Identification of signatures specific to the epithelial or mesenchymal phenotypes from the heterogeneous mammary epithelial HMLER cells

To understand the role of the epithelial-to-mesenchymal transition (EMT) and its reverse MET dynamics between health and disease, the identification of epithelial (E)- and mesenchymal (M)-specific genes is essential. The aim of the study was to derive genes that are characteristic E and M phenotypes in breast-derived cells. Their identification can help to understand the functional meaning of their expression and dynamic changes of individual cells, whole cell populations, and tumors, and even patients. Breast cancer stem cells (CSCs) are thought to drive recurrence and metastasis. Their identity has been linked to the epithelial-to-mesenchymal transition (EMT) but remains highly controversial since -- depending on the cell line studied -- either epithelial (E) or mesenchymal (M) markers, alone or together, have been associated with stemness. Using distinct transcript expression signatures characterizing the three different E, M and hybrid E/M cell types, our data support a novel model that links a mixed E/M signature with stemness in 1) individual cells, 2) luminal and basal cell lines, 3) in vivo xenograft mouse models, and 4) in all breast cancer subtypes. In particular, we found that co-expression of E and M signatures was associated with poorest outcome in luminal and basal breast cancer patients, as well as with enrichment for stem-like cells in both E and M breast cell lines. This link between a mixed E/M expression signature and stemness was explained by two findings: first, mixed cultures of E and M cells showed increased cooperation in mammosphere formation (indicative of stemness). Second, single-cell qPCR analysis revealed that E and M genes could be co-expressed in the same cell. These hybrid E/M cells were generated by both E or M cells and had stem-like character since they displayed increased plasticity, produced ALDH1+ progenies self-renewal and mammosphere formation compared to the more differentiated E and M cell types. Thus, the hybrid E/M state reflecting stemness and its promotion by E-M cooperation offers a dual biological rationale for the robust association of the mixed E/M signature with poor prognosis, independent of cellular origin. Together, our model explains previous paradoxical findings that breast CSCs appear to be M in luminal cell lines but E in basal breast cancer cell lines. Our results suggest that targeting E/M heterogeneity by eliminating hybrid E/M cells and cooperation between E and M cell types could improve breast cancer patient survival independent of breast cancer-subtype. We chose to identify genes characteristic of the E and M phenotypes isolated from the same individual, which are therefore independent of genetic differences that are the usual problem when comparing cancer cell lines derived from different individuals. HMLER cells are derived from normal mammary epithelial cells from a healthy individual (HMEC) and immortalized and transformed by transduction with hTERT, SV40LT and RAS oncogenes (Elenbaas et al., 2001). The parental HMLER cell line (HP) is heterogeneous as it usually contains both epithelial and mesenchymal subpopulations that can be identified and discriminated by inverse expression of CD24 and CD44 cell surface markers. Single HMLER cells were sorted using flow cytometry and expanded in adhesion cultures. Morphology with respect to cobble stone-like epithelial phenotype or fibroblast-like solitary growing mesenchymal phenotypes was assessed. Epithelial clones were designated as E clones and mesenchymal clones were designated as M clones. Expression of the cell surface markers CD24 and CD44 strongly correlated with the E (CD24+/CD44-) and M (CD24-/CD44+) phenotypes, and only clones with 95% homogeneity for CD24+/CD44- or CD24-/CD44+ were used for determination of E versus M genes. In order to determine genes specific for the respective epithelial or mesenchymal phenotypes, we generated gene expression arrays (Agilent) by comparing stable (with respect to morphology and cell surface CD24/CD44 profile) E clones (E1 to E6) with M clones (M2-M5), which were collected as biological replicates by at least two passages and usually a freezing cycle to test biological reproducibility of the gene expression profiles. Clones E1 to E3 and M2 and M3 were generated in a first batch of single cell-cloning, and after another freezing cycle of the parental HMLER cell line, clones E4 to E6 and M1, M4 and M5 were generated. Data normalization was performed in Genedata Analyst using central tendency followed by relative normalization. Transcripts showing differential expression (p value < 0.01) between two different morphological phenotypes (E clones E1, E2, E3, E4, E5, E6 and M clones M2, M3, M4, M5) were identified by Limma and ranked according to their effect size. As control for genes that are not specific for homogeneous single cell-derived E clones or M clones, we generated gene expression arrays from the adherent heterogeneous parental HMLER cells (=HP) that has also an epithelial phenotype and about 10-30% CD24+/CD44- cells. M1 was a CD44+ clone that also contained a stable subpopulation of CD24+/CD44+ cells, could change phenotype and had either E or M morphology, which was thus excluded from the E versus M comparison. As a control for genes that are not specific for adhesion cultures (supposedly differentiated cultures, _adh) but rather for stem cell-enriched populations, we analyzed gene expression profiles from mammospheres (_sus) from the epithelial HP cell line and the mesenchymal clone M4. Biological replicate adherent E cell lines - E1 cell line: E1_adh_0629, E1_adh_0826; E2 cell line: E2_adh_0616, E2_adh_0629, E2_adh_0826; E3 cell line: E3_adh, E3_adh_0629, E3_adh_0826; E4 cell line: E4_adh_0210, E4_adh_0607; E5 cell line: E5_adh_0210, E5_adh_0607; HP cell line: HP_adh_2210, HP_adh_0607c, HP_adh_0607sc, HP_adh_0629. Biological replicates adherent M cell lines - M1: M1_adh_0210, M1_adh_0607; M2 cell line: M2_adh, M2_adh_0629, M2_adh_2210; M3 cell line: M3_adh, M3_adh_2210, M3_adh_2810; M4 cell line: M4_adh_0607, M4_adh_0210. Biological replicates mammospheres - HP cells: HP_sus_0603_1, HP_sus_0603_2; M4 cells: M4_sus_0603_1, M4_sus_0603_2.