Curcumin activates a ROS/KEAP1/NRF2/miR-34a/b/c cascade to suppress colorectal cancer metastasis – Cell Death & Differentiation

p53-independent effects of curcumin on CRC cells

As shown previously [23], the amount of p53 protein increased after treatment of HCT116 cells with curcumin (Fig. 1A). In order to determine the relevance of p53 for the effects of curcumin on CRC cells, we treated HCT116 with wild-type p53 and isogenic HCT116 cells with homozygous deletion of p53 with increasing concentrations of curcumin for 48 hours. P53-proficient cells displayed an IC₅₀ of 21.49 μM, whereas p53-deficient cells had an IC₅₀ of 18.03 μM curcumin (Fig. 1B). Similar effects were observed in p53-proficient or p53-deficient RKO and SW48 CRC cell lines (Figs. S1A and 1D). A concentration of 15 μM curcumin, slightly below the IC₅₀ value of p53-deficient cells, was chosen for the following experiments. Exposure to curcumin strongly reduced the proliferation of both p53-deficient and p53-proficient HCT116 cells as determined by impedance measurements (Fig. 1C) and cell number alterations were confirmed at the final time point (Fig. 1D). Similar p53-independent effects of curcumin were observed in the CRC cell lines RKO and SW48 (Figs. S1B, 1C, 1E, and 1F). Interestingly, the curcumin-induced decrease in viability and proliferation was more pronounced in p53-deficient cells. Therefore, these effects of curcumin on CRC cells are independent of p53, although p53 accumulates after treatment with curcumin.

Next, we analysed which processes may underlie the curcumin-mediated suppression of proliferation. Curcumin resulted in an increase of cells with sub-G₁ DNA content, indicating increased apoptosis, irrespective of the p53 status (Fig. 1E). p53-deficient cells displayed more G₂/M-arrest 24 hours after curcumin treatment, whereas cells accumulated more in G₀/G₁ in WT p53 cells. Annexin V/PI detection revealed a more pronounced increase in apoptosis in p53-deficient when compared to p53-proficient cells (Figs. 1F and S1G). These results were confirmed by detection of cleaved-PARP, cleaved Caspase-3, Bcl-2, and Bax proteins (Figs. 1G, S1H, and S1I): an increase in BAX and cleaved Caspase 3 was detected by 24 hours in p53-deficient cells, whereas p53-proficient cells showed a delayed induction of these proteins by 48 hours. In addition, curcumin induced senescence in HCT116 cells to a similar degree in p53-proficient and p53-deficient HCT116 cells as determined by detection of β-gal pH 6 (Fig. 1H). Finally, curcumin suppressed migration and invasion in p53-proficient and p53-deficient HCT116 cells, with the latter displaying increased basal levels of migration and invasion (Fig. 1I, J). Taken together, curcumin suppressed cell viability, proliferation, migration and invasion, whereas it induced apoptosis and senescence of CRC cells in a p53-independent manner.

Curcumin activates NRF2 via ROS in CRC cells

Next, we aimed to determine the mechanism mediating the effects of curcumin on CRC cells. Although curcumin is mainly known for its anti-oxidative effects, it has also been shown to increase the production of reactive oxygen species/ROS [24]. Indeed, ROS levels of HCT116 cells were significantly increased when exposed to curcumin for 48 hours (Fig. 2A). The ROS-inducer tert-butyl hydroperoxide (tBHP) was used as a positive control. The effect of curcumin was suppressed by concomitant treatment with the antioxidant N-acetylcysteine (NAC) (Fig. 2A, left panel). Similar results were observed in RKO and SW48 cells (Figs. S2A and 2B, left panel). Interestingly, the levels of ROS were higher in p53-deficient HCT116 cells than in p53-proficient HCT116 cells, suggesting that p53 suppresses the production of ROS by curcumin (Fig. 2A, right panel). Similar results were obtained in RKO and SW48 cells (Figs. S2A and 2B, right panel). Notably, the negative effect of curcumin on cell viability was partially reversed by NAC in p53-proficient and p53-deficient HCT116 cells (Fig. 2B). Similar results were observed in RKO and SW48 cells (Figs. S2C and 2D). NAC completely suppressed cleavage of PARP and therefore apoptosis induced by curcumin (Fig. 2C). Similar results were observed in RKO and SW48 cells (Fig. S3B).

A Analysis of ROS formation in HCT116 cells treated as indicated, Left panel: representative IF pictures. Scale bar: 100 µM. Right panel: Quantification of fluorescence intensity. B Cell viability of HCT116 cells after treatment with curcumin and NAC for indicated periods was determined by MTT assays. C Apoptosis of HCT116 cells treated with curcumin and/or NAC for 48 h was determined by Western blot analysis of cleaved PARP protein levels. β-actin served as a loading control. D Immunofluorescence detection of NRF2 protein after treatment with DMSO, 15 μM curcumin, or 15 μM curcumin and 5 mM NAC for 48 hours. Nuclear DNA was detected by DAPI. Scale bar represents 20 µm. E Western blot analysis of NRF2 protein levels in cytoplasmic and nuclear cellular fractions after curcumin treatment. F qPCR analysis of NQO1 expression in HCT116 cells after treatment with DMSO, 15 μM curcumin, or 15 μM curcumin with 5 mM NAC for 48 hours. G Western blot analysis of NQO1 protein levels after treatment with DMSO, 15 μM curcumin, or 15 μM curcumin with 5 mM NAC for 48 hours. In panels A, B, and F mean values ± SD are shown (n = 3). *P <  0.05, **P <  0.01, ***P <  0.001, ****P  <  0.0001.

The KEAP1/NRF2-pathway is a central mediator of the cellular response to oxidative stress [25]. Upon exposure to ROS the transcription factor NRF2 is released from Keap1 and translocates from the cytoplasm to the nucleus to activate the transcription of its target genes [26]. Indeed, curcumin induced translocation of NRF2 to the nucleus in p53-proficient and p53-deficient HCT116 cells (Fig. 2D). Similar results were observed in RKO and SW48 cells (Fig. S3A). After exposure to curcumin, the NRF2 protein was increased in the nuclear fraction of HCT116 cells, whereas it decreased in the cytoplasmic fraction (Fig. 2E). NRF2 mRNA expression was not affected by curcumin (Fig. S3C). NQO1 is a conserved target gene of NRF2 that is commonly used to monitor the activity of the NRF2 pathway [27]. Consistent with an activation of NRF2, expression of NQO1 mRNA and protein was increased upon curcumin treatment (Figs. 2F, G). Similar results were observed in RKO and SW48 cells (Fig. S3B). Importantly, the induction of nuclear translocation and activation of NRF2 was largely reversed by the inhibition of ROS with NAC (Figs. 2D, F, G and S3A).

Curcumin up-regulates miR-34a and miR-34b/c independent of p53

As indicated in the introduction, previously published observations indicating that miR-34a is induced by curcumin sparked our interest in curcumin. Here we determined whether p53 is required for the induction of miR-34 genes by curcumin in a set of p53-deficient, isogenic CRC cell lines. Interestingly, curcumin induced the expression of the primary pri-miR-34a and pri-miR-34b/c transcripts in HCT116 cells independent of their p53 status (Figs. 3A, B, and  S4A, B). Similar results were obtained in the CRC cell lines SW48 and RKO (Figs. 3C–F, and S4C-F). The levels of mature miR-34a were also up-regulated by curcumin in p53-proficient and p53-deficient HCT116 cells (Figs. 3G and S4G). Since the induction of pri-miR-34a and pri-miR34b/c by curcumin was prevented by treatment with NAC, it was mediated by ROS (Fig. 3H, I). Taken together, these results showed that curcumin induces miR-34a and miR-34b/c expression in a p53-independent and ROS-dependent manner in CRC cell lines.

A–F qPCR analysis of pri-miR-34a and pri-miR-34b/c expression after treatment of the indicated cells with 15 μM curcumin for the indicated periods. G qPCR analysis of mature miR-34a expression in HCT116 cells after treatment with 15 μM curcumin for 48 hours. H, I qPCR analyses of pri-miR-34a and pri-miR-34b/c expression in HCT116 cells after treatment with curcumin and/or NAC for 48 hours. In (A)–(I) mean values ± SD are shown (n = 3). **P <  0.01, ***P <  0.001, ****P <  0.0001.

Curcumin-induced NRF2 directly activates miR-34a and miR-34b/c

Next, we hypothesized that NRF2, which is activated by curcumin-induced ROS as shown above, may represent a direct inducer of miR-34a and miR-34b/c expression. Indeed, suppression of NRF2 by a pool of 4 different siRNAs prevented the induction of pri-miR-34a and pri-miR-34b/c expression by curcumin in p53-deficient HCT116 cells (Figs. 4A, B, S4A, and S4B). Conversely, ectopic NRF2 expression increased pri-miR-34a and pri-miR-34b/c expression, as well as mature miR-34a, NQO-1 mRNA and protein expression in SW480 cells, which harbour mutant p53 (Figs. 4C, D). By examining the genomic sequence of the miR-34a and miR-34b/c promoter regions, we identified three potential NRF2 binding sites (TGAG/CnnnGC), so-called ARE (Antioxidant Response Elements), in the miR-34a locus and one potential NRF2 binding site in the miR-34b/c locus (Figs. 4E–G). A qChIP assays confirmed enhanced NRF2 occupancy at the four ARE sites within the miR-34a and miR-34b/c promoters after curcumin treatment for 48 hours (Fig. 4H). Therefore, we concluded that curcumin induces the expression of the miR-34a and miR-34b/c genes by activating NRF2, which directly binds to the miR-34a and miR-34b/c promoters and induces their transcription. Notably, this mode of miR-34 activation is p53-independent.

A, B qPCR analysis of A. pri-miR-34a and B. pri-miR-34b/c expression in HCT116 cells treated as indicated for 48 hours. C qPCR analysis of indicated mRNAs in SW480 cells after transfection with empty pcDNA3.1 or NRF2 pcDNA3.1 vectors for 72 hours. D Western blot analysis of NRF2 and NQO1 proteins in SW480 cells after transfection with pcDNA3.1 and NRF2 pcDNA3.1 for 72 hours. E ARE consensus sequence defined as 5ʹ-A/G TGA C/G NNNGC A/G-3ʹ, where “N” represents any nucleotide according to jaspar.genereg.net (MA0150.1). F Map of human miR-34a and miR-34b/c genomic regions with indicated NRF2 binding sites. G Sequences of NRF2 binding sites within the miR-34a and miR-34b/c genomic regions. H qChIP analysis of NRF2 occupancy at the human miR-34a and miR-34b/c genomic regions in HCT116 p53 −/− cells after treatment with curcumin or DMSO for 48 hours. NQO1 and 16q22 served as positive and negative controls, respectively. In (A)–(C), and (H) mean values ± SD are shown (n = 3). **P <  0.01, ***P <  0.001.

miR-34a and miR-34b/c mediate the effects of curcumin on CRC cells

To determine whether the induction of miR-34a and miR-34b/c expression is required for the effects of curcumin, we employed isogenic HCT116 cells that were rendered deficient for miR-34a and/or miR-34b/c genes using a CRSPR/CAS9 approach (their generation will be described elsewhere). The viability of miR-34a– or miR-34a/b/c-deficient HCT116 cells was significantly higher than miR-34-proficient HCT116 cells after treatment with curcumin (Fig. 5A). MiR-34b/c-deficient cells showed an intermediate gain of viability when compared to miR-34a/b/c-proficient HCT116 cells. In agreement, miR-34a/b/c-deficiency resulted in a decreased enhancement of apoptosis by curcumin when compared to miR-34a/b/c-proficient HCT116 cells (Figs. 5B and S6A). In miR-34a/b/c-proficient HCT116 cells, curcumin treatment induced strong cleavage of caspase 3, whereas cleaved caspase 3 was not detectable in miR-34-deficient cells (Fig. 5C). In miR-34a/b/c-proficient HCT116 cells, curcumin suppressed the expression of the anti-apoptotic protein Bcl-2, a known target of miR-34 [28], and induced the expression of the pro-apoptotic protein BAX (Fig. 5C). HCT116 cells with single or combined deletion of miR-34a and miR-34b/c displayed elevated expression of Bcl-2 and after treatment with curcumin the expression of Bcl-2 exceeded the levels detected in untreated WT cells. This effect was most pronounced in the miR-34a/b/c-deficient HCT116 cells. Furthermore, the induction of BAX by curcumin was diminished in HCT116 cells with miR-34a/b/c-deficiency (Fig. 5C). Therefore, it is conceivable that the repression of Bcl-2 by miR-34a/b/c contributes to the induction of apoptosis by curcumin.

A The indicated cell lines were treated with increasing concentrations of curcumin for 48 hours. IC₅₀ was determined by MTT assays. B Analysis of apoptosis after treatment with 15 μM curcumin for 48 hours determined by Annexin V-FITC and PI staining. C The expression of cleaved PARP, Bcl-2, Bax, and cleaved caspase 3 after treatment with 15 μM curcumin for 48 hours was determined by Western blot analysis. D Detection of senescent cells after curcumin treatment for 48 hours by pH 6 β-gal staining. E Evaluation of migration by wound healing assay 24 hours after treatment with curcumin. F Analysis of invasion using Boyden chamber assays 48 hours after treatment with curcumin or DMSO. G. Analysis of cell viability in HCT116 p53-deficient cells by MTT assay after transfection NRF2 siRNA pool or control siRNA pool 24 hours with curcumin for 48 hours. H Wound healing assay in HCT116 p53-deficient cells treated with curcumin for 24 h after transfection with NRF2-specific siRNA pool or control siRNA pool. I Invasion assay of HCT116 p53-deficient cells exposed to curcumin for 48 hours after transfection with NRF2-specific siRNA pool or control siRNA pool. J Evaluation of migration by wound healing assay 24 hours after transfection with pcDNA3.1 or NRF2 pcDNA3.1. K Analysis of invasion using Boyden chamber assays after transfection with pcDNA3.1 or NRF2 pcDNA3.1. In panels A, B, and D–H mean values ± SD (n = 3) are shown. *P <  0.05, **P  <  0.01, ***P  <  0.001, ****P <  0.0001.

miR-34a, miR-34b/c, and miR-34a/b/c-deficient cells exhibited no increase in senescence after curcumin treatment as determined by detection of β-galactosidase activity at pH 6.0, whereas WT HCT116 cells showed a signficiant increase in senescence (Fig. 5D). Finally, the curcumin-induced suppression of migration and invasion was significantly lower in miR-34a-, miR-34b/c-, and miR-34a/b/c-deficient cells than in miR-34-proficient HCT116 cells (Figs. 5E, F, S6B, and S6C). In all these assays cells with combined inactivation of miR-34a and miR-34b/c showed the strongest resistance to curcumin. Knockdown of NRF2 by siRNAs partially prevented the curcumin-induced suppression of cell viability (Fig. 5G), migration (Figs. 5H and S7A), and invasion (Figs. 5I and S7B). Furthermore, ectopic expression of NRF2 repressed the migration and invasion of miR-34-proficient HCT116 cells (Figs. 5J, K, S7C, and S7D) in the absence of treatment with curcumin. However, in miR-34a/b/c-deficient HCT116 cells ectopic NRF2 had no effect on migration and invasion, demonstrating that miR-34a and miR-34b/c are required mediators of NRF2 function. Taken together, our results demonstrate that the NRF2-mediated activation of miR-34a and miR-34b/c genes is a required mediator for the effects on curcumin on apoptosis, senescence, migration and invasion.

The induction of miR-34a/b/c by H₂O₂ and tBHP is mediated by NRF2

It has been previously shown, that the ROS induced by treatment with H₂O₂ induces the expression of miR-34a [29]. Here, we determined whether this regulation is mediated by NRF2. First, we analysed cell viability of CRC cell lines after treatment with H₂O₂ and determined IC₅₀ values of 92.61, 48.35, and 61.52 µM for p53-proficient HCT116 cells, p53-deficient HCT116 cells, and SW620 cells, which express mutant p53, respectively (Figs. S8A and S8B). Treatment of p53-proficient and p53-deficient HCT116 cells as well as SW620 cells with H₂O₂ or tBHP resulted in the induction of the NRF2 target gene NQO1, indicating that H₂O₂ and tBHP activate NRF2 (Figs. 6A and S8C). Importantly, H₂O₂ and tBHP induced the expression of miR-34a/b/c, which was prevented by siRNA-mediated suppression of NRF2 (Figs. 6B, C, S8D, and S8E).

A–C Expression of NQO1 (A), pri-miR-34a (B), and pri-miR-34b/c (C) after treatment with H₂0₂ and transfection with control or NRF2 siRNA for 24 hours. D, E Expression of pri-miR-34a (D) and pri-miR-34b/c (E) in p53-deficient HCT116 cells after the indicated transfections/treatments. F, G Expression of pri-miR-34a (F) and pri-miR-34b/c (G) in indicated cells after transfection with control or NRF2 siRNA and treatment with curcumin and/or IL-6 (200 ng/ml) for 48 hours. In (A)–(G) mean values ± SD are shown (n = 3)*P <  0.05, **P <  0.01, ***P <  0.001, ****P  <  0.0001.

The suppression of miR-34a/b/c by hypoxia or IL-6 is reversed by curcumin-induced NRF2

We have previously demonstrated that hypoxia [20] and IL-6-mediated STAT3 activation [18] repress miR-34a/b/c expression in p53-deficient CRC cells. To determine whether curcumin could reverse the hypoxia-mediated suppression of miR-34a/b/c, we cultured p53-deficient HCT116 cells under hypoxia and treated them with curcumin. Indeed, the hypoxia-mediated suppression of miR-34a/b/c expression was completely reversed by curcumin (Figs. 6D, E). However, this reversion was significantly weaker when NRF2 was suppressed by siRNAs (Fig. 6D, E), indicating that this effect of curcumin is largely mediated by NRF2 activation. Previous studies showed that IL-6/STAT3 signalling suppresses miR-34a expression via direct binding of STAT3 to the miR-34a promoter region [20]. Notably, the IL-6-mediated suppression of miR-34a and also that of miR-34b/c was completely reversed by treatment with curcumin (Fig. 6F, G). However, this was not the case when NRF2 was suppressed by siRNA (Fig. 6F, G), indicating that this effect of curcumin was mediated by NRF2.

Curcumin induces MET and inhibits lung-metastases formation via inducing miR-34a

Next, we characterized the effect of curcumin on SW620-Luc2 cells, which stably express luciferase, since we intended to study these cells after transplantation into mice. Curcumin inhibited the viability of SW620-Luc2 cells in a dose-dependent manner (Fig. S9A). The IC₅₀ was 14.49 μM as determined by an MTT assay. This concentration of curcumin was used subsequently. The expression of pri-miR-34a and mature miR-34a was upregulated after exposure of SW620-Luc2 cells to curcumin for 48 hours (Fig. 7A, B). Mature miR-34b and miR-34c were expressed at very low levels in SW620-Luc2 cells when compared to mature miR-34a, also after exposure to curcumin (Fig. S9B). Therefore, we focussed on the analysis of the role of miR-34a for the effects of curcumin on SW620-Luc2 cells. Our previous studies had shown that miR-34a critically contributes to p53-induced mesenchymal-epithelial-transition (MET) in CRC cells [17, 30]. Therefore, we asked whether curcumin induced miR-34a may mediate MET. Indeed, treatment of SW620-Luc2 cells with curcumin resulted in repression of the mesenchymal markers Vimentin (VIM), SNAIL, SLUG, and ZEB1 (Fig. 7E). Also this inhibitory effect of curcumin was abolished by miR-34a-specific antagomirs. Curcumin also suppressed invasion and migration, which presumably is a functional consequence of MET, in SW620-Luc2 cells in a miR-34a-dependent manner, as determined by silencing of miR-34a using antagomirs (Figs. 7C, D, S9C, and 9D). Finally, we performed mouse xenograft experiments to determine whether curcumin affects the capacity of CRC cells to form lung metastases. Therefore, we treated SW620-Luc2 cells ex vivo with curcumin or/and miR-34a-specific antagomirs for 48 hours. Subsequently, these cells were injected into the tail veins of NOD/SCID mice to assess the formation of lung metastasis. Longitudinal, non-invasive imaging showed that the treatment of SW620-Luc2 cells with curcumin completely abrogated metastasis formation within 5 weeks after injection (Fig. 7F, G). However, concomitant inhibition of miR-34a by antagomirs partially restored metastasis formation after curcumin treatment (Fig. 7F, G). Five weeks after injection, resected lungs were devoid of macroscopically visible metastases in mice injected with curcumin-treated SW620-Luc2 cells (Fig. 7H). Haematoxylin and eosin (H&E) staining confirmed the absence of metastatic nodules in mice injected with curcumin-treated cells (Fig. 7H, I). However, SW620-luc2 cells concomitantly treated with miR-34a-antagomirs and curcumin, formed lung-metastases (Fig. 7H, I). In summary, these results show that curcumin inhibits metastases formation via inducing miR-34a.

A, B Expression of pri-miR-34a (A) and mature miR-34a (B) in SW620-luc2 cells treated with curcumin or DMSO for 48 hours. C Analysis of SW620-luc2 cell migration after transfection with miR-34 antagomirs and/or curcumin for 72 hours. D Analysis of SW620-luc2 cell invasion after transfection with miR-34 antagomirs and/or curcumin for 72 h. E qPCR analyses of EMT-related genes in SW620-luc2 cells 72 hours after the indicated treatments/transfections. F–I Formation of lung metastases by SW620-Luc2 cells, which were treated as indicated for 48 hours and then injected into tail-vein of NOD/SCID mice. Representative images of luciferase signals at the indicated time points after xenografting (F) and the quantification of total photon flux (G). H right: representative lungs 5 weeks after tail vein injection. Arrows indicate metastatic tumor nodules. Left: representative images of H&E-stained resected lungs. Scale bar: 500 μm; 50 μm (insert). I. Quantification of metastatic nodules in the lungs of indicated mice. In (A)–(E) mean values ± SD (n = 3) are shown. *P <  0.05, **P <  0.01, ***P <  0.001, ****P  <  0.0001.

p53 and miR-34a/b/c modulate sensitivity towards curcumin and/or 5-FU

Finally, we investigated the effect of curcumin on CRC cell viability in combination with the chemotherapeutic drug 5-fluorouracil (5-FU), which is widely used for the treatment of CRC. For this we used clinically relevant and tolerable concentrations of 5-FU (2 mg/L) and curcumin (15 µM) [31, 32]. Combined treatment of HCT116 cells with curcumin and 5-FU showed a stronger suppression of cell viability when compared to single treatment with either compound (Fig. 8A). Compared to p53-proficient cells, p53-deficient cells were more sensitive to curcumin but more resistant to 5-FU. However, no difference was observed between p53-deficient and p53-proficient cells when treated with curcumin and 5-FU (Fig. 8A). Compared to wt cells, miR-34a– and miR-34a/b/c-deficient cells were more resistant to curcumin, but only marginally more resistant to 5-FU. However, miR-34a- and miR-34a/b/c-deficiency resulted in a markedly higher resistance to the combination of curcumin and 5-FU (Fig. 8B). Additional deletion of p53 in miR-34a- or miR-34a/b/c-deficient cells increased their resistance to 5-FU. Interestingly, inactivation of p53 reversed the resistance of miR-34a- or miR-34a/b/c-deficient cells to curcumin and the combination of curcumin and 5-FU (Fig. 8C, D). Human colonic epithelial cells (HCEC-1CT) and human intestinal fibroblasts (CCD-18Co cells) were less sensitive to curcumin and/or 5-FU than HCT116 cells (Fig. 8E, F). Taken together, these results suggest that curcumin may enhance the therapeutic effects of 5-FU on CRC cells. Interestingly, this effect is most pronounced in cells lacking p53 and miR-34a/b/c, which are also known to give rise to CRCs with the poorest prognosis [21].

A–D HCT116 cells with the indicated genotypes were exposed to curcumin (15 μM), 5-FU (2 mg/L), or curcumin combined with 5-FU. After 48 hours cell viability was determined by MTT assays. E, F The viability of HCEC-1CT and CCD-18Co cells was determined by MTT assay after the indicated treatments for 48 hours. G. Schematic model of the findings obtained in this study.