The atypical mitogen-activated protein kinase ERK3 is essential for establishment of epithelial architecture

Abstract

Epithelia contribute to physical barriers that protect internal tissues from the external environment and also support organ structure. Accordingly, establishment and maintenance of epithelial architecture are essential for both embryonic development and adult physiology. Here, using gene knockout and knockdown techniques along with gene profiling, we show that extracellular signal-regulated kinase 3 (ERK3), a poorly characterized atypical mitogen-activated protein kinase (MAPK), regulates the epithelial architecture in vertebrates. We found that in Xenopus embryonic epidermal epithelia, ERK3 knockdown impairs adherens and tight-junction protein distribution, as well as tight-junction barrier function, resulting in epidermal breakdown. Moreover, in human epithelial breast cancer cells, inhibition of ERK3 expression induced thickened epithelia with aberrant adherens and tight junctions. Results from microarray analyses suggested that transcription factor AP-2α (TFAP2A), a transcriptional regulator important for epithelial gene expression, is involved in ERK3-dependent changes in gene expression. Of note, TFAP2A knockdown phenocopied ERK3 knockdown in both Xenopus embryos and human cells, and ERK3 was required for full activation of TFAP2A-dependent transcription. Our findings reveal that ERK3 regulates epithelial architecture, possibly together with TFAP2A.

Keywords: Xenopus; cell biology; development; embryo; epidermis; epithelial cell; mitogen-activated protein kinase (MAPK).

Figure 1.

ERK3 expression in X. laevis embryos. A, real-time quantitative RT-PCR analysis. The expression levels of ERK3 were normalized to those of odc in two independent experiments (#1 and #2). The normalized ERK3 expression level at stage 1 was defined as 1.0 in each experiment. B, whole-mount in situ hybridization analysis of ERK3 expression. ec, ectoderm; nc, neural crest; s, somite; nt, neural tube; ep, epidermis; pn, pronephros; b, brain; e, eye. Scale bars, 400 μm. Anterior is to the left (stage (St.) 19–33/34). Dorsal is up for the lateral views (stage 19–33/34). Shown are representative images of 6–10 embryos from one experiment using the ERK3A probe. Essentially the same results were obtained for 6–9 embryos from another experiment using the ERK3B probe.

Figure 2.

ERK3 MOs block the translation of ERK3A and/or ERK3B mRNA(s). A, The target sequences of three antisense MOs for X. laevis ERK3 homeologs are shown in rectangles. ERK3 MO1 and MO2 were designed to target ERK3A alone and ERK3B alone, respectively. ERK3 MO3 was designed to target both ERK3A and ERK3B. B–E, the indicated sets of MOs (80 ng), ERK3 mRNAs (1.6 ng), and GFP mRNA (800 pg, control) were injected into the animal regions of all blastomeres at the 4-cell stage. C-terminally Myc-tagged ERK3A (ERK3A-Myc) and ERK3B (ERK3B-Myc) mRNAs comprise the 5′-UTR and the coding region to contain the entire MO target sequences. N-terminally Myc-tagged ERK3A (Myc-ERK3A) and ERK3B (Myc-ERK3B) mRNAs contain additional AUG and Myc-encoding sequences upstream of the coding region to avoid MO-mediated translational inhibition. To analyze the ERK3A-Myc and ERK3B-Myc proteins, which were unstable due to proteasome-dependent degradation, animal caps (n = 25, each sample) were dissected at stage 9, cultured with 10 μm MG132 (a proteasome inhibitor) until stage 12, and then lysed with 50 μl of lysis buffer. To analyze the Myc-ERK3A and Myc-ERK3B proteins, injected whole embryos (n = 10, each sample) were harvested at stage 12 and lysed with 200 μl of lysis buffer. The protein levels were examined by immunoblotting. The data are representative of two or three independent experiments. B, ERK3 MO1 blocked the translation of ERK3A-Myc mRNA but not that of ERK3B-Myc mRNA. C, ERK3 MO2 blocked the translation of ERK3B-Myc mRNA but not that of ERK3A-Myc mRNA. D, ERK3 MO3 blocked the translation of both ERK3A-Myc and ERK3B-Myc mRNAs. E, Myc-ERK3A and Myc-ERK3B mRNAs were MO-resistant.

Figure 3.

ERK3 is required for pronephros and epidermal development in X. laevis embryos. A–G, phenotypes of ERK3 knockdown embryos. Lateral views are shown with anterior to the left and dorsal up. A–E, control MO (60 ng in A, 40 or 80 ng in B, 15 ng in C, and 40 ng in E), ERK3 MO1/2 (30 ng in A or 7.5 ng in C each of ERK3 MO1 and ERK3 MO2), or ERK3 MO3 (40 or 80 ng in B and 40 ng in E) was co-injected with the lineage tracer dextran-fluorescein into both (A and B) or the left (C and E) V2 blastomere(s) at the 8-cell stage. Embryos with pronephric fluorescence were further analyzed. A and B, phenotypes at stage 42. The black arrowhead indicates edema. The graph indicates the percentage of embryos with edema from four (A) or two (B) independent experiments. **, p < 0.01 by z test. Scale bars, 1 mm. C–E, expression of the pronephric marker gene atp1b1 at stage 33/34. pn, pronephros. Scale bars, 200 μm. C and D, in the rescue experiment, Myc-ERK3A and Myc-ERK3B mRNAs (1 ng each) were co-injected with ERK3 MO1/2. D, percentage of embryos with normal or aberrant atp1b1 expression in pronephros in four independent experiments. Severe, no expression; Mild, weak expression. **, p < 0.01 by Mann–Whitney U test. E, expression of atp1b1 was inhibited by ERK3 MO3 (n = 23/23) but not by control MO (n = 1/28) in three independent experiments. F and G, control MO (60 ng), ERK3 MO1/2 (20 ng each of MO1 and MO2), or ERK3 MO3 (60 ng) was injected into the animal regions of ventral blastomeres at the 4-cell stage. In three independent experiments, the epidermal architecture was disrupted by ERK3 MO1/2 (n = 14/38 at stage 23 and n = 37/38 at stage 39) and ERK3 MO3 (n = 36/40 at stage 39) but not by control MO (n = 0/39 at stages (St.) 23 and 39). Black arrowheads indicate initially observed, local disintegration of the epidermis. Scale bars, 500 μm.

Figure 4.

ERK3 knockdown leads to defective adherens and tight junctions in the epithelia of the X. laevis embryonic epidermis. Control MO (60 ng in A and C–F or 40 ng in B), ERK3 MO1/2 (20 ng each of MO1 and MO2), ERK3 MO3 (60 ng), or TFAP2A MO1/2 (20 ng each of MO1 and MO2) was injected into the animal regions of both ventral blastomeres at the 4-cell stage. A, injected embryos were fixed at stage 13 for immunofluorescence using the anti-E-cadherin antibody. Apical surfaces of the epidermal epithelia were observed by confocal microscopy. Scale bars, 10 μm. CAAX-GFP mRNA (200 pg) was co-injected to visualize the plasma membrane. B, real-time quantitative RT-PCR analysis of E-cadherin and ZO-1 expression. Injected embryos were harvested at stage 14. The expression levels of E-cadherin or ZO-1 were normalized to those of odc. The bars represent the average ± S.D. (error bars). The normalized E-cadherin or ZO-1 expression level in control embryos was defined as 1.0 in each experiment. Shown are all data points from four independent experiments. **, p < 0.01 by t test. n.s., not significant by t test. C–E, injected embryos were fixed at stage 13 for immunofluorescence using the anti-ZO-1 antibody. Apical surfaces of the epidermal epithelia were observed by confocal microscopy. C, scale bars, 10 μm. The yellow arrowheads indicate gaps in ZO-1 signals. D, each point indicates the number of gaps in ZO-1 signals per microscopic field (control MO, n = 23; ERK3 MO1/2, n = 25; ERK3 MO3, n = 27; TFAP2A MO1/2, n = 27) from two independent experiments. The bars represent the average ± S.D. **, p < 0.01 by Tukey test. E, each point represents the average fluorescence intensity of junctional ZO-1 in one cell (control MO, 80 cells from 27 microscopic fields; ERK3 MO1/2, 80 cells from 27 microscopic fields; ERK3 MO3, 81 cells from 27 microscopic fields; TFAP2A MO1/2, 81 cells from 27 microscopic fields; from two independent experiments). The bars represent the average ± S.D. **, p < 0.01 by Tukey test. F, injected embryos after vitelline membrane removal at stage 23 were exposed to 1 mg/ml EZ-link Sulfo-NHS-LC-Biotin for 10 min at 15 °C and then fixed for transverse sectioning. EZ-link Sulfo-NHS-LC-Biotin was detected with streptavidin FITC. Scale bars, 100 μm. Dorsal is up. A and F, essentially the same results were obtained for 8–11 embryos from three independent experiments. The data are representative of 19–23 images.

Figure 5.

ERK3 knockdown induces thickened epithelia with defective adherens and tight junctions in MCF7 cells. A, two different siRNAs targeting human ERK3 were effective in MCF7 cells. Cells were transfected with 20 nm control siRNA, ERK3 siRNA1 or ERK3 siRNA2, cultured for 3 days, and then subjected to real-time quantitative RT-PCR analyses. The expression levels of human ERK3 were normalized to those of human GAPDH. Shown are all data points from three or seven independent experiments. The bars represent the average ± S.D. (error bars). The normalized ERK3 expression level in control cells was defined as 1.0 in each experiment. **, p < 0.01 by Dunnett's test. B, typical morphologies of MCF7 cells 3 days after transfection with the indicated siRNAs (20 nm). The data are representative of three images from three independent experiments. Scale bars, 100 μm. C, proliferation was assessed by cell counting. MCF7 cells (3.5 × 10⁵) were transfected with the indicated siRNAs (20 nm) and counted 3 days after transfection. Shown are all data points from three independent experiments. Bars, average ± S.D. n.s., not significant by Dunnett's test. D–H, MCF7 cells were transfected with the indicated siRNAs (20 nm), fixed 3 days after transfection, and subjected to triple staining with Hoechst (blue), Alexa Fluor 568 phalloidin (red), and an antibody against E-cadherin (D and E) or ZO-1 (F and G) (green). 3D images obtained by confocal microscopy were viewed from the apical (D and F), lateral (E), or apico-lateral (G) sides. D and E, representative images of 6–10 3D images from three independent experiments. F and G, representative images of 6–13 3D images from three independent experiments. D and F, higher magnification images of the regions outlined in white in the top panels are shown in the bottom panels. Images of Alexa Fluor 568 phalloidin staining are not shown. Scale bars, 20 μm. E, scale bars, 10 μm. H, the scatter plot represents the maximum thickness of the cell layers (or colonies) in the 3D images in D–G. The bars represent the average ± S.D. *, p < 0.05; **, p < 0.01 by Dunnett's test.

Figure 6.

ERK3 knockout induces thickened epithelia with defective adherens and tight junctions in MCF7 cells. A, genomic structure of the human ERK3 gene. Dark green, translated regions in exons; light green, untranslated regions in exons. sgRNA target sequences are underlined in blue. The black arrowhead indicates the predicted double-strand break (DSB) site between 3 and 4 nucleotides upstream of the PAM sequence (underlined in red). B, genomic sequencing of ERK3-knockout clones (Clone #1 and Clone #2) showing a one-base insertion in exon 3. C, ERK3 protein levels in WT and ERK3-knockout clones (#1 and #2) were determined by Western blotting. The data are from one experiment. D, typical morphologies of WT and ERK3-knockout cells (Clone #1 and Clone #2). The data are representative of two images from one experiment. Scale bars, 100 μm. E, proliferation was assessed by cell counting. WT and ERK3-knockout cells (3.5 × 10⁵) were seeded in 6-well plates and counted after 3 days. The data points are from four independent experiments. The bars represent the average ± S.D. (error bars). *, p < 0.05; **, p < 0.01 by Dunnett's test. F–J, WT or ERK3-knockout cells were seeded on coverglasses, cultured for 3 days, and subjected to triple staining with Hoechst (blue), Alexa Fluor 568 phalloidin (red), and an antibody against E-cadherin (F and I) or ZO-1 (G and H) (green). 3D images obtained by confocal microscopy were viewed from the apical (F and G), apico-lateral (H), or lateral (I) sides. F and I, shown are representative images of 4–7 3D images from three coverglasses used in one experiment. G and H, shown are representative images of three 3D images from three coverglasses used in one experiment. F and G, higher magnification images of the regions outlined in white in the top panels are shown in the bottom panels. Images of Alexa Fluor 568 phalloidin staining are not shown. Scale bars, 20 μm. I, scale bars, 10 μm. J, the scatter plot represents the maximum thickness of the cell layers (or colonies) in the 3D images in F–I. The bars represent the average ± S.D. *, p < 0.05; **, p < 0.01 by Dunnett's test.

Figure 7.

TFAP2A is a candidate factor for contributing to ERK3-dependent gene expression changes. A, sample preparation for the microarray experiment. Control MO (80 ng), ERK3 MO1/2 (40 ng each of MO1 and MO2), or ERK3 MO3 (80 ng) was injected into animal regions of all blastomeres at the 4-cell stage. B, Venn diagram for genes down-regulated >2-fold upon ERK3 knockdown. C, top 20 transcription factors in the transcription factor–binding motif enrichment analysis.

Figure 8.

tfap2a expression in X. laevis embryos. Whole-mount in situ hybridization analysis of tfap2a expression was performed. ve, ventral ectoderm; nc, neural crest; s, somite; ep, epidermis; pn, pronephros; b, brain. Anterior is to the left (stages (St.) 17–39). Dorsal is up in the lateral views (stages 17–39). Essentially the same results were obtained for 7–10 embryos from one experiment. Scale bars, 400 μm.

Figure 9.

TFAP2A MOs block the translation of tfap2a.L or tfap2a.S mRNA. A, target sequences of antisense MOs for X. laevis tfap2a homeologs are shown in rectangles. TFAP2A MO1 and MO2 were designed to target tfap2a.L and tfap2a.S, respectively. B and C, the indicated sets of MOs (80 ng), tfap2a mRNAs (1.6 ng), and GFP mRNA (400 pg, control) were injected into all blastomeres at the 4-cell stage. C-terminally V5-tagged tfap2a.L (tfap2a.L-V5) and tfap2a.S (tfap2a.S-V5) mRNAs comprise the 5′-UTR and the coding region to contain the entire MO target sequences. Injected embryos (n = 10 in each sample) were harvested at stage 12 and lysed with 200 μl of lysis buffer. The protein levels were examined by immunoblotting. The data are representative of two independent experiments. B, TFAP2A MO1 blocked the translation of tfap2a.L-V5 mRNA but not that of tfap2a.S-V5 mRNA. C, TFAP2A MO2 blocked the translation of tfap2a.S-V5 mRNA but not that of tfap2a.L-V5 mRNA.

Figure 10.

Transcriptomic changes induced by ERK3 knockdown are similar to those induced by tfap2a knockdown. A, sample preparation for the second microarray experiment. Control MO (80 ng), ERK3 MO1/2 (40 ng each of MO1 and MO2), or TFAP2A MO1/2 (40 ng each of MO1 and MO2) was injected into animal regions of all blastomeres at the 4-cell stage. B, Venn diagrams for genes down-regulated or up-regulated >1.5-fold upon ERK3 or tfap2a knockdown. C and D, scatter plots represent the -fold changes of all (C) or cell junction–related (D) gene expression levels in ERK3 or tfap2a morphants relative to those in control morphants. r, correlation coefficient.

Figure 11.

tfap2a knockdown phenocopies ERK3 knockdown in X. laevis embryos. A and B, control MO (40 (A) or 20 (A and B) ng) or TFAP2A MO1/2 (20 (A) or 10 (A and B) ng each of MO1 and MO2) were co-injected with dextran-fluorescein into both (A) or the left (B) V2 blastomere(s) at the 8-cell stage. Embryos with pronephric fluorescence were further analyzed. A, top panels, stage 42 embryos injected with control MO (40 ng) or TFAP2A MO1/2 (20 ng each of MO1 and MO2). The black arrowhead indicates edema. The graph shows the percentage of embryos with edema from two independent experiments. **, p < 0.01 by z test. B, atp1b1 expression at stage 33/34 was inhibited by TFAP2A MO1/2 (n = 21/21) but not by control MO (n = 0/18) in three independent experiments. pn, pronephros. C, control MO (80 ng) or TFAP2A MO1/2 (40 ng each of MO1 and MO2) was injected into animal regions of both ventral blastomeres at the 4-cell stage. At stage 39, epidermal disintegration was induced by TFAP2A MO1/2 (n = 31/31) but not by control MO (n = 0/39) in three independent experiments. A–C, lateral views with anterior to the left and dorsal up. Scale bars, 1 (A), 0.2 (B), or 0.5 (C) mm. D–F, control MO (60 ng in D and F or 40 ng in E) or TFAP2A MO1/2 (20 ng each of MO1 and MO2) was injected into the animal regions of both ventral blastomeres at the 4-cell stage. D, injected embryos were fixed at stage 13 for immunofluorescence using the anti-E-cadherin antibody. Apical surfaces of the epidermal epithelia were observed by confocal microscopy. Scale bars, 10 μm. CAAX-GFP mRNA (200 pg) was co-injected to visualize the plasma membrane. E, real-time quantitative RT-PCR analysis of E-cadherin and ZO-1 expression. Injected embryos were harvested at stage 14. The expression levels of E-cadherin or ZO-1 were normalized to those of odc. The bars represent the average ± S.D. (error bars). The normalized E-cadherin or ZO-1 expression level in control embryos was defined as 1.0 in each experiment. Shown are all data points from four independent experiments. **, p < 0.01 by t test. n.s., not significant by t test. F, injected embryos after vitelline membrane removal at stage 23 were exposed to 1 mg/ml EZ-link Sulfo-NHS-LC-Biotin for 10 min at 15 °C and then fixed for transverse sectioning. EZ-link Sulfo-NHS-LC-Biotin was detected with streptavidin FITC. Scale bars, 100 μm. Dorsal is up. D and F, essentially the same results were obtained for 8–10 embryos from two independent experiments. The data are representative of 13–15 images.

Figure 12.

TFAP2A knockdown phenocopies ERK3 knockdown in MCF7 cells. A, two different siRNAs targeting human TFAP2A were effective in MCF7 cells. Cells were transfected with 20 nm control siRNA, TFAP2A siRNA1, or TFAP2A siRNA2, cultured for 3 days, and then subjected to real-time quantitative RT-PCR analyses. The expression levels of human TFAP2A were normalized to those of human GAPDH. Shown are all data points from three independent experiments. The bars represent the average ± S.D. (error bars). The normalized TFAP2A expression level in control cells was defined as 1.0 in each experiment. **, p < 0.01 by Dunnett's test. B, typical morphologies of MCF7 cells 3 days after transfection with the indicated siRNAs (20 nm). The data are representative of three images from three independent experiments. Scale bars, 100 μm. C, proliferation was assessed by cell counting. MCF7 cells (3.5 × 10⁵) were transfected with the indicated siRNAs (20 nm) and counted 3 days after transfection. Shown are all data points from three independent experiments. The bars represent the average ± S.D. n.s., not significant by Dunnett's test. D–H, MCF7 cells were transfected with the indicated siRNAs (20 nm), fixed 3 days after transfection, and subjected to triple staining with Hoechst (blue), Alexa Fluor 568 phalloidin (red), and an antibody against E-cadherin (D and E) or ZO-1 (F and G) (green). 3D images obtained by confocal microscopy were viewed from the apical (D and F), lateral (E), or apico-lateral (G) sides. D and E, shown are representative images of 6–8 3D images from three independent experiments. F and G, shown are representative images of 6–7 3D images from three independent experiments. D and F, higher-magnification images of the regions outlined in white in the top panels are shown in the bottom panels. Images of Alexa Fluor 568 phalloidin staining are not shown. Scale bars, 20 μm. E, scale bars, 10 μm. H, the scatter plot represents the maximum thickness of the cell layers (or colonies) in the 3D images in D–G. The bars represent the average ± S.D. **, p < 0.01 by Dunnett's test.

Figure 13.

Phenotypes induced by ERK3 knockdown are partially rescued by tfap2a overexpression in X. laevis. A and B, control MO (20 ng) or ERK3 MO1/2 (10 ng each of MO1 and MO2) was injected with or without tfap2a.L mRNA (400 pg) into the left V2 blastomere at the 8-cell stage. All injections were carried out using dextran-fluorescein as a lineage tracer. Embryos with pronephric fluorescence were fixed at stage 33/34 for whole-mount in situ hybridization. A, lateral views with anterior to the left and dorsal up. pn, pronephros. Scale bars, 200 μm. B, the percentage of embryos with normal or aberrant atp1b1 expression in pronephros was scored in five independent experiments. Severe, no expression; Moderate, weak and discontinuous expression; Mild, weak but continuous expression. *, p < 0.05 by Mann–Whitney U test. C and D, control MO (40 ng) or ERK3 MO1/2 (20 ng each of MO1 and MO2) was injected with or without tfap2a.L mRNA (800 pg) into the animal regions of both ventral blastomeres at the 4-cell stage. CAAX-GFP mRNA (200 pg) was also co-injected in all conditions to visualize the plasma membrane. Injected embryos were fixed at stage 13 for immunofluorescence using anti-E-cadherin. Apical surfaces of the epidermal epithelia were observed by confocal microscopy. Scale bars, 10 μm. D, each point represents the average fluorescence intensity of junctional E-cadherin in one cell (control MO, 49 cells from 16 microscopic fields; ERK3 MO1/2, 51 cells from 20 microscopic fields; ERK3 MO1/2 plus tfap2a.L mRNA, 58 cells from 19 microscopic fields; from two independent experiments). The bars represent the average ± S.D. (error bars). *, p < 0.05 by Mann–Whitney U test.

Figure 14.

ERK3 regulates TFAP2A-dependent transcription. A, real-time quantitative RT-PCR analysis of tfap2a expression in X. laevis embryos. Control MO (40 ng), ERK3 MO1/2 (20 ng each of MO1 and MO2), or ERK3 MO3 (40 ng) was injected into both ventral blastomeres at the 4-cell stage. Injected embryos were harvested at stage 14. The expression levels of tfap2a were normalized to those of odc. All data points from four independent experiments are shown. The bars represent the average ± S.D. (error bars). The normalized tfap2a expression level in control embryos was defined as 1.0 in each experiment. n.s., not significant by Dunnett's test. B, the TFAP2-Luc reporter (3xAP2-Luc) contains three tandem repeats of the consensus binding sequence (GCCNNNGCC) of the TFAP2 family to drive firefly luciferase expression. C, luciferase assays using X. laevis embryos. The indicated combinations of control MO (40 ng), ERK3 MO1/2 (20 ng each of MO1 and MO2), and tfap2a.L mRNA (1 ng) were co-injected with the 3xAP2-Luc plasmid (20 pg) and the pGL4.74 Renilla luciferase plasmid (20 pg) into the animal regions at the 2-cell stage. Animal caps dissected at stage 9 or whole embryos were lysed at stage 13 for the Dual-Luciferase assay, in which firefly luciferase activity was normalized to the internal control Renilla luciferase activity. Shown are all data points from four independent experiments. The bars represent the average ± S.D. The normalized relative light unit for firefly luciferase in control animal caps or embryos was defined as 1.0 in each experiment. *, p < 0.05; **, p < 0.01 by t test. D, 3xAP2-Luc (150 ng) and pGL4.74 (10 ng) plasmids were transfected with the indicated combinations of the expression plasmids for ERK3A WT (300 or 600 ng), ERK3A K49R/K50R (KR) (600 ng), or Myc-tagged tfap2a.L (600 ng) into HepG2 cells. The total amount of plasmids was kept constant by adding the mock vector. After 24 h, the cells were lysed for Dual-Luciferase assays. Shown are all data points from 5–8 independent experiments. The bars represent the average ± S.D. The normalized relative light unit for firefly luciferase in control cells was defined as 1.0 in each experiment. *, p < 0.05 by Dunnett's test. n.s., not significant. E, real-time quantitative RT-PCR analysis of ERK3 expression. HepG2 cells were transfected with 20 nm control siRNA, ERK3 siRNA1, or ERK3 siRNA2, cultured for 48 h, and then subjected to analyses. The mock plasmid was also transfected to examine the efficiency of siRNAs in the presence of co-delivering plasmid DNAs. Shown are all data points from three or four independent experiments. The expression levels of human ERK3 were normalized to those of human GAPDH. The bars represent the average ± S.D. The normalized ERK3 expression level in control cells was defined as 1.0 in each experiment. **, p < 0.01 by Dunnett's test. F, the 3xAP2-Luc plasmid (1 μg), the pGL4.74 plasmid (40 ng), and the indicated siRNA (20 nm) were simultaneously transfected into HepG2 cells with or without the expression plasmid for Myc-tagged tfap2a.L (0.96 μg). The total amount of plasmids was kept constant by adding the mock vector. After 48 h, the cells were harvested for the Dual-Luciferase assay. Shown are all data points from four independent experiments. The bars represent the average ± S.D. The normalized relative light unit for firefly luciferase in control cells was defined as 1.0 in each experiment. **, p < 0.01 by Tukey's test. n.s., not significant by Tukey's test.