Looking in the Rear-View Mirror: How Lineage Tracing Experiments Redefined Cell Hierarchy in the Mammary Gland

Abstract

Understanding the cell hierarchy of the mammary gland has been a major focus of research for several decades. This review retraces the key advances enabled by lineage tracing approaches to study stem cells and cell hierarchy in the mammary gland. Although not exhaustive, this review highlights the paradigm shifts brought about by lineage tracing studies, which have, in the author’s view, profoundly reshaped our understanding of mammary epithelial organization. In particular, these studies have challenged traditional models of cell hierarchy and the concept of terminally differentiated cells in the mammary gland.

Introduction

Understanding how the different lineages are maintained during development and homeostasis has been a central question in biology for decades. Identifying the cellular hierarchy underlying tissue maintenance is also of relevance for cancer research, as deciphering tissue organization is essential to determine what becomes deregulated in cancer and to identify the cellular origin of tumors.

The mammary gland plays an essential role in early postnatal life by producing milk, which provides nutrients, water, electrolytes and immune protection to offspring until they are able to feed independently. Mammary glands are epidermal appendages specified along the ventral epidermis around embryonic day 12 (E12) during mouse development. Most mammary gland development occurs postnatally in two main phases. During puberty, the ductal network expands to fill the entire fat pad, whereas during pregnancy, the gland undergoes further expansion and differentiation to form milk-producing alveoli. After each cycle of pregnancy and lactation, the mammary gland undergoes involution, characterized by extensive apoptosis of alveolar cells, after which it returns to a morphology and status similar to that observed prior to pregnancy [1].

The glandular compartment consists of ducts composed of two main epithelial cell types: basal cells, also known as myoepithelial cells, which form the outer layer and possess contractile properties, and luminal cells, which line the lumen. Luminal cells can be further subdivided into hormone sensing cells, which are estrogen receptor (ER) positive (also referred to as mature luminal cells), and ER- negative luminal cells, which can differentiate into milk-producing alveolar cells during pregnancy and are also referred to as alveolar or secretory cells, or luminal progenitors [1]. A recent consensus has proposed renaming these populations based on their functional properties. Accordingly, the three principal epithelial cell types of the mammary gland are now referred to as basal-myoepithelial (B-Myo), luminal hormone-sensing (LHS) and luminal adaptive secretory precursor (LASP) cells.

In this review, we focus on key advances enabled by lineage tracing approaches to investigate stem cells and cellular hierarchy in the mammary gland, with particular emphasis on studies that have redefined our understanding of mammary epithelial organization.

Major breakthrough: transplantation assays to study stemness properties in the mammary gland

The regenerative capacity of mammary epithelial cells in adult life is well illustrated by the ability of the gland to undergo extensive expansion and differentiation during successive pregnancy cycles. A major technological advance in the study of stemness in the mammary gland was achieved by Deome and colleagues in 1959. They developed the first transplantation assays, in which small fragments of mammary epithelium were grafted into the mammary fat pads of prepubescent female mice from which the endogenous epithelium was surgically removed (cleared fat pad). These experiments demonstrated that transplanted cells could regenerated a fully functional mammary outgrowth, regardless of the age or parity status of the donor tissue [2].

Moreover, the progeny of transplanted cells could be serially transplanted, demonstrating their self-renewal capacity. More than sixty five years later, this technique remains widely used to assess the stemness of mammary epithelial cells. While these pioneering experiments established that regenerative potential is broadly distributed throughout the mammary gland and maintained over time, they did not identify which specific cell types were capable of repopulating the fat pad or their potential to give rise to different cell types.

Later on, in 1998, Kordon and Smith demonstrated that an entire mammary outgrowth could be generated from a single mammary epithelial cell, implying this cell has multipotent properties giving rise both to basal and luminal cells [3]. In 2006, the groups of Jane Visvader and of Connie Eaves used cell surface markers and fluorescence-activated cell sorting to isolate distinct populations of mammary epithelial cells. They both showed that luminal cells were unable to reconstitute mammary outgrowths, whereas basal cells could generate functional outgrowths comprising all epithelial cell types of the gland. Furthermore, transplantation of single basal cells, although with low efficiency, was sufficient to produce complete outgrowths.

Together, these findings suggested the presence of rare multipotent mammary stem cells within the basal population, at the top of the cellular hierarchy of the breast epithelium [4, 5].

Mammosphere assays as an in vitro complementary approach to study stemness properties of mammary epithelial cells

Concomitant with these seminal transplantation studies, additional methods were developed to assess the stemness properties of mammary epithelial cells in vitro. In particular, the mammosphere assay was established based on the ability from non-adherent cells to form spherical cell clusters. In the original study using human cells, single cells derived from mammospheres were able to generate multilineage colonies under differentiating culture conditions, suggesting that mammospheres are enriched in mammary stem cells capable of differentiating into all mammary epithelial lineages [6].

Findings from mammosphere and transplantation assays, demonstrating that single cells can give rise to all mammary epithelial lineages, led to a model in which multipotent mammary stem cells maintain the gland throughout life, which became the commonly accepted paradigm in the field [7].

Lineage tracing to trace cell fate in situ

Lineage tracing has become the gold standard for studying the renewal and differentiation potential of stem cells in intact tissues. This approach relies on genetic labeling of specific cell populations to follow their fate and progeny in vivo, thereby avoiding potential artifacts associated with ex vivo culture and transplantation assays [8].

These experiments are typically performed in mice expressing a recombinase under the control of a lineage-specific promoter, together with a reporter gene that is activated upon recombinase-mediated excision of a stop cassette. Once recombination has occurred, the reporter is permanently expressed in the targeted cells and all their progeny (Fig. 1).

Fig. 1.

Lineage tracing. Schematic Illustration of the lineage tracing principle using the Cre/loxP recombination system. Upon tamoxifen administration, the STOP cassette flanked by loxP site is excised, leading to irreversible expression of a fluorescent reporter in the targeted cells and all their progeny. Created in BioRender. Van Keymeulen, A. (2026) http://BioRender.com/gacwvkn

The most commonly used system is the Cre/loxP system, in which Cre recombinase can be either constitutively active or inducible upon tamoxifen or doxycyline administration. Inducible Cre systems provide temporal control of the labelling, and allow flexibility on the level of chimerism achieved: by adjusting the dose of tamoxifen or doxycyline, the proportion of labelled cells can range from less than 1% to almost 100% of the targeted population.

While a low mosaicism is mandatory to perform clonal analysis, enabling the tracing of progeny derived from single cells, it may not always reflect the heterogeneity of the entire cell population, for which higher chimerism is more appropriate [8].

Cellular hierarchy in the mammary gland revisited

The first lineage tracing study in the mammary gland used a constitutive Wap-Cre in 2002. As the Wap gene is specifically expressed during alveolar differentiation and lactation, and returns to undetectable levels upon involution, this system enabled the targeting of alveolar cells during pregnancy to assess their fate during involution. Prior to this study, it was widely accepted that differentiated alveolar cells were terminally differentiated and were undergoing apoptosis upon involution, with alveolar cells in subsequent pregnancies arising from undifferentiated stem or progenitor cells. This first lineage tracing study revealed that a subset of differentiated alveolar cells not only survived involution but also contributed to the formation of new alveoli in subsequent pregnancies. These findings challenged the concept of terminal differentiation and revised the model in which alveolar cells arise exclusively from undifferentiated stem cells [9].

A second major paradigm shift emerged from studies using lineage-specific promoters to specifically label luminal or basal cell populations. K5CreER, K14rtta/TetOCre and Lgr5CreER mouse models were used to label basal cells at different developmental stages. While Lgr5CreER also labelled rare luminal cells, K5 and K14 promoters specifically targeted basal cells. These basal cells were shown to self-renew over time, as the proportion of labeled cells remained stable, and did not give rise to luminal cells, even after multiple pregnancy cycles. Conversely, luminal cells labeled using K8CreER or K18CreER were able to self-renew and to give rise to alveolar cells during pregnancy. Following involution, these labeled luminal cells persisted and contributed to luminal cell expansion during subsequent rounds of pregnancy and lactation. Together, these results demonstrated that mammary epithelial lineages are maintained by lineage-restricted stem cells during puberty, adulthood and rounds of pregnancy [10]. However, this study could not exclude the existence of rare multipotent stem cells with minimal contribution to lineage maintenance that were not targeted by the Cre lines used.

These findings were quite unexpected and contrasted with the results obtained in transplantation assays. To resolve this discrepancy, experiments combining lineage tracing and transplantation assays were performed. First, to determine whether the transplantation procedure itself could promote the differentiation potential of stem cells, labelling of basal cells was performed in mice with K14rttaTetOcre, and transplantation of unsorted cells was performed. In these conditions, labelled basal cells contributed to mammary outgrowth but remained restricted to basal cells and did not contribute to luminal cells. Similarly, luminal cells labeled with K8CreER contributed exclusively to luminal compartments. These results demonstrated that transplantation per se does not alter lineage restriction. However, when isolated cell populations were transplanted, only basal cells were capable of regenerating a complete mammary outgrowth, giving rise to both basal and luminal lineages, as previously described [4, 5]. This demonstrated that K14-traced basal cells, which behave as unipotent in situ, can acquire bipotent potential upon transplantation in the absence of luminal cells. This study demonstrated that transplantation assays, although extremely informative about the differentiation potential of stem cells, do not always reflect the physiological conditions [10].

Another paradigm shift came from the demonstration that myoepithelial cells possess stem cell properties. Indeed, these contractile myoepithelial cells were previously considered a population of fully differentiated cells. The basal compartment was thought to include both differentiated myoepithelial cells and less differentiated stem cells capable of generating an outgrowth upon transplantation [4, 5]. To assess the self-renewal and stemness potential of fully differentiated myoepithelial cells expressing alpha-smooth muscle actin, lineage tracing with Acta2-CreER mice was performed. Surprisingly, these cells were able to expand within the basal layer during puberty, adulthood and pregnancy, persisted after involution and contributed to the basal layers of ducts and alveoli during successive pregnancies [11].

Finally, a paradigm shift also occurred concerning the hierarchy within luminal cells. It was long assumed that a common luminal progenitor population was giving rise to both ER +/LHS and ER-/LASP luminal cells [12]. Moreover, for a long time, ER +/LHS luminal cells were also considered terminally differentiated with low proliferative potential [13, 14], although later studies demonstrated that these cells were able to incorporate nucleotide analogs suggesting they undergo cell division [15, 16]. Three distinct studies, using different promoters to target the different luminal cell subtypes, were able to follow the fate of these cells by lineage tracing. In the postnatal mammary gland, Notch1-CreER targets ER-/LASP luminal cells which only gave rise to ER- luminal cells and were able to self-renew throughout life and to differentiate into alveoli upon pregnancy [17]. Using Sox9-CreER, which targets basal cells and ER-/LASP luminal cells, and Prominin1-CreER, which targets a fraction of ER +/LHS lineage, Wenjun Guo and colleagues achieved to independently trace the two luminal populations [18]. Using the doxycycline inducible ERrtta coupled to TetOcre, Cédric Blanpain and colleagues were able to label most of the ER +/LHS luminal cells and showed that this lineage is specified early after birth and remains lineage-restricted [19].

Collectively, these studies demonstrated that ER +/LHS and ER-/LASP luminal cell lineages are self-sustained and independent maintained by distinct unipotent stem cells, thereby revising the model of mammary epithelial hierarchy (Fig. 2).

Fig. 2.

Selection of key studies that drove conceptual shifts in mammary epithelial cell hierarchy. Time line representing first key experiments, in the author’s perspective. Created in BioRender. Van Keymeulen, A. (2026) http://BioRender.com/kgojgvf

Multipotent stem cells reside at the top of the hierarchy during embryonic development

Lineage tracing experiments using different Cre drivers during embryogenesis have shown that, at the population level, labeled embryonic cells are multipotent and give rise to all mammary epithelial lineages [10, 17, 20]. To determine whether individual embryonic stem cells are multipotent in the mammary gland, clonal analyses combined with statistical approaches was performed in lineage-traced mice.

Using two distinct promoters, K14 and Notch1, two separate studies demonstrated that early embryonic development of the mammary gland is driven by multipotent stem cells, which progressively switch to become lineage-restricted unipotent stem cells during embryogenesis [21, 22].

More recently, an elegant study combining lineage tracing with cellular barcoding estimated that approximately 120 bipotent embryonic stem cells, with similar long-term proliferative potential, contribute to the maintenance of the adult mammary gland [23].

Controversies and limitations of lineage tracing studies

Since these pioneering studies, lineage tracing using a variety of promoters has been widely employed to label distinct mammary epithelial cell populations. While many reports have confirmed that the postnatal mammary gland is predominantly maintained by unipotent, lineage-restricted stem cells [20, 24–32], others have concluded, based on lineage tracing experiments, that bipotent basal [33–36] or luminal [37] stem cells exist. These discrepancies, sometimes arising from studies using the similar promoters to target cell lineages, have generated considerable confusion and controversy in the field, leading to skepticism regarding the interpretations of lineage tracing experiments.

Lineage tracing approaches indeed have limitations. A critical parameter is the choice of promoter used to drive Cre recombinase (or alternative systems such as rtTA/TetOcre) to target specific lineages. Identifying a promoter that is uniformly active across all cells within a given lineage is challenging, as gene expression is often heterogeneous, both in terms of expression levels and temporal dynamics. Moreover, temporal variations in gene expression can further complicate interpretation, as different cell types may express the same gene at distinct developmental stages.

A clear example is Lgr5 expression, which is observed in luminal cells during early postnatal stages and later in basal cells, resulting in the labeling of different populations depending on the timing of induction of Lgr5-CreER mice [28]. In addition, the specificity and efficiency of recombination driven by a given promoter must be rigorously validated. Different mouse models using the K14 promoter illustrate this issue: while K14rtTA [38] drives labelling specifically in basal cells and can achieve near-complete labeling upon prolonged doxycycline administration [10, 39], the K14CreER [40] line generated by the Chambon laboratory labels basal cells as well as a small fraction of luminal cells, potentially leading to the erroneous conclusion of bipotency [33, 39]. In contrast, the K14CreER [41] line developed by the Fuchs laboratory does not induce recombination in the mammary gland (our unpublished observations). Another example is K19, thought to be expressed only in luminal cells, for which the K19CreER [42] drives recombination mainly in luminal cells but in a small subset of basal cells as well [39]. Finally, the level of induction may also influence the cell population targeted. Very low induction levels, often required for clonal analysis (less than 1% of cells) may preferentially label cells with the highest expression of the marker, which may not be representative of the entire lineage of interest.

Strategies to resolve controversies and limitations of lineage tracing

To overcome the limitations and caveats of lineage tracing approaches, as well as discrepancies between studies, several approaches have been proposed. First, careful characterization of the initial labeling- too often overlooked- should be systematically performed at each time point of induction. Varying the extent of induction is essential to determine which cell types are targeted at high versus low doses of induction and to assess whether each level of induction is representative of the entire cell population of interest.

To rigorously assess the multipotent potential of traced cells, clonal induction levels should be used to follow the fate of single cells, as the initial labeling of two neighboring unipotent stem cells may be mistakenly interpreted as evidence of multipotency. Moreover, clonal analysis should be combined with robust statistical pipeline to estimate the likelihood that two adjacent labeled cells originate from a single clone rather than from distinct clones [22, 39].

Whenever possible, multiple promoters targeting the same lineage should be employed to minimize biases resulting from promoter-specific labeling of subpopulations. Alternatively, unbiased approaches using ubiquitous promoters could be used. For example, this approach was used in the mammary gland to resolve the discrepancy between the different studies. Using Rosa26-CreER, Christine Watson and colleagues were able to perform clonal analysis in the mammary gland, without any bias related to the promoter and confirmed that the mammary gland was maintained by lineage-restricted stem cells during postnatal development [25].

A remaining concern is that rare multipotent stem cells may escape the labeling when performing clonal analysis or when specific promoters are used. To address this issue, near-saturation labeling strategies have been employed. Using doxycycline-inducible K8rtTA [43], it was possible to label more than 99% of luminal cells. If luminal cells were replenished over time by basal (or other) bipotent stem cells, the proportion of labelled luminal cells would be expected to decrease. However, the absence of any decrease in labelling of luminal cells after several months or after pregnancy definitively demonstrated that the luminal lineage is self-sustained and maintained by luminal stem cells that are not replaced over time by putative bipotent stem cells [39].

Together, these approaches highlight the importance of rigorous experimental design to accurately interpret lineage tracing data and define cell hierarchy.

Perspectives

As our understanding of mammary gland cell hierarchy has improved, lineage tracing has been combined with genetic perturbations, including oncogenic mutations and loss of tumor suppressor genes, as well as experimental manipulations such as cell ablation and transplantation, to address key questions related to the cell of origin of breast cancer, cellular plasticity, and the reactivation of multipotency programs. These topics have been recently reviewed by Mohamed Bentires-Alj and colleagues [44]. Lineage tracing has also enabled the investigation of clonal dynamics during development and in the context of field cancerization [45, 46].

Over the years, numerous technical advances have led to the development of a wide range of mouse models targeting specific mammary epithelial cell populations. Table 1 summarizes the different drivers used in the mammary gland, along with their induction system and targeted cell population. In addition, alternative approaches such as intraductal injection of adenovirus expressing Cre under lineage-specific promoters have been developed, as recently reviewed by Zhe Li and colleagues [47].

Table 1.

Main transgenes used to target Cre and perform lineage tracing in the mammary gland

Transgene	Induction system	Targeted cells	References in MG
Wap-Cre	Constitutive CRE	Alveoli	[9, 27, 48]
MMTV-Cre	Constitutive CRE	Luminal cells	[48]
K14-Cre	Constitutive CRE	All MECs	[10, 23]
K14-rtta/TetOCre	Doxycycline	Basal	[10, 21, 39, 49]
K5-CreERT2	Tamoxifen	Basal	[10, 21, 22, 33]
Lgr5-GFP-CreERT2	Tamoxifen	Basal/luminal	[10, 21, 28, 33, 39]
K8-CreERT2	Tamoxifen	Luminal	[10]
K18-CreERT2	Tamoxifen	Luminal	[10]
Lgr6-GFP-CreERT2	Tamoxifen	Basal/luminal	[29, 39]
K14-CreERT2	Tamoxifen	Basal + few luminal	[33, 39]
Sox9-CreERT2	Tamoxifen	Basal/ER- LASPluminal	[18, 36, 39]
K19-CreERT	Tamoxifen	Luminal + few basal	[39]
K8-rtta/TetOCre	Doxycycline	Luminal	[21, 39, 49]
Rosa26-CreERT2	Tamoxifen	All	[25, 45]
Notch1-CreERT2	Tamoxifen	ER- LASP luminal	[17, 22]
ERrtta/TetOCre	Doxycycline	ER + LHS luminal	[19]
Prominin1-CreERT2	Tamoxifen	ER + LHS luminal	[18]
Acta2-CreERT2	Tamoxifen	Myoepithelial	[11, 22]
K5-Cre	Constitutive CRE	Basal	[49]
Axin2-CreERT2	Tamoxifen	Wnt responsive	[20]
Prdm1-CreERT2	Tamoxifen	Subset of ER- luminal	[30]
P63-CreERT2	Tamoxifen	Cap cells	[31]
Elf5-rtta/TetOCre	Doxycycline	ER- LASP luminal	[33]
K5rtta/TetOCre	Doxycycline	Basal	[33]
Procr-CreERT2	Tamoxifen	Rare Procr +	[34]
PDL1-CreERT2	Tamoxifen	Rare PDL1 +	[35]
K8-CreER BAC	Tamoxifen	luminal	[37]

Reporter systems have also undergone substantial improvements, including multicolor labeling strategies that enable in situ clonal analysis, as well as genetic barcoding approaches that allow high-throughput tracking of cell populations by sequencing. Recent advances in barcoding technologies and their applications in mammary gland biology and breast cancer have recently been reviewed by Silvia Fre and colleagues [50].

Looking ahead, lineage tracing approaches, particularly when combined with barcoding combined and sequencing technologies, will continue to play a central role in addressing major outstanding questions in mammary gland biology and breast cancer research, including tumor heterogeneity and the hierarchical relationships between primary tumors and metastases.

Author’s contributions

A.V.K. conceived and wrote the review.

Funding

This work was supported by the FNRS, the Télévie, the Fondation contre le Cancer and the Fonds Genicot.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.