Discovery of Novel Biomarkers in Oral Submucous Fibrosis by Microarray Analysis

Oral submucous fibrosis (OSF) is a chronic, insidious, and progressive oral mucosal disease that primarily affects any part of the oral cavity (1). It is characterized by a juxta-epithelial inflammatory reaction followed by progressive fibrosis of the lamina propria and the underlying submucosal layer, with associated epithelial atrophy. This always leads to the stiffness of the oral mucosa and the restriction of mouth opening, eventually impairing the ability to eat, the ability to speak, and dental care (2).

Although the etiology of OSF is obscure, evidence has shown that it is a precancerous disorder related to the habit of chewing areca nut, either alone or as a component of betel quid (3). There are ∼600 million people worldwide amounting to 10% to 20% of the world's population who chew raw areca nut or in any processed form (4). Several case-control studies provide overwhelming evidences that areca nut is the main risk factor for OSF in Hunan of China (5, 6). Thus, the disease is now a public health issue in many parts of the world.

OSF carries a high risk of transition to oral cancer. In an epidemiologic study in India, the malignant transformation rate was 7.6% over a period of 17 years (7). Moreover, more than 2,400 new cases of oral cancer arising from OSF are diagnosed every year in Taiwan due to the prevalent use of betel quid (8). In another study by Jian, three cases of oral cancer were found in 147 cases of OSF in Hunan (9). Therefore, early diagnosis of this potentially malignant oral lesion is very important and effective.

Much efforts have been devoted into researching the underlying molecular mechanisms of OSF. However, understanding the differences in gene expression between OSF and normal tissue is important for the research. High-throughput oligonucleotide microarrays can perform analysis of such differences at a transcriptional level to know expression profiles for thousands of genes simultaneously and to characterize the biological behaviors of cell in a single experiment. To explore OSF biology and search for genes with differential expression to represent diagnostic as well as therapeutic biomarkers for OSF, we used microarray analysis to screen genes deregulated in OSF and shed some light on the molecular mechanism for its etiology and pathogenesis. Five novel genes identified in our study may be clinically useful for detection of OSF. Their functional implication of pathogenesis could provide a basis for better understanding of the molecular mechanisms underlying the development of OSF.

Under an ethical guideline of the Central South University Ethics Committee, 29 patients with clinically defined OSF lesions were recruited from the Department of Oral and Maxillofacial Surgery, Xiangya Hospital, Changsha, China. Among them, eight patients who had previous local treatments for oral mucosa or underwent systemic diseases (hepatitis B, diabetes) were excluded, and two patients refused to accept the biopsy. Eventually, from the remaining 19 untreated primary patients, we obtained 19 biopsy mucosa samples of the buccal lesion area, which is the mainly affected site of OSF. All the patients had the habit of areca nut chewing. Because there is no way to collect the matched sample in the same patient because of the special characteristics of OSF, 14 unmatched normal buccal mucosa (NBM) tissues were procured from 14 healthy volunteers without the habit of areca nut chewing and systemic diseases, who underwent surgery for third molar impactions or road traffic accidents. Informed consent was obtained from all donors. Immediately after surgical removal and wash, each sample was divided into two parts: one part was fixed in 10% neutral-buffered formalin for 24 to 48 h and embedded in paraffin for pathologic review by an experienced investigator according to the concept of Pindborg and Sirsat (2); the residual sample was snap frozen with liquid nitrogen and stored in a −80°C refrigerator after we removed partial submucous tissue to ensure the same thickness of all biopsies (∼2 mm). In this procedure, the removal of part connective tissue could lead to losing some information of genes in deeper connective tissues but still remain the key parts (epithelium, lamina propria, and adjacent connective tissue) of OSF mucosa for our study. After a final pathologic review, 14 of 19 biopsies were pathologically confirmed as OSF lesion, two samples were oral lichen planus, two were inflammatory tissues, and the last was scored for dysplasia. All control samples were assessed as NBM tissues. In addition, 66 paraffin-embedded buccal mucosa specimens of OSF patients with the habit of areca nut chewing were randomly drawn and reconfirmed from the files of the Department of Oral Pathology between March 2007 and June 2007. Patient age, gender, duration of areca nut chewing, duration of OSF disease, and histopathologic grade were used as clinicopathologic variables (Supplementary Table S1).

Total RNA was extracted from OSF and normal biopsies using Trizol reagent (Invitrogen). RNA quality and purity were assessed by 1% agarose gel electrophoresis. Total RNA with A₂₆₀/A₂₈₀ >1.8 was used for microarray experiments.

A global expression analysis of buccal mucosa samples from 4 OSF patients (4 males; age range 22-36 years old; median age 29.5 years old; the histopathologic stage early stage in 1 case, moderately advanced stage in 2 cases, advanced stage in 1 case), and 4 healthy volunteers (4 males; age range 16-30 years old; median age 20.7 years old) was done in microarray analysis. Briefly, total RNA from buccal mucosa of either the 4 volunteers or the 4 OSF patients was prepared to cRNA, biotin labeled, and hybridized to commercially available high-density oligonucleotide microarrays (human genome U133A 2.0 Gene Chips, Affymetrix, Inc.), containing ∼18,400 transcripts and 14,500 genes. Then, the arrays were scanned using GeneChip Scanner 3000 (Affymetrix). The hybridization data were analyzed using GeneChip Operating software (GCOS 1.4). The raw signals of individual probes for the eight arrays were normalized against the chip with median raw signal intensity using the dChip software (dChip2006).

To identify genes with the most differentially altered expression in OSF, we focused on the results of supervised analysis with the Significance Analysis of Microarrays (SAM) software. Normalized expression values from dChip analysis were used for a two class paired SAM analysis. The SAM software estimated the false discovery rate and generated a q value for each gene, which can indicate a more significantly differential expression than the P value. In the present study, the SAM false discovery rate accepted in the analysis was <0.01, and those showing at least a 2-fold induction or 0.5-fold repression (q < 0.01) in their expression were selected as deregulated genes in our study. Three up-regulated genes with the most significantly differential changes and top two down-regulated genes in OSF versus NBM were selected for next validation. Besides, clustering analysis was done using Gene Cluster 3.0 software and TreeView 1.6 with average linkage method and correlation.

For semiquantitative reverse transcription-PCR (RT-PCR), 1 μg of total RNA each from 10 normal and 10 OSF samples was used as template for reverse transcription using PrimeScrip 1st Strand cDNA synthesis Kit (Takara Bio, Inc.). We then used gene-specific primers for five genes: Loricrin, forward 5′-CAGGTCACTCAGACCTCGT-3′, reverse 5′-CCTCCAGAGGAACCACCT-3′, product sizes 200 bp; COMP, forward 5′-ACGTGGTCTTGGACACAAC-3′, reverse 5′-TCATAGTCCTCTGGGATGGT-3′, product sizes 121 bp; CXCL9, forward 5′-CTGTGGCCAGAATTTAAACC-3′, reverse 5′-ATGCAAGGTAAGTGGGTCAC-3′, product sizes 201 bp; KRT19, forward 5′-AGGGTCTTGAGATTGAGCTG-3′, reverse 5′-CTCACTATCAGCTCGCACAT-3′, product sizes 161 bp; CYP 3A5, forward 5′-ACACCCTTTGGAACTGGAC-3′, reverse 5′-GAAGAAGTCCTTGCGTGTCT-3′, product sizes 154 bp. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) amplification from the same cDNA samples served as internal controls, whose specific primer was forward 5′-ACCACAGTCCATGCCATC-3′, reverse 5′-TCCACCACCCTGTTGCTG-3′, product sizes 452 bp. The PCR products were visualized by UV illumination on a 1.5% agarose gel. Signals were captured with the Multi Genus Bio Imaging System and signal intensity was analyzed by the GeneTools software (Synoptics, Ltd.).

For Western blotting, proteins (30 μg) of OSF samples and normal controls were separated by 12% SDS-PAGE and transferred onto a polyvinylidene fluoride membrane. After being blocked, filters were incubated with the following primary antibody: rabbit polyclonal anti-Loricrin (Abcam; 1:1,000 dilution), rat monoclonal anti-COMP (Genetex; 1:1,000 dilution), mouse monoclonal anti-CXCL9 (R&D; 2 μg/mL dilution), mouse monoclonal anti-KRT19 (Abcam; 1:1,000 dilution), and rabbit polyclonal anti-CYP3A5 (Biomol; 1:2,000 dilution). Then, the horseradish peroxidase–conjugated secondary antibody (Santa Cruz Biotechnology) was applied onto the filter at 1:2,000 dilution. As an internal control, samples were probed with mouse monoclonal anti–β-actin antibody (BD Biosciences). Bands were visualized using the ECL system (Amersham) and signal intensity was analyzed by the Bandscan software (Glyko).

Immunohistochemical studies were done using the avidin-biotin-peroxidase method. Briefly, serial 3-μm-thick sections were mounted on silanized slides. After being deparaffinized and rehydrated, the sections were subjected to 2 min heat-induced antigen retrieval or 15 min pepsin antigen retrieval (only for KRT19). After being blocked by 3% hydrogen peroxide, the sections were incubated with primary antibodies: polyclonal anti-Loricrin (Abcam; 1:500 dilution), monoclonal anti-COMP (Genetex; 1:10 dilution), monoclonal anti-CXCL9 (R&D; 10 μg/mL dilution), monoclonal anti-KRT19 (Abcam; 1:1,000 dilution), and polyclonal anti-CYP3A5 (Biomol; 1:2,000 dilution). Then, the slides were incubated for 30 min with the biotinylated IgG (Santa Cruz Biotechnology). Antigen-antibody complexes were visualized with 3,3′-diaminobenzidine. Subsequently, the slides were counterstained with Mayer's hematoxylin, differentiated, dehydrated, and mounted. The known Loricrin-positive hyperkeratotic skin, COMP-positive synovium of articular genu, CXCL9-positive oral mucosa of lichen planus, KRT19-positive breast carcinoma, and CYP 3A5-positive lung cancer served as positive controls. A routinely processed OSF section without the primary antibody served as a negative control in each staining series.

The slides were examined and scored independently by two investigators blinded to the clinicopathologic data. We defined such criteria for each antigen: Given that Loricrin is normally undetectable by immunohistochemistry in nonkeratinized mucosa (10), any expression was considered positive, regardless of the number of positive cells and staining intensity; tissues showing immunostaining of CXCL9, KRT19, and CYP 3A5 were considered positive irrespective of staining intensity if more than 10% of cells displayed staining; tissues showing immunostaining of COMP was considered positive irrespective of staining intensity if more than 10% of lamina propria had continuous and unfragmented staining because COMP staining was present in areas with very few cells.

Statistical analysis of RT-PCR and Western blotting intensity data was done using Student's t test. The associations between expression of any two proteins as well as between a immunohistochemical result and a clinicopathologic variable were tested by χ² test. Pearson contingency coefficient (r) shows the strength of correlation. A P value of <0.05 was considered significant. The statistical analysis was carried out by using the SPSS for Windows 13.0 program (SPSS, Inc.).

Using a q value of <0.01, the expression pattern of genes either with 2-fold induction or with 0.5-fold repression in OSF specimens relative to normal controls was considered as differential regulation. Based on these selection criteria, 661 up-regulated and 129 down-regulated probe sets in OSF with statistically different expression levels were selected by SAM analysis (Supplementary Table S2). The top 30 genes with increased expression and the top 20 ones with decreased expression in OSF relative to NBM were listed in Tables 1 and 2. Loricrin (46.990-fold, q = 0.006), COMP (45.764-fold, q = 0.000), and CXCL9 (29.330-fold, q = 0.005) showed the most significantly up-regulated changes, and the top two differentially down-regulated genes were KRT19 (0.063-fold, q = 0.004) and CYP 3A5 (0.129-fold, q = 0.009) as revealed by microarray analysis.

The eight samples were grouped using hierarchical clustering with the 790 probe sets identified the differential expression genes in OSF. The resulting expression map was visualized with Treeview 1.6. As expected, four OSF samples clustered separately from the normal samples, meaning that the differential expression genes could distinguish OSF from normal tissues (Fig. 1).

To determine the validity of the microarray analysis, we did RT-PCR to confirm the mRNA level of five genes. In concordance with microarray results, OSF tissues expressed significantly more transcript for Loricrin (14.606-fold, P = 0.001), COMP (4.755-fold, P = 0.002), and CXCL9 (3.054-fold, P = 0.001) and less for KRT19 (0.184-fold, P = 0.001) and CYP 3A5 (0.321-fold, P = 0.002) compared with the normal controls (Fig. 2A and B). Because the tissue samples we used in RT-PCR were different from those in the array, heterogeneity of tissues led to the some differences between array data and RT-PCR results, mainly in the fold values.

To verify whether the alterations of genes at the level of transcription ultimately result in the alterations at the level of translation, we conducted Western blotting for the five proteins. In agreement with differential mRNA expression, staining for COMP (P = 0.032) and CXCL9 (P = 0.001) revealed a stronger protein level in OSF than in normal control, whereas KRT19 (P = 0.022) and CYP 3A5 (P = 0.019) protein levels were significantly weaker in OSF than in normal mucosa. However, we did not get a statistically significant result of Loricrin staining (P > 0.05), possibly because of its unique insolubility and limited expression characteristics in established cell culture systems and any epithelial tissue (10, 11) or limited sample sizes in the study. A representative Western blotting result was presented in Fig. 2C and D.

No detectable expression of Loricrin was shown (Fig. 3A) in all normal samples. However, 42 (63.6%) of 66 OSF cases exhibited intensively brown staining for Loricrin almost limited to the upper spinous epithelial layer and sometimes in the keratinocyte layer. The staining was predominantly in the cytoplasm, whereas occasionally both cytoplasmic and nuclear immunoreactivity were observed (Fig. 3B). A significant increase in Loricrin expression was observed in OSF lesions in comparison with NBM specimens (χ² = 18.756, P = 0.149 × 10⁻⁴; Supplementary Table S3).

Little to no staining of COMP was found in 14 normal biopsies (Fig. 3C); however, intense signal for COMP was observed in 36 (54.5%) of the 66 OSF cases, which was present in the juxta-epithelial lamina propria with or without deeper connective tissue. However, COMP staining was shown in areas with very few cells, most likely cytoplasm of fibroblasts as judged by the shape and location of the cells (Fig. 3D). A significant increase in COMP immunopositivity was observed in OSF lesions compared with NBM specimens (χ² = 13.884, P = 0.194 × 10⁻³; Supplementary Table S3).

The staining of CXCL9 protein was observed mainly in the cytoplasm of inflammatory cells and endothelial cells throughout the superficial layer of connective tissue, especially in lamina propria. The connective tissue in 1 (7.1%) of 14 NBM showed very faint CXCL9 expression (Fig. 3E). Forty-three (65.2%) of 66 OSF samples exhibited intense staining of CXCL9 protein (Fig. 3F). The expression of CXCL9 protein was significantly stronger in OSF lesions than in NBM specimens (χ² = 15.703, P = 0.741 × 10⁻⁴; Supplementary Table S3).

All NBM stained continuously and strongly for KRT19 in virtually the cytoplasm of basal cells (Fig. 3G). Only 7 (10.6%) of 66 OSF samples showed faint and fragmented staining in the cytoplasm of basal cells or no staining (Fig. 3H). A significant decrease in KRT19 staining was observed in OSF lesions versus NBM specimens (χ² = 47.677, P = 0.503 × 10⁻¹¹; Supplementary Table S3).

CYP 3A5 staining was detected in the membrane and/or cytoplasm of spinous epithelial cells and cytoplasm of endothelial cells. Twelve (85.7%) of 14 NBM specimens showed very intense CYP 3A5 staining in spinous epithelial cells, and all normal controls showed intense staining in endothelial cells (Fig. 3I). In OSF cases, only 5 of 66 OSF samples (7.6%) showed faint staining in the cytoplasm of spinous epithelial cells; 33 of 66 OSF samples (50%) showed very faint staining of endothelial cells (Fig. 3J). The expression of CYP 3A5 protein was weaker in OSF lesions than in NBM specimens (χ² = 5.986, P = 0.014; Supplementary Table S3).

As shown in Table 3, there is a significantly positive correlation in OSF between Loricrin-positive expression and histologic grade (early stage and moderately advanced stage, χ² = 0.230, P = 0.030, r = 0.295) between positive expression of COMP and duration of areca nut chewing (χ² = 10.000, P = 0.002, r = 0.379) as well as histologic grade (moderately advanced stage and advanced stage, χ² = 7.580, P = 0.006, r = 0.397). A significantly inverse correlation was found between CYP 3A5 expression and histologic grade (early stage and moderately advanced stage, χ² = 4.180, P = 0.040, r = −0.281; moderately advanced stage and advanced stage, χ² = 4.370, P = 0.040, r = −0.302). However, no significant association of the expression of positive CXCL9 or KRT19 with clinicopathologic variables was shown by the χ² test.

A significant inverse correlation was found only between COMP and CYP 3A5 expression in OSF (χ² = 3.911, P = 0.048, r = −0.243; Supplementary Table S4). No clear correlation between the expression of any other two proteins was found in OSF (data not shown), possibly because of the limited sample size in our immunohistochemical study.

To the best of our knowledge, there are limited reports that focus on OSF by means of microarray technology. Kao et al. (12) revealed 18 up-regulated genes and 10 down-regulated genes in OSF compared with normal controls by using microarrays with 1,316 genes, but the limited genes detected could lead to ignoring many significant genes of OSF. Meanwhile, any relationship between individual gene and clinicopathologic factor was never clarified in their experiment. Toward this end, in our study, a microarray with 14,500 genes was first used to identify a list of 790 genes that best represented the gene expression differences between OSF and NBM. Among these deregulated genes, some were previously reported by other groups (13-15); some were not yet observed in OSF but within oral squamous cell carcinoma (16-18); and a number of genes had not been described in any reports on both OSF and oral squamous cell carcinoma. However, there are still some different results between our data and other researchers' reports (12). We consider that microarray analysis for human tissue is very complex, which has extremely numerous biological informations hard to be fully discovered by a single experiment. A systemic collection and analysis for the sorted complementary data from various research groups will benefit us in making global profiles of OSF. Moreover, the difference of races and region distributions, the different processed methods of betel quid, as well as the different procedure of tissue collection and management may contribute to the distinction among various laboratories.

Notable genes in our present study were those with the largest expression changes in OSF compared with NBM, which were thought to be represented diagnostic and therapeutic biomarkers for OSF. The top three differentially up-regulated genes (Loricrin, COMP, and CXCL9) and the top two significantly down-regulated genes (KRT19 and CYP 3A5) were selected as the candidate genes. However, the long lists of data of microarray analysis help little in the understanding of clinical characteristics of OSF. Analysis of gene expression in correlation with clinical or phenotypic variations by our immunohistochemical experiment may indicate biologically meaningful changes of the five novel genes in OSF patients. Inferences based on our findings and evidence from others studies are discussed below.

Loricrin, as a protective barrier to protect from environmental hazards, is one of the differentiation markers and the major protein of the cornified envelope of terminally differentiated keratinocyte (19). The abundance of Loricrin in keratinizing epithelia subject to considerable mechanical stress, such as human foreskin epidermis (20), has led to the assumption that expression of Loricrin is essential for the function of these tissues. Obviously, the coarse fibers of areca nut are the key mechanical stress to the epithelium of oral mucosa of OSF patients; thus, it is plausible to hypothesize to find corresponding increase of Loricrin expression in OSF samples. Although Loricrin is normally incorporated into the cornified envelope within 2 hours of synthesis in the differentiating keratinocyte (21), the enhanced expression of extractable Loricrin (not cornified envelope incorporated) in protein extracts from OSF, in the present study, indicated that it was a protective effect of oral mucosa against the persistently mechanical irritation of areca nuts. Moreover, Loricrin in OSF was identified to be localized in the upper spinous epithelial layer and keratinocyte layer corresponding to its location in keratinizing epidermis (22), which could also provide support for our above-mentioned hypothesis. A significant difference of Loricrin staining between early stage and moderately advanced stage of OSF was found but not between moderately advanced stage and advanced stage. The reason for this result might be, on the one hand, the limited number (n = 13) of advanced-stage OSF samples we used in the present study; on the other hand, it is possible that Loricrin would be a kind of protein with limited capacity against the persistently mechanical stress of areca nut in the advanced stage of OSF lesion; however, additional experiments will be necessary to get a definitive answer.

Cartilage oligomeric matrix protein (COMP) is an abundant, noncollagenous, extracellular matrix protein as a member of the thrombospondin gene family that modulates the cellular phenotype during tissue genesis and remodeling (23), which was previously found likely to bind to several extracellular matrix proteins, including type I, type II, type IX collagen, and fibronectin (24, 25). COMP was shown to be localized mainly in the extracellular matrix of chondrocytes, synovium, tendons, and ligaments (26). However, other authors also assigned COMP to dermal fibroblasts in vitro (27), in skin wound tissue but not normal skin (28), and in scleroderma dermal fibroblasts (29). In the present study, COMP was verified to be up-regulated significantly in the submucous tissue of OSF, whereas immunohistochemical analysis also showed that COMP in OSF was stained in the area with the kinds of collagens deposited consistent with the staining region of COMP in scleroderma (29), possibly pointing to the similar mechanisms of collagen deposition between the two diseases. Thus, it might be excessive COMP secreted by fibroblasts that subsequently facilitate the deposition of collagen in OSF and promote the development of OSF by binding to matrix components. Moreover, the increased expression of fibrogenic cytokines, namely TGF-β, was found in OSF tissues (30), whereas a previous study also indicated that COMP could be induced to overexpress by TGF-β treatment (29), suggesting that increased TGF-β might also provoke the synthesis of COMP in OSF. In addition, autoimmunity has been examined as an etiologic factor for OSF (31), whereas COMP has been cited as potential autoantigens in patients with rheumatoid arthritis (32). It may thus raise the possibility that there is an autoimmune response to COMP within the subepithelial connective tissue of some patients with OSF. Nonetheless, additional studies will be necessary to elucidate the question of whether COMP is synthesized by fibroblasts in the connective tissue of OSF rather than other types of cell.

Chemokines are constitutively expressed or stimulated by inflammatory processes (33), which play important roles as regulators of leukocyte migration, adhesion, and activation during inflammatory diseases, angiogenesis, tumor rejection, rejection of organ transplants, and HIV infection (34, 35). Among a variety of chemokines, CXCL9 (monokine induced by IFN-γ, MIG) functions as a potent chemoattractant for tumor-infiltrating lymphocytes (36), activated natural killer cells, and T_H1 lymphocytes of peripheral blood lymphocytes (37), and has been shown to inhibit neovascularization (38) and contribute to a variety of inflammatory disorders. Strong expression of CXCL9 was shown in both basal keratinocytes and dermal mononuclear cells of lichen planus lesions (39), as well as exclusively restricted to macrophages and vessel-associated cells in the papillae tips of psoriasis lesion (40). One of the hallmarks of OSF is a juxta-epithelial and diffuse mononuclear cell infiltration in the lamina propria. Accordingly, in present study, the enhanced expression of CXCL9 was found in the inflammatory cells of lamina propria of OSF, indicating that CXCL9 might contribute to an ascending chemotactic gradient in the subepithelial connective tissue and the pronounced recruitment of inflammatory cells in OSF. Furthermore, our data showed that CXCL9 protein was stained in endothelial cells of OSF microvessels, hypothesizing that the expression of CXCL9 in endothelial cells, on the one hand, contributes to the infiltration of inflammatory cells to adhere to endothelial cells and, on the other hand, might have a detrimental effect in the repair process of the vasculature and result in the eventual atresia of OSF microvessels. Interestingly, another study showed that IFN-γ, which is considered as an exclusive regulator to induce CXCL9, was shown to have little or no expression in biopsies from OSF tissues compared with normal controls (41). This raises the question whether there might be other extracellular stimuli in OSF to modulate the expression of CXCL9. However, the hypothesis has been supported by recent studies (42, 43). Indeed, this is a topic that deserves investigation on the concretely regulating mechanisms of CXCL9 gene in OSF with reduced IFN-γ expression.

KRT19, which is expressed exclusively by epithelial cells and cancers derived from them, is the smallest member of the cytoplasmic intermediate filament protein family and has a wide tissue distribution (44). Previous studies have suggested that KRT19 expression might be linked to the retention of proliferative (stem cell) potential or undifferentiated character in oral nonkeratinized mucosa, hair follicle, and skin (45-47). In normal nonkeratinized mucosa, KRT19 was detectable in the basal cell layer, whereas there was no detectable KRT19 in normal keratinized mucosa (45). In the present study, the expression of KRT19, whether in the mRNA or protein level, was shown to be significantly weaker in the OSF basal cell layer, which is the proliferative invasive layer, than in normal buccal nonkeratinized mucosa, indicating that the self-renewing capacity of the basal cell layer of OSF through stem cells was obviously inhibited, likely by the chemical irritation of areca nut that subsequently depressed the constant regeneration of OSF mucosa and promoted the atrophy of oral epithelium, which is known as one of characteristic pathologic features of OSF. Eventually, the atrophic epithelium of OSF with the decreased repair rate would potentially be more vulnerable and subject to facilitating the more hazardous chemical substances of areca nut to permeate through this narrow epithelial strip into the connective tissue and causing further deterioration of OSF lesion. In the present immunochemical data, no significant correlation was found between the KRT19 expression and the histopathologic grade of OSF, which might suggest that the proliferative potential of oral epithelium would be obviously depressed by some unknown mechanism from the early stage of the lesion. These results lend strong support to the idea that enhancing the renewing ability of mucosa from the early stage would be a new field for the therapy of OSF.

The cytochrome P450 enzymes are a very large gene family of constitutive and inducible mono-oxygenase enzymes that metabolize many lipophilic, biologically active endogenous and xenobiotic substrates, including a large number of therapeutic drugs and toxic environmental chemicals (48, 49). CYP 3A5 enzyme is one of the main P450 families involved in xenobiotic metabolism, including the metabolism of various carcinogens and several current anticancer drugs, which has been identified in several normal tissues including colon, lung, and anterior pituitary gland. Among numerous chemical constituents of areca nut, alkaloid is the most important agent to undergo nitrosation and give rise to N-nitrosamines, which might have cytotoxic and mutagenic effects on cells of the oral mucosa, whereas CYP 3A5 enzyme activities can just metabolize the activation of these procarcinogens (50). However, in the present study, it seems to be irrational that the expression of CYP 3A5 in OSF was found significantly depressed versus NBM, whereas the reason would be the possibility that the central capacity of CYP 3A5 to metabolize these procarcinogens in OSF mucosa was depressed likely by the persistently and excessively chemical stress whose strength has exceeded the defending capacity of NBM from areca nut constituents. Thus, the depressed ability of OSF mucosa to defend the stress of active endogenous and xenobiotic substrates would open more doors for more toxicities of areca nut to affect more cells in the OSF mucosa and further deteriorate the disease. Moreover, in the present study, the presence of CYP 3A5 protein in the spinous epithelial cells showed an apparent depression from the initial stage; the endothelial cells revealed a gradual reduction correlated with the histopathologic grade, which is possible because the epithelium of mucosa is the first place to respond to the xenobiotic toxicity of areca nut, whereas the endothelial cells are not so sensitive to the toxicity as the epithelium because of their deeper location.

In conclusion, the techniques of microarray analysis provide a dramatic tool to screen novel pathopoiesis-associated genes in OSF research. In the present microarray analysis, Loricrin, COMP, CXCL9, KRT19, and CYP 3A5 were the genes with the most significantly differential deregulation and their functional implications could provide more novel and valuable explanations on the distinctive pathologic changes and pathopoiesis of OSF. Moreover, based on the above-mentioned notions, we presume that some important changes of the mucosa epithelium in OSF might play a vital role in the development of OSF and subsequently result in the typical pathologic changes of the subepithelial connective tissue of OSF. However, we may need more clinical studies involving larger numbers of clinical specimens and more quantitative analysis to explore the potential application of these markers in clinical management of OSF patients. Otherwise, in the present study, we were unable to completely eliminate the contribution of alcohol and smoking to the alteration of gene expression because a great majority of areca nut chewers here were also heavy drinkers and/or smokers. Thus, in a further study, OSF patients without alcohol and smoking habits will be collected to address the roles of the deregulated genes only caused by areca nut chewing.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Prof. Weixin Hu and Dr. Haitao Zeng (Research Center of Molecular Biology of Central South University) for their technical assistance.