Introduction

Oral squamous cell carcinoma (OSCC) is the most common malignant tumor in the head and neck region [1], and represents over 90% of oral cancers [2]. More than half of all oral cancers occur in the tongue, floor of the mouth, and gingiva, with areas of mechanical stimulation being the most frequently affected [3,4,5]. OSCCs have the same genetic profile beyond the sites including the tongue, floor of the mouth, and gingiva. Globally, the 5-year survival rate for oral cancer is 55–60%, while tumor diameter, presence or absence of lymph node metastasis, and presence or absence of distant metastasis are regarded to be the most important prognostic factors [6]. Additionally, perineural invasion [7, 8], vascular invasion [8], and bone infiltration [9] are also cited as prognostic factors and it is important that they each be evaluated histopathologically. On the other hand, as regards recurrence, oral epithelial dysplasia (OED) may progress to OSCC, and OED located in the proximity of OSCC is reportedly associated with an increased risk of local recurrence and progression to OSCC [10, 11]. The 5-year survival in such cases has been reported to be as low as 30% [12]. Since OED is closely involved in the development of OSCC, it is important to examine the mechanism of OED occurrence using a molecular pathological approach, and to identify the key molecules involved in carcinogenesis.

Approximately 12% of all OED progress to OSCC [13]. Tumor protein p53 (TP53) [14, 15], cyclin-dependent kinase inhibitor 2A (CDKN2A) [15], phosphatase and tensin homolog deleted from chromosome 10 (PTEN) [15], Harvey rat sarcoma (HRAS) [15], phosphatidylinositol-4, and 5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) [15] are all reported to be important genes linked to oncogenesis potentially; however, cases not dependent on these known factors are also known to exist, so it is important to search for other factors.

In this study, we extracted RNA from normal epithelium, dysplasia, and squamous cell carcinoma components and performed a microarray analysis to identify the key molecules involved in OSCC carcinogenesis and elucidate the molecular mechanism of OSCC pathogenesis. Of the potential key molecules identified in the assay, we focused on SUOX. Until now, there have been no reports on the role that SUOX plays in OSCC. In this study, SUOX expression was further evaluated in detail at the protein level, and its relationship with other clinicopathological features was examined.

Materials and methods

Microarray analysis

For the extraction of RNA, formalin-fixed paraffin-embedded (FFPE) blocks from 3 OSCC cases were used. The FFPE blocks were sectioned to a thickness of 7 μm using a Leica RM2245 microtome (Leica Microsystems K.K., Tokyo, Japan), with an RNase-free water-treated blade. For each case, one section was stained with hematoxylin eosin stain (H.E.) while three sections were used for extracting RNA. The normal epithelium, dysplasia, and squamous cell carcinoma components were marked using the previously mentioned H.E-stained specimen, superimposed on the unstained specimen, and each component was collected by scraping with a scalpel treated with RNase-free water. RNA from each component was isolated, linearly amplified, hybridized to the Affymetrix GeneChip Human X3P Array (Affymetrix, Santa Clara, CA, USA) and labeled in accordance with the manufacturer’s instructions for the Arcturus Paradise PLUS Reagent System (Life Technologies, Grand Island, NY, USA) and GeneChip 3′ IVT Express Reagent kit (Affimetrix). The extracted RNA was measured using Nano drop ® ND 1000 (Thermo Fisher Scientific, Waltham, MA, USA). Affymetrix array CEL files were processed by the RNA algorithm [16] to obtain probe set-level gene expression data, using the Expression Console software (Affymetrix).

Patients

We selected patients with primary tongue cancer who had not undergone preoperative treatments such as chemotherapy and/or radiotherapy at Kurume University Hospital between 2010 and 2015. Fifty-eight patients filled this condition. Of these 58 patients, 30 were male and 28 were female. The mean age was 63.4 ± 15.3 (range 28–90). The mean tumor size was 18.3 ± 9.32 mm (range 2–45 mm). All cases accompanied dysplastic components adjacent to carcinoma components. Twenty-six patients were classified T1, 31 were T2 and 1 was T3. Regarding histological grading, 47 were well-differentiated, 10 were moderately-differentiated, and 1 was poorly-differentiated. Thirteen cases had lymphatic vessel invasion, while four cases had vascular invasion (Table 1).

Table 1 Clinicopathological features of the 58 oral squamous cell carcinoma cases
Full size table

The excised tissues were fixed using 10% buffered formalin, sectioned at 4-µm-thickness, followed by HE staining. Histopathological evaluations were performed by three pathologists (K.N., J.A. and H.Y.). Pathological diagnosis was performed according to the WHO classification of Head and Neck Tumors 4th Edition [6].

This study was approved by the ethics committee of Kurume University (#330).

Immunohistochemistry

We performed immunohistochemistry (IHC) on paraffin-embedded sections using the SUOX antibody (ab88346, dilution 1: 300, Abcam plc., Cambridge, UK). IHC was performed using the Ventana Benchmark (Ventana, Tucson, AZ). All IHC analyses were evaluated by two experienced observers who were unaware of the patients’ clinical conditions. We considered only nuclear expression of SUOX as positive. We used the Allred score system [17] in the staining evaluation to calculate a total score (TS) from a population score (PS) and an intensity score (IS). Allred score is usually used in breast carcinoma, and is in various carcinomas including OSCC [18, 19]. This system is easy to learn and highly reproducible [20]. Briefly, a PS was assigned representing the estimated proportion of positive staining cells (0 = none; 1 = < 1/100; 2 = 1/100 to < 1/10; 3 = 1/10 to < 1/3; 4 = 1/3 to < 2/3; 5 = > 2/3). Average estimated intensity of staining in positive cells was assigned an IS (0 = none; 1 = weak; 2 = intermediate; 3 = strong). PS and IS were added to obtain a total score that ranged from 0 to 8. We compared the expression of SUOX in each component and examined the correlation between the SUOX staining intensity of the carcinoma component and clinicopathological factors, such as tumor size, differentiation (well / moderately to poor), T classification, lymphatic vessel invasion, and vascular invasion.

Statistical analysis

The inter-rater reliability coefficient for TS was calculated for two doctors (K.N and M.N). The inter-rater reliability coefficient was κ = 0.8 (excellent). Additionally, the scores for (K.N) were statistically analyzed and the results were compared with those for the other pathologist (M.N). It was confirmed that there was no significant difference between the resulting scores. Differences in expression intensity for normal epithelium vs dysplasia components as well as for dysplasia vs squamous cell carcinoma components were evaluated by T-test, with p < 0.05 considered significant. Correlation between TS and clinicopathological features (tumor size: mm, T classification, histological grade, lymphatic vessel invasion, vascular invasion) in the carcinoma component was evaluated using a bivariate logistic regression model (p < 0.05). The statistical software used was JMP® Pro 13 (SAS Institute Inc., Cary, NC, USA).

Results

Microarray analysis

Table 2 shows the clinicopathological features and RNA yield for each component of 3 cases. The expression of 61,298 molecules was confirmed by microarray analysis, and a heat map was prepared (Fig. 1). Among these molecules, we focused on 22 that showed significant differences between normal epithelium vs dysplasia component and between dysplasia component vs squamous cell carcinoma component. Carcinogenesis was associated with decreased expression in four molecules and increased expression in 18 molecules (Table 3). Of the 22 molecules, SUOX showed the most significant differences in expression between each group. Exact p-values of SUOX for normal epithelium vs dysplasia component, normal epithelium vs carcinoma and dysplasia component and carcinoma were p = 0.023, 0.018 and 0.014, respectively. Moreover SUOX expression decreased gradually in the order of normal epithelium to dysplasia component to squamous cell carcinoma component (Table 3). These changes were the highest in SUOX and statistically significant. Therefore, we selected SUOX for further examination.

Table 2 Clinicopathological features and the amount of RNA in each component of 3 cases
Full size table
Fig. 1
figure 1

Heat map based on microarray analysis. Molecules with different levels of expression in normal epithelium (N), dysplasia (D) and squamous cell carcinoma (Ca), was identified by microarrary analysis

Full size image
Table 3 Molecules that showed significant differences in expression and their differences among normal epithelium (N), dysplasia (D) and squamous cell carcinoma (Ca) components by RNA microarray analysis
Full size table

Immunohistochemical analysis

Immunohistochemistry was performed using samples from the aforementioned 58 tongue cancer patients. In normal sections, SUOX was mainly expressed in the nuclei of intermediate and basal layer squamous cells, but there was little expression on the surface layer.

On the other hand, in the dysplastic epithelia, expression was not readily seen in the basal layer, but was mainly observed in cells from the intermediate to surface layer, while in the carcinoma component nucleic expression was almost completely absent (Fig. 2a). Further, the intensity of SUOX expression by Allred score was normal epithelium = 7.2 ± 0.107 (range 6–8), dysplasia = 5.27 ± 0.153 (range 3–7), squamous cell carcinoma = 3.17 ± 0.274 (range 0–6), and there was a significant difference (p < 0.01) between each component (Fig. 2b).

Fig. 2
figure 2

A Immunohistochemical findings of SUOX in normal epithelium, dysplasia and carcinoma components. Expression of SUOX was observed in the nuclei from the basal layer to the intermediate layer of the normal squamous epithelium (a, d). In the dysplasia component, expression was observed mainly in the nucleus of atypical cells from the intermediate layer to the surface layer, but was not readily observed in nuclei of the basal layer (b, e). Conversely, almost no expression was observed in the carcinoma component (c, f). (Hematoxylin eosin a–c ×200, SUOX d–f ×200) Scale bar 50 μm. b Comparison of total score (TS) of SUOX expression among normal epithelium (N), dysplasia (D) and squamous cell carcinoma (Ca) components. TS of SUOX of each component was significantly and gradually decreased as carcinogenesis progressed (*p < 0.01). TS of N, D and Ca was 7.29 ± 0.107, 5.27 ± 0.153 and 3.17 ± 0.274, respectively

Full size image

As for carcinoma, nucleic expression was scatteringly observed in T1 squamous carcinoma but rarely in ≥ T2 squamous carcinoma (Fig. 3a). TS in the carcinoma component was 3.84 ± 0.39 in T1, and 2.62 ± 0.35 ≥ in ≥ T2 squamous carcinoma, and this decrease in SUOX expression from T1 to ≥ T2 was significant (p = 0.025) (Fig. 3b).

Fig. 3
figure 3

A Immunohistochemical findings of SUOX in T1 and ≥ T2 oral squamous cell carcinoma. Tumor cells with nuclear SUOX expression are scatteringly observed in a case of T1 squamous cell carcinoma (a, c), while a case of ≥ T2 squamous cell carcinoma shows almost no SUOX expression (b, d). (Hematoxylin eosin: a, b ×200, SUOX: c, d ×200) Scale bar indicates 50 μm. b Comparison of total score (TS) of SUOX expression between T1 and ≥ T2 oral squamous cell carcinoma cases. TS average was 3.84 ± 0.39 in T1 squamous cell carcinoma, and 2.62 ± 0.35 in ≥ T2 squamous cell carcinoma. This difference in SUOX expression by T classification was significant (p = 0.025)

Full size image

In addition, when clinicopathological features and TS were examined in the carcinoma component, a significant negative correlation was found between SUOX expression and tumor diameter. However, TS in carcinoma component showed no significant correlation with any other factors, including degree of differentiation or vascular invasion.

Discussion

Many studies have examined the relationship between various genes and carcinogenesis in OSCC. However, most of these reports have focused either on expression occurring exclusively in the carcinoma [21,22,23,24,25] or on comparisons between non-cancerous and cancerous components [26,27,28]; few have investigated for factors that showed significant gradual differences among normal, dysplastic, and carcinoma components [29]. For this reason, we extracted and comprehensively examined RNA from normal, dysplastic, and carcinoma components using microarray analysis. Based on our results, we isolated SUOX, a factor that exhibited a significant and gradual decrease in expression from normal epithelium to dysplasia component to squamous cell carcinoma component, and evaluated its clinicopathological significance.

SUOX is a metallo-enzyme present in the mitochondria of all eukaryotes that utilizes, as coenzymes, molybdenum and heme, [30]. Via cytochrome c, it transfers electrons, produced through the oxidation of sulfurous acid into sulfuric acid, to the electron transport for use in ATP synthesis through oxidative phosphorylation [31,32,33]. This is the final step in the metabolism of sulfur-containing compounds, and sulfite is excreted [34]. In general, SUOX deficiency is the most common disease associated with SUOX [30, 35, 36] and there are few studies of SUOX in malignant tumors. In particular, though there have been no reports examining the expression of SUOX in OSCC, it has been reported in recent years that carcinogenesis of liver tumors is accompanied by diminished SUOX expression [37]. Jin GZ et al. demonstrated that the expression of SUOX decreased in a stepwise fashion along with the carcinogenic process and, furthermore, the expression of SUOX decreased along as tumor size increased [37]. Although SUOX in OSCC shows similar kinetics in liver tumor, it is difficult to compare the role of SUOX between OSCC and liver tumor as the mechanisms of carcinogenesis and the processes of tumor progression are different. Out of 22 genes, 17 were reported as cancer-related genes [19, 37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53]. However, the number of reports on each gene is limited. Extensive investigation on these genes has not been conducted so far.

Many immunohistochemical studies on OSCC have been conducted and many biomarkers have been reported to be associated with various clinicopathological factors, including prognosis. Most of the biomarkers demonstrated that higher expressions were associated with aggressive clinicopathological factors. These types of biomarkers have been listed as follows: epithelial cell adhesion molecule, CD44s, cyclooxygenase-2, autophagy-related 16-like 1, glucose-regulated protein 78, cysteine-rich 61, Aurora B, urokinase-type plasminogen activator receptor [54,55,56,57,58,59,60,61]. On the other hand, the expressions of SUOX decreased with tumor progression and showed an inverse association with tumor size. These types of biomarkers on OSCC were relatively limited. Deleted in liver cancer, keratin 13 and matriptase-2 were reported as these types of biomarkers [29, 62, 63].

In this study, we demonstrated that the reduction of SUOX expression was significantly associated with the stage but not with the tumor differentiation. Most biomarkers described before were reported to be related with several worse clinicopathological factors, including the stage and the tumor differentiation. A few markers, such as cysteine-rich 61 and Aurora B, were associated with only the stage and not with the tumor differentiation as well as SUOX [59, 60]. On the other hand, the expression of urokinase-type plasminogen activator receptor showed significant correlations with the tumor differentiation but not the tumor size. The reasons for these differences to the findings of our study are still unclear. Patient background, race, smoking and/or alcohol intake and cohort size might influence these differences.

The mechanisms by which down-expression of SUOX affected the carcinogenesis remain poorly understood. The following possibilities were considered. First, SUOX is an enzyme present in all normal eukaryotic mitochondria and is involved in ATP synthesis through oxidative phosphorylation. On the other hand, in cancer cells oxidative phosphorylation is suppressed and ATP is produced by the glycolytic pathway [64, 65]. Second, other known or unknown molecules that are altered with carcinogenesis in OSCC may regulate SUOX expression. Indeed, we found that the expression level of some molecules was correlated with SUOX alteration in our microarray analysis; however, the exact mechanisms are still unclear as we did not confirm the correlation between them by IHC.

It is possible that the difference in metabolic pathways between these normal cells and cancer cells may be involved in the reduction of SUOX expression in cancer cells.

The fact that SUOX acts to suppress carcinogenesis in tongue cancer suggests that SUOX may be a key molecule useful in clarifying patient condition and diagnosis, or as a potential target for treatment. There are many aspects of the role of SUOX in cancer that are yet unclear. Although not a detailed examination, our present microarray study identified multiple molecules either positively or inversely correlated with SUOX. Detailed in vitro and in vivo studies, involving factors related to SUOX, and not limited to OSCC, but targeting multiple carcinomas, are necessary to further clarify the specific functions of SUOX in relation to cancer.