KLF4, Slug and EMT in Head and Neck Squamous Cell Carcinoma

Abstract

Epithelial to mesenchymal transition (EMT) is clinically relevant in head and neck squamous cell carcinoma (HNSCC). We hypothesized that EMT-transcription factors (EMT-TFs) and an anti-EMT factor, Krüppel-like-factor-4 (KLF4) regulate EMT in HNSCC. Ten control mucosa and 37 HNSCC tissue samples and three HNSCC cell lines were included for investigation of EMT-TFs, KLF4 and vimentin at mRNA and protein levels. Slug gene expression was significantly higher, whereas, KLF4 gene expression was significantly lower in HNSCC than in normal mucosa. In the majority of HNSCC samples, there was a significant negative correlation between KLF4 and Slug gene expression. Slug gene expression was significantly higher in human papilloma virus (HPV) negative HNSCC, and in tumor samples with irregular p53 gene sequence. Transforming-growth-factor-beta-1 (TGF- β1) contributed to downregulation of KLF4 and upregulation of Slug. Two possible regulatory pathways could be suggested: (1) EMT-factors induced pathway, where TGF-β1 induced Slug together with vimentin, and KLF4 was down regulated at the same time; (2) p53 mutations contributed to upregulation and stabilization of Slug, where also KLF4 could co-exist with EMT-TFs.

Keywords: FaDu; SCC-25; TWIST; TissueFaxs; UPCI-SCC090; ZEB.

Figure 1

Comparison of relative quantification of Slug (a) and KLF4 (b) gene expression in normal mucosa and in HNSCC. Ten normal mucosa and 37 HNSCC samples were used for real-time PCR analysis. On the Y-axis the fold change difference to the mean reference value of normal mucosa is presented. None of the compared sample sets showed normal distribution. The medians of control and HNSCC samples were compared using Mann-Whitney-test. Slug gene expression was significantly higher in HNSCC than in normal mucosa (p = 0.009), whereas, KLF4 gene expression was significantly lower in HNSCC than in normal mucosa (p = 0.041).

Figure 2

In HNSCC where KLF4 is reduced (red box) compared to normal mucosa from UPPP (blue box), Slug gene expression is upregulated. HNSCC with reduced KLF4 gene expression have a negative correlation between KLF4 and Slug gene expression.

Figure 3

Comparison of relative quantification of Slug (a,c) and KLF4 (b,d) gene expression in p16-positive and negative (a,b) HNSCC, as well as in HNSCC with regular and irregular p53 gene expression (c,d). Ten HPV-positive and 27 HPV-negative HNSCC samples were used for real-time PCR analysis. On the Y-axis the fold change difference to the mean reference value of normal mucosa is presented. None of the compared sample sets showed normal distribution. The median of Slug gene expression in p16-negative HNSCC samples was significantly higher (p = 5 × 10⁻⁴) than in p16-positive samples (a). In contrast, KLF4 gene expression was not significantly different in p16-positive or negative cases (b). Eleven HNSCC samples were available with regular (wild type) and 26 with irregular (mainly mutated) p53 gene background. The median of Slug gene expression in HNSCC samples with irregular p53 was significantly higher (p = 5.5 × 10⁻³) than in samples with regular p53 (c). In contrast, KLF4 gene expression was not significantly different in cases with regular or irregular p53 (d). All datasets were not normal distributed, and Mann-Whitney-test median comparison was used.

Figure 4

Comparison of immunohistochemical labeling of KLF4 (red), pan-cytokeratin (green) and vimentin (light blue) (a,c,e,g) with enzyme immunohistochemistry of Slug (b,d,f,h) in normal layered epithelium (a,b); in HPV⁺, p53 wild type HNSCC (c,d); in HNSCC with loss of p53 gene product (e,f) and in HNSCC with p53 gain of function mutation (g,h). The normal layered epithelium showed dispersed intensive green cytokeratin (CK) reaction, the nuclei of the CK⁺ cells contained KLF4, the CK⁻ stroma region contained vimentin-positive (VIM⁺) cells (a). In this tissue no traces of Slug staining were detected (b). The HPV⁺ p53 wild type HNSCC showed limited green cytokeratin (CK) reaction, the nuclei in cancer cell nests contained KLF4, also if they were CK⁻, the stroma region contained numerous vimentin-positive (VIM⁺) cells (c). In this tissue no traces of Slug staining were detected (d). In the HNSCC with loss of p53 gene product diffuse CK reaction was detected at different levels in cells of the cancer cell nests, only few scattered nuclei in cancer cell nests contained red KLF4 reaction, the light blue reaction of VIM was spread in the stroma, but traces of VIM were also visible in the cancer cell nests (e). In this tissue intensive diffuse Slug reaction was detected (f). In HNSCC with p53 gain of function mutation diffuse CK reaction was detected at different levels in cells of the cancer cell nests, the majority of the nuclei in cancer cell nests contained red KLF4 reaction, and in this tissue section the vimentin reaction was scattered (g). Several cell nuclei at border areas of cancer cell nests were Slug⁺ (h).

Figure 5

Comparison of enzyme immunohistochemical labeling of Slug (a) and TGF-β1 (b) with immunofluorescence labeling of KLF4 (red), pan-cytokeratin (blue) and vimentin (green) combined (c) in HNSCC with p53 gain of function mutation. The Slug⁺ cells formed a cluster in the cancer cell nest in the tumor-stroma interface (a). In this location scattered stroma cells, and at a lower extent also tumor cells reacted positively with antibody against TGF-β1 (open big arrows) (b). Using increased magnification acquisition of the combined immunofluorescence labeling of CK, vimentin and KLF4 (c), more cells with combined CK, KLF4 and vimentin were detected (c) (white arrows). Bars: 50 µm in brightfield and 10 µm in fluorescence. The colors in fluorescence were chosen to present the co-localizations optimally.

Figure 6

Comparison of enzyme immunohistochemical labeling of Slug (a) and TGF-β1 (b) with immunofluorescence labeling of KLF4 (red), pan-cytokeratin (blue) and vimentin (green); combined (c) in HNSCC with loss of p53 gene product. The Slug⁺ cells were diffusely present in the cancer cell nest (a). In this location scattered stroma cells reacted positively with antibody against TGF-β1 (b). Using increased magnification acquisition of the combined immunofluorescence labeling of CK, vimentin and KLF4 (c), more cells with combined KLF4, vimentin and CK (c), were detected (white arrows). Bars: 100 µm in brightfield and 20 µm in fluorescence. The colors in fluorescence were chosen to present the co-localizations optimally.

Figure 7

Western blot detection of Slug (a), KLF4 (b) and vimentin (e) proteins in SCC-25 cells in control, 1 ng/mL TGF-β1 and in 50 ng/mL IL-6 treated conditions. Both Slug (c) (p = 10⁻⁴ by Kruskal-Wallis test and Dunn’s multiple comparison) and vimentin (f) (p = 0.0083 Kruskal-Wallis test and Dunn’s multiple comparison) showed significant upregulation by TGF-β1 treatment (a,c,e,f), whereas the KLF4 protein levels (b,d) were statistically significantly reduced (p = 0.017 by unpaired t-test compared to control). IL-6 upregulated Slug (c) and downregulated KLF4 (b,d), but it was not statistically significant. Vimentin showed a significant increase (p = 0.0083 Kruskal-Wallis test and Dunn’s multiple comparison) by IL-6 treatment (e,f). Four western blot membranes were acquired digitally, and the band optical densities (ODs) of proteins of interest and of loading control (GAPDH) were measured using the Image Studio Lite of Li-cor. The ODs of proteins of interest were normalized to loading control. Mean normalized optical densities of control samples were set to “1”, and this control mean was used as reference (c,d,f). Loading control by western blot membrane reacted with GAPDH antibody is presented on panel (g).

Figure 8

Representative western blots of Slug (a), KLF4 (b) and vimentin (e) proteins in UPCI-SCC-90 cells in control, 1 ng/mL TGF-β1 and in 50 ng/mL IL-6 treated conditions. TGF-β1 induced a moderate but significant increase (p = 0.0499, by Mann Whitney test compared to control) in Slug protein level. Il-6 did not statistically upregulated Slug in UPCI- SCC-90 cells (a,c). KLF4 was not regulated by TGFβ1, but it was upregulated by IL-6 (p = 0.021 by Student’s t-test compared to control) (b,d). TGF-β1 upregulated vimentin at protein levels, but it was not statistically significant. IL-6 induced significant upregulation of vimentin (p = 0.012 by Student’s t-test compared to control) (e,f). Four western blot membranes were acquired digitally, and the band optical densities (ODs) of proteins of interest and of loading control (GAPDH) were measured using the Image Studio Lite of Li-cor. The ODs of proteins of interest were normalized to loading control. Mean normalized optical densities of control samples were set to “1”, and this control mean was used as reference (c,d,f). Loading control by western blot membrane reacted with GAPDH antibody is presented on panel (g).

Figure 9

Western blot detection of p-stat3 (a), stat3 (b), p-p38 (c), p38 (d) proteins in SCC-25 cells in control, 1 ng/mL TGF-β1 and in 50 ng/mL IL-6 treated conditions. GAPDH was used as loading control in both SCC-25 and UPCI-SCC90 cells (e,f). Phospho-stat3 (p-stat3) and stat-3 (a,b) were upregulated by IL-6 treatment. TGF-β1 induced phosphorylation and activation of the estimated p38 mitogen-activated protein kinase (c,d).

Figure 10

Enzyme immunohistochemistry for Slug (a), and KLF4 (b) in HNSCC tumor tissue. Staining differences between the core and the border of the tumor cell nest were analyzed by TissueFax^® (Tissuegnostics^TM). Slug IHC reaction was detected at the border of the tumor cell nest (a). The tumor cell nest boundary showed a drastic decline or loss of the epithelial marker KLF4 in Slug positive HNSCC tumor tissue samples (b). The decreasing gradient of KLF4 from the middle to the border of the tumor cell nest (b), was presented together with the increasing gradient of the nuclear reaction of Slug, localized at the border of the tumor cell nest (a).