Association of cytokeratin 17 expression with differentiation in oral squamous cell carcinoma

Purpose The aim of this study was to confirm the expression profile of cytokeratin (CK)17 in comparison with that of CK13 in oral squamous cell carcinoma (OSCC) and leukoplakia and to clarify an association of CK17 with the OSCC differentiation. Materials The expression of CK17 and CK13 was immunohistochemically examined in 105 patients with OSCC and 108 patients with leukoplakia. A correlation of CK expression with clinicopathological variables was carried out. The over-expression levels of CK17 mRNA were analyzed by real-time RT-PCR in 5 OSCC cell lines (HSC-2, HSC-3, SAS, SQUU-A, SQUU-B). Results CK17 and CK13 were detected in 101 (96.2 %) and three (2.9 %) of the 105 OSCCs, respectively. CK17 was significantly expressed in well-differentiated OSCC compared to moderately/poorly differentiated OSCC (p < 0.01). As detected in 19 of the 34 dysplastic leukoplakias (55.9 %) and 36 of the 74 hyperplastic leukoplakias (48.6 %), CK17 was significantly expressed in dysplastic leukoplakias (p < 0.01). As detected in 11 of the 34 dysplastic (32.4 %) and 52 of the 74 hyperplastic leukoplakias (70.3 %), CK13 was significantly expressed in hyperplastic leukoplakias (p < 0.01). The relative expression of CK17 mRNA in HSC-2 was significantly higher than in HSC-3 and SAS (p < 0.05). Moreover, the relative expression of CK17 mRNA in SQUU-A was significantly higher than in SQUU-B (p < 0.05). Conclusion CK17 expression could be associated with the differentiation and the malignancy of OSCC. A combination pattern of CK17/CK13 might be a suitable marker of malignant transformation.


Introduction
Despite diagnostic and therapeutic advances by introducing combination therapy including surgery, radiotherapy, and chemotherapy, the 5-year survival rate of oral squamous cell carcinoma (OSCC) remains 70-80 % (Gorsky et al. 2004;Sasaki et al. 2011), because of the late-stage diagnosis and the resistance to radiotherapy or chemotherapy. For early diagnosis and individual therapy of the patients, it is of great importance to Wnd out diagnostic markers of OSCC. Cytokeratins (CKs), intermediate Wlament of the cytoskeletons, are candidates for diagnostic markers of OSCC, as they are over-expressed in OSCC compared to normal mucosa (Xu et al. 1995). Biochemical and immunohistochemical studies have indicated that OSCC expresses a wider range of CKs than normal epithelium, because cancerization of normal oral epithelium could lead to variation of diVerentiation degree (Wetzels et al. 1992).
Among CKs, CK17 should be focused as a diagnostic marker of OSCC, since several studies have reported that the expression of CK17 could be detected in malignant tissues compared to normal tissues in squamous cell carcinoma of lung (Wetzels et al. 1992), cervix (Carrilho et al. 2004;Ikeda et al. 2008), larynx (Cohen-Kerem et al. 2004), and esophagus (Takahashi et al. 1995;Luo et al. 2004). In oral cavity, our previous study based on the microarray and the real-time RT-PCR analyses have reported that the sig-niWcant up-regulation and the strong over-expression of CK17 could exhibit the utility as diagnostic marker of OSCC (Toyoshima et al. 2008). Focused on diVerentiation of cells expressing CK17, CK17 is also considered as a marker of basal cell diVerentiation to distinguish columnar epithelium from squamous epithelium in cervix (Martens et al. 2004). Moreover, CK17 is markedly expressed in well-diVerentiated squamous cell carcinoma (Carrilho et al. 2004). However, to our knowledge, a paucity of material is available on a correlation between the CK17 expression and the diVerentiation in OSCC.
CK13 is mucosa speciWc and expressed in the suprabasal layers of non-keratinized stratiWed epithelia. In normal tissue, CK13 is usually expressed in the conjunctival and limbal epithelium (Ramirez-Miranda et al. 2011). On the contrary, decreasing of CK13 expression is found in dysplastic lesions in the oral epithelium and associated with the development of OSCC and oral leukoplakia (Ohta et al. 2010). Because of the loss of stratiWcation, CK13 is downregulated in malignant lesion (Bloor et al. 2000). Carrilho et al. (2004) have reported that the expression of CK17 and the loss of CK13 are appropriate markers of malignant transformation in cervix carcinoma. In addition, Whipple et al. (2004) have reported that CK17 mRNA is up-regulated, while CK13 mRNA is down-regulated in OSCC by the microarray analyses. Moreover, it has been reported that CK17 is signiWcantly up-regulated in leukoplakia and is absent in normal oral epithelium (Mikami et al. 2011). In contrast, CK13 is down-regulated in leukoplakia and Wrmly expressed in normal oral epithelium. However, little is known about an association of CK17 and CK13 expressions with the diVerentiation of OSCC and leukoplakia.
Therefore, the aim of this study was to conWrm the immunohistochemical proWle of CK17 in comparison with that of CK13 in OSCC, leukoplakia, and normal oral epithelium from a diVerentiation point of view. Secondly, we focused on CK17 mRNA expression in OSCC cell lines in order to clarify an association of CK17 with the diVerentiation of OSCC.

Patients
One hundred and Wve patients with primary OSCC and 108 patients with primary leukoplakia, diagnosed at the Department of Oral and Maxillofacial Surgery, Kyushu University Hospital, from 2005 to 2008, were enrolled in this study.
Ten control cases (normal oral epithelium) were also enrolled. Patients' informed consent and the approval of the local ethical committee were given. Following the initial biopsy, all specimens were Wxed in 4 % buVered formalin solution and embedded in paraYn blocks. Subsequently, the paraYn-embedded specimens were processed to 5-mthick sections, stained with hematoxylin-eosin (HE), and examined by three experienced pathologists to conWrm the diagnoses and the histological grades. The histological grade in the OSCC and the degree of epithelial dysplasia in the leukoplakia were assessed according to the World Health Organization classiWcation (Gale et al. 2005;Wahi et al. 1977). The tumor extent and the clinical stages were evaluated according to the TNM classiWcation established by the American Joint Committee on Cancer and the International Union Against Cancer (UICC) (Sobin and Wittekind 2002). The mode of invasion was also determined on  (Yamamoto et al. 1984). Medical records were reviewed to collect the information concerning the clinical characteristic. The detailed clinical data of the patients with OSCC and leukoplakia are presented in Tables 1 and 2.

Immunohistochemistry
Immunohistochemical staining was performed on 5-mthick sections sliced serially from paraYn-embedded blocks after formalin Wxation of the excised specimens. The sections were deparaYnized in xylene and rehydrated in a graded series of ethanol/water concentrations (100, 95, 90, 85, and 75 %). For antigen retrieval, the sections were immersed in DAKO Target Retrieval Solution (Dako Cytomation, Denmark) and autoclaved at 120°C for 5 min. The endogenous peroxide activity was then eliminated with 1 % hydrogen peroxide for 30 min, and the section was rinsed twice for 10 min with phosphate-buVered saline (PBS) at pH 7.4. Non-speciWc protein binding was blocked by incubation for 1 h with 10 % goat serum, and then, the sections were incubated with each primary antibody for 3 h at room temperature. The following primary antibodies were used: anti-human monoclonal CK17 antibody (clone E3, Dako Cytomation, Denmark; diluted 1:60) and anti-human monoclonal CK13 antibody (clone KS-1A3, Novocastra, Germany; diluted 1:1000). The sections were rinsed twice for 10 min with PBS and incubated with secondary antibodies conjugated with peroxidase-labeled amino acid polymer for 1 h at room temperature. After rinsing with PBS twice for 10 min, the immunoreactivity was visualized by immersing the sections in 3, 3Ј-diaminobenzidine and 0.6 % hydrogen peroxide (DAB substrate kit, Nichirei, Japan). Subsequently, the sections were counterstained with Mayer's hematoxylin, dehydrated in graded ethanol (75,85,90,95 and 100 %), cleared with xylene, and Wnally mounted with permanent mounting medium (Mount-Quick, Daido Sangyo, Japan). Negative controls were prepared by substituting PBS for each primary antibody. To evaluate the expression of CK17 and CK13 in OSCC and leukoplakia, positively stained cells were counted in at least three randomly selected areas at £200 magniWcations, and then, each percentage of these positive cells was calculated as a BIOREVO Xuorescent microscope (Keyence, Japan). The stained sections with <5 % reactive cells were considered to be negative, and those with more than 5 % reactive cells were deWned as positive. The cutoV points were established at 30 and 60 %, and 5-30 %, 30-60 %, and more than 60 % reactive cells deWned as +, ++, and +++, respectively. The sections were divided into two groups as follows: over 60 % positive cells were deWned as "strong" case; less than 60 % positive cells were deWned as "weak" case.

Cell lines
Five human OSCC cell lines (HSC-2, HSC-3, SQUU-A, SQUU-B, and SAS) were used in the study. Human epithelium cell line (HaCaT) was used as a calibrator. HSC-2 was established from well-diVerentiated OSCC of mouth Xoor, while HSC-3 and SAS were established from poorly diVerentiated OSCC of tongue. SQUU-A and SQUU-B cell lines were established from well-diVerentiated OSCC of tongue with local recurrences. Cells were maintained in Dulbecco's modi-Wed Eagle's minimum essential medium (DMEM) supplemented with 10 % fetal bovine serum (FBS) and incubated at 37°C in a 5 % CO 2 atmosphere. Culture media was changed on alternate days during the experiments.

Real-time RT-PCR
Total RNA was extracted from these 6 cell lines with the TRIzol ® reagent (Invitrogen, USA). cDNAs from total RNA were synthesized using the GeneAmp RNA PCR kit (Applied Biosystems, USA) according to the manufacturer's instructions. Real-time RT qPCR analyses were done using QuantiTect Primar Assay (200)   PCR run included a 15-min activation time at 95°C as required by the instrument. The three-step cycle included denaturing (94°C, 15 s), annealing at 55°C, and extension at 72°C. At the end of each PCR run, melting curve analysis was performed from 60 to 95°C for detecting non-spe-ciWc PCR product and primer-dimer co-ampliWcation. CK17 mRNA quantities were analyzed in duplicate, normalized GAPDH as an internal control gene and expressed in relation to mRNA from HaCaT as a calibrator. Results are expressed as relative gene expression using the Ct method (Livak and Schmittgen 2001).

Statistical analysis
In the real-time RT-PCR, mean value of duplicate CK17 mRNA RQs was deWned as positive if it was higher than 2.0-fold. The statistical signiWcance of the diVerences between the groups was determined by Student's t-test and the chi-square test for univariate analyses. The interaction between CK17 expression and the other variables in patients with OSCC was determined by multivariate logistic regression analysis. p values of <0.05 were considered to be signiWcant.

Immunohistochemical analyses of CK17 and CK13 expression in OSCC
A representative case of CK17 and CK13 expressions is shown in Fig. 1. CK17 and CK13 were predominantly expressed in the cellular cytoplasm. In well-diVerentiated OSCC, CK17 was strongly expressed in the majority of tumor cells. In the cancer nest, CK17 was expressed in the inner layers and not expressed in the outer layers (Fig. 1d). In moderately diVerentiated OSCC, CK17 was weakly expressed in the majority of tumor cells (Fig. 1e). In poorly diVerentiated OSCC, CK17 was absent in the majority of tumor cells but expressed in a few of tumor cells (Fig. 1f). CK13 was detected in three of the 105 OSCCs (Fig. 1). In these cases, CK13 was expressed in the center of cancer pearls in well-diVerentiated OSCC. There were no signiWcant diVerences between CK13 expression and clinicopathological variables (data not shown).
CK17 was detected in 101 of the 105 OSCC cases (96.2 %). The "strong" cases with over 60 % reactive cells were 23 of all 72 well-diVerentiated OSCCs (31.9 %). The "strong" case was one of all 33 moderately and poorly diVerentiated OSCCs (3.0 %). CK17 was signiWcantly expressed in well-diVerentiated OSCC compared to moderately and poorly diVerentiated OSCC (p < 0.01, chi-square test; Table 3), though there were no signiWcant diVerences for other clinical variables (Table 3). All variables were further analyzed in a multivariate logistic regression model. The result of the analysis showed that the OSCC diVerentiation was the signiWcant factor of CK17 "strong" cases (p < 0.01, multivariate logistic regression analysis; Table 3). Gender, age, alcohol, smoking, TNM classiWcation, and localization of the lesions were not signiWcant (Table 3).
A representative case of CK17 expression in moderately diVerentiated OSCC is shown in Fig. 2. CK17 was markedly expressed in the dysplastic epithelium adjacent to the tumor, while CK17 was not completely expressed in the normal oral epithelium (Fig. 2d). In the cancer nest, CK17 was obviously expressed except for the outer layer and the center of cancer pearl (Fig. 2e). In the invasion front of the tumor, CK17 was absent in the inWltrating cells (Fig. 2f).

Discussion
The carcinogenesis mechanism of OSCC is very complex, and a wider range of CKs is expressed in OSCC (Wetzels et al. 1992). CK17 may play an important role in the diagnosis of OSCC, since several studies have reported that the over-expression of CK17 could be detected in malignant tissues compared to normal tissues in squamous cell carci- Our immunohistochemical results indicate the signiWcant percentage of CK17 expression in OSCC. This is in harmony with the detection of CK17 in breast carcinoma by immunohistochemical staining and microarray analyses (van de Rijn et al. 2002). Moreover, CK17 mRNA is sig-niWcantly over-expressed compared to CK19 mRNA and CK20 mRNA in OSCC (Toyoshima et al. 2009). Further- Table 3 Uni-and multivariate analyses for association between CK17 expression and clinicopathological variables in 105 patients with OSCC a CK17 was detected by immunohistochemical staining, and the degree of expression was estimated as described in the "Materials and Methods" b The number of cases is indicated c The histological grade of diVerentiation was determined according to the criteria of WHO d The histological grade of tumor invasion was determined according to Yamamoto  more, up-regulation of CK17 mRNA is observed by microarray analyses in OSCC compared to normal oral epithelium (Ye et al. 2008). In the view of these reports, CK17 is markedly expressed in malignant tissue and thus could be a diagnostic marker of OSCC. The results of CK17 indicate the signiWcant expression in well-diVerentiated OSCC and the decreasing expression in moderately and poorly diVerentiated OSCC. This is in disharmony with the CK17 expression in the most of all OSCCs despite clinicopathological variables, considered as a discriminating marker of OSCC (Wei et al. 2009). But the number of cases was too small for a statistical evaluation, since only 30 OSCCs were examined in the study. Therefore, CK17 could be a precise candidate for a diagnostic marker of well-diVerentiated OSCC.
Focusing on the diVerentiation of OSCC cell individual, our results show that CK17 is not expressed in the outer layer of cancer nest in well-diVerentiated OSCC and inWl-trating OSCC cells in poorly diVerentiated OSCC. This is in line with the absence of CK17 expression in the basal area of carcinoma in situ and the outer layer of cancer nest in OSCC (Mikami et al. 2011). It has been suggested that the outer layer of cancer nest is considered as poorly diVerentiated keratinocyte, since this area is the counterpart of basal and parabasal cells in normal oral epithelium (Bloor et al. 2000;Smedts et al. 1992). In addition, since poorly diVerentiated OSCC has few well-diVerentiated keratinocytes and poorly diVerentiated keratinocytes are observed at the invasive front (Willen et al. 1975), inWltrating OSCC cells could be also considered as poorly diVerentiated keratinocyte. Therefore, CK17 could be considered as a speciWc marker of well-diVerentiated OSCC cells. Since poorly diVerentiated OSCC has poor outcome because of high malignancy (Pekkola-Heino et al. 1991), it might be possible that CK17 is not strongly expressed in high malignancy of OSCC. Other immunohistochemical results indicate the strong expression of CK17 in mildly, moderately, and severely dysplastic leukoplakia. This conXicts with the absence of CK17 in mildly and moderately dysplastic leukoplakia (Mikami et al. 2011), but no severely dysplastic leukoplakia was examined in the study. On the other side, our results are compatible with the expression of CK17 mRNA in severely dysplastic leukoplakia (Ohkura et al. 2005). Moreover, CK17 is expressed in mildly, moderately, severely dysplasia of uterine cervix (Martens et al. 2004). Indeed, a marker of the premalignant lesion is important to predict the malignant transformation. Nevertheless, CK17 is expressed in a portion of dysplastic leukoplakias and thus not enough to diagnose individually premalignant lesion. In contrast, CK13 is coordinately expressed in normal oral epithelium and disappeared consistently in dysplastic leukoplakia and OSCC in the view of our results. Therefore, combination of CK17 and CK13 might be a useful marker to diagnose premalignant lesion with high potential of transformation. In addition, over-expression of CK17 and absence of CK13 might be associated with malignant transformation. OSCC tissues change their histological characters and express various CKs (Wetzels et al. 1992). These pathological changes involve the terminally diVerentiating cell compartment, but it is still unclear about correlation between the changes and the gene expressions of CKs (Debus et al. 1984;Morgan et al. 1987;Vigneswaran et al. 1989;Vaidya et al. 1989). Further studies are needed to clarify what operates to CKs genes.
Real-time RT-PCR results indicate that the quantity of CK17 mRNA in HSC-2 derived from well-diVerentiated OSCC is signiWcantly higher than that in HSC-3 and SAS derived from poorly diVerentiated OSCC. This conXicts with the detection of CK17 mRNA in OSCC cell lines by  CK17 "strong" c /CK13 "weak" 8 3 11 CKlT "weak"/CK13 "strong" 0 23 23 p < 0.01 real-time RT-PCR despite their diVerentiation (Ohkura et al. 2005). In fact, CK17 is synthesized at the stage of cell fate speciWcation within developing skin epithelium (McGowan et al. 2002). Moreover, onset of CK17 expression reXects the determination of the lineage in skin epithelium (McGowan and Coulombe 1998). Furthermore, CK17 is not expressed in the basal cell, while it is expressed in the suprabasal cells that have been already diVerentiated. This indicates that there is a diVerence of CK17 expression pattern according to the cell diVerentiation. CK17 mRNA quantity was signiWcantly higher in SQUU-A than in SQUU-B, though both are derived from well-diVerentiated OSCC. The in vitro formation of intermediate Wlaments is prominently observed in SQUU-A compared to SQUU-B (Morifuji et al. 2000). Furthermore, piling up of SQUU-A cells was partially seen, whereas Immunohistochemical staining of CK17 and CK13 in the border between dysplastic leukoplakia and hyperplastic leukoplakia. CK17 was clearly expressed in dysplastic leukoplakia, while it was disappeared in hyperplastic leukoplakia (a). CK13 was disappeared in dysplastic leukoplakia, while it was clearly expressed in hyperplastic leukoplakia (b). A reciprocal expression of CK17 and CK13 could be found between the two diVerent lesions (original magni-Wcation£100, scale bars 100 m) Fig. 6 Relative expression of CK17 mRNA in OSCC cell lines and human epithelium cell line. The relative expression of CK17 mRNA in HSC-2 was signiWcantly higher than in HSC-3 and SAS. Moreover, the relative expression of CK17 mRNA in SQUU-A was signiWcantly higher than in SQUU-B. The relative expression of CK17 mRNA in HaCaT was signiWcantly lower than the others. Statistical analyses were performed by Student's t-test SQUU-B cells developed multiple layers throughout. Combined with the ability of muscle invasion and cervical lymph node metastasis in vivo, SQUU-B resembles to HSC-3 closely, because HSC-3 also has the ability (Kawahara et al. 1995). Therefore, SQUU-B might have feature of poorly diVerentiated OSCC compared to SQUU-A.
In conclusion, CK17 expression could be associated with the diVerentiation and the malignancy of OSCC. A combination pattern of CK17/CK13 might be a suitable marker of malignant transformation.