Serially transplantable mammary epithelial cells express the Thy-1 antigen

C57BL/6 and FVB mice were purchased from The Jackson Laboratory, Bar Harbor, ME, USA. pCx-GFP founder mice were kindly provided by Dr. Irving Weissman. All animals were maintained at the Stanford Animal Facility in accordance with the guidelines of both Institutional Animal Care Use Committees.

Six- to 10-week-old mice were euthanized and all fat pads surgically resected. Tissue was digested in Media 199 + 10 mM HEPES + PSA or L-15 for 2 h, and single-cell suspension was obtained as described previously [2] and then processed as described by the manufacturer’s instructions for Epicult (Stemcell Technologies, Vancouver, BC, Canada). For all experiments, cells were > 99% viable as assessed by Trypan Blue dye exclusion. Cells were then resuspended at a concentration of 1 × 10⁷ per ml and subjected to staining for flow cytometry. The antibodies CD24-PE, Thy-1.1-APC, Thy-1.1-PE-CY7, Thy-1.2-APC, and Thy-1.2-PE-CY7 were obtained from eBioscience (San Diego, CA, USA), CD45-BIO, Ter119-BIO, CD31-BIO, and CD140α-BIO were obtained from BD Pharmingen (San Jose, CA, USA), and Streptavidin-Pacific Blue was obtained from Invitrogen (Carlsbad, CA, USA). FACS for all experiments was performed on BD FACSAria II equipped with a UV laser. All gating to distinguish positive and negative expression was based upon IgG-isotype color-specific control staining.

Ultra-low attachment 96-well plates (BD, Franklin Lakes, NJ, USA) were prepared with a feeder layer of irradiated L1-WNT3a mixed with 60 μl of growth factor-reduced Matrigel (BD) per well. Sorted cells were then plated into liquid media as previously described [10, 11] 10% FBS, 250 ng/ml Rspo I (R&D Systems, Minneapolis, MN, USA) and 2.5% growth factor-reduced Matrigel were added as supplements. 5 ng/ml purified mouse TGFβ1 ligand (R&D Systems) was added to wells as indicated. Colonies were counted after 7 days in culture in a 5% CO₂ incubator.

Sorted cell populations were collected in HBSS + 2% HICS and resuspended at the correct concentration before being injected into the cleared fat pads of 21–28-day-old recipient C57Bl/6 mice as previously described [2]. For all injections of 600 cells and below, cell counts were verified using either a nuclear staining count (1% Trypan Blue/0.1% Triton-X 100 in PBS) or GFP⁺ cell count. Cells were injected in either 10 or 5 ul volumes of PBS using a 25 ul Hamilton syringe. All transplants were allowed to grow for at least 5 weeks, but not more than 10 weeks before analysis. In the case of secondary and tertiary transplants, whole glands were dissected under fluorescence to obtain 1–2 mm pieces of tissue that contained GFP⁺ ductal structures that were transplanted into recipient mice.

Whole mounts stained with carmine alum were processed according to a standard protocol [12]. For immunocytochemistry, mammary glands were fixed in formalin and embedded in paraffin. Three-micrometer sections were dewaxed, hydrated and microwaved for 20 min in Tris-EDTA (0.01 M; pH 9) for antigen retrieval. Tissue sections were incubated overnight at 4 °C with primary antibodies in TBS + 1% BSA. Antibodies were CK6 (Covance, Princeton, NJ, USA), CK5 (Covance), Troma-I (DSHB, Iowa City, IA, USA) and Troma-III (DSHB). Sections were then incubated with anti-rat–Alexa A488 antibody and anti-rabbit–A594 antibody (both from Invitrogen) for 30 min at room temperature. Secondary antibodies were applied at 1:200 dilution. Samples were stained with DAPI and mounted in ProLong before pictures were taken. All images were produced with a Leica (Wetzlar, Germany) microscope and Image Pro Software (Media Cybernetics, Rockville MD, USA).

Sorted cell populations were collected in staining media directly and then centrifuged at 5000 rpm for 5 min at 4 °C. Supernatant was then carefully removed from the cell pellet, which was immediately frozen in liquid N₂ and stored at −80 °C until RNA extraction. RNA was extracted from frozen cell pellets by the Trizol method or MirVana kit (Life Technologies, Carlsbad, CA, USA). For RT-PCR, RNA was then converted to cDNA using the Superscript III Reverse Transcriptase system (Invitrogen). qPCR was then performed on fresh cDNA with 2 × SYBR Green Master Mix (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions with primers to amplify specific genes (available upon request).

RNA was then processed for microarray hybridization by the Stanford Protein and Nucleic Acids core facility. RNA was applied to Affymetrix (Santa Clara, CA, USA) GeneChip Mouse Genome 430 2.0 oligonucleotide arrays. The resulting CEL images were then processed using the TIGR TM4 [13, 14] suite of bioinformatic software. Arrays were pre-processed by MIDAS using robust multichip average (RMA) normalization across all samples. Using Cluster 3.0 [15], probes that expressed less than 6.229 (in Log₂ space) in any two of the arrays and whose value was not at least fourfold different in any two arrays were excluded from the analyses. Probes intensities were then mean centered across all arrays as a set. Using MeV (TIGR), Rank products algorithm was then used to compare different populations using two-class unpaired comparisons with a critical p value of 0.01 for differentially expressed probes. Heatmaps are scaled from −3.0 (low) to 3.0 (high) in Log₂ space. Microarray data may be found at GenBank under accession GSE89720.

Previous studies have shown CD49f and CD24 are cell surface markers that are able to distinguish self-renewing MRUs (CD24^medCD49f^high) from basal myoepithelial cells (CD24^lowCD49f^med), CD24^highCD49f^med colony-forming cells, and CD24^medCD49f^−/low luminal cells [2, 3]. We found that Thy-1 was expressed on ~30% of luminal cell populations and ~50% of CD24^medCD49f^high and myoepithelial cells (Fig. 1a, b, c, d). We also found that Thy-1 was expressed by CD24^med/highCD133⁺Sca-1⁺ hormone-sensing luminal progenitor cells [16] as well as CD61⁺ luminal progenitor cells [17] (Additional file 1: Figure S1A). Real-time PCR confirmed the basal or luminal identity of sorted populations based on the expression of Smaα and Krt14 (basal marker) and Krt18 and Krt19 (luminal marker) genes, respectively (Fig. 1e and f). To test if our enrichment procedures dramatically altered the physiology of the sorted populations, we exposed sorted populations to TGFβ1, a well-studied ligand that is associated with epithelial-to-mesenchymal transition [18]. Sorted populations that expressed basal or luminal gene markers shared similar responses to TGFβ1 exposure (Additional file 1: Figure S1B). Therefore, the sorting paradigm and related procedures effectively enriched for physiologically distinct phenotypic populations.

Numerous studies have linked the transcriptional expression of key genes to the function of cells enriched for by cell surface demarcation or lineage-tracing strategies [16, 19]. For example, Tbx3 is a gene that regulates the development of mammary hormone-sensing cells [20]. Interestingly, we found that all populations except for the Thy-1 expressing CD24^medCD49f^high cells also expressed Tbx3 (Fig. 1g). This population also had a low transcriptional level of Cdkn1a (Fig. 1h), whose expression is associated with halted cell cycle progression, comparable to luminal CD24^highCD49f^med cells that have been previously characterized to be highly proliferative. Furthermore, single-cell RNA-Seq data from the Marioni and Khaled laboratories showed that Thy-1 is most highly expressed in cells that are enriched for Procr, a Wnt target gene that has been linked to multipotent mammary stem cells [21] (Fig. 1i). Taken together, these results suggested that Thy-1 does indeed segregate a functionally distinct subpopulation of CD24^medCD49f^high cells.

To further investigate the transcriptional differences that define each population, we performed gene expression microarrays on sorted cell populations. We found that Thy-1⁺CD24^medCD49f^high, Thy-1⁻CD24^medCD49f^high, and CD24^lowCD49f^med cells had unique global gene expression profiles in both C57BL6 and FVB mice (Fig. 2a, b, Additional file 2: Figure S2A and B). There were both strain-specific and distinct genes expressed by sorted populations (Additional file 3). Notably, the Thy-1⁺CD24^medCD49f^high cells expressed the highest level of genes that have previously been linked to mammary gland stem cells such as Rarres2 and Aldh1a1 [22, 23] (Fig. 2a). The enriched expression of Rarres1 and Aldh1a1 in both mouse strains’ microarray data in the Thy-1⁺CD24^medCD49f^high cells was confirmed by real-time PCR and in Procr-enriched cells by single-cell RNA-seq from the Marioni and Khaled laboratories [24] (Additional file 1: Figure S1C, D, E, F). The Thy-1⁺CD24^medCD49f^high population was also enriched for the expression of Procr by real-time PCR (Additional file 1: Figure S1E), with the Procr-enriched population expressing high levels of both Procr and Thy-1 by single-cell RNA-seq (Fig. 1i and Additional file 1: Figure S1H). We also observed that CD24^highCD49f^med luminal cells expressed genes that distinguished this population from CD24^medCD49f^−/low luminal cells in both the C57BL6 and FVB mouse strains (data not shown, Additional file 2: Figure S2C and Additional file 3). In the BL6 strain, the CD24^medCD49f^−/low population was enriched for the expression of Elf5, a transcription factor that suppresses mammary stem cell activity [25] (Additional file 3).

Murine mammary stroma depleted of endogenous epithelium provides an excellent in vivo model system to assess the repopulating potential and frequency of implanted cell populations. We leveraged this system and transplanted sorted cell populations from pCx-GFP transgenic mouse cells that ubiquitously express GFP to distinguish donor outgrowths from any residual endogenous epithelium. We found that 100 Thy-1⁺CD24^medCD49f^high and 100 Thy-1⁻CD24^medCD49f^high cells were able to produce GFP⁺ ductal outgrowths that looked morphologically similar to those produced by 100 bulk mammary epithelial cells (MECs) (Fig. 3a). However, neither luminal CD24^highCD49f^med progenitor cells, CD24^medCD49f^−/low luminal cells, nor CD49f^medCD24^low differentiated myoepithelial cells were able to produce extensive outgrowths up to 10,000 cells transplanted (Fig. 3a and data not shown). Limiting dilution transplantation studies revealed that the frequency of engrafting cells in the total CD24^medCD49f^high population was 1 in 254 cells, consistent with previous reports (Fig. 3b, c, Additional file 4: Figure S3A). Strikingly, we observed that the majority of the outgrowth-forming ability was confined to CD24^medCD49f^high cells that also expressed Thy-1 (Fig. 3b, c, Additional file 4: Figure S3A). Single-cell transplantation revealed that one in eight cells from the Thy-1⁺CD24^medCD49f^high population were able to produce clonal outgrowths (Fig. 3d).

Previous reports using limiting dilution transplantation studies revealed that the MRU frequency within the CD24^medCD49f^high phenotype might be influenced by the strain of mice employed [2, 3]. Therefore, we performed another set of limiting dilution transplantation experiments in wild-type (WT) FVB mice (Additional file 4: Figure S3B). Again, we found that CD24^medCD49f^high cells that also expressed Thy-1 contained most of the outgrowth-forming cells (Additional file 4: Figure S3C). In FVB mice, the Thy-1⁻CD24^medCD49f^high cells gave rise to smaller ductal outgrowths that were largely devoid of side branching structures (Additional file 4: Figure S3D). The frequency of outgrowth-forming cells in the Thy-1⁺CD24^medCD49f^high population was 1 in 156 cells in FVB mice, similar to the frequency in C57BL6 mice (Additional file 4: Figure S3E).

MRUs are defined as cells that have both the ability to self-renew and to differentiate into mature luminal and basal cells that give rise to functional milk-secreting epithelium. The presence of MRUs can be detected by serially passaging transplanted cells or tissue fragments in vivo since only cells with extensive proliferative capacity will be able to continually produce outgrowths [3, 26]. Our initial transplantation data showed that the Thy-1⁺CD24^medCD49f^high phenotype contained most of the outgrowth-forming cells. We serially transplanted 2 mm³ fragments of GFP⁺ donor epithelium from the initial primary outgrowths that were produced by C57BL6 Thy-1⁺CD24^medCD49f^high, or Thy-1⁻CD24^medCD49f^high GFP⁺ cells and found these outgrowths grew secondary GFP⁺ structures (Fig. 4a). A total of 23 out of 33 Thy-1⁺CD24^medCD49f^high primary GFP⁺ epithelial fragments engrafted upon secondary transplantation, giving an engraftment efficiency of about 67% (Additional file 5: Figure S4A). We found that Thy-1⁺CD24^medCD49f^high cells were able to give rise to secondary ductal outgrowths that resembled the outgrowths formed from bulk MECs and the phenotypic CD24, CD49f and Thy-1 diversity of WT glands (Figs. 1 and 4c). Thy-1⁺CD24^medCD49f^high primary outgrowths were also able to give rise to tertiary outgrowths similar to bulk MECs (Additional file 5: Figure S4B). Strikingly, we found that only 11% of the GFP⁺ fragments produced by Thy-1⁻CD24^medCD49f^high transplanted cells were able to form secondary outgrowths (Additional file 5: Figure S4A). In addition, the secondary outgrowths were much smaller than those produced by initially engrafted GFP⁺ Thy-1⁺CD24^medCD49f^hi cells (Fig. 4a). Unlike the Thy-1⁺CD24^medCD49f^high epithelium that produced tertiary outgrowths like bulk MECs, the few GFP⁺ secondary outgrowths formed by the initially engrafted GFP⁺Thy-1⁻CD24^medCD49f^hi cells were unable to re-engraft a third time (Fig. 4a). These data demonstrated that Thy-1⁺CD24^medCD49f^high phenotype enriched for MRUs and that the Thy-1⁻CD24^medCD49f^hi phenotype cells were enriched for basal short-term MRUs (ST-MRUs) with less proliferative capacity.

Functionally competent mammary epithelium will be able to give rise to alveolar milk-secreting cells upon pregnancy. We took mice that were engrafted with secondary GFP⁺ outgrowths produced by bulk MECs and Thy-1⁺CD24^medCD49f^high cells at 10 weeks posttransplantation and mated them to determine the milk-producing capability outgrowths. Hematoxylin and eosin staining showed extensive expansion of the ductal epithelium (Additional file 5: Figure S4C). Cytokeratin staining showed CK14, CK8, CK19, and CK6 localization was indistinguishable between WT (non-transplanted), primary, and secondary GFP⁺ transplanted epithelium produced by Thy-1⁺CD24^medCD49f^high cells (Additional file 5: Figure S4C). In addition, Thy-1⁺CD24^medCD49f^high secondary outgrowths underwent pregnancy morphogenesis in a similar fashion to WT epithelium as evaluated by cytokeratin expression, with visual confirmation of milk inside of the lumen of ducts in hematoxylin and eosin staining (Fig. 4c).