PMC42, a breast progenitor cancer cell line, has normal-like mRNA and microRNA transcriptomes

Abstract

Introduction: The use of cultured cell lines as model systems for normal tissue is limited by the molecular alterations accompanying the immortalisation process, including changes in the mRNA and microRNA (miRNA) repertoire. Therefore, identification of cell lines with normal-like expression profiles is of paramount importance in studies of normal gene regulation.

Methods: The mRNA and miRNA expression profiles of several breast cell lines of cancerous or normal origin were measured using printed slide arrays, Luminex bead arrays, and real-time reverse transcription-polymerase chain reaction.

Results: We demonstrate that the mRNA expression profiles of two breast cell lines are similar to that of normal breast tissue: HB4a, immortalised normal breast epithelium, and PMC42, a breast cancer cell line that retains progenitor pluripotency allowing in-culture differentiation to both secretory and myoepithelial fates. In contrast, only PMC42 exhibits a normal-like miRNA expression profile. We identified a group of miRNAs that are highly expressed in normal breast tissue and PMC42 but are lost in all other cancerous and normal-origin breast cell lines and observed a similar loss in immortalised lymphoblastoid cell lines compared with healthy uncultured B cells. Moreover, like tumour suppressor genes, these miRNAs are lost in a variety of tumours. We show that the mechanism leading to the loss of these miRNAs in breast cancer cell lines has genomic, transcriptional, and post-transcriptional components.

Conclusion: We propose that, despite its neoplastic origin, PMC42 is an excellent molecular model for normal breast epithelium, providing a unique tool to study breast differentiation and the function of key miRNAs that are typically lost in cancer.

Figure 1

Analysis of mRNA gene expression in breast cell lines and normal tissue. (a) Analysis of differentially expressed mRNAs. Clustering heatmap for genes with analysis of variance B values of greater than 10 after either a median or a quantile normalisation. Cell type origin (B, basal; L, luminal; N, normal; S?, possible stem cell) and status of oestrogen receptor/progesterone receptor (ER/PR) are indicated on the left. mRNA clusters are indicated by purple letters (below, with their key enriched process) and purple arrows (above and below). ECM, extracellular matrix; VR, viral response. (b) Analysis of luminal/basal mRNA markers. The average log₂ intensity of expression of 252 luminal and 360 basal annotated non-redundant markers is shown.

Figure 2

Analysis of microRNA (miRNA) gene expression in breast cancer cell lines and normal tissue. (a) Global miRNA heatmap. Log₂ ratios of spots with intensity (A values) greater than 6 were used in CIMminer clustering analysis using correlation distances. (b) Heatmap of differentially expressed miRNAs. miRNA genes for which at least one value differed with log₂ greater than 2.5 from the pool (either upregulated or downregulated) were used in CIMminer clustering analysis using correlation distance (upper). The PN miRNA cluster is indicated by a star and is enlarged (lower). The cell type origin (B, basal; L, luminal; N, normal; S?, possible stem cell) and the status of oestrogen receptor/progesterone receptor (ER/PR) are indicated on the left. PN miRNA, microRNA specifically expressed in PMC42 and Normal breast tissue.

Figure 3

Validation of microRNA (miRNA) expression by Luminex bead arrays and real-time quantitative reverse transcription-polymerase chain reaction (qRT-PCR). The levels of 26 miRNAs measured using the methods indicated on the left are shown. qRT-PCR experiments were performed three times, each in triplicate, except for miRNAs 7, 126, and 126* (red diamond), which were measured once in triplicate; error bars indicate one standard deviation. On Luminex bead arrays, miRNAs 127 and 17-3p were below detection (red star). As units of expression in all methods are arbitrary, data were scaled to the highest-expressing sample for each miRNA. Normalised slide microarray data are presented on a non-logarithmic scale identical to the other two methods, for which the data were not normalised in any way. PN miRNAs appear in boldface. miRNAs with at least one negative value for correlation between the methods are shown in red. PN miRNA, microRNA specifically expressed in PMC42 and Normal breast tissue.

Figure 4

Real-time reverse transcription-polymerase chain reaction validation of PN miRNA loss in breast and lymphoblastoid (LB) cell lines. (a) Expression of the listed miRNA genes in 11 additional breast cancer and 2 immortalised normal cell lines alongside 6 of the initial panel of tested cell lines (red). The luminal/basal nature of each cell line is indicated on the right. Values were normalised to the GEOmean of miRNAs 93 and 191 [65]. Data were scaled to the highest-expressing sample for each miRNA. PN miRNAs appear in boldface. (b) The expression of the listed miRNA genes in B cells from 3 healthy blood donors and 3 LB cell lines. Values were normalised to the levels of 5S rRNA and scaled as above. The levels of miRNAs 127, 221, 199a-s, 199a-as, 143, and 145 were independently tested twice. PN miRNAs appear in boldface. miR-214 was below the detection threshold in all samples. PN miRNA, microRNA specifically expressed in PMC42 and Normal breast tissue.

Figure 5

Mechanisms of PN miRNAs loss. (a) Examination of genomic copy number. Genomic copy numbers of the chromosomal regions surrounding candidate miRNAs from the PN cluster (grey highlight) were examined using bacterial artificial chromosome (BAC) and oligonucleotide array comparative genomic hybridisation and are listed by genomic location. Normalised log₂ (copy numbers) of the two nearest clones on each array (names in square brackets) are reported unless unavailable due to the miRNA's proximity to the end of the chromosome. Gains are highlighted in orange; losses are highlighted in green. The mRNAs in whose introns the corresponding miRNAs reside are highlighted in yellow; where available, the log₂ (mRNA expression) is given as measured by beadchip arrays. ND, not determined. (b) Real-time reverse transcription-polymerase chain reaction examination of pre-miRNA levels. As units of expression are arbitrary, data were scaled to the highest-expressing sample for each miRNA and were not otherwise normalised. Experiments were performed twice in triplicate; bars indicate one standard deviation. PN miRNA, microRNA specifically expressed in PMC42 and Normal breast tissue.