Objectives: Cell polarity and epithelial morphology are peculiar features of cells forming the terminal ductal lobular unit, and they are early lost during neoplastic transformation because of an epithelial-mesenchymal transition (EMT). To understand these early events we analyzed a set of 125 genes related to cell polarity, EMT and cell-fate decision in 26 breast cancer specimens and corresponding patient-matched normal tissue. Methods: The difference of gene expression was explored by t-paired test. In addition, to evidence latent variables accounting for genes correlations, a Factor Analysis was applied as exploratory technique. Results: Among the 90 differentially expressed genes, those coding for cell polarity complexes, apical-junctional components and luminal cytokeratins were overexpressed in tumor samples (suggesting a terminally differentiated phenotype) whereas those coding for stemness-associated features or related with EMT were expressed in normal tissues but not in tumor samples, suggesting the persistence of stem/progenitor cells. Factor analysis confirmed these findings and indicated that the difference between tumors and normal tissues can be synthesized in three main features representative of specific molecular/morphological alterations. Conclusions: The a priori definition of a selected panel of genes and the application of an exploratory statistical approach, greatly contribute to reduce the intrinsic biological complexity of tumor specimens and to describe the difference between tumor specimens and corresponding histologically normal tissues.
Breast cancer is a biologically heterogeneous disease arising at the terminal ductal lobular units of the mammary gland. According to the widely accepted model of breast cancer progression, normal epithelial cells of the terminal ductal lobular unit acquire increasingly aberrant molecular and morphological abnormalities, resulting in different subtypes of invasive cancer [
To better understand breast cancer biology we focused on a peculiar feature of cells forming the terminal ductal lobular unit: cell polarity. Indeed, epithelial mammary cells are characterized by an asymmetric distribution of cytoplasmic and membrane proteins, termed apico-basolateral cell polarity that is essential to maintain the normal tissue specificity and a correct directional secretion of milk into the lumen [
As reported in the original paper [
As some genes were recognized by more than a single probe-set, each of which characterized by an individual specificity and sensitivity that differently contribute to gene expression value, a gene expression mean value was calculated after weighting each probe-set for its own sensitivity and specificity score. Specifically, each expression value (already log2-transformed in the original dataset) was multiplied for the semi-sum of sensitivity and specificity scores of the corresponding probe-set. Given the patient-matched samples study design, all statistical analyses were performed considering the difference of gene expression between tumor and corresponding noncancerous adjacent tissue as a new variable. The difference of gene expression was explored by t-paired test. To correct for multiple testing, the false discovery rate (FDR) was estimated [
To evaluate the impact of the selected genes in distinguishing tumor from the corresponding histologically normal tissue with respect to non-selected genes, we ranked the overall probe-sets group according to t-paired test on the whole 18382 probe-sets available and computed the FDR. One hundred and forty-five out of the 220 selected probe-sets had an FDR < 0.01 and ranked among the first 9600 probe-sets. Statistical analysis indicated that 90 of 125 selected genes were differentially expressed between tumors and corresponding histologically normal tissues with an estimated FDR < 0.01. Of the differentially expressed genes, 42 (47%) were overexpressed (
epithelial architectures; TGFB1, coding for transforming growth factor-β (TGF-β), known for its antiproliferative effects; TGFBR1, coding for type I TGF-β receptor; ERBB2, coding for ErbB2/neu. Conversely, among the 48 underexpressed genes, we found those coding for basal markers (EGFR, KRT5, KRT14, KRT17 and VIM), many genes associated with EMT (AKT2, FOXC1, HOXB7, ID1, ID3, ID4, KLF8, SMAD2, SMAD3, SMURF1, TGFBR2, TGFBR3, TWIST1, TWIST2, ZEB1 and ZEB2) or cell-fate decision (ALDH1A1, ALDH1A3, JAG1, JAG2, NOTCH1, NOTCH2, NOTCH3, NOTCH4). Notably, according to the inverse relation between differentiation and tumor aggressiveness we found as underexpressed the genes involved in endothelium proliferation (ASAM, ESAM, JAM2, JAM3, VEGFB and VEGFC).
When we applied a factor analysis to investigate the latent variables intrinsically associated with the 125 genes, three factors were retained according to the scree plot. Genes associated with the epithelial phenotype principally characterized the first factor (
The second factor (
presence of VIM coding for vimentin, a typical mesenchymal marker, and by the presence of genes involved in endothelial cells proliferation (AMOTL1, JAM2, JAM3 and VEGFC) or cell-fate decision (ALDH1A1, ALDH1A3, JAG1, JAG2, NOTCH2, NOTCH3, NUMB).
Finally, the third factor (
In the original paper, Pedraza et al. [
patient-matched samples and we investigated a selected set of genes involved in cell polarity and apical junction assembly [11,12], EMT activation [3,4] cell fate-decision and maintenance of a self-renewal state in tumorigenic adult tissues [13,14]. Our results indicated that 72% of the selected 125 genes were differentially expressed between tumor and corresponding histologically normal tissue with an estimated FDR < 0.01. In particular, we found overexpressed the genes associated with a welldifferentiated luminal phenotype; complementally we found underexpressed those associated with a basal phenotype, EMT activation or stemness-like morphology. With respect to the corresponding histologically normal tissue, tumor samples differentially overexpressed also the genes coding for estrogen receptor and for the regulators of cellular polarity and adhesion. Conversely, histologically normal tissues expressed the genes coding for some transcription factors involved in EMT and stemness-associated features. Such an apparently paradoxical finding, i.e., the positive association between histologically normal tissues and EMTand stemness-related genes (more likely expected in neoplastic samples) is not a peculiar feature of breast epithelium since we recently observed a similar behavior in normal pleura with respect to pleural mesothelioma [
Normal mammary gland development and differentiation imply tissue remodeling. Therefore, the expression, in histologically normal samples, of genes involved in EMT and, in particular, the panel of genes related to TGF-β signaling pathway is not at all surprising. TGF-β, in fact, is the recognized major inducer of EMT during embryogenesis and cancer progression [40,41]. According to the conventional model, TGF-β binds to a heteromeric complex formed by type I and type II receptors and leads to the activation of Smad2 and Smad3 that, after forming a trimer with Smad4, translocate to the nucleus where they bind to Smad-binding elements and activate/repress a plethora of target genes. Through Snail, ZEB and bHLH transcription factors the Smad complex represses the expression of several epithelial markers, including claudins, ZOs proteins, occludin, E-cadherin and cytokeratins, and induces the expression of mesenchymal markers including N-cadherin, fibronectin and vitronectin. Consistent with such a model, in normal tissue we found expressed genes coding for TGF-β receptors (TGFBR2 and TGFBR3), Smad proteins (SMAD2 and SMAD3), ZEB (ZEB1 and ZEB2) and bHLH (TWIST1, TWIST2, ID1, ID3 and ID4) transcription factor family members. Concomitantly, we found underexpressed all genes coding for epithelial markers. In agreement with recent acquisition providing evidence of a negative transcriptional regulation of CD24 expression by Twist [