Epithelial-to-mesenchymal transition: immunohistochemical investigation of related molecules in canine cutaneous epithelial tumours
Background – Epithelial-to-mesenchymal transition (EMT) is a multistep process, important in tumour invasion and metastasis, characterized by loss of epithelial markers, redistribution of β-catenin and gain of mesenchymal markers.
Hyposthesis/Objectives – Our aim was to investigate the immunohistochemical aberrant expression of cytokeratin, vimentin, survivin and heat shock protein 72 (Hsp72) in canine cutaneous epithelial tumours, to understand the association of expression of these molecules with features of malignancy and their role in the EMT phenotype.
Methods – Ten canine squamous cell carcinomas (SCCs; one with lymph node metastasis), 30 canine hair follicle tumours (six pilomatricomas, eight infundibular keratinizing acanthomas, six trichoepitheliomas and 10 trichoblastomas) and five normal skin samples were investigated by immunohistochemistry using specific anti-vimentin, -cytokeratin, -survivin and -Hsp72 antibodies. A semi-quantitative method was used to analyse the results, as follows: 0 to <5%; ≥5 to <10%; ≥10 to <25%; and ≥25% of positive cells. Immunofluorescence was performed to investigate survivin-vimentin and survivin-Hsp72 colocalization in selected SCCs.
Results – In malignant hair follicle tumours and SCCs, a reduced intensity of cytokeratin and increased survivin and Hsp72 expression were observed. In SCCs, loss of cytokeratin expression and vimentin immunolabelling, suggestive of the EMT phenotype, were evident in <5% of neoplastic cells in the front of tumour invasion. In the same areas, strong nuclear survivin and cytoplasmic Hsp72 staining was evident, often colocalizing. Only a few neoplastic cells in the front of tumour invasion showed vimentin-survivin colocalization.
Conclusions and clinical importance – A possible simultaneous involvement of survivin and Hsp72 in tumour invasion and the multistep process of EMT of cutaneous epithelial tumours of dogs is suggested.
In order to grow, infiltrate and invade, epithelial tumour cells need to acquire new properties, allowing them to survive without cell-to-cell contact, to pass through basal membranes, to interact with stromal cells and molecules and, finally, to gain access to the vessels for metastatic dissemination. All these cellular proprieties have recently been grouped under the name of epithelial-to-mesenchymal transition (EMT).1 The process of EMT is believed to be one of the key steps towards metastatic dissemination.2 Epithelial-to-mesenchymal transition is characterized by diminished epithelial characteristics and increased mesenchymal attributes3 induced by several pathways, such as transforming growth factor-β (TGF-β)4 and Wnt5 pathways, using the transcription factors Snail, Slug and Twist.6 The exact phenotypic changes associated with EMT are difficult to define. The exact contribution of EMT to the metastatic cascade is still a subject of debate,7,8 although the expression of EMT markers in circulating tumour cells has been demonstrated.9 It has been suggested that many intermediate phenotypes, ranging from epithelial to mesenchymal differentiation, are likely to coexist in a tumour population.10 The EMT phenotype has been associated with several features, as follows: increased invasive ability; improved resistance to apoptotic signals; and augmented ability to promote angiogenesis.3 Cells undergoing EMT are characterized by weakness of cell-to-cell adhesion, which favours the ability to degrade the matrix and modify the cell cytoskeleton, facilitating cell motility.3 During EMT, many cytokeratins are downregulated, whereas vimentin is upregulated.3 Vimentin can be upregulated by central EMT transcription factors, such as Snail11 and Twist,12 as well as the β-catenin pathway,13 showing a functional role in both epithelial and mesenchymal cell migration.14–16 Vimentin expression by epithelial neoplastic cells has been considered a hallmark of the EMT phenotype, as a marker of mesenchymal differentiation and as a useful marker of carcinomas with a more aggressive behaviour in several types of human cancers.17–19
Activation of the Wnt/β-catenin pathway has been related to the EMT process.6 As a consequence of adherens junction reorganization, β-catenin can accumulate in the cytoplasm and translocate into the nucleus, where it activates the expression of EMT target genes, such as vimentin, or EMT regulators, such as Snail and Twist.6,13,20 In canine cutaneous squamous cell carcinomas (SCCs), we have recently found an altered subcellular β-catenin distribution21 and co-expression with vimentin in SCCs with a more aggressive behaviour.22 These findings suggest that EMT molecular modifications might take place in a subset of neoplastic cells in these canine skin neoplasms.
Based on these previous results and on knowledge of the functions and properties of survivin and heat shock proteins (HSPs),23,24 we hypothesized a possible association of these molecules with tumour-associated EMT, because several links are evident between their roles in neoplastic cells and the EMT phenotype.
Survivin is a member of the inhibitor of apoptosis (IAP) protein family. It has low or undetectable expression in most adult tissues and is upregulated in the majority of cancers, suggesting that it is one of the most specific cancer proteins identified to date. It plays a fundamental role in cell proliferation and survival, with anti-apoptotic functions.25 Survivin has been implicated in several aspects of tumour malignancy. In skin tumours, its expression can correlate with tumour severity, metastases, neoangiogenesis and decreased patient survival and it has been inversely correlated with sensitivity to cytotoxic agents used in anticancer therapy.23 Upregulation of survivin has been demonstrated in both human23,26-28 and canine28 cutaneous SCCs.
The HSPs are a wide group of highly conserved molecules that act as molecular chaperones and play an important role in the cellular maintenance of homeostasis, in both physiological conditions and during stress.29 Altered expression of HSPs has been reported in several human30 and canine24 tumours, because they are involved in tumour proliferation, impaired apoptosis, tumour-associated angiogenesis, invasion and metastasis.29 Several HSPs have been demonstrated to exert anti-apoptotic roles,31 particularly Hsp72, the expression of which has been observed in canine cutaneous SCC;32,33 however, their exact role in EMT is still poorly understood. Heat shock protein 72 has recently been implicated in EMT and seems to act by protecting cells from TGFβ-induced EMT in different types of cellular processes, through an interaction with the Smad proteins.34–36
The aims of the present study were to perform a comparative immunohistochemical evaluation of the pattern and levels of expression of cytokeratin, vimentin, survivin and Hsp72 in canine cutaneous epithelial tumours, in order to understand the association of expression of these molecules with features of malignancy and to discuss their possible role in the EMT phenotype.
Materials and methods
In this retrospective study, 40 samples of canine cutaneous epithelial tumours (10 canine SCCs, including one SCC lymph node metastasis, and 30 canine hair follicle tumours) and five normal skin samples were selected from the database of the Departments of Comparative Biomedical Sciences, University of Teramo, Italy, and Biopathological Sciences and Hygiene of Animal and Alimentary Productions, University of Perugia, Italy. Table 1 summarizes the tumour location and signalment of affected animals. Tissue samples were unrelated to those used in previous studies by the authors.28,32,33
All specimens were fixed in 10% neutral buffered formalin, embedded in paraffin, sectioned at 4-5 μm, stained with haematoxylin and eosin and examined by light microscopy. Tumours were classified according to the World Health Organization37 criteria for canine cutaneous neoplasms and according to the classification of Gross et al.38 by two pathologists (L.B. and C.B.) in an independent way. Squamous cell carcinomas were subclassified into well, moderately and poorly differentiated. For each sample, the degree of invasiveness into the surrounding tissue and the degree of inflammatory reaction were evaluated as additional features.
Deparaffinized and rehydrated tissue sections were immunostained by the streptavidin-biotin peroxidase complex method using commercially available antibodies, including the following: full-length survivin (rabbit polyclonal, diluted 0.7 μg/mL; Novus Biologicals, Littleton, CO, USA), vimentin (mouse monoclonal, V-9, diluted 1:200; DakoCytomation, Ely, UK), pancytokeratin (mouse monoclonal, AE1/AE3, diluted 1:50; Dako, Glostrub, Denmark) and Hsp72 (mouse monoclonal, C92 F3A-5, diluted 1:100; StressGen, Kampenhout, Belgium). Endogenous peroxidases were blocked with 3% hydrogen peroxide in absolute methanol for 45 min. Antigen retrieval was undertaken by heat-treating sections in citrate buffer at pH 6 in a pressure cooker for 20 min for survivin and in a microwave oven (3 × 5 min) for vimentin, pancytokeratin and Hsp72. To reduce nonspecific binding, slides were incubated in 5% bovine milk (Bio-Rad, Deeside, UK) in Tris-buffered saline for 15 min at room temperature. Overnight incubation with primary antibodies was performed in a humidified chamber at 4°C. Slides were treated with secondary biotinylated goat anti-mouse + rabbit antibodies (Biospa, Milan, Italy), and detected with streptavidin-peroxidase (Kit Vectstain ELITE ABC; Vector Laboratories, Burlingame, CA, USA), incubated at room temperature for 30 min. The reaction was visualized with 3-3′-diaminobenzidine (D5905; Sigma-Aldrich, St Louis, MO, USA) solution, which was applied for 5 min, and finally lightly counterstained with Mayer’s haematoxylin (Merk, Darmstadt, Germany) for 2 min. Samples of canine cutaneous SCC were used as positive controls. A negative control was performed in all instances by incubating tissue sections with an antibody directed against an unrelated antigen (mouse anti-human desmin monoclonal antibody; Dako) dissolved in Tris-buffered saline instead of the primary antibody.
Quantification of immunolabelling
Immunolabelling was classified based on the subcellular localization (cytoplasmic/nuclear) and the percentage of positive cells found on analysis of 10 high-power fields (x40 magnification; semi-quantitative method), as follows: 0 to <5%; ≥5 to <10%; ≥10 to <25%; and ≥25% of positive cells. Cytokeratin and vimentin expression was also evaluated as positive (+), negative (-) or with a reduced intensity of expression (-/+).
Immunofluorescence was also used to investigate the colocalization of survivin-vimentin and survivin-Hsp72 expression in four selected canine infiltrative SCC samples and one SCC lymph node metastasis. Tissue samples were treated as described for the immunohistochemical procedure. A mixture of primary antibodies against survivin-vimentin (diluted 0.5 μg/mL and 1:100, respectively) and survivin-Hsp72 (diluted 0.5 μg/mL and 1:50, respectively) was applied overnight at 4°C. The first secondary antibody, biotinylated goat anti-rabbit (Vector Laboratories), diluted 1:200, was applied and incubated for 30 min at room temperature, and slides were then incubated with fluorescein-conjugated avidin (Vector Laboratories) diluted 1:100 in sodium bicarbonate buffer 0.1 mol/L, NaCl 0.15 mol/L (pH 8.2-8.5) for 10 min at room temperature. A blocking step was performed by incubating slides for 15 min with avidin and then biotin, at room temperature. The second secondary antibody, biotinylated goat anti-mouse (Vector Laboratories), diluted 1:200, was applied and incubated for 30 min at room temperature, and slides were then incubated with Texas Red-conjugated avidin (Vector Laboratories) diluted 1:100 in sodium bicarbonate buffer 0.1 mol/L, NaCl 0.15 mol/L (pH 8.2-8.5) for 10 min at room temperature. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories).
Histologically, the 40 canine cutaneous tumour samples were identified as 10 canine SCCs, all with an infiltrative growth, of which seven were well differentiated, three poorly differentiated and one was associated with lymph node metastasis; four benign and two malignant pilomatricomas; eight infundibular keratinizing acanthomas; three benign and three malignant trichoepitheliomas; and 10 trichoblastomas.
Normal skin (skin of control dogs, as well as epidermis and hair follicles adjacent to the neoplastic tissue).
In normal epidermis, all keratinocytes were negative for vimentin and positive for pancytokeratin. Vimentin immunostaining was noted in scattered cells, compatible with melanocytes and epidermal dendritic cells.
Weak Hsp72 cytoplasmic immunolabelling was observed in all layers of the epidermis, except the stratum corneum. Scattered survivin-positive nuclei were found in the basal cell layer.
In normal hair follicles, vimentin was negative in epithelial cells, while positive cells were observed in the dermal papilla. Diffuse cytokeratin staining was observed, excluding the matrix cells and dermal papilla. The Hsp72 staining was diffuse in the cytoplasm of the outer root sheath, with scattered positive nuclei in the basal cell layer.
Survivin-positive nuclei were present in the basal cell layer of the outer root sheath and matrix cells.
Benign hair follicle tumours.
All immunohistochemical results are shown in Table 2.
Neoplastic cells were positive for cytokeratin, although cells with matrical differentiation (pilomatricomas and trichoepitheliomas) were negative.
Vimentin was not expressed, other than in scattered cells representing melanocytes and epidermal dendritic cells accompanying the tumour growth. In three of 10 cases of trichoblastomas, small clusters of spindle-shaped cells (so-called ‘follicular papillary mesenchymal bodies’, similar to abortive dermal papillae)38 accompanying the tumour were vimentin positive (Figure 1a).