Repression of tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) but not its receptors during oral cancer progression

Background TRAIL plays an important role in host immunosurveillance against tumor progression, as it induces apoptosis of tumor cells but not normal cells, and thus has great therapeutic potential for cancer treatment. TRAIL binds to two cell-death-inducing (DR4 and DR5) and two decoy (DcR1, and DcR2) receptors. Here, we compare the expression levels of TRAIL and its receptors in normal oral mucosa (NOM), oral premalignancies (OPM), and primary and metastatic oral squamous cell carcinomas (OSCC) in order to characterize the changes in their expression patterns during OSCC initiation and progression. Methods DNA microarray, immunoblotting and immunohistochemical analyses were used to examine the expression levels of TRAIL and its receptors in oral epithelial cell lines and in archival tissues of NOM, OPM, primary and metastatic OSCC. Apoptotic rates of tumor cells and tumor-infiltrating lymphocytes (TIL) in OSCC specimens were determined by cleaved caspase 3 immunohistochemistry. Results Normal oral epithelia constitutively expressed TRAIL, but expression was progressively lost in OPM and OSCC. Reduction in DcR2 expression levels was noted frequently in OPM and OSCC compared to respective patient-matched uninvolved oral mucosa. OSCC frequently expressed DR4, DR5 and DcR1 but less frequently DcR2. Expression levels of DR4, DR5 and DcR1 receptors were not significantly altered in OPM, primary OSCC and metastatic OSCC compared to patient-matched normal oral mucosa. Expression of proapoptotic TRAIL-receptors DR4 and DR5 in OSCC seemed to depend, at least in part, on whether or not these receptors were expressed in their parental oral epithelia. High DR5 expression in primary OSCC correlated significantly with larger tumor size. There was no significant association between TRAIL-R expression and OSSC histology grade, nodal status or apoptosis rates of tumor cells and TIL. Conclusion Loss of TRAIL expression is an early event during oral carcinogenesis and may be involved in dysregulation of apoptosis and contribute to the molecular carcinogenesis of OSCC. Differential expressions of TRAIL receptors in OSCC do not appear to play a crucial role in their apoptotic rate or metastatic progression.


Background
TRAIL (tumor necrosis factor-related apoptosis-inducing ligand; Apo2L) is a type II transmembrane protein which selectively induces apoptosis in tumor cells but not normal cells [1,2]. Because of this differential sensitivity TRAIL is considered an ideal anticancer drug [3,4]. It interacts with four distinct surface receptors, TRAIL-R1 (DR4), -R2 (DR5), -R3 (DcR1), and R4 (DcR2) and with the soluble receptor osteoprotegerin [1,2]. DR4 and DR5 act as transmembrane signaling death receptors with cytoplasmic death domains (DD) which respond to ligand binding and activate the extrinsic cell death pathway by facilitating interaction between the specific adapter protein (FAS-associated DD) and proapoptotic effector proteins (caspases 8 and 10) [5]. DcR1 and DcR2 (decoy receptors), conversely, cannot mediate apoptosis because they lack functional DD [6]. It has been suggested that the differences in the expression levels of death (DR4 and DR5) versus decoy (DcR1 and DcR2) receptors can determine the sensitivity of tumor cells to TRAIL-induced apoptosis [6]. However, susceptibility to TRAIL-induced apoptosis does not always correlate well with the cell surface expression levels of death-inducing and decoy TRAIL receptors suggesting a regulatory role for downstream signaling and effector molecules [7,8]. Furthermore, recent studies indicate that TRAIL susceptibility of tumors in vivo is modified by tumor microenvironmental factors and tumor hypoxia [9,10].
Recombinant TRAIL protein (rTRAIL) induces significant tumor regression in mice bearing human tumor xenografts without producing any serious systemic effects in the host [11,12]. Importantly, radiation and certain chemotherapeutic drugs increase the susceptibility of tumor cells to rTRAIL and agonistic TRAIL-R antibodies [13][14][15][16][17]. Hence, combinations of rTRAIL + chemotherapy or TRAIL + radiation produce synergistic anti-tumor effects both in vitro and in mice bearing human tumor xenografts [18]. Human rTRAIL (PRO1762) and monoclonal antibodies that induce trimerization of DR4 (HGS-ETR 1) and DR5 (HGS-ETR 2) are currently undergoing phase Ib and II clinical evaluations, respectively [19]. Moreover, use of DR4 (HGS-ETR1) agonist antibody as a single agent or in combination with chemotherapy has been shown to stabilize disease in patients with advanced head and neck cancer [20]. In addition to its therapeutic potential, endogenously expressed TRAIL is an effector molecule important for the host's antitumor immune response [2,21,22].
Despite various treatment approaches, most patients with advanced oral squamous cell carcinomas (OSCC) develop local or regional recurrences (50-60%) and metastatic disease (~20%) leading to poor survival rate. Therefore, there is an unmet need for more efficacious and less toxic molecularly targeted therapies for treating OSCC. Recent preclinical and clinical data have shown the potential utility of TRAIL-R targeted therapies in advanced cancers, including OSCC [4,19,23]. Currently, there are only limited data available pertaining to the baseline expression levels of TRAIL and its receptors in normal oral mucosa, oral premalignancies (OPM) and primary or metastatic oral squamous cell carcinoma (OSCC) [24,25]. These data are critical to analysis and interpretation of clinical trial data involving rTRAIL and TRAIL receptor agonist antibodies in OSCC patients, as well as the understanding of the role of TRAIL and its receptors during oral carcinogenesis. In the present study we determined changes in the expression of TRAIL and TRAIL-R during OSCC development and progression by comparing expression patterns in normal oral mucosa (NOM), oral premalignancies (OPM), primary OSCC and metastatic OSCC. We also correlated the expression levels of TRAIL-receptor in OSCC with various clinicopathologic prognosticators. Finally we correlated the apoptosis rate of tumor cells and tumor-infiltrating lymphocytes (TIL) in OSCC with their respective TRAIL and TRAIL-R expression patterns.
Archival tissue specimens from primary OSCC (n = 45) and corresponding lymph node metastases (n = 11) were used for light microscopic and immunohistochemical studies (Table 1). Oral leukoplakias with moderate to severe dysplasia (n = 25) were used as OPM specimens. Five μm serial sections were cut from each specimen and processed for H&E and immunohistochemical staining. The histological grading of OPM and OSCC was determined by two pathologists using H&E stained sections according to published criteria [30]. All tissue specimens and appropriate clinical information were obtained under the guidelines and approval of the Institutional Review Boards of The University of Texas Health Science Center at Houston.

Microarray data sets
We interrogated the expression levels of TRAIL and its receptors in our pre-existing microarray data sets on the above oral cell lines representing different phenotypes. Relative RNA expression levels were determined using Affymetrix U133A oligonucleotide microarrays (33,000 annotated genes) as described [27,28]. Either two or four biological replicate experiments were done independently by two different investigators. The U133A array included three different probe sets for TRAIL (NM_003810), DR5 (TRAIL-R2; NM_003842) and DcR1 (TRAIL-R3; NM_003841) and one probe set for DcR2 (TRAIL-R4; NM_003840), but no probe sets for TRAIL receptor DR4 (TRAIL-R1; NM_003844). Signal intensity for each gene was calculated using the Affymetrix MAS 5.0 probe level algorithm. Phenotypic-specific gene expression values for TRAIL, DR5, DcR1 and DcR2 genes were calculated by averaging signal intensities across sample replicates for cell lines belonging to each phenotypic group.
Western blotting for TRAIL protein TRAIL protein levels were examined by Western blotting in cell extracts of NOM cells from two different patients (NOM-1 & 2), OPM cell line Leuk1, and from primary (1386Tu) and metastatic (1386Ln) OSCC cell lines. These cellular extracts were separated by 12% SDS-PAGE and transferred to nitrocellulose membrane (Novex, San Diego, CA). The membranes were incubated in blocking solution (5% milk powder in 10 mM TRIS-HCl, pH 8.0, 150 mM NaCl, 0.1% Tween 20) at room temperature for 1 h, then immunoblotted with polyclonal goat antihuman TRAIL (1: 200, Santa Cruz Biotechnology, CA, USA) or monoclonal mouse anti-βactin (Sigma-Aldrich Co, St. Louis, MO) antibody followed by horseradish peroxidase-conjugated secondary antibody and chemiluminescence detection.

Immunohistochemistry
We investigated TRAIL and its receptor expression by semiquantitative immunohistochemistry in 25 OPMs (leukoplakias with moderate to severe dysplasia) and in 45 primary OSCCs and 9 patient-matched metastatic OSCCs. Tissue sections were deparaffinized, rehydrated, and sub- Immunoreactive sites were visualized using the standard Streptavidin-Biotin-HRP detection method, using diaminobenzidine tetrachloride as chromogenic substrate.

Pathological and immunohistochemical evaluations
Immunoreactive patterns of TRAIL and its receptors were compared with uninvolved mucosa adjacent to the epithelial dysplasias and invasive tumors. Expression levels of TRAIL and its receptors were evaluated semiquantitatively by two independent examiners as described [32]. A score from 0 (no staining) to 4 (strong immunoreactivity) was assigned to staining intensity, and the percentage of positive cells for each staining intensity in tumors and dysplastic areas of leukoplakias was determined. Immunoreactive scores (range 0 -4) were calculated by multiplying percentage of positive cells times staining intensity score. Cleaved caspase-3 positive tumor cells and TIL in randomly selected five to seven high-power (200 ×) fields (HPF) across the tumor section were counted using Image Pro Plus V (Media Cybernetics, Silver Spring, MD, USA). Caspase 3 labeling index (LI Casp ) for each tumor is expressed as average # positive cells/HPF.

Statistical analysis
Two-sample t-test and one-way ANOVA were used to determine the statistical significance of the differences in TRAIL and its receptors expression levels in OPM and OSCC with various clinicopathologic parameters. Pearson's correlation was used to test the relationship between TRAIL-R expression patterns of uninvolved oral mucosa, and primary and metastatic tumor from the same patient.

TRAIL, DR5, DcR1 and DcR2 RNA expression levels in normal, premalignant and malignant oral epithelial cells in vitro
TRAIL mRNA was detectable in all samples; however its expression level was significantly (p ≤ 0.001) higher in NOM cells (mean ± SD = 364.9 ± 20.4) compared to immortalized oral epithelia (HOK-16B; mean ± SD = 23.3 ± 10.6), OPM (Leuk1 and Leuk 2 cells mean ± SD = 48.2 ± 20.1), malignant primary (686Tu, 1386Tu, HOK-16B-Bap-Tu &101A; mean ± SD = 25.3 ± 11.3) and metastatic (686Ln and 1386Ln; mean ± SD = 32.5 ± 4.8) oral epithelial cells ( Figure 1). Among the TRAIL receptors, DR5 mRNA was consistently higher (> 5-fold) in all cell lines compared to low mRNA levels of DcR1 and DcR2 ( Figure  2). DR4 expression levels were not available in our microarray database. There were no significant differences in the expression levels of all three TRAIL receptors among normal, immortalized, premalignant, and malignant primary and metastatic oral epithelial cells ( Figure 2). Western blotting analysis showed that TRAIL protein is detectable as a 24 kD band only in NOM cells but not in OPM and OSCC cells ( Figure 3). We previously reported that expression of DR4, DR5, DcR1 and DcR2 proteins in primary and metastatic OSCC cells were not significantly different [7].

Expression of TRAIL and its receptors in OPM specimens
All OPM specimens (leukoplakias/erythroplakias with moderate to severe dysplasia) used in this study had normal oral mucosa (NOM) adjacent to the dysplasia. Of the 25 cases of oral leukoplakias with dysplasia, 56% and 44% of them had moderate and severe dysplasia, respectively. Stratified squamous epithelium of the NOM exhibited membrane and cytoplasmic immunoreactivity for DR4, DR5, DcR1 and DcR2 (Figures 4, 5, 6 , 7, 8, 9, 10). Cytoplasmic TRAIL protein was constitutively expressed by NOM (Figures 4, 5). Moreover, DR4, DR5 and DcR1 were expressed in all dysplastic oral epithelia (Figures 4,6,  7). On the other hand, TRAIL expression was lost in dysplasia compared to adjacent NOM (Figures 4, 5, 8). Staining intensity for DcR2 was also lower in dysplasia than in adjacent NOM (Figures 4 &7).

Expression of TRAIL and TRAIL-R in primary and metastatic OSCC specimens
Forty-two primary OSCC and nine patient-matched cervical lymph node metastases, including NOM adjacent to the primary OSCC, were examined for TRAIL and TRAIL-R expression patterns ( DR4 and DR5 immunohistochemical staining of the lesion shown in figure 5 (× 100) Figure 6 DR4 and DR5 immunohistochemical staining of the lesion shown in figure 5 (× 100). Immunoreactivity for either DR4 (top) or DR5 (bottom) is not altered in dysplasia compared to adjacent uninvolved mucosa.
The expression levels of TRAIL and DcR2 are markedly reduced in oral premalignancy compared to normal oral mucosa Figure 4 The expression levels of TRAIL and DcR2 are markedly reduced in oral premalignancy compared to normal oral mucosa. Graphic depiction of the mean expression levels of TRAIL and TRAIL-R in patient matched normal oral mucosa (NOM), oral premalignancies (OPM: leukoplakias with dysplasia). Mean expression levels of DR4, DR5 and DcR1 did not differ significantly between normal oral mucosa (NOM) and oral premalignancies (OPM).
Loss of TRAIL expression in oral leukoplakia with moderate epithelial dysplasia Figure 5 Loss of TRAIL expression in oral leukoplakia with moderate epithelial dysplasia. A: Hematoxylin and eosin staining; B: Immunohistochemical staining for TRAIL (× 100). Note that immunoreactivity for TRAIL is significantly reduced in dysplasia (arrows) compared to adjacent uninvolved mucosa (N).

TRAIL-R expression levels and clinicopathologic parameters
We examined the relationship between TRAIL-R expression and various prognostic factors. There was no signifi- Loss of TRAIL expression pattern in oral leukoplakia with severe dysplasia Figure 8 Loss of TRAIL expression pattern in oral leukoplakia with severe dysplasia. A: Hematoxylin and eosin staining (× 100); B: Immunohistochemical staining for TRAIL (× 100). Note that TRAIL expression is mostly lost in dysplastic epithelium (arrow). In contrast, mononuclear immune cell infiltrate (MNC) associated with dysplasia reveals intense staining for TRAIL.
cant association of either DR4 or DcR1 or DcR2 expression with either tumor size (T) or nodal status (N) or histologic grade of the primary tumor (Table 2). Expression levels of DR5 in primary OSCC correlated positively with tumor size (p = 0.03) but failed to show any significant correlation with either nodal status or histologic tumor grade (Table 3).

TRAIL-R expression levels and apoptosis rates in tumor cells and tumor infiltrating lymphocytes (TIL)
We determined the apoptosis rates of tumor cells and TIL by measuring their cleaved caspase 3 labeling indices (LI Casp ) in primary OSCC ( Figure 12). The overall LI Casp of tumor cells was 3.4/HPF (LI Casp -TU = 3.4 ± 2.6; range = 0.2 -12.2), and LI Casp of TIL was 6.9/HPF (LI Casp -TIL = 6.9 ± 5.2; range = 0.6-25). There were no significant associations between the tumor cells or TIL apoptosis rates and tumor size (T) or nodal status (N) ( Table 3). However, basaloid squamous cell carcinoma, which is a highly aggressive variant of OSCC, exhibited significantly higher LI Casp -TIL compared to the conventional OSCC (Table 3). We also examined the correlation between TRAIL and TRAIL-R expression levels and apoptosis rate among tumor cells and tumor infiltrating lymphocytes (Table 3). There was no significant relationship between the expression levels of all four TRAIL-R in tumor cells and their apoptosis rate (Table 3). Moreover apoptosis rates of TIL in OSCC did not differ significantly between TRAIL-positive and -negative tumors (Table 3).

Discussion
We and others have demonstrated that OSCC cells are susceptible to TRAIL-induced apoptosis [7,10,33]  The loss of individual TRAIL-R expression in malignant tumors has been attributed to chromosomal gene mutation/deletion (DR4 and DR5) or promotor methylation (DcR1 and DcR2) events [4]. However, gene mutation or promotor methylation is not implicated for the malignancy-specific down-regulation of TRAIL expression. Recent experimental findings indicate that loss of TRAIL expression during malignant transformation is not mediated by genetic aberration but by dysregulation of signal transduction pathways common to various cancers. TNFα increases the susceptibility of breast cancer cells to chemotherapy by up-regulating TRAIL expression by promoter activation [39]. Similarly, retinoids and interferons exert their anticancer activity in breast cancer cells by enhancing TRAIL expression in these cells [40]. Hyperactivation of the phosphatidylinositol 3-kinase (PI-3K)/Akt pathway has been implicated in suppressing TRAIL expression in colon cancer cells [41]. Treatment of colon cancer cells with the PI-3K inhibitor Wortmannin rescues TRAIL expression in these cells and induces enterocytelike differentiation [41]. Interestingly, hyperactivation of PI-3K/AKT pathway via dysregulated EGFR signaling is an important and early event in the pathogenesis of OSCC [42]. Hence, it is plausible that there is a cause-and-effect relationship between dysregulated prosurvival PI-3K/AKT signaling and loss of TRAIL expression in OPM and OSCC.
Interestingly, oral carcinogens such as tobacco smoke frequently inactivate the tumor suppressor gene p53, thus depriving its protective role against oral carcinogeninduced DNA damage [43]. TRAIL-induced apoptosis of transformed/dysplastic cells is independent of p53 status [34,44], hence, TRAIL may act as a substitute guardian against malignant transformation by eliminating transformed cells during the initial genesis OPM. Thus, TRAIL down-regulation may allow clonal expansion of transformed cells by protecting them from apoptosis, thereby increasing the risk of malignant progression.
TRAIL protein detection was mostly negative or only focally positive in isolated tumor cells in both primary and metastatic OSCC tumors, and our Western blot analysis clearly demonstrated the lack of TRAIL protein in primary and metastatic OSCC cell lines. Our microarray data also confirmed low levels of TRAIL mRNA in primary and metastatic OSCC cell lines compared to NOM. Our findings are not, however, in agreement with a previous study reporting the expression of TRAIL protein in primary OSCC tumor specimens and TRAIL mRNA expression in established OSCC cell lines [24,25]. That study used RT-PCR, which can presumably detect much lower levels of TRAIL mRNA compared to our microarray analysis. Moreover, that study did not compare the TRAIL mRNA levels of OSCC cell lines with NOM to determine significant differences in their respective expression levels. The microarray chip used in our study included three different probe sets for TRAIL which further increased the specificity and reliability of the data. Nevertheless, tumor heterogeneity   and the differences in the specificity of the TRAIL antibody used in the two investigations may account for the discrepancies in the OSCC TRAIL expression data.

Expression of DR4, DR5 and DcR1 receptors are not significantly altered during oral cancer progression
Relative expression levels of TRAIL death and decoy receptors are critical for exploiting therapeutic possibilities based on TRAIL/TRAIL-R interaction in OPM and OSCC [7,10,33,45]. Expression of these three receptors by OSCC seems to depend at least in part on whether or not they are expressed by the parental oral epithelium from which the tumors arose. Moreover, expression of all three receptors in the tumor cells were higher than in adjacent NOM, a feature also reported for melanomas and breast and colon carcinomas [46][47][48]. In contrast, DcR2 expression was significantly lower in OSCC compared to the adjacent NOM. Loss of DcR2 expression in malignant tumor cells has been attributed to aberrant promotor methylation [49].
DcR1, DR4 and DR5 are very frequently overexpressed in dysplastic oral mucosa, as well as primary and metastatic OSCC tumor specimens. All primary OSCC specimens examined by us expressed either DR4 or DR5 or both at high levels. Among the death-inducing receptors, DR4 was expressed more frequently in OSCC than DR5, as also observed by others in head and neck squamous cell carcinomas [25]. Therefore, it would seem that the DR4-targeted humanized agonist antibody (HGS-ETR1) would be a better choice than the DR5-targeted antibody (HGS-ETR2) for OSCC clinical trials.
TRAIL death-inducing receptors DR4 and DR5 are considered tumor suppressors because their loss is expected to provide a survival advantage to the tumor cells. Conversely, overexpression of the anti-apoptotic decoy receptors DcR1 and DcR2 is expected to promote malignant progression. Therefore, we evaluated the association between the expression levels of these receptors and certain accepted prognostic indicators such as tumor size, nodal status and histologic tumor grade. Expression levels of these receptors failed to show any significant correlation with either nodal status or tumor grade, and there was no significant association between DR4, DcR1 and DcR2 expression levels and tumor size. However, DR5 expression correlated positively with tumor size. Such an association is in conflict with its proposed pro-apoptotic and tumor-suppressive function yet a similar observation is also made in breast cancers, in which high DR5 expression has correlated with nodal status, tumor size, and poor survival rate [48]. Similarly, high DR5 but not DR4 expression correlated with decreased survival in patients with non-small-cell lung cancers [50]. DR5 expression is induced by DNA damage and the induction is both wildtype p53 dependent and independent [51]. Activation of endogenous NF-kB factors by TNF-α also induces DR5 expression in tumor cells [52]. Thus, anti-apoptotic NF-kB activation may explain the association between increased DR5 expression and larger tumor size.

TRAIL and TRAIL-R expression in OSCC does not influence tumor cell and tumor infiltrating lymphocyte apoptosis rates
Susceptibility of OSCC cells to TRAIL-mediated apoptosis is dependent on their relative expression levels of deathinducing receptors DR4 and DR5 versus decoy receptors DcR1 and DcR2. OSCC expressing high levels of DR4 and/ or DR5 will be more susceptible to apoptosis than tumors expressing low levels of these receptors. Conversely, OSCC expressing high levels of decoy receptors often shows reduced tumor cell apoptosis rates. However, our data did not show any significant difference in the tumor cell apoptosis rates for primary OSCC with high and low levels of DR4 and/or DR5 expression, or for differential expression levels of decoy receptors DcR1 and DcR2.
TIL represent one of the manifestations of host immune response against malignancy [53], and several clinical studies have documented the prognostic significance of Immunohistochemical staining for cleaved caspase 3 to detect apoptotic cells among tumor cells and tumor-infiltrat-ing lymphocytes Figure 12 Immunohistochemical staining for cleaved caspase 3 to detect apoptotic cells among tumor cells and tumor-infiltrating lymphocytes. Examples showing the detection apoptotic cells among tumor cells (LI-Cas-Tu)) and TIL (LI-Cas-TIL) based on their immunoreactivity for cleaved caspase-3 (× 400).
TIL in various malignancies, including OSCC [54,55]. Apoptosis of TIL in OSCC leads to the depletion of lymphocytes and negatively affects their prognosis [54]. This is further supported by our finding that TIL apoptosis rates in basaloid squamous cell carcinomas, a very aggressive variant of OSCC [56,57], were 5-fold higher than in conventional OSCC. However, TRAIL/TRAIL-R expression patterns or tumor cell apoptosis rates were not significantly different between basaloid squamous cell carcinoma and conventional OSCC. Depletion of TIL in malignant tumors is mediated by a counterattack against Fas-bearing lymphocytes by FasL-expressing tumor cells [58]. Therefore, the potential of tumor-cell-derived TRAIL cytotoxicity against TIL has been proposed by some studies [59]. However, our study did not find any significant difference in the TIL apoptosis rate between focal TRAILpositive and TRAIL-negative OSCC.

Conclusion
Our investigation has shown that TRAIL is constitutively expressed in normal oral mucosa but its expression is gradually lost in oral premalignant and malignant epithelia. OSCC more frequently expresses DcR1, followed by DR4 and DR5. Moreover, expression of DR4, DR5 and DcR1 are up-regulated in premalignant and malignant oral epithelia compared to normal oral epithelium. A partial loss of DcR2 expression is also noted in premalignant and malignant oral epithelia compared to normal oral epithelia, and DR5 expression is significantly associated with larger tumor size. Expression levels of TRAIL receptors show no significant correlation with nodal status and apoptosis rates of tumor cells and TIL.