The Crosstalk between Nrf2 and TGF-β1 in the Epithelial-Mesenchymal Transition of Pancreatic Duct Epithelial Cells

Abstract

Nrf2 and TGF-β1 both affect tumorigenesis in a dual fashion, either by preventing carcinogen induced carcinogenesis and suppressing tumor growth, respectively, or by conferring cytoprotection and invasiveness to tumor cells during malignant transformation. Given the involvement of Nrf2 and TGF-β1 in the adaptation of epithelial cells to persistent inflammatory stress, e.g. of the pancreatic duct epithelium during chronic pancreatitis, a crosstalk between Nrf2 and TGF-β1 can be envisaged. By using premalignant human pancreatic duct cells (HPDE) and the pancreatic ductal adenocarcinoma cell line Colo357, we could show that Nrf2 and TGF-β1 independently but additively conferred an invasive phenotype to HPDE cells, whereas acting synergistically in Colo357 cells. This was accompanied by differential regulation of EMT markers like vimentin, Slug, L1CAM and E-cadherin. Nrf2 activation suppressed E-cadherin expression through an as yet unidentified ARE related site in the E-cadherin promoter, attenuated TGF-β1 induced Smad2/3-activity and enhanced JNK-signaling. In Colo357 cells, TGF-β1 itself was capable of inducing Nrf2 whereas in HPDE cells TGF-β1 per-se did not affect Nrf2 activity, but enhanced Nrf2 induction by tBHQ. In Colo357, but not in HPDE cells, the effects of TGF-β1 on invasion were sensitive to Nrf2 knock-down. In both cell lines, E-cadherin re-expression inhibited the proinvasive effect of Nrf2. Thus, the increased invasion of both cell lines relates to the Nrf2-dependent downregulation of E-cadherin expression. In line, immunohistochemistry analysis of human pancreatic intraepithelial neoplasias in pancreatic tissues from chronic pancreatitis patients revealed strong Nrf2 activity already in premalignant epithelial duct cells, accompanied by partial loss of E-cadherin expression. Our findings indicate that Nrf2 and TGF-β1 both contribute to malignant transformation through distinct EMT related mechanisms accounting for an invasive phenotype. Provided a crosstalk between both pathways, Nrf2 and TGF-β1 mutually promote their tumorigenic potential, a condition manifesting already at an early stage during inflammation induced carcinogenesis of the pancreas.

Fig 1. Interference of TGF-β1 with Nrf2 activation in premalignant and malignant pancreatic duct cells.

HPDE or Colo357 cells either left untreated or treated for 8h with 50 μM tBHQ, 10 μM SFN or 10 ng/mL TGF-β1, either alone or in combination, were analysed by Nrf2 western blot (A,D) using nuclear extracts (lamin-A served as loading control), by ARE-luciferase assay (B,E) using an empty vector as control (co) and the pARE vector (ARE), or by real time PCR (C,F) for detection of the established Nrf2 target genes NQO1 and GCLC (TBP served as control). In A) and D) representative results from three independent experiments are shown. A densitometric band intensity evaluation is provided in Figs. A and B in S1 File. In B) and E) data represent the mean ± SD of four independent experiments performed in duplicate, and in C) and F) data represent the mean ± SD of six independent experiments. *p<0.05 (treated versus untreated).

Fig 2. HPDE cell morphology and wound healing after Nrf2 activation by tBHQ or TGF-β1 treatment.

A) HPDE cells were treated with 50 μM tBHQ or 10 ng/mL TGF-β1, or were left untreated for 24h. Then, cells were analysed by microscopy (at 200x magnification) and photographs were taken. B) Confluently grown HPDE cells in a two-chamber insert were treated with 50 μM tBHQ or 10 ng/mL TGF-β1, either alone or in combination, or were left untreated. Then, the insert was removed (t = 0h) and selected areas were analysed by microscopy (at 100x magnification) and photographed at the indicated periods. *marks the intitial wound edges.

Fig 3. Effect of Nrf2 activation and TGF-β1 on the invasiveness of premalignant HPDE and malignant Colo357 pancreatic duct cells.

Modified Boyden chamber assays on collagen-I coated transwells were performed with A) HPDE or C) Colo357 cells treated either alone or in combination with 50 μM tBHQ, 10 μM SFN or 10 ng/mL TGF-β1, or left untreated for 24h. Data are expressed as percentage of invaded cells and represent the mean ± SD from eight independent experiments, *p<0.05 (+tBHQ and +SFN versus–tBHQ and–SFN, respectively). Boyden assays were also conducted with B) HPDE cells or D) Colo357 cells subject to Nrf2 or control siRNA treatment for 48h (westernblot verification shown in Figs. A and B in S2 File) prior to further treatments with 50 μM tBHQ and/or 10 ng/mL TGF-β1. Data are expressed as percentage of invaded cells and represent the mean ± SD from six independent experiments, *p<0.05 (treated versus untreated).

Fig 4. Effect of Nrf2 activation and TGF-β1 on the expression of EMT markers in premalignant HPDE and malignant Colo357 pancreatic duct cells.

HPDE or Colo357 cells incubated with 50 μM tBHQ, 10 μM SFN or 10 ng/mL TGF-β1 alone or in combination or without for 24h. Then, either nuclear extracts (n.e.) or total cell lysates (t.c.l.) were analysed by westernblot (A,C), or RNA samples were analysed by real-time PCR (B,D) for the expression of Slug, L1CAM, E-cadherin or vimentin. Lamin-A/C and Hsp90 were used as loading controls for the westernblots of nuclear extracts and total cell lysates, respectively (A,C), and for normalization of Slug, L1CAM, E-cadherin and vimentin mRNA level TBP was analysed in parallel (B,D). Either a representative result from three independent experiments (A,C) or the mean ± SD from six independent experiments (B,D) are shown, *p<0.05 (+ tBHQ and + SFN versus–tBHQ and–SFN, respectively); (A,C) a densitometric band intensity evaluation is provided in Figs. A and B in S3 File.

Fig 5. Nrf2 dependency of the tBHQ effects on EMT marker expression in premalignant HPDE and malignant Colo357 pancreatic duct cells.

HPDE or Colo357 cells incubated with Nrf2 or control siRNA for 48h were further treated for 24h with tBHQ or TGF-β1, either alone or in combination, or were left untreated. Then, either nuclear extracts (n.e.) or total cell lysates (t.c.l.) were analysed by westernblot (A,C), or RNA samples were analysed by real-time PCR (B,D) for the expression of Slug, L1CAM, E-cadherin or vimentin. Lamin-A/C and Hsp90 were used as loading controls for the westernblots of nuclear extracts and total cell lysates, respectively (A,C), and for normalization of Slug, L1CAM, E-cadherin and vimentin mRNA level TBP was analysed in parallel (B,D). Either a representative result from three independent experiments (A,C) or the mean ± SD from six independent experiments (B,D) are shown, *p<0.05 (+tBHQ versus–tBHQ). (A,C) a densitometric band intensity evaluation is provided in Figs. A and B in S4 File.

Fig 6. Nrf2 activation affects basal and TGF-β1 dependent Smad and JNK signalling pathways in premalignant and malignant pancreatic duct cells.

A) HPDE or B) Colo357 cells incubated with 50 μM tBHQ, 10 μM SFN or 10 ng/mL TGF-β1 alone or in combination for the indicated periods. Then total cell lysates were analysed by westernblot (A,C) for expression of P-Smad2, P-Smad3, Smad2/3, P-JNK and JNK using Hsp90 as loading control. Representative results from three independent experiments are shown and a densitometric band intensity evaluation is provided in Figs. A and B in S5 File. B) HPDE or D) Colo357 cells were transfected with pGL3 (control) or p6SBE (SRE) together with ptkRL followed by incubation with 50 μM tBHQ and/or 10 ng/mL TGF-β1 or without. After 16h, cell lysates were analysed for firefly and renilla luciferase expression and firefly luciferase units were normalized to those of renilla luciferase. Data represent the mean ± SD of six independent experiments performed in duplicate, p<0.05 (+tBHQ versus-tBHQ). E) HPDE or F) Colo357 cells were treated with Nrf2 or control siRNA for 48h, followed by incubation with 50 μM tBHQ, 10 μM SFN and/or 10 ng/mL TGF-β1, or without. Then, total cell lysates were analysed by westernblot for expression of P-Smad2, P-Smad3, Smad2/3, P-JNK and JNK using Hsp90 as loading control. Representative results from three independent experiments are shown, and a densitometric band intensity evaluation is provided in Figs. C and D in S5 File.

Fig 7. A Nrf2 binding site in the human E-cadherin promoter exerts transcriptional repression.

HPDE (A) or Colo357 (B) cells were transfected with firefly luciferase reporter gene constructs containing the ARE like site (Ecad[–1189]) from the E-cadherin promoter, or not (Ecad[–1153]), or with the empty reporter gene vector. Cells were left untreated or were treated with tBHQ or SFN for 8h. Then, firefly luciferase activity was measured and normalized to renilla luciferase. Data represent the mean of 4 independent experiments. A scheme of the cloned E-cadherin promoter fragments is provided in Figs. A and B in S6 File.

Fig 8. Maintained expression of E-cadherin affects the inducing effect of Nrf2 activation on the invasion of premalignant and malignant pancreatic duct cells.

HPDE (A) or Colo357 (B) cells were transfected with a constitutive E-cadherin expression vector or an empty pcDNA3.1 vector (mock). Afterwards, cells were incubated with 50 μM tBHQ or 10 ng/mL TGF-β1 for 24h, or were left untreated. Then, cells were submitted to the modified Boyden assay on collagen-I coated transwells (A,B, right panels). In parallel, total cell lysates were analysed by E-cadherin western blots (A,B, left panels) using Hsp90 as loading control. A densitometric band intensity evaluation is provided in Figs. A and B in S7 File. The westernblots show representative results from four independent experiments. The invasion data (A,B, right panels) represent the mean ± SD of four independent experiments performed in duplicate, *p<0.05 (treated versus untreated).

Fig 9. Immunohistochemistry analysis for activated Nrf2 and E-cadherin expression in premalignant pancreatic tissues.

Formalin fixed and paraffin embedded pancreatic tissues from CP patients was subjected to immunostaining for P-Nrf2 and E-cadherin. Representative images are shown of (A) normal ducts and PanIN lesions exhibiting high expression of P-Nrf2 or (B) PanIN lesions with low expression of P-Nrf2 that display the respective reciprocal expression level of E-cadherin. Usage of the isotype matched control antibodies revealed no or only weak background staining (not shown). Images were taken at 400x magnification. Arrows indicate ductal regions of reciprocal P-Nrf2 and E-cadherin expression within the same lesion.