Epithelial to mesenchymal transition and microRNA expression are associated with spindle and apocrine cell morphology in triple-negative breast cancer

Triple negative breast cancers (TNBC) are a morphologically and genetically heterogeneous group of breast cancers with uncertain prediction of biological behavior and response to therapy. Epithelial to mesenchymal transition (EMT) is a dynamic process characterized by loss of typical epithelial phenotype and acquisition of mesenchymal characteristics. Aberrant activation of EMT can aggravate the prognosis of patients with cancer, however, the mechanisms of EMT and role of microRNAs (miRNAs) in EMT activation is still unclear. The aim of our study was to analyze miRNA expression within areas of TNBCs with cellular morphology that may be related to the EMT process and discuss possible associations. Out of all 3953 re-examined breast cancers, 460 breast cancers were diagnosed as TNBC (11.64%). With regard to complete tumor morphology preservation, the tissue samples obtained from core—cut biopsies and influenced by previous neoadjuvant therapy were excluded. We assembled a set of selected 25 cases to determine miRNA expression levels in relation to present focal spindle cell and apocrine cell morphology within individual TNBCs. We used descriptive (histological typing and morphology), morphometric, molecular (microdissection of tumor and non-tumor morphologies, RNA isolation and purification, microchip analysis) and bioinformatic analysis (including pathway analysis). The results were verified by quantitative real-time PCR (RT-qPCR) on an extended set of 70 TNBCs. The majority of TNBCs were represented by high—grade invasive carcinomas of no special type (NST) with medullary features characterized by well-circumscribed tumors with central necrosis or fibrosis and frequent tendency to spindle-cell and/or apocrine cell transformation. Apocrine and spindle cell transformation showed a specific miRNA expression profile in comparison to other tumor parts, in situ carcinoma or non-tumor structures, particularly down-regulated expression of hsa-miRNA-143-3p and hsa-miRNA-205-5p and up-regulated expression of hsa-miR-22-3p, hsa-miRNA-185-5p, and hsa-miR-4443. Apocrine cell tumor morphology further revealed decreased expression of hsa-miR-145-5p and increased expression of additional 14 miRNAs (e.g. hsa-miR-182-5p, hsa-miR-3135b and hsa-miR-4417). Pathway analysis for target genes of these miRNAs revealed several shared biological processes (i.e. Wnt signaling, ErbB signaling, MAPK signaling, endocytosis and axon guidance), which may in part contribute to the EMT and tumor progression. We provide the first miRNA expression profiling of specific tissue morphologies in TNBC. Our results demonstrate a specific miRNA expression profile of apocrine and spindle cell morphology which can exhibit a certain similarity with the EMT process and may also be relevant for prognosis and therapy resistance of TNBC.


Triple negative breast cancers (TNBC) are a morphologically and genetically heterogeneous group of breast cancers with uncertain prediction of biological behavior and response to therapy. Epithelial to mesenchymal transition (EMT) is a dynamic process characterized by loss of typical epithelial
phenotype and acquisition of mesenchymal characteristics. Aberrant activation of EMT can aggravate the prognosis of patients with cancer, however, the mechanisms of EMT and role of microRNAs (miRNAs) in EMT activation is still unclear. The aim of our study was to analyze miRNA expression within areas of TNBCs with cellular morphology that may be related to the EMT process and discuss possible associations. Out of all 3953 re-examined breast cancers, 460 breast cancers were diagnosed as TNBC (11.64%). With regard to complete tumor morphology preservation, the tissue samples obtained from core-cut biopsies and influenced by previous neoadjuvant therapy were excluded. We assembled a set of selected 25 cases to determine miRNA expression levels in relation to present focal spindle cell and apocrine cell morphology within individual TNBCs. We used descriptive (histological typing and morphology), morphometric, molecular (microdissection of tumor and nontumor morphologies, RNA isolation and purification, microchip analysis) and bioinformatic analysis (including pathway analysis). The results were verified by quantitative real-time PCR (RT-qPCR) on an extended set of 70 TNBCs. The majority of TNBCs were represented by high-grade invasive carcinomas of no special type (NST) with medullary features characterized by well-circumscribed tumors with central necrosis or fibrosis and frequent tendency to spindle-cell and/or apocrine cell transformation. Apocrine and spindle cell transformation showed a specific miRNA expression profile in comparison to other tumor parts, in situ carcinoma or non-tumor structures, particularly downregulated expression of hsa-miRNA-143-3p and hsa-miRNA-205-5p and up-regulated expression of hsa-miR-22-3p, hsa-miRNA-185-5p, and hsa-miR-4443. Apocrine cell tumor morphology further revealed decreased expression of hsa-miR-145-5p and increased expression of additional 14 miRNAs (e.g. hsa-miR-182-5p, hsa-miR-3135b and hsa-miR-4417). Pathway analysis for target genes of these Molecular biological methods. In order to examine specific tumor (spindle cell, apocrine cell, ductal in situ carcinoma-DCIS, invasive front of tumor) and non-tumor areas (normal ducts and lobules, tumorinfiltrating lymphocytes-TILs) of TNBC we used PALM MicroBeam laser capture microdissection (LCM) with PALM Robo Software version 4.6 (Cat. No. 415109-2620-102; Carl Zeiss Microscopy GmbH, Jena, Germany). The procedure was performed on 10 µm breast cancer sections mounted onto 6 glass slides with RNAse-free conditions. After mounting, the slides were dried overnight in a drying oven at 56 °C and deparaffinized. For high quality RNA we used freshly prepared and precooled staining solution of Cresyl Violet. Table 1. Morphological and molecular characteristics of EMT.

Bioinformatics.
For the microarray analysis, the resulting CEL files were read and processed by R, ver.
3.5.0 (2018-04-23) (R Core Team, 2018). Data are available at the Gene Expression Omnibus under the accession number GSE162670. The linear models for microarrays (limma) were applied for comparison of resulting miRNA expression in tumor and non-tumor morphologies mentioned above. We considered the Benjamini-Hochberg adjusted p-value less than 0.05 to be statistically significant. MiRNA expression profiles of microdissected morphologies (Supplementary Table 1a) were compared by both paired and unpaired analysis. Selection of 20 miRNAs for Table 3 was based primarily on paired analysis, nevertheless 15 of them were found by unpaired analysis, too (Supplementary Tables 1b and 1c, respectively).
We have also performed pathway analysis by KEGG and Gene Ontology resources. Target genes of candidate miRNAs were defined by miRDB 20 or TargetScan 21 databases. Pathway analyzes were performed by the KEGGprofile 22 and clusterProfiler 23 R-packages.
For the RT-qPCR analysis, relative quantification was carried out according to the ΔCt method using a reference gene (ΔCt = Ct target miRNA -Ct reference U6) and inverse values of ΔCt (-ΔCt) were used for subsequent statistical analysis and visualization. For the direct comparison of microarray and RT-qPCR results from the same RNA samples, a transformation to the value of 40 was performed (ΔCt = 40 − Ct target miRNA). The data Table 3. Significantly changed levels of miRNAs in selected tumor morphologies. AveExpr average miRNA expression calculated from all samples, log2FC log2 of fold change miRNA expression in the respective tumor morphology in comparison to the normal breast epithelium.

Results
Tumor morphology. The majority of tumors were classified as invasive carcinomas of no special type (NST) with medullary features (69/81; 85.2%). These tumors were predominantly characterized by well-defined and pushing borders and predominantly solid growth pattern with syncytial architecture (Fig. 1a)

Dysregulated miRNAs.
We initially performed the laser capture microdissection of cells from areas with spindle-cell (Fig. 1b) and apocrine transformation (Fig. 1c), areas of dense TILs, ductal in situ carcinoma and epithelium of normal ducts and lobules on the set of 25 TNBCs with medullary features. After miRNA extraction and purification, we analyzed 74 microdissected samples by microarrays with 2578 human miRNAs. MiRNAs with the highest differential expression are listed in Table 3 (see also Supplementary Table 1a-c). The relation of miRNA expression profiles to specific tumor morphology and normal tissue is shown by heatmap (Fig. 2, Supplementary Fig. 1). Besides miRNAs from Table 3, hsa-miR-200c-3p and hsa-miR-155-5p were also included. Hsa-miR-200c-3p was the most downregulated miRNA in tumor areas with lymphocytic infiltration, while hsa-miR-155-5p was highly upregulated in these areas (Supplementary Table 1c). Hsa-miR-155-5-p was also upregulated in spindle cell morphology. We selected several candidate miRNAs for validation by RT-qPCR. First, the same RNA samples used for Affymetrix microarray analysis were measured by RT-qPCR for two miRNAs and results from both methods were highly correlated ( Supplementary Fig. 2). Next, the localization of six miRNAs was verified by RT-qPCR on the extended set of of 94 newly microdissected tissue areas from 70 TNBC patients (Fig. 3, Supplementary Table 5). The best concordance with microarray analysis was observed for hsa-miR-143-3p, hsa-miR-205-5p and hsa-miR-4417, which also had highly significant differences in Table 3 (adjusted p values < 0.005). The other three miRNAs (hsa-miR-182-5p, hsa-miR-185-5p and hsa-miR-155-5p) had less significant differences both in the Affymetrix microarray and RT-qPCR analyses which is probably caused by higher variability of their expression.
Epithelial mesenchymal transition and pathway analysis. We also analyzed the available data concerning EMT-related miRNAs, their target genes and compared them with our results (see Tables 4,5). The foci of spindle-cell morphology in breast cancer were found in 45.7% of cases. Apocrine cell transformation, considered to be a possible variant of EMT-related morphology, was recorded in 48.6% cases. Downregulation of E-cadherin and upregulation of vimentin were found also in our set of TNBC (Fig. 1d,e). Apocrine cell transformation was associated with increased expression of GCDF-15 (Fig. 1f).

Discussion
TNBCs are considered to be a morphologically and genetically heterogeneous molecular subtype of breast cancers with specific and variable response to chemotherapy 24 . The mechanism of arising chemoresistance lies in the induction of rare pre-existing subclones (adaptive resistance), new mutations (acquired resistance) or www.nature.com/scientificreports/ cancer-associated metabolic reprogramming (altered metabolism of glucose, lipids and amino acids) 25 . Our study provides detailed miRNA analysis of the unique set of chemotherapeutically non-influenced TNBCs from a morphological point of view. MiRNAs are involved in many cancer-related pathways, such as DNA damage response, cell cycle, apoptosis, autophagy, tumor cell proliferation, migration and invasion or immune response, reported as the concept of cancer immunoediting 15,26 . They are also associated with the patients' overall survival and recurrence. Each miRNA has multiple targets which can be simultaneously modulated by several miRNAs. The mechanisms deregulating miRNA expression in TNBC involve genomic and epigenetic alterations (chromosomal abnormalities), defects Our results demonstrate the importance of specific miRNAs in TNBC morphogenesis. We focused on the description and isolation of several forms of tumor cell differentiation, including apocrine and spindle cell morphology within individual TNBCs. These morphological changes are probably a manifestation of incipient EMT, corresponding to the specific changes in microRNA expression levels. With respect to EMT-related morphology, we proved a significant increased expression of hsa-miRNA-185-5p, hsa-miRNA-155-5p, hsa-miR-885-3p, hsa-miR-3687, considered as oncogenic miRNAs and down-regulation of tumor suppressive hsa-miR-205c-3p and hsa-miR-143-3p. However, the suppressive effect of miRNA-185-5p in breast cancer via regulation of S100A8/ A9, nuclear factor-κB/Snail signaling pathway and programmed cell death was also reported in the literature 27 . miR-155 is a well-known miRNA with both oncogenic and tumor suppressive character. Hsa-miRNA-155-5p induces cell proliferation via activation of the STAT3 gene and reduces bufalin-induced apoptosis in TNBC cells 28 . On the other hand, this miRNA is highly expressed in immune cells, which may be related to better survival of TNBC patients due to higher presence of tumor infiltration lymphocytes 29,30 . The activation of CIAPIN1 protein which may depresshsa-miR-143-3p expression can be one of the potential mechanisms leading to therapeutic resistance 31 . Demethylation of the miR-200c promoter was found to be associated with tamoxifen reversed EMT and inhibition of cell migration in TNBCs 32 . Tumor suppressive miRNA-205-5p reduces TGF-β-induced EMT 33 . Its inhibition indicates a resistance to chemotherapy. www.nature.com/scientificreports/ Recently published data indicate that EMT in breast cancer has typical morphological correlates such as changes of cellular shape and features resulting from spindle, giant and apocrine metaplasia 8 . Usually they are accompanied by changes of expression profile such as the downregulation of epithelial markers E-cadherin and cytokeratins as well as upregulation of mesenchymal markers N-cadherin, vimentin and transcription factor twist 18,34,35 . We also analyzed the available data on characteristic EMT-related miRNAs and compared them with our results 30, . In accordance with published data, our study showed the down-regulation of tumor suppressive hsa-miR-145-5p and hsa-miR-143-3p. Tumor suppressive EMT-related hsa-miR-145 promotes TNFα-induced apoptosis through targeting cIAP1and may serve as a potential biomarker for an early diagnosis of TNBC. In our study, hsa-miRNA-205-5p was decreased in spindle cell and apocrine cell tumor morphologies. Increased expression of hsa-miR-182-5p in apocrine cell tumor morphology can confirm its oncogenic potential together with its relation to the EMT process. In relation to apocrine cell transformation we proved the significant increased expression of hsa-miR-182-5p, hsa-miR-4417, hsa-miR-3687, hsa-miR-1225-5p and down-regulation of hsa-miR-145-5p. We also found a significant profile for both spindle cell and apocrine tumor morphology which displayed simultaneous increased expressions of oncogenic hsa-miR-3135b, hsa-miR-3648, hsa-miR-4443, hsa-miR-7110-5p, hsa-miR-574-3p, hsa-miR-4632-5p, hsa-miR-22-3p, hsa-miR-1268a, hsa-miR-185-5p, hsa-miR-6802-5p, hsa-miR-885-3p and decreased expression of hsa-miR-143-3p and hsa-miR-205-5p.
We have also performed pathway analysis with the aim to elucidate the role of relevant miRNAs in EMT and other biological processes 82 . Although we cannot prove direct involvement of these miRNAs in the EMT process, there are several pathways (i.e. Wnt signaling, ErbB signaling, MAPK signaling, endocytosis and axon guidance) commonly affected by all these groups. The eight novel miRNAs might contribute to the cancer transformation and EMT process, but a more detailed mechanistic link will require additional functional experiments and investigation. www.nature.com/scientificreports/ Among upregulated miRNAs, we have found two miRNAs (hsa-miR-3687 and hsa-miR-4417) which are not included in miRbase and assigned as non-confidential in TargetScan database. Nevertheless, several recent studies support their relevance by different methods. Benoist et al. 83 identified hsa-miR-3687 as a novel prognostic marker for response in patients with castration-resistant prostate cancer treated with enzalutamide. This miRNA may also confer aggressiveness of oesophagus cancer through its target gene PGRMC2 84 . Using in-vitro and in-vivo models 85 found that upregulation of hsa-miR-4417 and its target genes contribute to nickel chloridepromoted lung epithelial cell fibrogenesis and tumorigenesis. On the other hand, low expression of hsa-miR-4417 was significantly associated with worse prognosis in TNBC patients, while its overexpression was sufficient to inhibit migration and mammosphere formation of TNBC cells. We have observed upregulation of this miRNA both by microarray analysis and RT-qPCR in TNBC, in particular in apocrine cell morphology. Others also report increased expression of this miRNA in TNBC 86 . Further investigation is needed to clarify the validity of these two candidates which are currently not among confidentially identified miRNAs.

Conclusions
Our findings indicate the importance of detailed description of morphological changes in TNBCs since they reflect miRNA expression which can be related to EMT as well as to prognosis and therapy resistance. Generally, it is shown that morphology of cells in cancers should be closely related to expression of specific miRNAs. This also demonstrate that previously detected heterogeneity of miRNAs in breast cancers needn´t be necessary accidental but it may be related to morphological character of cells. Understanding the function of miRNAs implicated in cancer pathogenesis, including the EMT process, also extends the range of new potential diagnostic and therapeutic options and can provide information on prognosis and response to therapy. Table 4. Available data on miRNAs involved in EMT and their target genes-oncogenic miRNAs. + = significant expression; ± = non-significant expression; − = non-changed expression; 0 = not analysed.

Data availability
All important data generated or analyzed during this study are included in this article (and its Supplementary  Information files). Raw data of miRNA expression profiling will be provided upon request.