Background: Periostin is a secreted adhesion protein, normally expressed in mesenchime-derived cells. Aberrant expression of the periostin gene in epithelial tumours seems to play a role in angiogenesis and metastases.
Aims: To investigate periostin expression in a consecutive series of breast carcinomas and correlate it with established biological and prognostic factors.
Methods: A consecutive series of 206 breast carcinomas was investigated by immunohistochemistry with a specific antiperiostin antibody. Immunohistochemical expression of oestrogen and progesterone receptors, Ki-67 (MIB-1), HER-2/neu, VEGF-A, VEGFR-1 and VEGFR-2 was analysed. Periostin expression was also investigated in MCF-7 and MDA-468 cell lines by immunohistochemistry, western blot and quantitative RT-PCR. Localisation of periostin was investigated in MCF-7 cells by the green fluorescent protein (GFP) approach.
Results: Periostin was highly expressed in carcinoma cells, but not in normal breast tissues. The pattern of expression was mainly cytoplasmic. However, in 12% of cases a nuclear reactivity was observed. Nuclear periostin significantly correlated with tumour size, and with expression of oestrogen receptor, progesterone receptor, VEGF-A, VEGFR-1 and VEGFR-2. A nuclear localisation of periostin was also observed in MCF-7 and MDA-468 cell lines. In MCF-7 cells the nuclear localisation of periostin was also shown by transfection of a vector expressing a GFP-periostin chimeric protein.
Conclusions: Results indicate that the aberrant gene expression of periostin in breast cancer cells is associated with an abnormal nuclear localisation of the protein. The nuclear localisation of periostin in breast cancer may induce significant biological effects.
Statistics from Altmetric.com
The identification of a set of genes whose expression might help to classify breast cancers into biologically distinct subtypes is considered the starting point in molecular profiling of these tumours.1 2 By comparing normal breast tissue with breast cancer cells, some investigators observed six different categories, designated basal-like, ErbB-2 positive, normal basal-like, and luminal types A, B and C.2 Subsequently, several research groups applied these concepts to identify specific “genetic signatures” that could have clinical implications,3 especially with the aim of predicting survival or benefit from therapy.2–4
In a recent study, highly enriched cell populations of malignant neoplastic epithelial cells from primary breast cancers and luminal and myoepithelial cells from normal breast tissue were subjected to analysis of the epithelial-specific transcriptome based on massively parallel signature sequencing (MPSS) and multiple genome-wide microarrays.5 6 Using this technique, more than 6000 genes have been found to be differentially expressed between the pool of normal luminal cells and that of primary tumours substantially enriched for epithelial cells.6 In addition, microarray gene signatures identified a list of transcripts whose expression differed between luminal epithelial cells and myoepithelial cells. Some of them belong to the group of genes that are involved in skeletal development and are associated with the myoepithelial/basal cells. One of the most highly overexpressed genes in this category is periostin, a gene characterised by a sequence homology with the insect adhesion molecule fasciclin I.7 The product of the gene has been hypothesised to play a role in the process of adhesion and migration of epithelial cells.8 9 Furthermore, periostin seems to have a role in tumour angiogenesis through the up-regulation of vascular endothelial growth factor receptors in endothelial cells.10
Periostin expression is undetectable in normal breast epithelium and, in accordance with its known mesenchymal expression, it is expressed only in the stroma.6–10 Notably, some breast carcinomas were found to express the protein in epithelial neoplastic cells, supporting the concept that during cancer progression, transition from an epithelial to a mesenchymal phenotype of tumour cells could be a key event.11 12
This hypothesis has been corroborated by the observation that stable expression of a periostin transgene in tumourigenic but not metastatic 293T cells, induces a fibroblast-like transformation characterised by increased expression of specific proteins such as vimentin (a hallmark for mesenchymal cells), epidermal growth factor receptor, and matrix metalloproteinase-9 (MMP-9).12 Consequently, periostin has been hypothesised to have a role as a marker for tumour aggressiveness with potential clinical implications in multiple human cancers, including breast cancer.6–14
In the present study we investigated the immunohistochemical expression of periostin in breast cancer and its correlation with established biological and prognostic factors.
PATIENTS AND METHODS
This study was performed in accordance with the Declaration of Helsinki. It was based on a consecutive series of 206 primary early invasive breast cancers from 200 patients observed at the University Hospital of Udine between 1992 and 2004. Representative paraffin blocks from surgical excision were reviewed by a pathologist; two 1.5 mm cores, from the centre and the periphery of the tumour, were incorporated in tissue microarray (TMA) blocks. In each TMA block were arrayed 21 cases with a precision instrument (Beecher Instruments, Silver Spring, Maryland, USA).
Immunohistochemistry of periostin, vascular endothelial growth factor (VEGF-A), vascular endothelial growth factor receptor 1 (VEGFR-1) and 2 (VEGFR-2), oestrogen (ER) and progesterone receptors (PR), HER-2, Ki-67 and vimentin (table 1) was performed using a polymer-based immunohistochemical detection system (Super Sensitive Polymer-HPR IHC Detection System, BioGenex, San Ramon, California, USA).
In each experiment, a negative control was included in which primary antibody was replaced by non-immune serum. Positive and negative controls for each marker were performed according to the supplier’s indications. In each tumour, only staining of the invasive malignant cells was evaluated. For periostin, the intensity of immunoreactivity was scored as 0, 1, 2 or 3, denoting negative, weak, moderate and strong staining, respectively. Therefore, for statistical analysis, the periostin expression level for each case was putatively defined as positive if the predominant intensity was ⩾1. ER, PR and MIB-1 were evaluated as percentage of positive nuclei. Vimentin was reported as percentage of cytoplasmic immunoreactive cells. Results were analysed as continuous variables.
For VEGF-A, VEGFR-1 and VEGFR-2 the cytoplasmic staining intensity was evaluated semi-quantitatively using a classification from 0 to 3, representing lack of staining (grade 0), low staining intensity (grade 1), intermediate staining intensity (grade 2), and high staining intensity (grade 3), as previously reported.15
Cell lines and transfection
MCF-7 and MDA-468 cell lines were cultured in DMEM supplemented with 10% fetal bovine serum (Gibco). MCF-7 cell line was transfected with pcDNA-GFP/periostin vector, which produces a periostin-green fluorescent protein (GFP) and with pcDNA-GFP as control. Both vectors have been kindly provided by Professor Inoue of Shiga University, Japan.16 The lipofectamine procedure used for transfections was performed according to the instruction manual (Invitrogen). MCF-7 cells were plated at 4×105 cells/35 mm culture dish 20 hours prior to transfection. For either plasmids, pcDNA-GFP/periostin vector and pcDNA-GFP, transfection was performed with 0.5 μg per dish. Cells were then analysed 48 hours after transfection with a fluorescence microscope, counterstaining nuclei with Hoechst 33258.
Quantitative PCR and western blot
Quantitative PCR analysis of periostin mRNA expression was performed as previously described.17 Real time PCR reactions were performed using the ABI Prism 7300 Sequence Detection System (Applied Biosystems, Foster City, California, USA). Oligonucleotide primers and probes for periostin were purchased from Applied Biosystems as Assays-on-Demand Gene Expression Products. Oligonucleotide primers and probe for the endogenous control β-glucuronidase (GUS) is described by Beillard et al.9 The ΔCT method, by means of the SDS software (Applied Biosystems), was used to calculate the mRNA levels.
For western blot analysis, 20 μg of either nuclear or cytoplasmic extracts obtained from MCF-7 and MDA-468 cell lines, were electrophoresed on 12% SDS-PAGE. Proteins were then transferred to nitrocellulose membranes and these were saturated by incubating for 1 hour with 5% non-fat dry milk in PBS/0.1% Tween 20. The membranes were then incubated with the rabbit polyclonal anti-periostin antibody overnight. After three washes with PBS/0.1 Tween 20, membranes were incubated with anti-rabbit immunoglobulin coupled to peroxidase (Sigma-Aldrich). After 2 hours of incubation the membranes were washed three times with PBS/0.1% Tween 20, and the blots were developed using the chemiluminescence procedure (Amersham Bioscence). Polyclonal anti-actin antibody was used as a control.
The χ2 and Fisher’s exact tests were used to compare the baseline characteristics (categorical variables) between the subgroups. Spearman’s rank test was used to assess the correlation between continuous variables. The Mann–Whitney test was used to evaluate if the sums of the rankings for two groups were different from an expected number (Kruskal–Wallis test was used in case of more than two groups). A value of p<0.05 was considered statistically significant.
In a set of preliminary experiments we first tested the specificity of periostin immunodetection by staining a 19-week-old human fetus and a 15-day-old mouse embryo. In both organisms the antibody stained the developing heart, specifically in the endocardial cushions (data not shown). Staining was also observed in osteoblasts; numerous chondrocytes showed staining on their surface and in the surrounding extracellular matrix; strong staining was observed within the perichondrium (data not shown). Thus, the antibody markedly stains structures where high expression of periostin has been previously reported,18 19 indicating the specificity of our procedure.
A total of 189 tumour samples were examined for periostin expression. The remaining 17 cases (8.25%), were not suitable for accurate analysis, mainly because of bad status of tumour blocks. Table 2 presents details of the patients and tumour characteristics.
In addition to the stroma, periostin was localised in carcinoma cells, but was absent in normal breast tissues (fig 1A). Cytoplasmic expression was found in 108/189 (57%) evaluable cases. However, in 22 cases (12%), a nuclear reactivity was observed. All cases with nuclear expression of periostin also showed cytoplasmic expression.
A significant correlation was found between cytoplasmic periostin expression and tumour size (ρ = 0.2, p = 0.001), progesterone receptor expression (ρ = 0.2, p = 0.008), VEGF-A (ρ = 0.36, p<0.001) and VEGFR-1 (ρ = 0.25, p = 0.001). No significant correlation or association was found between cytoplasmic periostin and other variables (tumour grade, nodal status, vimentin, ER status, HER-2 status, Ki-67 expression). Notably, when periostin expression was examined in the nuclear compartment, a significant correlation was observed with tumour size (ρ = 0.38, p<0.001), oestrogen receptor (ρ = 0.31, p<0.001), progesterone receptor (ρ = 0.33, p<0.001), VEGF-A (ρ = 0.46, p<0.001), VEGFR-1 (ρ = 0.48, p<0.001) and VEGFR-2 (ρ = 0.43, p<0.001).
To reinforce the notion that periostin may localise in the nucleus of breast cancer cells, continuous cell lines were investigated. A nuclear periostin localisation was evident in MCF-7 and MDA-468 cell lines (fig 1B,C). Immunocytochemical results were confirmed by western blot analysis of MCF-7 and MDA-468 cell extracts, fractionated into nuclear and cytoplasmic components. As shown in fig 1D, in both cell lines, periostin was localised either in nuclear and cytoplasmic fractions.
In order to corroborate immunohistochemical and western blot data, cellular localisation of periostin was investigated by the GFP approach. MCF-7 cells were transfected with plasmids pcDNA-GFP/periostin which produces a periostin-green fluorescent protein (GFP) or with pcDNA-GFP as control. As shown in fig 2, while GFP alone localises only in cytoplasm (fig 2A), the fusion protein periostin-GFP is evident in nucleus and in cytoplasm (fig 2B).
Periostin mRNA levels were also evaluated in MCF-7 and MDA-468 cell lines by a quantitative RT-PCR approach. As shown in table 3, both MCF-7 and MDA-468 cell lines express periostin mRNA. In MCF-7 cells periostin mRNA levels were about 50-fold lower than in MDA-468 cells, in agreement with immunohistochemical data (see fig 1). The low levels of periostin mRNA in MCF-7 cells may explain why other authors have not detected it in this cell line.20
Epithelial-mesenchymal transition (EMT) is a cellular event characterised by the conversion of epithelial cells to a mesenchymal phenotype.21 EMT has recently been proposed as a crucial process involved in the invasion and metastasis of some tumour cells. Specifically, expression of proteins that are characteristic of mesenchymal cells (vimentin, stromelysin-3, etc) and loss of epithelial markers (E-cadherin) has been described after tumour progression and development of metastasis.22
Periostin is a mesenchyme-specific protein that is implicated in regulating adhesion and differentiation of osteoblasts.23 24 In human tumours, periostin appears to be involved in the EMT and clinical studies have suggested a correlation between this protein, tumour angiogenesis and metastasis. In particular, periostin seems to play a role in different molecular mechanisms such as adhesion and migration of epithelial cells8 9 and angiogenesis, through regulation of the VEGF system.10
Recently, Grigoriadis et al6 found a significant association between periostin expression and poor outcome in ER-positive tumours, suggesting a possible explanation of the worse prognostic behaviour of luminal B breast cancer phenotype.
In our study, we investigated periostin expression in breast cancer and analysed potential correlations between periostin and some biological characteristics of these tumours. A positive correlation between periostin and hormonal receptors was found. In addition, periostin expression correlated with tumour size. Taken together, these observations may suggest that periostin expression in hormonal receptor positive tumours has an unfavourable prognostic role, as previously observed by Gregoriadis et al.6 Moreover, the correlation between periostin and molecules of the VEGF signalling system reinforce the notion that expression of this mesenchyme-specific gene may be a crucial step in tumour angiogenesis.11
We show that epithelial breast cancer cells contain the periostin protein, confirming previous data from other groups.6 This finding contrasts to that produced by Baril et al.25 These authors identified epithelial tumour cells as the unique source of periostin mRNA; however, they detected the protein only in the juxtatumoral stroma.
A major observation of our study was the nuclear localisation of periostin in a fraction of breast cancers as well in breast cancer cell lines. Moreover, nuclear periostin expression significantly correlated with tumour size, and with expression of oestrogen receptor, progesterone receptor, VEGF-A, VEGFR-1 and VEGFR-2. Since periostin has been uniquely considered to be a secreted protein, our finding indicating a nuclear localisation of this protein is surprising. However, we immunostained mouse and human embryos as control and detected strong periostin signal in locations where high periostin expression has been previously reported (see the Results section). Therefore, our results are unlikely to be due to cross-reactions of the antibody with unknown proteins. In agreement with our data, Yoshioka et al,16 using GFP-periostin fusion proteins in the COS7 cell line, have clearly shown that periostin is not only secreted but also localised in cytoplasm and nucleus. Our GFP-periostin fusion protein experiments strongly confirm the possibility of a nuclear localisation of periostin in breast cancer cells. Among factors that correlate with periostin expression in our patients, the relationship with VEGFR-2 has been previously shown to be causally related. In fact, Shao et al have shown that periostin secreted by MCF-7 cells is able to up-regulate VEGFR-2 in human microvascular endothelial cells.11 We found, however, a correlation between nuclear periostin and VEGFR-2 expression. Thus, our data may suggest that, in addition to a mechanisms based on extracellular signalling, periostin could induce VEGFR-2 expression more directly, by acting in the nucleus at the level of transcription or RNA maturation.
Modification of cell localisation is not uncommon in human tumours. It has been shown, for example, that the matrix metalloproteinase MMP3, which normally is a secreted protein, shows a nuclear localisation in hepatocarcinomas.26 Moreover, the extracellular matrix organiser and TGF-β partner decorin shows an aberrant nuclear localisation in dysplastic oral epithelial cells.27
Yoshioka et al have shown that periostin has a tumour suppressor activity in a human kidney cell line, which is mediated by the protein not secreted in the culture medium.16 Nevertheless, exposure of colorectal cancer cells to antiperiostin antibodies inhibits the colony forming ability of the cells and induces apoptosis.20 The latter observation suggests that several biological effects of this protein are mediated by the extracellular fraction. Altogether, these findings indicate that molecular mechanisms by which periostin exerts functional roles in human tumours should be re-evaluated.
This study indicates overexpression of periostin in breast cancer. It also shows that periostin can be aberrantly localised in the nucleus of breast cancer cells.
Nuclear periostin significantly correlated with tumour size, and with expression of the oestrogen receptor, progesterone receptor, VEGF-A, VEGFR-1 and VEGFR-2, suggesting that the nuclear localisation of this protein may induce relevant biological effects.
Further studies are required to clarify the role of the periostin nuclear localisation in breast cancer as well to reveal the causes of this phenomenon.
We thank Dr Parmjit S Jat for sharing data before publication. We also thank Professor Hirokazu Inohue for providing the GFP-periostin expression vector. This work was performed under the auspices of Associazione di Ricerca Traslazionale In Senologia (ARTIS).
Competing interests: None.
Funding: This work is funded by grants from MIUR (PRIN no. 2005060778-002) and Regione Friuli Venezia Giulia to GD. CP is supported by the Fondazione Italiana Ricerca sul Cancro (FIRC) fellowship.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.