsive cells [68].

p53 regulates stem cell number.

An additional tumor‐suppressive function for p53 emerges from the control it exerts on the expansion of the number of SCs through regulation of the modality of cell division. In mammary epithelial cells, this was first proposed by Sherley and colleagues who reported a change in in vitro growth kinetics from exponential to linear upon p53 upregulation in an inducible non‐tumorigenic mouse cell line. The authors reasoned that this shift in the growth pattern was compatible with asymmetric cell divisions accompanied by the generation of a quiescent stem cell pool [127]. Later studies placing Numb as an upstream regulator of p53 created further speculation on the potential role of p53 in the control of ACD in mammary epithelial cells [128].

The question was formally addressed by Cicalese and colleagues, who showed that p53 signal attenuation in the ErbB2 model of mammary tumorigenesis is responsible for the unlimited self‐renewal potential of breast CSCs and their expansion in numbers through symmetric self‐renewing divisions. The shift of the modality of mitoses from mainly asymmetric to symmetric was directly demonstrated by time‐lapse microscopy and Numb immunostaining of dividing PKH^high isolated cells. The authors further documented an increased proliferative capacity of CSCs, which was shown through limiting dilution transplantation assays of both bulk and PKH subsets of WT and tumor epithelial cells. The same observations also held true for the bulk epithelial cells and mammospheres derived from the mammary glands of p53‐null mice, suggesting that loss of p53 signaling is responsible for the increased frequency of symmetric divisions. Pharmacological restoration of p53 in ErbB2 cells by Nutlin‐3 administration resulted in a reduction in the self‐renewal potential of tumor mammospheres in vitro and tumor growth in vivo. Altogether, these data led to the conclusion that functional p53 signaling limits the expansion of MaSC numbers by imposing an asymmetric mode of division [70].

The role of p53 in governing ACD in MaSCs was further illustrated by studies from the same group, in which p53 levels were modulated in normal mammary epithelial cells using DNA damage (DD) induction by X‐ray irradiation. Under these conditions, MaSCs (PKH^high) exhibited a p21‐dependent DD response that inhibited p53 activation. Similar to p53^−/− PKH^high cells, low levels of p53 activity in p53^+/+ MaSCs (PKH^high) upon DD skewed the modality of division toward symmetric mitoses, resulting in increased self‐renewal and expansion of the functional stem cell pool as seen in the mammosphere assay and upon transplantation in vivo [129].

The above studies established p53 as a key player in the regulation of ACD and, by extension, in the maintenance of tissue homeostasis. The control that p53 exerts on the size of the MaSC pool could be seen as part of its tumor‐suppressive function, underlining the need to identify potential downstream effectors. Members of the miR34 family have been appointed as candidate p53 direct targets associated with tumor appearance and progression. Interestingly, in vitro, ectopic expression of miR34a or miR34c was found limiting breast CSC expansion in the mammosphere assay, while in vivo, restoration of miR34 in donor cells reduced tumor growth upon xenotransplantation. In both cases, the effects were mediated by Notch1 and Notch4, respectively [130], [131]. Collectively, these studies implicate miR34 members in the suppression of breast CSC self‐renewal capacity, albeit without a direct demonstration of their effect on the modality of division.