Genetic labeling of parity-identified mammary epithelial cells
To investigate the cellular identity and lineage potential of PI-MECs in intact mammary glands, we chose to label them with a fluorescent reporter. The labeling of PI-MECs was first described by Wagner and colleagues [15, 23] and makes use of the Cre/lox system to permanently label cells by genetic recombination. They created a transgenic mouse line expressing the Cre recombinase under the control of the WAP gene promoter [23]. We crossed these with a reporter strain that contains the YFP gene separated from the constitutive Rosa26 promoter by a stop sequence flanked by loxP sites (Rosa26-lox-stop-lox-YFP, or Rosa-lsl-YFP) [24]. Transcriptional activation of the WAP promoter induces expression of Cre recombinase [16], which excises the transcriptional stop sequence in the reporter construct under control of the constitutively active Rosa26 promoter (Figure 1). From that point onwards, the cell and all of its progeny permanently express the YFP reporter gene.
We crossed WAP-Cre mice with the Rosa-lsl-YFP strain and collected mammary glands from double transgenic female mice before, during, and after pregnancy (Figure 2). Thoracic mammary glands were examined by wholemount fluorescence microscopy (Additional file 4) and by cryosections (Figure 2) to confirm the proper functioning of the PI-MEC labeling system. The cryosections were counterstained with phalloidin to visualize the alveolar structures independent of their YFP status. At day 7 of the first pregnancy, we detected no evident YFP expression by wholemount analysis (Additional file 4A) and very few YFP-positive (YFPpos) cells on cryosections (Figure 2A, see Additional file 5 for close-ups of individual cells). In only one mouse out of five harvested at this stage did we observe extensive YFP labeling of epithelial structures (later determined by FACS to represent approximately 6% of epithelial cells; Additional file 6), which may be attributable to minor alveolar development as part of normal estrus cycling, as previously described [15]. At 14 days of pregnancy, we detected unambiguous activation of the YFP reporter gene (Figure 2B), with about half of alveoli containing at least one YFP-expressing cell and a small percentage of alveoli consisting of more than 90% YFPpos cells. This timing precedes reports of labeling using the LacZ reporter, which was undetectable at 14 days and started at only 15 days [11] or 18 days of pregnancy [15]. This could be due to more efficient floxing of the YFP reporter compared with the LacZ reporter or a difference in the sensitivity of detection of the labeled cells. We could detect WAP mRNA, and in parallel Cre mRNA, in primary MEC preparations at day 7 of the first pregnancy (Figure 2G). Consistent with the observed Cre expression and appearance of more YFPpos cells, we could detect recombination of the genomic Rosa26 locus by PCR at pregnancy day 7 (Additional file 3) which is followed at day 14 by detectable accumulation of mRNA expression for YFP (Figure 2G). As expected, WAP (and Cre) expression reached maximum levels during lactation. Expression of the alveolar specification gene Elf5 [26] reflects the increase in the number of alveolar cells generated during pregnancy and precedes the appearance of YFP, which is dependent on the accumulation of sufficiently high levels of the Cre recombinase in the individual cells (Figure 2G).
At 3 days post-partum, virtually every alveolus contained only YFP-expressing cells (Figure 2C, see also Additional file 5C), consistent with complete recombination of the Rosa-lsl-YFP locus at this stage and underscoring the efficiency of the WAP-Cre/YFP labeling system. Larger ducts of glands at 3 days lactation often remained negative for YFP, in contrast to secondary ducts or side branches that contain multiple YFPpos cells (Figure 2C, top panel and Additional file 7). Six weeks after weaning, the alveolar structures had been cleared by involution. Only the ductal network and regressed terminal branches remained in the post-involution glands, and YFPpos cells were readily detectable in involuted mammary epithelium (Figure 2D). These YFPpos cells are referred to as PI-MECs, cells present in parous tissue that survived involution. In contrast, a large proportion of cells that were initially labeled during pregnancy terminally differentiated and were removed. Thus, not all YFPpos cells that are present during pregnancy and lactation will become PI-MECs. In parous epithelium, WAP and Cre expression dropped by several orders of magnitude, consistent with loss of detectable WAP protein at this stage [19]. Notably, the baseline WAP mRNA expression remains higher after involution than in virgin mice (Figure 2G), consistent with a recent study from the Bentires-Alj laboratory [27]. Overall, these results were concordant with observations with previously published LacZ and GFP reporters [11, 15, 22, 23, 28].
At day 7 of the second pregnancy, alveoli emerged from the involuted ductal network. In contrast to the first pregnancy, some of these contained mostly YFPpos cells, whereas others were unlabeled (Figure 2E). The majority of these YFPpos cells are likely progeny generated by PI-MECs (this will be addressed later). By day 14 of the second pregnancy, most alveoli were completely labeled by YFP (Figure 2F) but with a marked proportion of alveoli that were still partially labeled or even unlabeled. This suggests that, in addition to PI-MEC-derived alveoli which are already YFP-labeled, some alveoli were derived from unlabeled progenitors that progressively undergo YFP recombination and activation, perhaps at a somewhat accelerated pace compared with the first pregnancy (compare Figure 2B and F). The increase in cells with a de novo recombined YFP reporter coincides with the renewed induction of WAP (and in parallel Cre) during the second pregnancy (Figure 2G, note the logarithmic scale).
Parity-identified mammary epithelial cells are luminal cells that express markers of alveolar progenitor cells
After validation of the reporter system, we set out to characterize the cellular identity of PI-MECs by FACS analysis by using cell surface markers for the various epithelial populations of the mammary gland. Six weeks after weaning, involuted mammary glands were processed to single cells, and luminal and basal MEC populations were identified by staining for CD24 and α6-integrin (CD49f) [21, 29], after exclusion of doublets, dead cells, and lymphocytes (Figure 3A and Additional file 1B). Analysis of four individual WAP-cre;Rosa-lsl-YFP animals showed that the YFPpos cells (PI-MECs) fall squarely within the luminal gate (Figure 3B). Plotting the luminal and basal population on separate histograms for YFP further highlights the lack of PI-MECs in the basal population (Figure 3C). Within the luminal population, PI-MECs represent roughly half of the population (Figure 3C and D). Immunofluorescence staining on sections of involuted mammary glands confirmed the exclusive localization of PI-MECs to the luminal layer (Figure 3E and F). Cells identified with an antibody recognizing YFP also expressed the luminal cell-specific marker CK8 but never the basal cell-specific marker SMA. In contrast to the previously reported basal identity of cultured PI-MECs [22], these data definitively establish PI-MECs as a luminal cell type in intact mammary glands.
To further evaluate the cellular identity of PI-MECs within the luminal population, we separated the luminal cells into hormone-sensing cells (Sca1hiCD49blo) and estrogen receptor (ER)-negative cells (Sca1loCD49bhi) [21, 29] (Figure 4A). The ER-negative Sca1loCD49bhi population contains most of the progenitor activity in the luminal population as measured by colony-forming ability [29] and likely contains progenitor cells in different stages of lineage commitment. For clarity, and based on our results described below and the enrichment for markers such as beta-casein and Elf5 [21, 30], we refer to the ER-negative Sca1loCD49bhi population as ‘alveolar progenitor cells’.
FACS analysis of MECs isolated from involuted mammary glands and stained for Sca1 and CD49b showed that about a quarter of luminal cells fell within the gate for hormone-sensing cells, whereas the remaining three quarters were found within the alveolar progenitor gate (Figure 4B). YFP-negative (YFPneg) cells belonged to both cell types, but the vast majority of YFPpos cells were part of the alveolar progenitor population (Figure 4C). Only around 6% of YFPpos cells were part of the hormone-sensing cell population (Figure 4D). This rare population expresses ERα mRNA to the same extent as the collective hormone-sensing cell population (Additional file 8), suggesting that their FACS profile truly reflects a hormone-sensing identity of these cells. Notably, even though almost all YFPpos cells belonged to the alveolar progenitor population, about 40% of the alveolar progenitor population in the involuted mammary gland lacked YFP expression (Figure 4D). Since secretory alveoli were close to 100% YFPpos at lactation (Figure 2C), this YFPneg population may represent alveolar progenitors that resided in the primary ducts, where most cells are YFPneg even during lactation (Additional file 7). In summary, there appear to be three major populations of luminal cells in involuted mammary epithelium: (a) YFPneg hormone-sensing cells, (b) YFPneg alveolar progenitor cells, and (c) YFPpos alveolar progenitor cells (PI-MECs). Immunofluorescence analysis on tissue sections confirmed the presence of all of these three types of cells and showed their juxtaposition within the luminal layer of ducts in involuted mammary glands (Figure 4E).
We have optimized a direct lysis method for qPCR analysis of limited numbers of cells [25] and used this method to validate the molecular identity of the cell populations identified by FACS. Estrogen receptor alpha (ERα) and progesterone receptor (PR) expression was largely limited to hormone-sensing cells (Figure 4F), whereas both YFPpos and YFPneg alveolar progenitor populations exclusively expressed the alveolar marker genes Elf5 and β-Casein [26] (Figure 4G). PI-MECs thus clearly belong to the luminal alveolar lineage within the involuted mammary epithelium.
Initially, we performed a microarray on YFPpos and YFPneg luminal cells in order to identify unique cell surface markers of PI-MECs to facilitate future studies of this cell population without the need for generating parous double transgenic mice. However, the presence of a significant proportion of YFPneg alveolar progenitor cells that express Elf5 and β-casein to the same extent as the YFPpos cells explains why we were unable to identify unique markers for PI-MECs. Apart from YFP, which was enriched approximately 30-fold and validates the sorting procedure, no other transcripts were significantly enriched in the YFPpos population (data not shown). This suggests that the YFPpos and YFPneg alveolar progenitor populations share their transcriptome profile beyond Elf5 and β-Casein.
Taken together, our data show that in involuted mammary epithelium apart from PI-MECs there exists an unlabeled alveolar progenitor population with a similar transcriptional and cell surface profile. To address the question of whether this unlabeled alveolar progenitor population is functionally equivalent to PI-MECs, we examined the relative contribution of these populations toward lobuloalveologenesis in a second pregnancy (Figure 5).
An unlabeled pool of alveolar progenitor cells is equipotent to parity-identified mammary epithelial cells
To evaluate whether PI-MECs or the unlabeled alveolar progenitor cells were the main source for the development of new alveoli, we analyzed WAP-Cre;Rosa-lsl-YFP females at day 7 of their second pregnancy. In typical lineage-tracing techniques, Cre is activated for a short time period following administration of a chemical inducer [31]. In contrast in PI-MEC labeling, Cre expression is controlled by WAP promoter activity and therefore the progeny of PI-MECs can be traced only during developmental time periods when WAP is not expressed. We chose the 7-day second pregnancy time point because alveologenesis is already apparent but WAP expression is not significantly induced yet ([15, 16] and Figure 2G). Thus, YFPpos alveoli at this time point likely originate from PI-MECs and are not the result of de novo activation of the reporter, which occurs again later in the second pregnancy (Figure 2F).
In involuted glands, roughly half of the alveolar progenitor population was YFPpos (Figure 4D). Strikingly, at day 7 of the second pregnancy, newly forming alveoli stained for YFP either almost completely or not at all (Figure 2E and Figure 5A-C, more examples in Additional file 9). Therefore, besides the PI-MECs that give rise to YFPpos alveoli, there is a significant contribution of unlabeled cells that can form morphologically indistinguishable alveoli. Notably, the number of alveoli derived from PI-MECs or unlabeled alveolar progenitor cells varied widely between animals. To quantify this effect, a cryosection of an entire mammary gland from each of the three mice was scanned on a confocal microscope and reconstituted digitally (see Additional file 10 for an example). Each individual alveolus from a section was then scored into one of three categories: YFPneg, YFPpos, or partially labeled by YFP. To be considered YFPpos, at least 90% of the cells in the alveolus must express YFP, since we had noticed that alveoli frequently contain a few ER-positive hormone-sensing cells which are YFPneg (Figure 6). In all three 7-day pregnant samples, the vast majority of alveoli were scored as either YFPpos or YFPneg (Figure 5A-D and Additional file 11). This all-or-nothing distribution suggests a clonal contribution from either an unlabeled alveolar progenitor or a PI-MEC rather than a mixture of alveolar progenitors contributing to the same alveolus. The small proportion of alveoli that had a mixture of YFPpos and YFPneg cells could signify the mixed contribution of a labeled and an unlabeled progenitor, but these partially labeled alveoli could also reflect early de novo activation of the reporter. Indeed, reporter activation due to induction of WAP expression at 14 days of the first pregnancy shows a similar stochastic pattern and the majority of alveoli that become labeled at this point belong to the YFPpartial category (Figure 5E and Additional file 11). Together with the radically different ratio of partial-versus-totally labeled alveoli (1:3 at day 7 of the second pregnancy compared with 5:1 at day 14 of the first pregnancy), these data fit best with a model whereby at day 7 of the second pregnancy the frequency of de novo floxing is still low and the majority of YFPpos cells are generated by PI-MECs.
Overall, the relative contribution of PI-MECs to alveologenesis, as measured by the fraction of total YFPpos alveoli, varied from 4% to 79%. A similar variation was found when four independent animals were analyzed by FACS (27% to 77%, Figure 6H). It is currently unclear whether this variation is stochastic or whether certain conditions favor one population over the other. It should be noted, however, that the females in Figure 5A and B, which show the most divergence in PI-MEC contribution, are littermates who shared nearly identical histories. The proportion of YFPpos alveoli was roughly similar in the right and left mammary glands of each individual mouse (data not shown). This suggests that the relative contribution of PI-MECs and unlabeled alveolar progenitors to alveologenesis is due to systemic regulation rather than independent gland-autonomous effects.
In summary, even though PI-MECs contribute significantly to the generation of new alveoli, a substantial proportion of newly developing alveoli are unlabeled. Therefore, the unlabeled alveolar progenitor cells that are present after involution not only are very similar to PI-MECs in their transcriptome and cell surface markers but also have the ability to contribute to alveologenesis to the same extent as PI-MECs.
Basal and hormone-sensing cells of developing alveoli are not generated by parity-identified mammary epithelial cells
At low magnification, clusters of alveoli appeared to be clonally derived from PI-MECs (for example, Figure 5B and Additional file 9); however, at a higher magnification, it became apparent that even though the majority of cells in these alveoli were YFPpos, they consistently contained a small fraction of YFPneg cells. To investigate the cellular identity of these YFPneg cells, we stained paraffin sections with antibodies against YFP, ER, and CK8. Figure 6A shows a typical example from the 7-day second pregnancy time point, whereby all luminal cells in an alveolus are YFPpos, except the few cells that express the ER (in red) (more examples in Additional file 12). CK8-negative basal cells are also YFPneg, like hormone-sensing cells (Figure 6A). The small proportion of alveoli that are almost fully recombined at 14 days of the first pregnancy also contain a low proportion of YFPneg cells (Figure 6B). We noted that the signal intensity for ER is already reduced by 7 days of pregnancy and becomes virtually undetectable at 14 days, and therefore we used an antibody against progesterone receptor for sections from 14-day pregnant animals. Again, the majority of YFPneg cells are hormone-sensing cells (Figure 6B), similar to the 7-day second pregnancy time point (Figure 6A). Thus, even though all luminal alveolar cells derive from a common progenitor cell (based on the clonal appearance at 7 days of the second pregnancy), developing alveoli contain cells from different lineages; PI-MECs are able to contribute all ER-negative luminal cells to a cluster of alveoli, but the hormone-sensing and basal cells are derived from different lineages (see Figure 7 for a schematic representation of these observations).
Since PI-MECs can give rise to all mammary epithelial cell types in transplantation experiments [11, 15, 22], we further evaluated their lineage potential in vivo during a normal pregnancy by FACS analysis (Figure 6C). Separating the mammary epithelial population of WAP-Cre;Rosa-lsl-YFP mice at 7 days of the second pregnancy into YFPneg and YFPpos subpopulations showed that the majority of YFPpos cells remained restricted to the luminal cell population (Figure 6D). Only 0.29% ± 0.03% of PI-MEC-derived cells were found in the basal cell compartment, indicating that PI-MECs cannot be the source of the expansion of the basal layer during pregnancy. A closer examination of the luminal subpopulations showed that during pregnancy the hormone-sensing cells lose their CD49b and especially their Sca1 expression (Figure 6E versus Figure 4A), consistent with a previous report [29]. The cellular identity of the FACS populations was again validated by qPCR. Even during pregnancy, more than 98% of PI-MECs remained firmly within the luminal alveolar progenitor population (Figure 6F and G). These data show that, as a rule, PI-MECs remain restricted to the luminal alveolar lineage during pregnancy and do not contribute significantly to the other lineages.
Out of all the hormone-sensing cells analyzed by FACS at the 7-day second pregnancy time point, only 1.7% ± 1.2% were YFPpos (Figure 6H). Even though the percentage of YFPpos cells in the hormone-sensing cell gate was small, it is not noise in the FACS data, because rare cells that were positive for both ER and YFP were detectable by confocal microscopy of tissue sections from pregnant mice (Figure 6I).
Taking these findings as a whole, we conclude that, in contrast to the clear multi-lineage potential of PI-MECs in transplantation assays, the lineage potential of PI-MECs in vivo is almost entirely restricted to luminal ER-negative cells, even during pregnancy.