Expression of ABCG2 and Bmi-1 in oral potentially malignant lesions and oral squamous cell carcinoma

Early diagnosis is vital for effective treatment of oral squamous cell carcinoma (OSCC). The optimal time for clinical intervention is prior to malignancy when patients present with oral potentially malignant lesions such as leukoplakia or erythroplakia. Transformation rates for oral dysplasia vary greatly and more rigorous methods are needed to predict the malignant potential of oral lesions. We hypothesized that the expression of two putative stem cell markers, ABCG2 and Bmi-1, would correlate with disease severity for non diseased, potentially malignant and OSCC specimens and cell lines derived from an equivalent range of tissues. We compared immunoreactive protein and relative gene expression of ABCG2 and Bmi-1 in eight cell lines derived from source tissues ranging in disease severity from normal (OKF6-TERT2) through mild and moderate/severe dysplasia (DOK, POE-9n) to OSCC (PE/CA-PJ15, SCC04, SCC25, SCC09, SCC15). We also analyzed immunoreactive protein expression of ABCG2 and Bmi-1 in 189 tissue samples with the same range of disease severity. A trend between oral lesion severity to ABCG2 and Bmi-1 immunostain intensity was observed. Flow cytometry of oral cell lines confirmed this trend and gave good correlation with RT-PCR results for ABCG2 (r = 0.919, P = 0.001; Pearson) but not Bmi-1 (r = −0.311). The results provide evidence of increased density of ABCG2 and Bmi-1-positive populations in malignant and oral potentially malignant lesions and derived cell lines, but that intragroup variability within IHC, flow cytometry, and RT-PCR results compromise the diagnostic potential of these techniques for discriminating oral dysplasia from normal tissue or OSCC.


Introduction
Early diagnosis is of paramount importance for effective treatment of oral squamous cell carcinoma (OSCC), the incidence of which is rising in the developed world among a younger demographic that have less association with known risk factors such as excessive alcohol or tobacco consumption [1]. The optimal time for clinical intervention is prior to malignancy when patients present with oral potentially malignant lesions (OPML) such as oral leukoplakia or erythroplakia. The presence and severity of dysplasia in OPMLs is associated with poorer clinical outcomes [2,3], but more rigorous methods are needed to predict the malignant potential of oral lesions as rates of malignant progression from oral dysplasia vary greatly [4].
Tumors often contain hierarchical arrangements of genetically or epigenetically dissimilar tumorigenic and non tumorigenic cells [5][6][7]. In these cases, a theoretic subpopulation of tumorigenic cells termed Cancer Stem Cells have been identified through detailed investigations with disaggregated tumor-derived cells, NOD/SCID mouse transplant models [8] and antigenic [7,9,10] markers of cell phenotype. The acquisition of a genetic or epigenetic predisposition to tumorigenicity by cancer stem cells is a cumulative process [5] that may be evident in premalignant lesions. The association of putative stem cell antigens with oral potentially malignant lesions

273
Cancer Medicine Open Access [11,12] and oral epithelial tumorigenicity [10] is the subject of a recent review [13].
A recent long-term follow-up study of 135 patients associated two stem cell markers, ATP-binding cassette G2 subfamily (ABCG2) and Bmi-1 with transformation of oral leukoplakia to cancer [2]. ABCG2 belongs to the ATP-binding cassette transporter superfamily that contributes to multiple drug resistance in cancers and permits cells in vitro to export the fluorescent dye Hoechst 33342, a technique used to define a side population of purported stem cells [13][14][15][16]. Bmi-1 (B lymphoma Mo-MLV insertion region 1 homolog) contributes to the epigenetic regulation of cell cycle and senescence of tissue stem cells via the INK4A loci [17]. Bmi-1 is oncogenic and has been associated with purported cancer stem cells and cell lines derived from head and neck squamous cell carcinoma [10,13,16,[18][19][20][21][22].
We hypothesized that ABCG2 and/or Bmi-1 expression would correlate with disease severity for a collection of cell lines derived from non diseased, potentially malignant, and OSCC specimens, and that these correlations would also hold for patient samples with a similar range of oral disease severity. Our aim was to ascertain whether in vitro analysis with established cell lines provides a valid model for further investigation of the mechanistic involvement of ABCG2 and/or Bmi-1 in neoplastic transformation of oral dysplasia to OSCC, and to evaluate whether immunohistochemical analysis of ABCG2 and/or Bmi-1 has diagnostic potential for discriminating oral dysplasia from normal tissue or OSCC. Accordingly, we compared ABCG2 and Bmi-1 antigen positive population densities among eight cell lines by flow cytometry, correlated this with RNA expression analysis of the gene products, and conducted a confirmatory retrospective cohort study of ABCG2 and Bmi-1 antigen stain intensity in 189 patient samples.

Patient samples
Of 189 formalin-fixed paraffin-embedded (FFPE) specimens (80 male, 61 female and 48 gender not recorded; mean age 54 years, range 18-86 years) comprised: 74 normal oral mucosa (referred to herein as normal), 17 mild dysplasia (MD), 17 moderate to severe dysplasia (SD), and 81 oral squamous cell carcinoma (OSCC). Sample diagnosis was confirmed retrospectively by an oral pathologist (CSF) according to the World Health Organization (WHO) classification system [24]. The study received ethical review and approval from Hospital and University Human Research Ethics Committees (HREC/10/QRBW/ 336 and UQ/2007001478).

IHC Scoring
Ten random fields were scored at 9400 magnification, counting >500 cells. Sections were scored for stain intensity as described previously [12,25] (scaled as: 0 = no stain to 4 = dark brown) and the percentage of positively stained epithelial cells (categorized as: 0% = score 0; <25% = score 1; 25-49% = score 2; 50-74% = score 3; and 75-100% = score 4) was calculated. The final index score, which ranged from 0 to 16, was the product of categorized percentage positive score and the scaled stain intensity score. (IQR) or mean ( x) and standard deviation (SD) have been presented. Immunolocalization score data for disease groups were compared to normal using Mann-Whitney U tests with Bonferroni correction for n = 3 comparisons (P < 0.016). Pearson product-moment correlation coefficients were used to compare average real-time PCR results with average flow cytometry results. With the exception of SCC09 and SCC15, all of the cell lines exhibit characteristic epithelial morphology and growth features, forming monolayers with squamous appearance, restricted intercellular space, regimented intercellular cohesion, and clonal growth. SCC09 and to a lesser extent SCC15 exhibited more of a mesenchymal morphology, being multipolar with greater intercellular space and exhibited a dispersed growth pattern (Fig. 1).
ABCG2 and Bmi-1 gene expression in normal, dysplastic, and OSCC-derived cell lines Semi-quantitative real-time PCR analysis of ABCG2 and Bmi-1 gene expression in eight oral cell lines originating from normal tissue (OKF6-TERT2), mild dysplasia (DOK), severe dysplasia (POE-9n), and oral squamous cell carcinoma (PE/CA PJ15, SCC04, SCC25, SCC09 and SCC15) are provided in Figure 2 and Table 2. Good correlation between relative quantity of mRNA for ABCG2 ( Table 2) and mean percent ABCG2-positive cell populations (Table 1) was attained across the eight cell lines (r = 0.919, P = 0.001; Pearson product-moment correlation coefficient). There was, however, no correlation between relative quantity of mRNA for Bmi-1 and mean percent Bmi-1-positive cell populations (r = À0.311, P = 0.453; Pearson). The mild dysplasia-derived cell line (DOK) and an OSCC-derived cell line (PE/CA PJ15) were exceptional in having very high ABCG2 gene expression relative to control (OKF6-TERT2); another OSCC-derived cell line (SCC04) gave the lowest relative expression of ABCG2 (Table 2). Overall, there were no discernible trends between ABCG2 expression and source tissue disease severity (Fig. 2). For Bmi-1 gene expression, only the OSCC-derived SCC04 cell line exhibited lower Bmi-1 expression than control (OKF6-TERT2), all other cell lines exhibited two or three times higher Bmi-1 expression relative to control (Table 2), without displaying discernible trends between Bmi-1 expression and source tissue disease severity (Fig. 2).

ABCG2 and Bmi-1 antigen stain intensity in normal, dysplastic, and OSCC biopsies
Representative immunostaining for ABCG2 and Bmi-1 are shown in Figure 3, and nonparametric descriptive statistics for ABCG2 and Bmi-1 immunoreactivity index scores are shown in Figure 4.

Discussion
In this study, we analyzed immunoreactive protein and relative gene expression of ABCG2 and Bmi-1 in a panel of eight cell lines derived from oral tissues with a spectrum of diseases ranging from normal through mild and moderate/severe dysplasia to oral squamous cell carcinoma (OSCC). In addition, we analyzed immunoreactive protein expression of ABCG2 and Bmi-1 gene products in an archive of 189 formalin-fixed paraffin-embedded (FFPE) patient oral tissue samples bearing the same range of disease severity as the cell lines. Our aim was to ascertain whether in vitro analysis with established cell lines provides a valid model for further investigation of the mechanistic involvement of ABCG2 and Bmi-1 in neoplastic transformation of oral dysplasia to OSCC, and to evaluate whether immunohistochemical analysis of ABCG2 and Bmi-1 has diagnostic potential for discriminating oral dysplasia from normal tissue or OSCC. Overall, the results indicate that cell populations that are positive for the putative cancer stem cell markers ABCG2 and Bmi-1 can occur with increased density in malignant or oral potentially malignant lesions, and cell lines derived from them. This indicates that carefully selected and maintained oral cell lines provide a valid model for investigating ABCG2 and Bmi-1 involvement in neoplastic transformation of oral dysplasia to OSCC, but that   (14) intragroup variability within IHC, flow cytometry, and RT-PCR results could compromise the diagnostic potential of the techniques for discriminating oral dysplasia from normal tissue or OSCC. Flow cytometric analysis of oral lesion-derived cell lines gave a convincing correlation between source tissue disease severity and increased ABCG2 and Bmi-1 expression, yet even among five OSCC-derived cell lines there was considerable variability in the expression of ABCG2 and Bmi-1. Semi-quantitative real-time PCR results correlated well with mean percent positive cell population flow cytometry data for ABCG2 (r = 0.919, P = 0.001; Pearson) but not with Bmi-1 (r = À0.311, P = 0.453; Pearson) for the eight oral cell lines. Immunohistochemical study of FFPE patient biopsies demonstrated a limited trend between the diagnostic grading of oral lesion severity to ABCG2 and Bmi-1 immunostain intensity scores, the power of which was compromised by considerable intragroup variability. Association of ABCG2 and Bmi-1 with transformation of oral potentially malignant lesions to cancer has been demonstrated in a recent long-term follow-up study of 135 patients with oral leukoplakia (mean (SD) duration follow-up: 5.6 (3.7) years) [2]. This earlier study reported increased risk of malignant transformation for lesions that were either ABCG2 positive (3.24-fold increase) or Bmi-1 positive (4.03-fold increase), and a statistically significant, although numerically negligible (r = 0.195), correlation between ABCG2 and Bmi-1 immunoscore (P = 0.24; Pearson correlation coefficient, 0.195) [2]. Overall, the pattern of staining reported previously shows the same trend as reported here, with increased ABCG2 and Bmi-1 immunoscore coinciding with increased disease severity; and a statistically significant, although numerically negligible, correlation between ABCG2 and Bmi-1 expression across the entire cohort (r = 0.144, P = 0.048; Pearson product-moment correlation coefficient). We are unable to comment on the ability of ABCG2 and Bmi-1 to predict transformation of oral dysplasia tissues to OSCC because few of the patients were biopsied more than once, and we are unable to comment upon survival rates for the OSCC patient groups as these data lie beyond our ethical constraints.
Irrespective of the tissue diagnosis, immunoreactivity index scores were generally higher for Bmi-1 than ABCG2. The opposite trend was predominant for cell lines analyzed by flow cytometry, with five lines (OKF6-TERT2, POE9n, DOK, PE/CA PJ15 and SCC25) exhibiting larger proportions of ABCG2 positive than Bmi-1 positive populations, two lines (SCC09 and SCC15) exhibiting equivalent ABCG2 and Bmi-1 populations, and just one line (SCC04) exhibiting predominance of Bmi-1 over ABCG2. Difference between ex vivo and in vitro predominance of one antigen over another may reflect either an inherent feature of the cell lines or an influence of the cell culture conditions. Routine submerged monolayer culture tends to suppress oral epithelial cell differentiation, has the propensity to select for proliferative subpopulations and may preferentially select for cells with repressed differentiation pathways. Alternatively, organotypic culture of oral cell lines permits the establishment of tissue hierarchy and generates patterns of putative stem  cell marker expression that are more representative of ex vivo tissues [23]. In this study, cells for flow cytometry were harvested from subconfluent cultures in post logarithmic growth phase at which point the influence of contact inhibition on cellular growth and differentiation are minimal. Future related work should include study of organotypic cultures by IHC and rtPCR, noting that organotypic cultures are not readily compatible with flow cytometry. This study was successful in using a diverse panel of oral cell lines to demonstrate a trend between source tissue disease severity and expression of the putative stem cell markers: ABCG2 and Bmi-1. Presence and absence of mRNA to protein correlation for ABCG2 and Bmi-1, respectively, indicated that rtPCR could be used in place of flow cytometry in future oral cell line-based studies of ABCG2 but not Bmi-1 expression. Confirmation of the trend between putative stem cell marker expression and oral lesion severity was achieved with a cohort of 189 FFPE patient biopsies. We observed discrepancy in ABCG2 versus Bmi-1 predominance between cell line and patient biopsy investigations which lead us to speculate whether organotypic [23] rather than monolayer culture may be preferable for future cell line-based investigations. Immunoreactivity index scoring of 189 patient biopsies revealed a trend between diagnostic grading of oral lesion severity and expression of the putative stem cell markers: ABCG2 and Bmi-1. These findings complement the previously reported potential for ABCG2 and Bmi-1 immunostaining to predict malignant transformation from oral leukoplakia [2], but our data indicate that intragroup variability within IHC, flow cytometry and RT-PCR results may compromise the diagnostic potential of these techniques for discriminating oral dysplasia from normal tissue or OSCC. Overall, the results provide evidence of increased density of ABCG2 and Bmi-1 positive populations in malignant and oral potentially malignant lesions and cell lines derived from them.