Human oral cancer cells, HSC-3 (a human tongue squamous cell carcinoma cell line) and Ca9.22 (a human gingival squamous cell carcinoma cell line), and human oral normal cells, human gingival fibroblast (HGF) and periodontal ligament (PDL) cells, were cultured at 37°C with 5% CO₂ in DMEM (Nissui Seiyaku, Tokyo, Japan) containing 10% foetal bovine serum with penicillin/streptomycin (Sigma, St Louis, MO, USA). Some of these cells were treated with heat shock (45°C, 1 h) and some were treated with LMB (Sigma) (5 ng ml⁻¹) for 7 h.

Cell fractionation was carried out by separating the cells into cytoplasmic and nuclear fractions, as previously described (Weigel and Dobbelstein, 2000), by harvesting and re-suspending the cells in a fractionating buffer (10 mM Tris-HCl, pH 7.6; 150 mM NaCl; 1.5 mM MgCl₂ and 0.5% Nonidet P-40 (Sigma-Aldrich, St Louis, MO, USA), protease inhibitor cocktail), followed by vigorous shaking for 5 min and centrifugation at 12 000 r.p.m. for 30 s. The supernatant was used as the cytoplasmic fraction. To estimate the accuracy of cell fractionation, cytoplasmic protein, β-tubulin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and nuclear protein, Lamin A/C (Santa Cruz), were detected by immunoblotting.

Tissue samples were obtained from excised tongue carcinomas of patients at the Hokkaido University Dental Hospital, Hokkaido, Japan, and from normal tongue tissues. Informed consent was obtained from the patients before the samples were used. Immunohistochemical detection of HuR was carried out using the avidin–biotin complex method as previously described (Shindoh et al, 1996).

Cells were fixed with 4% formaldehyde, permeabilised with 0.1% Triton X-100 and incubated with antibodies specific to HuR (Santa Cruz). After incubation with HuR-specific antibodies, the cells were again incubated with FITC-conjugated secondary antibodies (Molecular Probes, Carlsbad, CA, USA). The cells were observed using an OLYMPUS IX71 (Olympus, Tokyo, Japan) fluorescence microscope.

Western blot analysis was performed using antibodies specific to HuR (Santa Cruz), β-tubulin, Lamin A/C, CRM1 (BD Biosciences, San Jose, CA, USA) and β-actin (Sigma), as previously described (Aoyagi et al, 2003).

To determine the quantity of ARE-mRNAs, the cells were treated with TRI REAGENT (Sigma), after which their RNA was subjected to reverse transcription using Rever Tra Ace (TOYOBO, Osaka, Japan). For quantitative real-time RT–PCR analysis, amplification was performed in a DNA Engine-Opticon 2 PCR machine (MJ Research, Waltham, MA, USA) with SYBR Green PCR master mix (DyNAmo SYBR Green qPCR Kit, MJ Research) as previously described (Higashino et al, 2005).

To evaluate the half-life of c-myc mRNA, HSC-3 and HGF cells were treated with actinomycin D (Act.D) (Sigma) (5 μg ml⁻¹) for 30 or 60 min. The extracted RNA was subjected to quantitative real-time RT–PCR.

In situ hybridisation was performed according to a previously described method (Higashino et al, 2005). The cells were fixed in cold 3% formaldehyde and permeabilised with cold 0.5% Triton X-100. Hybridisation was performed overnight at 37°C in 20 μl of a mixture containing 2.5 μg of tRNA, 10 μg of salmon sperm DNA, 2 × SSC, 0.2% BSA, 1 mM vanadyl ribonucleoside complexes, 50% formamide, 10% dextran sulphate and 10–30 ng of 3′-digoxigenin (DIG)-labelled anti-sense deoxyoligonucleotide probe (Hokkaido System Science, Sapporo, Hokkaido, Japan) for c-fos or c-myc mRNA. The coverslips were washed thrice with 2 × SSC (Invitrogen, Carlsbad, CA, USA) at 37°C and thereafter with 1 × SSC at room temperature. After washing, they were incubated for 60 min at room temperature with a dilution of 1 : 50 of anti-DIG fluorescein Fab fragments (Roche, Basel, Switzerland) in 0.2% Triton X-100/PBS containing 1% BSA (Sigma). After incubation, the coverslips were washed twice with 0.2% Triton X-100/PBS and thereafter with only PBS. The probes (sense and anti-sense) used were complementary to the nucleotides 288–328 of c-fos and to the nucleotides 6278–6311 of c-myc. Nuclei of the cells were stained using DAPI (4′,6′diamidino-2-phenylindole) (Roche).

To knockdown HuR, siRNA was transfected using RNAiMAX (Invitrogen) as per the manufacturer's instructions. The knockdown level of HuR mRNA was analysed by western blot analysis. HuR siRNA was 5′-uuaccaguuucaauggucatt-3′, and the control siRNA was Silencer Negative Control #1 siRNA (Invitrogen).

To determine the subcellular localisation of HuR in oral cancer cells, we carried out an immunohistochemical analysis of HuR using oral cancer cells, HSC-3 (tongue carcinoma) and Ca9.22 (gingival carcinoma) cells, and normal oral cells, HGF and PDL cells. In HSC-3 and Ca9.22 cells, both the cytoplasm and the nucleus were positively stained by HuR antibody, whereas in HGF and PDL cells, only the nucleus was stained (Figure 1A). These data suggest that HuR is localised in the nucleus and cytoplasm of the oral cancer cells, whereas almost all of it is localised in the nucleus of normal cells.

Thereafter, we stained the oral cancer tissue (tongue carcinoma) with the same antibody, which showed a cytoplasmic staining of HuR (Figure 1B, arrows) as against the nuclear staining observed in the adjacent normal cells. In contrast, HuR was localised in the nucleus in normal tissues (Figure 1B). These data indicate the fact that HuR is present in the cytoplasm of oral cancer cells.

To confirm the cytoplasmic localisation of HuR, the cells were separated into cytoplasmic and nuclear fractions, and the HuR of each fraction was detected by western blotting. The amounts of HuR in the cytoplasm of HSC-3 and Ca9.22 cells were much higher than those observed in the cytoplasm of normal cells (Figure 1C). These results also suggest that HuR is localised in the cytoplasm of oral cancer cells.

Together, these results indicate that HuR is present in the nucleus of normal cells as well as in both the nucleus and the cytoplasm of oral cancer cells, which further suggest that HuR is exported to the cytoplasm in oral cancer cells.

We examined the export of ARE-mRNA in oral cancer cells, as ARE-mRNA is known to be exported to the cytoplasm with its binding partner HuR (Brennan et al, 2000; Higashino et al, 2005). The subcellular localisation of c-fos and c-myc mRNAs in oral cancer cells (HSC-3 and Ca9.22) and in normal cells (HGF) was confirmed by in situ hybridisation. These mRNAs were detected in the nucleus and cytoplasm of HSC-3 and Ca9.22 cells, but were localised only in the nucleus of HGF cells (Figure 2A). These data suggest the export of ARE-mRNAs to the cytoplasm in oral cancer cells.

It has been previously reported that the exported ARE-mRNA is stabilised in the cells transformed with adenovirus E4orf6 (Higashino et al, 2005). In this study, we also examined the stabilisation of ARE-mRNA in oral cancer cells. The amount of c-myc mRNA expressed in oral cancer (HSC-3 and Ca9.22) and normal (HGF) cells was measured by quantitative real-time RT–PCR. Accumulation of the ARE-mRNAs was greater in the HSC-3 and Ca9.22 oral cancer cells than in the normal cells (Figure 2B). Moreover, to compare the half-life of c-myc mRNA, HSC-3 and HGF cells were treated with Act.D, and then the quantity of c-myc mRNA was measured by real-time RT–PCR. The half-life of the mRNA in HSC-3 cells was longer than that of HGF cells (Figure 2B). These results suggest the stabilisation of ARE-mRNA in oral cancer cells.

To explore the role of HuR for the export and stabilisation of ARE-mRNA in cancer cells, HSC-3 cells were subjected to HuR knockdown. In HuR-knockdown cells, c-myc mRNA was in the nucleus or in the perinuclear region, although the mRNA existed in both the cytoplasm and the nucleus (Figure 2C). In addition, the quantity of c-myc mRNA decreased in the HuR-knockdown cells compared with that in the cells transfected with the control siRNA (Figure 2C). These results indicate that the export and the increased accumulation of c-myc mRNA are indeed because of HuR in oral cancer cells.

HuR is known to be exported to the cytoplasm in a manner dependent on CRM1, which is a member of the exportin family of nuclear transporters, when cells are stimulated by heat shock or serum stimulation (Brennan et al, 2000; Gallouzi and Steitz, 2001). On the other hand, HuR and ARE-mRNAs are known to be exported to the cytoplasm through a CRM1-independent pathway in cells transformed by an adenovirus oncogene product (Higashino et al, 2005). To study the export pathway of HuR in oral cancer cells, an immunofluorescence analysis of the cancer cells, to examine the localisation of HuR, was carried out after their treatment with LMB, an inhibitor of CRM1. In normal cells, heat shock treatment induced stress granules (SGs), which include HuR, and LMB treatment reduced HuR-included SGs (Figure 3A). Conversely, the accumulation of HuR to the cytoplasm in the oral cancer cells was not inhibited, even after the cells were treated with LMB for 7 h (Figure 3A).

To further confirm the effect of LMB, the existence of HuR in the cytoplasm was examined by western blotting. In normal HGF cells, HuR was observed in the cytoplasm after heat shock treatment (45°C, 1 h), but the HuR band was not visible in the lysate of the cells treated with LMB before the heat shock. On the other hand, in HSC-3 cells, HuR was observed in the cytoplasm even after the cells were treated with LMB (Figure 3B). Together, these data indicate that, in oral cancer cells, the quantity of HuR in the cytoplasms appeared to be unchanged at least 7 h treatment of LMB, and that HuR is exported to the cytoplasm by perturbing the physiological CRM-dependent export machinery.