NFE2L2/NRF2 Activity Is Linked to Mitochondria and AMP-Activated Protein Kinase Signaling in Cancers Through miR-181c/Mitochondria-Encoded Cytochrome c Oxidase Regulation

Increased expression of miR-181c in NFE2L2/NRF2-silenced colon cancer cells

We previously found that cellular O₂ consumption was lower in NFE2L2/NRF2-knockdown HT29 and HCT116 cells than in control cells (31). Therefore, we sought to elucidate the molecular mechanisms underlying the altered mitochondrial function in NFE2L2/NRF2-knockdown cells by focusing on miRNA signatures. Our established NFE2L2/NRF2-knockdown stable HT29 cell line (shNRF2-HT) showed lower levels of NFE2L2/NRF2 transcript (0.55-fold) and protein (0.6-fold), and repressed target gene expressions compared with scRNA-expressing control cells (scHT) (Supplementary Fig. S1A–C; Supplementary Data are available online at www.liebertpub.com/ars). The stable knockdown of NFE2L2/NRF2 was also confirmed in HCT116 cells, and levels of transcripts and protein were 0.28- and 0.56-fold of the control cells, respectively (Supplementary Fig. S1D, E).

Then, miRNA expression patterns were analyzed in these cell lines with the Affymetrix GeneChip miRNA 3.0 system. Through comparative analyses, miRNAs exhibiting increases in expression greater than 1.5-fold after NFE2L2/NRF2 knockdown were collected, and 23 miRNAs were commonly increased in both HT29 and HCT116 cells (Table 1). The pathway analysis (miRSystems; http://mirsystem.cgm.ntu.edu.tw/) showed that, among 23 miRNAs, miR-146a-5p, miR-181c-5p, miR-200a-3p, and miR-93-3p were connected with the mitochondrial pathway, including fatty acid β-oxidation, OXPHOS, and/or apoptosis. In particular, miR-181c-5p (miR-181c) was associated with OXPHOS: A recent report showed that miR-181c was identified as mitochondria-enriched miRNA in rat cardiocytes, and it was demonstrated that miR-181c regulated MT-CO1 expression through the binding to the 3-UTR of the rat mt-co1 gene (11). Microarray analysis indicated that miR-181c levels were 2.01- and 1.59-fold higher in NFE2L2/NRF2-knockdown HT29 and HCT116 cells, respectively, than in the corresponding controls (Table 1). Real-time reverse transcription-polymerase chain reaction (RT-PCR) analysis confirmed these results (Fig. 1A).

OPEN IN VIEWER

FIG. 1.

miR-181c and MT-CO1 expression in NFE2L2/NRF2-knockdown colon cancer cells. (A) MiR-181c levels in shNRF2-HT and shNRF2-HCT cells were determined by relative real-time PCR. (B) The 3′-untranslated region (3′-UTR) of the human MT-CO1 gene was fused with the luciferase expression plasmid. The SV40-Luc-MT-CO1 3′ UTR plasmid was transfected into sc and shNRF2-HT cells, and luciferase activity was assessed. (C) MT-CO1, MT-CO2, MT-CO3, VDAC1, and β-tubulin protein levels were determined by Western blotting in total cell lysates from sc and shNRF2-HT cells. The quantified results were from three independent experiments. (D) MT-CO1 and β-tubulin levels were assessed in mitochondrial and cytosolic fractions from sc and shNRF2-HT cells. (E) Immunocytochemical analysis of MT-CO1 in sc and shNRF2-HT cells. Nuclear staining was performed with H342. Results are representative of three independent experiments, and cellular fluorescent intensities were quantified (1890 × magnification). (F) MT-CO1, MT-CO3, VDAC1, and β-tubulin protein levels were determined by Western blotting in cell lysates from sc and shNRF2-HCT cells. The quantified results were from three independent experiments. (G) To confirm the effect of NFE2L2/NRF2-silencing, different sequence shRNAs (Nos. 2–4) targeting NFE2L2/NRF2 were transiently transduced into HCT116 and levels of NFE2L2/NRF2 and MT-CO1 were monitored. Data are mean ± standard errors (SEs) of three to four experiments. ^a P < 0.05 compared with scRNA-expressing control cell line. Ratio to sc: ratios to the nonspecific scRNA-expressing control cell line (sc). Unedited blots of Figure 1C, D, F, and G are shown in Supplementary Blots (Fig. SW1C, SW1D, SW1F, and SW1G). PCR, polymerase chain reaction.

Decreased MT-CO1 levels in NFE2L2/NRF2-knockdown colon cancer cells

Next, we examined whether the increased miR-181c levels affected MT-CO1 expression in NFE2L2/NRF2-knockdown cells. Although MT-CO1 mRNA levels were not altered in NFE2L2/NRF2-silenced HT29 (shNRF2-HT) cells (Supplementary Fig. S2), luciferase activity derived from the 3′-untranslated region (3′-UTR) of the human MT-CO1 gene was significantly low in shNRF2-HT than that in the control cells (Fig. 1B). This result indicated the association of the NFE2L2/NRF2-mediated miR-181c elevation with 3-UTR of the MT-CO1 gene, and translational inhibition of MT-CO1. In agreement with this, MT-CO1 protein expression was reduced by 43% in shNRF2-HT compared with that in scHT cells (Fig. 1C). Unlike MT-CO1, protein levels for MT-CO2, MT-CO3, and VDAC1 were not altered after NFE2L2/NRF2 knockdown (Fig. 1C). MT-CO1 protein was also downregulated in isolated mitochondrial fractions (Fig. 1D) and in immunocytochemical analyses (Fig. 1E). Similarly, MT-CO1 protein levels were decreased by 23% in NFE2L2/NRF2-silenced HCT116 (shNRF2-HCT) cells compared with that in scHCT cells (Fig. 1F). The transient introduction with three additional shRNAs (NRF2 shRNA Nos. 2–4) could confirm the correlation of NFE2L2/NRF2 silencing and MT-CO1 repression (Fig. 1G). Thus, increased miR-181c levels may suppress MT-CO1 expression in shNRF2-HT and shNRF2-HCT cells.

In addition, we assessed expression levels of other ETC components in shNRF2-HT29. There were no significant changes in the expression of nucleus-encoded ETC genes (NADH dehydrogenase 1α subunit 5 [NDUFA5], NADH dehydrogenase iron-sulfur protein-8 [NDUFS8], succinate dehydrogenase iron-sulfur subunit [SDHB], cytochrome b-c1 complex subunit-1 [UQCRC1], and ATP synthase subunit-5O [ATP5O]) after NFE2L2/NRF2 knockdown (Supplementary Fig. S3A). When transcript levels for mitochondria-encoded ETC subunits (NADH-ubiquinone oxidoreductase chains [NDs], cytochrome b [CYB], MT-CO2, MT-CO3, and ATP synthase-6 [ATP6]) were monitored, several subunits, including ND1, ND2, ND3, ND6, and ATP6, were elevated by 1.36–1.54-fold compared with the control cells, indicating that the expression of other ETC subunits was not repressed significantly by NFE2L2/NRF2 knockdown (Supplementary Fig. S3B).

Reduced levels of cytochrome c oxidase activity, MMP, mitochondrial respiration rate, and ATP contents in NFE2L2/NRF2-knockdown colon cancer cells

Altered MT-CO1 expression can cause MMP disturbance and ETC dysfunction in NFE2L2/NRF2-silence cancer cells. Indeed, cytochrome c oxidase (complex IV) activity was significantly lower in shNRF2-HT cells (Fig. 2A). In addition, when ETC function was assessed by monitoring MMP levels by using fluorescence probes, cellular fluorescence intensity of MitoRed was reduced by 50% in shNRF2-HT cells compared with that in scHT cells (Fig. 2B). To exclude the involvement of differential uptake capacity of fluorescent dye, cellular MitoRed intensities were monitored by using a high-throughput cell imager after incubation with MitoRed for 10, 20, and 40 min. Results indicated that cellular MitoRed intensities were significantly lower in shNRF2-HT throughout the incubation time (Fig. 2C). In addition, MitoRed accumulation was greatly suppressed by carbonyl cyanide m-chlorophenyl hydrazone (CCCP), an inhibitor of MMP maintenance, in both cell lines. The accumulation of TMRM, another MMP-sensitive fluorescent dye, was also lower in shNRF2-HT cells than in scHT cells, which confirms MMP reduction in NFE2L2/NRF2-silenced colon cancer cells (Fig. 2D). Moreover, as direct evidence of diminished mitochondrial function, the levels of mitochondrial respiration were significantly low in NFE2L2/NRF2-silenced HT29. The basal oxygen consumption rate (OCR), which reflects the basal mitochondrial respiration, was reduced by 0.66-fold by NFE2L2/NRF2 knockdown, and spare respiratory capacity was also significantly lower in knockdown cells (Fig. 2E). As a result of MMP disturbance and respiration diminishment, cellular ATP levels were decreased by 23% in shNRF2-HT compared with that in the sc control (Fig. 2F). However, the activity of complex I (NADH:ubiquinone oxidoreductase) was not affected by NFE2L2/NRF2 silencing, indicating that altered tricarboxylic acid (TCA) cycle or differential NADH availability was not associated with MMP reduction and respiratory dysfunction in NFE2L2/NRF2-silenced cells (Fig. 2G). Similar to HT29, NFE2L2/NRF2-knockdown HCT116 cells showed reductions in MitoRed intensity and basal OCR (Fig. 2H, I).

OPEN IN VIEWER

FIG. 2.

Reductions in MMP, mitochondrial OCR, and ATP contents in NFE2L2/NRF2-knockdown colon cancer cells. (A) Enzymatic activity of cytochrome c oxidase (complex IV) in cell lysates from sc and shNRF2-HT cells. Quantification results were from three independent experiments. (B) MMP levels were monitored in sc and shNRF2-HT cells with MitoRed staining. Images were visualized by using a confocal microscope (1890 × magnification). Values are mean ± SEs from three to four independent experiments. (C) Time-dependent cellular MitoRed accumulation was monitored by using a Cell Insight Image analyzer (10, 20, and 40 min). As a negative control, cells were treated with an MMP disruptor CCCP (10 μM) for 6 h. Quantification results were from 11 independent assays. (D) MMP levels were assessed by using TMRM dye (1890 × magnification; n = 3 independent experiments). (E) OCR was determined in scHT and shNRF2-HT cells by using a Seahorse analyzer with the addition of oligomycin, FCCP, and rotenone & antimycin A. A representative measurement was shown, and quantified basal OCR and spare respiratory capacity OCR were from four independent experiments. (F) ATP levels in shNRF2-HT cells. Values are mean ± SEs from four independent experiments. (G) Enzymatic activity of NADH-ubiquinone oxidoreductase (complex I) in isolated mitochondria from sc and shNRF2-HT cells. (H) MMP levels in NFE2L2/NRF2-knockdown HCT116 cells were determined by MitoRed staining (100 × magnification). Quantification results were obtained from six independent experiments. (I) Basal levels of OCR were determined in scHCT and shNRF2-HCT cells by using a Seahorse analyzer. Quantification results were from four independent experiments. ^a P < 0.05 compared with scRNA-expressing control cell line (A–H). (J) scHT and shNRF2-HT cells were incubated with normal glucose or glucose-free medium (Glu(-)) for 48 h, and viable cell numbers were determined by using an automated cell counter. Data represent ratios with respect to the normal glucose scHT control. (K) scHCT and shNRF2-HCT cells were incubated with normal glucose or glucose-free medium (Glu(-)), and viable cell numbers were determined. Values are mean ± SEs from three independent experiments. ^a P < 0.05 compared with each vehicle group. ^b P < 0.05 compared with glucose-free medium incubated scRNA control cells (J, K). ATP, adenosine triphosphate; MMP, mitochondrial membrane potential; OCR, oxygen consumption rate.

Based on diminished ATP production, NFE2L2/NRF2-silenced cancer cells may be dependent on the nutrient supply. Indeed, when shNRF2-HT and shNRF2-HCT cells were cultured in glucose-free medium, the viable cell number was significantly lower than that in the sc control (Fig. 2J, K), whereas no differences were observed in cells grown in normal glucose medium. These results demonstrate that NFE2L2/NRF2-silenced colon cancer cells retain mitochondria with ETC dysfunction, thereby producing lower levels of ATP, which causes vulnerability to glucose limitation.

Involvement of miR-181c in MT-CO1 reduction and MMP disruption

High ROS level is one of the factors affecting MMP (18). As shNRF2-HT cells retain high levels of ROS (Fig. 3A), reduced MMP level might be associated with the ROS increase. However, 1 mM NAC (24 and 72 h) incubation could not restore the repressed MMP levels in shNRF2-HT cells (Fig. 3B and Supplementary Fig. S4), despite decreasing ROS levels with NAC treatment (Fig. 3A). In addition, the incubation of shNRF2-HT with mitochondrial ROS inhibitor (Mito-TEMPO) completely diminished mitochondrial superoxide levels; however, MMP level was not affected by MitoTEMPO (Fig. 3C, D). These data suggest that ROS may not be a major factor contributing to MMP disturbance in NFE2L2/NRF2-knockdown cells.

OPEN IN VIEWER

FIG. 3.

Involvement of miR-181c in MT-CO1 expression. (A) Total ROS levels were monitored with DCF-DA staining. scHT and shNRF2-HT cells were incubated with NAC (1 mM) for 24 h, and DCF-DA staining was performed (100 × magnification). (B) Cells were incubated with NAC (1 mM) for 24 h, and MMP levels were monitored after MitoRed staining (100 × magnification). Data represent ratios with respect to vehicle-treated scHT and are expressed as mean ± SEs of six experiments. (C) Superoxide levels were determined by using MitoSOX staining. sc-HT and shNRF2-HT cells were incubated with MitoTEMPO (20 nM) for 24 h, and fluorescent images were visualized by using a confocal microscope after the staining with MitoSOX and H342 (1000 × magnification). (D) MMP levels were assessed by using MitoRed staining (100 × magnification) with a Cell Insight analyzer. MitoTEMPO (20 nM) was incubated for 24 h as a mitochondria-specific ROS inhibitor. ^a P < 0.05 compared with vehicle group of scRNA-expressing control cell line. ^b P < 0.05 compared with vehicle group of NRF2-silenced cell line (A–D). (E) MT-CO1 expression was determined by Western blotting after transfection of scHT cells with miR-181c mimic or nonspecific NC. Data are from four independent experiments. (F) MT-CO1 levels in miR-181c mimic-transfected scHT were monitored by immunocytochemical analysis (1000 × magnification). (G) MMP levels in the miR-181c mimic group were quantified after incubation with MitoRed. Data are mean ± SEs of six experiments. ^a P < 0.05 compared with NC-transfected scHT cells (E–G). Unedited blots of Figure 3E are shown in Supplementary Blots (Fig. SW3E). NC, negative control; ROS, reactive oxygen species.

Next, to evaluate the direct relationship between miR-181c and MT-CO1/MMP, we transfected scHT cells with miR-181c mimic and examined MT-CO1 protein levels. After transfection, MT-CO1 protein expression was reduced by 20% compared with that in negative control nucleotide-transfected cells (Fig. 3E). These results were confirmed by immunocytochemical analyses (Fig. 3F). In addition, the MitoRed intensity decreased by 40% after miR-181c mimic transfection (Fig. 3G), indicating that miR-181c disrupted the MMP.

Then, shNRF2-HT cells were transfected with miR-181c inhibitor to confirm the role of miR-181c in NFE2L2/NRF2-silencing-induced MMP reduction (Fig. 4A). Western blot and immunocytochemical analyses showed that MT-CO1 levels in shNFE2L2/NRF2 cells were recovered to the levels in control cells after miR-181c inhibition (Fig. 4B, C). Similarly, MMP levels in knockdown cells were enhanced by miR-181c inhibition (Fig. 4D), supporting the regulation of MT-CO1 and MMP by miR-181c in NFE2L2/NRF2-knockdown cells.

OPEN IN VIEWER

FIG. 4.

miR-181c as a mediator for MT-CO1 regulation in NFE2L2/NRF2-knockdown cells. (A) shNRF2-HT cells were transfected with miR-181c inhibitor (inhibitor) or nonspecific NC, and miR-181c levels were assessed by real-time PCR. (B) After transfection of shNRF2-HT cells with miR-181c inhibitor or negative control, MT-CO1 protein expression was assessed by Western blotting. Data are from four experiments. (C) Immunocytochemical analysis of MT-CO1 was performed after transfection of shNRF2-HT with miR-181c inhibitor. (D) MMP levels in miR-181c inhibitor-transfected cells were quantified after incubation with MitoRed. Data are mean ± SEs of eight experiments. Values are ratios with respect to the negative control-transfected scHT29. ^a P < 0.05 compared with NC-transfected scHT cells. ^b P < 0.05 compared with NC-transfected shNRF2-HT cells. Unedited blots of Figure 4B are shown in Supplementary Blots (Fig. SW4B).

In the human miR-181 family, miR-181a-5p and miR-181c share common core sequences for binding to the 3′ untranslated region of the MT-CO1 gene (Supplementary Fig. S5A). Microarray analyses showed that miR-181a-5p levels were 1.45- and 1.90-fold higher in NFE2L2/NRF2-knockdown HT29 and HCT116 cells, respectively (data not shown), and real-time RT-PCR confirmed that miR181a-5p levels were higher in shNRF2-HT cells (Supplementary Fig. S5B). However, MT-CO1 expression was not repressed after transfection with miR-181a mimic (miR-181a-1 or miR-181a-2; Supplementary Fig. S5C), indicating that miR-181a was not involved in MT-CO1 regulation.

Association of nuclear factor-κB signaling with the miR-181c increase

Next, we investigated the molecular mechanism underlying miR-181c upregulation. First, in KEAP1-knockdown HT29 cells, which exhibit constitutive NFE2L2/NRF2 activation, MT-CO1 level was not altered (Supplementary Fig. S6 and Fig. 5A), excluding the direct association of NFE2L2/NRF2 with miR-181c regulation. In line with the effect of ROS on MMP (Fig. 3B, D), treatment of cells with H₂O₂ (400 μM, 6 h) or NAC (1 mM, 24 h) did not affect miR-181c levels (Fig. 5B, C).

OPEN IN VIEWER

FIG. 5.

Involvement of NF-κB signaling in miR-181c/MT-CO1 regulation. (A) Level of MT-CO1 was measured in sc-HT and shKEAP1-HT cells. (B) sc-HT cells were incubated with H₂O₂ (400 μM) for 6 h, and levels of miR-181c were quantified. (C) shNRF2-HT cells were incubated with NAC (1 mM) for 6 h, and miR-181c levels were determined. Quantification results were obtained from three independent experiments. (D) NF-κB-driven luciferase reporter activities were monitored in scHT and shNRF2-HT cells. ^a P < 0.05 compared with scHT cells. (E) Levels of p50 and p65 were determined in scHT and shNRF2-HT cells. (F) p50 and p65 subunits were overexpressed in scHT cells, and miR-181c levels were assessed. Quantification results were from four independent experiments. ^a P < 0.05 compared with pCMV4 control plasmid-transfected scHT cells. (G) p50-specific shRNA was introduced into shNRF2-HT cells, and levels of MT-CO1, p50 and NQO1 were measured. Unedited blots of Figure 5A, E, F, and G are shown in Supplementary Blots (Fig. SW5A, SW5E, SW5F, and SW5G). NF-κB, nuclear factor-κB.

Several reports have indicated that miR-181c is associated with inflammatory diseases (21, 55). Therefore, we hypothesized that shNRF2-HT cells may retain altered nuclear factor (NF)-κB activity, which could be related to the upregulation of miR-181c. Indeed, NF-κB reporter activity was higher in shNRF2-HT cells (Fig. 5D), and p65 and p50 proteins were upregulated in these cells (Fig. 5E). Crucially, forced expression of p65 and p50 in sc-HT elevated miR-181c levels (Fig. 5F), and p50-silencing in shNRF2-HT enhanced MT-CO1 levels to a level comparable to that of the sc control (Fig. 5G). These results suggest that NFE2L2/NRF2 knockdown-induced NF-κB activation is associated with miR-181c/MT-CO1 repression.

Activation of AMPKα signaling in NFE2L2/NRF2-silenced cancer cells through miR-181c

Since the proliferation of colon cancer cells was not affected by NFE2L2/NRF2-silencing (Fig. 2J, K), cellular adaptation to ETC dysfunction and ATP reduction can be hypothesized. Particularly, selective inhibition of cancer NFE2L2/NRF2 has been recognized as a promising strategy to control drug resistance (52) and therefore, elucidation of the molecular basis of cancer adaptation is crucial to develop this anticancer strategy. As a hypothetical adaptive pathway, AMPKα signaling can be activated in response to lower cellular ATP contents. Indeed, shNRF2-HT cells exhibited higher p-AMPKα/AMPKα levels compared with that in scHT cells (Fig. 6A). As a consequence of AMPK activation, NFE2L2/NRF2-knockdown cells exhibited changes in metabolic proteins. First, p-ACC/ACC ratio was increased in these cells, indicating the inhibition of fatty acid synthesis and the promotion of fatty acid oxidation (Fig. 6B). Moreover, p-HMGCR levels were higher in knockdown cells, suggesting repression of the sterol biosynthetic pathway. Second, as the glucose metabolism pathway, marginal increases were observed in GLUT1, GLUT3, GLUT4, (Tre-2/USP6, BUB2, Cdc16) domain-1 (TBC1D1), and 6-phosphofructo-2-kinase/fructose-26-biphosphatase 3 (PFKFB3) transcripts in knockdown cells (Fig. 6C). In accord with these, GLUT4 protein expression was higher in shNRF2-HT cells (Fig. 6D), implying enhanced glucose utilization.

OPEN IN VIEWER

FIG. 6.

Activation of AMPKα signaling in NFE2L2/NRF2-knockdown cancer cells. (A) Levels of p-AMPKα (Thr172) and AMPKα were measured in sc and shNRF2-HT cells. (B) Levels of p-ACC (Ser79), ACC, and p-HMGCR were determined in scHT and shNRF2-HT cells. (C) GLUT1, GLUT3, GLUT4, TBC1D1, and PFKFB3 transcripts were monitored in scHT and shNRF2-HT cells. (D) GLUT4 protein expression was assessed by Western blotting. Unedited blots of Figure 6A, B, and D are shown in Supplementary Blots (Fig. SW6A, SW6B, and SW6D).

To confirm whether miR-181c is a molecular mediator for AMPK activation in NFE2L2/NRF2-silenced cancer, effects of miR-181c mimic or inhibitor treatment on AMPK have been examined. The introduction of miR-181c mimic into scHT cells increased AMPKα and ACC phosphorylation (Fig. 7A); conversely, miR-181c inhibitor transfection in shNFE2L2/NRF2-HT29 cells reduced AMPKα and ACC phosphorylation (Fig. 7B). Similarly, AMPKα activation was also observed in NFE2L2/NRF2-knockdown HCT116 cells (Fig. 7C), and this activation was blocked by miR-181c inhibitor (Fig. 7D). All together, these results showed that AMPKα signaling is stimulated in NFE2L2/NRF2-silenced cancer cells via miR-181c/MT-CO1 axis.

OPEN IN VIEWER

FIG. 7.

Involvement of miR-181c in AMPK activation. (A) Effects of miR-181c mimic on AMPK, p-AMPKα, ACC, and p-ACC in scHT cells. ^a P < 0.05 compared with NC-transfected scHT cells. (B) Effects of miR-181c inhibitor on p-AMPKα and p-ACC in NFE2L2/NRF2 knockdown HT29 cells. Data are from three independent experiments. ^a P < 0.05 compared with NC-transfected shNRF2-HT cells. (C) Levels of AMPKα and p-AMPKα were assessed in NFE2L2/NRF2-knockdown HCT116 (shNRF2-HCT) cells. (D) Effects of miR-181c mimic or inhibitor on AMPKα and p-AMPKα in scHCT or shNRF2-HCT cells, respectively. Data are from three independent experiments. Unedited blots of Figure 7A–D are shown in Supplementary Blots (Fig. SW7A, SW7B, SW7C, and SW7D).

Sensitization of NFE2L2/NRF2-silenced cells to AMPK inhibition

The physiological implication of activated AMPKα in NFE2L2/NRF2-silenced cells could be supported by the effects of the potent AMPK inhibitor compound C. After treatment with compound C (5 μM), p-ACC levels were completely abolished in both scHT and shNRF2-HT cells, indicating the effective inhibition of AMPKα by compound C (Supplementary Fig. S7A). In this condition, the viable cell number was significantly lower in shNRF2-HT and shNRF2-HCT cells than in the control cells (Fig. 8A, B). As recent findings indicated nonspecific effects of compound C other than AMPK inhibition (3, 35), genetic inhibition of AMPKα was adopted as well. The administration of AMPKα_1/2-specific siRNA repressed viable cell numbers in NFE2L2/NRF2 knockdown HT29 and HCT116 cells but not in all control cells (Fig. 8C, D). These results suggest that NFE2L2/NRF2-knockdown cells were more dependent on AMPK signaling to promote survival than control cells.

OPEN IN VIEWER

FIG. 8.

Sensitization of cancer cells to AMPK inhibitor treatment by NFE2L2/NRF2 knockdown. (A) scHT and shNRF2-HT cells were incubated with CC (5 μM) for 48 h, and relative cell numbers were quantified. (B) Viable cell numbers were quantified in scHCT and shNRF2-HCT cells after compound C incubation for 48 h. ^a P < 0.05 compared with each vehicle group of cancer cell line. ^b P < 0.05 compared with CC-treated scRNA-expressing cell line (A–B). Data represent ratios with respect to the vehicle-treated sc control. (C) scHT and shNRF2-HT cells were transfected with a mixture of AMPKα₁- and AMPKα₂-specific siRNAs (200 nM) or control-siRNA (siC) for 48 h. Relative cell numbers were quantified. (D) Viable cell numbers were quantified in scHCT and shNRF2-HCT cells after transfection with a mixture of AMPKα₁- and AMPKα₂-specific siRNAs for 48 h. ^a P < 0.05 compared with control siRNA-treated NRF2-knockdown cells. ^b P < 0.05 compared with mixture of AMPKa1- and AMPKa2-siRNAs treated scRNA-expressing cell line (C–D). Data represent ratio with respect to each of the nonspecific control siRNA (siC)-treated cells. CC, compound C.

Altered MT-CO1/AMPKα in NFE2L2/NRF2-silenced tumors and effects of AMPK inhibition

As tumor cells are located within a specific environment with hypoxic stress and nutrient deficiency, our in vitro observation might differ from an in vivo system. To ask this possibility, sc-HT and shNRF2-HT cells were inoculated into nude mice to investigate the role of NFE2L2/NRF2 in MT-CO1/AMPK within the tumor environment. At the same time, the effect of dual inhibition of NFE2L2/NRF2 and AMPK was evaluated by treating mice with compound C. In the scHT group, tumors developed 7 days after inoculation, and the tumor volume increased continuously till 30 days (Fig. 9A). Compound C treatment (5 mg/kg, intraperitoneal injection) was started on day 11 after inoculation and continued till 28 days (Fig. 9A). After termination of the experiment on day 31, we found that the growth of shNRF2-HT-derived tumors, which was assessed by tumor volume and tumor weight, was significantly retarded compared with that of scHT cell-derived tumors (Fig. 9A, B), confirming our previous observations (31). In agreement with the in vitro results, compound C treatment suppressed the growth of NFE2L2/NRF2-knockdown tumors; compound C induced significant reductions in tumor weights in the shNRF2-HT group, but not in the scHT group (Fig. 9B).

OPEN IN VIEWER

FIG. 9.

Sensitization of NFE2L2/NRF2-silenced tumors to AMPK inhibitor treatment. (A) scHT and shNRF2-HT cells were inoculated in 5-week-old BALBc (nu/nu) mice. From day 11, mice were treated with saline or compound C (5 mg/kg) every other day (arrows). Tumor volumes were assessed twice a week (n = 5–6 mice/group). (B) At day 31, mice were terminated and tumor weights were measured. ^a P < 0.05 compared with saline-treated scHT tumor. ^b P < 0.05 compared with saline-treated shNRF2-HT tumor (B). (C) NQO1, MT-CO1, MT-CO3, and VDAC1 protein levels were measured in tumor lysates. For quantification, NQO1 was normalized to β-tubulin, and MT-CO1 was normalized to VDAC1. (D) p-AMPKα, AMPKα, p-ACC, and ACC levels were determined. Ratios of p-AMPKα to total AMPKα (p-AMPKα/AMPKα) and ratios of p-ACC to total ACC (p-ACC/ACC) were quantified. ^a P < 0.05 compared with saline-treated scHT tumor. ^b P < 0.05 compared with CC-treated scHT tumor. ^c P < 0.05 compared with CC-treated shNRF2-HT tumor (C–D). Values are mean ± SEs of three determinations of pooled samples. Unedited blots of Figure 9C and D are shown in Supplementary Blots (Fig. SW9C and SW9D).

Notably, within the tumor environment, the link between NFE2L2/NRF2 and MT-CO1/AMPK axis was retained. MT-CO1 levels were lower in shNRF2-HT cell-derived tumors (Fig. 9C), and the levels of p-AMPKα/AMPKα ratio were significantly higher in shNFE2L2/NRF2 tumors (Fig. 9D). Accordantly, NFE2L2/NRF2-knockdown tumors retained increased p-ACC/ACC ratio, p-HMGCR, and GLUT4 levels compared with control tumors. These results confirmed the association of NFE2L2/NRF2 with MT-CO1/AMPK signaling in an in vivo environment. In addition, obtained data demonstrated the advantageous antitumor effect of dual inhibition of NFE2L2/NRF2 and AMPK.

MT-CO1 repression and AMPK activation in NFE2L2/NRF2-knockdown breast cancer cells

Next, we examined this mechanism in other types of cancer cells. NFE2L2/NRF2-knockdown MCF7 breast cancer cells (shNRF2-MCF) showed decreased levels of MT-CO1 and increased levels of p-AMPKα (Fig. 10A). MMP levels and cellular ATP content were diminished in shNRF2-MCF (Fig. 10B, C). Similar to colon cancer cells, shNRF2-MCF cells were more sensitive to compound C-induced growth inhibition, indicating the beneficial antitumor effect of dual inhibition of NFE2L2/NRF2 and AMPK (Fig. 10D). Similar results were obtained in NFE2L2/NRF2-knockdown MDA-MB-231 breast cancer cells (Fig. 10E–H).

OPEN IN VIEWER

FIG. 10.

Decreases in MT-CO1 and MMP, and increases in p-AMPK in NFE2L2/NRF2-silenced breast cancer cells. (A) MT-CO1, AMPK, and p-AMPKα levels were measured in control MCF7 (scMCF) and NFE2L2/NRF2-knockdown MCF7 (shNRF2-MCF) cells. (B) MMP levels were assessed in scMCF and shNRF2-MCF cells by using TMRM staining (1890 × magnification). (C) Cellular ATP levels were measured in scMCF and shNRF2-MCF cells. Values are mean ± SEs from four independent experiments. (D) scMCF and shNRF2-MCF cells were incubated with compound C (5 μM) for 48 h, and viable cell numbers were determined. Values are mean ± SEs of four experiments. (E) Levels of MT-CO1, AMPK, and p-AMPKα were monitored in NFE2L2/NRF2-silenced MDA-MB-231 cells (shNRF2-MDA). (F) MMP levels were assessed in shNRF2-MDA cells by using MitoRed staining (1890 × magnification). (G) Cellular ATP contents in shNRF2-MDA. (H) scMDA and shNRF2-MDA cells were incubated with compound C (5 μM) for 48 h, and viable cell numbers were determined. Values are mean ± SEs of six experiments. ^a P < 0.05 compared with scRNA-expressing control cell lines (C, G). ^a P < 0.05 compared with each vehicle group. ^b P < 0.05 compared with CC-treated scRNA expressing control cell lines (D, H). Unedited blots of Figure 10A and E are shown in Supplementary Blots (Fig. SW10A and SW10E).

Materials

Carbonyl cyanide 3-chlorophenyl hydrazone, Compound C, d-glucose, l-glutamine, hydrogen peroxide (H₂O₂), N-acetylcysteine (NAC), 2-(2,2,6,6-Tetramethylpiperidin-1-oxyl-4-ylamino)-2-oxoethyl)triphenylphosphonium chloride (MitoTEMPO), propidium iodide (PI), and puromycin were purchased from Sigma-Aldrich Co. (St. Louis, MO). MitoTracker Red CMXRos (MitoRed), tetramethylrhodamine methyl ester (TMRM), MitoSOX Red (MitoSOX), Hoechst 33342 (H342), carboxy-2′,7′-dichlorofluorescein diacetate (DCF-DA), and Alexa Fluor 488- or 549-conjugated secondary antibodies were purchased from Thermo Fisher Scientific, Inc. (Waltham, MA). Anti-AMPKα (No. 5832), anti-phosphorylated AMPKα (p-AMPKα, Thr172; No. 2535), anti-ACC (No. 3662), and anti-phospho-ACC (p-ACC, Ser79; No. 3661) antibodies were obtained from Cell Signaling Technology, Inc. (Danvers, MA). Anti-NFE2L2/NRF2 (sc-722), anti-NAD(P)H quinone oxidoreductase-1 (NQO1; sc-16464), anti-MT-CO2 (sc-23984), anti-MT-CO3 (sc-23986), anti-voltage-dependent anion channel-1 (VDAC1; sc-8828), anti-GLUT4 (sc-7178), p50 (sc-7151) and RelA (p65; sc-7938) of NF-κB, and anti-β-tubulin (sc-9104) antibodies were purchased from Santa Cruz Biotechnology, Inc. (Dallas, TX). MT-CO1 (ab14705) and phospho-HMGCR (p-HMGCR, Ser872; No. 09-356) antibodies were from Abcam (Cambridge, United Kingdom) and EMD Millipore Corporation (Billerica, MA), respectively.

Cell culture

HT29 (No. HTB-38) and HCT116 cells (No. CCL-247) (American Type Culture Collection [ATCC], Manassas, VA) were maintained in RPMI 1640 (GE Healthcare Life Sciences, Logan, UT) and Dulbecco's modified Eagle's medium (DMEM; GE Healthcare Life Sciences), respectively, with 10% fetal bovine serum (FBS; GE Healthcare Life Sciences) and penicillin/streptomycin (WelGene, Inc., Daegu, Republic of Korea). MCF-7 (No. HTB-22) and MDA-MB-231 (No. HTB-26) human breast cancer cells (ATCC) were maintained in DMEM with 10% FBS.

NFE2L2/NRF2-knockdown cancer cells

NFE2L2/NRF2-knockdown cancer cell lines were previously established (9, 31, 45) as shNRF2-HT, shNRF2-HCT, shNRF2-MDA, and shNRF2-MCF cells by using NFE2L2/NRF2-specific shRNA No. 1 (Supplementary Table S1); the corresponding scrambled RNA (scRNA)-transfected control cell lines were scHT, scHCT, scMDA, and scMCF cells, respectively.

Transient transduction with lentiviral particles containing shRNA

For transient silencing of NFE2L2/NRF2 or p50, the HEK 293T cells were transfected with pLKO.1-NFE2L2/NRF2 shRNA (Nos. 2, 3, and 4), pLKO.1-p50 shRNA, or pLKO.1-scrambled RNA (scRNA) with the Mission Lentiviral Packaging Mix (Sigma-Aldrich Co. LLC), and lentiviral particles were obtained for transient transduction. The shRNA sequences are indicated in Supplementary Table S1.

miRNA isolation and real-time RT-polymerase chain reaction

Small RNA-rich samples were isolated from cells with a mirVana miRNA Isolation kit (Thermo Fisher Scientific, Inc.). Briefly, cell lysates were extracted with acid phenol:chloroform and purified over a glass-fiber filter to yield miRNA-enriched fractions. cDNA was synthesized with an miScript RT kit (Qiagen GmbH, Hilden, Germany), and PCR was performed with an miScript SYBR green PCR kit (Qiagen GmbH) and LC480 LightCycler (Roche Diagnostics Deutschland GmbH, Mannheim, Germany). Forward primers were synthesized by Bioneer Corporation (Daejeon, Republic of Korea). The sequences for miR-181a, miR-181c, and U6 snRNA are described in Supplementary Table S2.

Total RNA isolation and real-time RT-PCR

Total RNA was isolated from cells with TRIzol reagent (Thermo Fisher Scientific, Inc.). After cDNA synthesis, PCR was performed with an LC480 LightCycler and a TaKaRa SYBR Premix ExTaq system (TaKaRa Bio, Inc., Shiga, Japan) as previously described (28). Primer sequences are shown in Supplementary Table S2.

MMP measurement

Cellular MMP levels were determined by using MitoRed and TMRM. Briefly, cells on cover glass (SPL Life Sciences Co., Ltd., Gyeonggi-do, Republic of Korea) were incubated with MitoRed (1 μM) or TMRM (0.5 nM) for 20 min at 37°C. After washing, fluorescent images were analyzed immediately with an LSM 710 confocal microscope (Carl Zeiss AG, Jena, Germany). MitoRed- or TMRM-derived fluorescence was excited at 579 or 514 nm, respectively, and emission was detected at 548–685 nm. The intensities were quantified with ZEN2011 software (Carl Zeiss AG). To analyze cellular fluorescent images with a Cell Insight Personal Cell Imager (Thermo Fisher Scientific, Inc.), cells were seeded in 96-well plates and incubated with MitoRed (0.5 nM) for 20 min. The fluorescent images were quantified after nuclear staining with H342 (2 μg/ml, 350/461 nm).

Mitochondrial OCR measurement

Cells (3 × 10⁴/well) were cultured on XF-24 plates (Seahorse Bioscience, Billerica, MA) in a CO₂ incubator at 37°C. After a 24 h incubation, growth medium from each well was exchanged with the pre-warmed pH 7.4 XF-base medium (Seahorse Bioscience) that was supplemented with d-glucose (11 mM) and l-glutamine (2 mM). The cells were incubated in a non-CO₂ incubator for 30 min and then, oligomycin, carbonyl cyanide 4-(tifuloromethoxy)phenylhydrazone (FCCP), and rotenone & antimycin A (XF Cell Mito Stress Test Kit; Seahorse Bioscience) were loaded into the injector ports A, B, and C of the sensor cartridge, respectively. After calibration of the sensor cartridge, basal and post-inhibitor OCR were monitored. Each injection was proceeded at the final concentration as follows: oligomycin (1 μM), FCCP (0.5 μM), and rotenone & antimycin A (0.5 μM). The protein concentration was measured after the assay by using BCA assay (Thermo Fisher Scientific, Inc.) and served as normalization values for each well.

ROS measurements

Total ROS levels were measured by staining with DCF-DA (20 μM) for 30 min at 37°C (46). DCF-DA fluorescence was determined with a Cell Insight Personal Imager (485/521 nm) after nuclear staining with H342. Mitochondrial superoxide levels were determined by using MitoSOX. Briefly, cells on cover glass (SPL Life Sciences Co.) were incubated with MitoSOX (5 μM) for 10 min at 37°C. After washing, the fluorescent images were analyzed immediately with an LSM 710 confocal microscope (Carl Zeiss AG) after nuclear staining with H342. MitoSOX-derived fluorescence was excited at 510 nm, and emission was detected at 580 nm.

Measurement of NADH-ubiquinone oxidoreductase (complex I) activity

Complex I activity was determined by using a colorimetric assay kit (BioVision, CA). In accordance to the manufacturer's introduction, isolated mitochondria samples were added to each well containing complex I assay mixture, which included complex I dye and decylubiquinone. Next, freshly prepared NADH solution was added to each well, and the decrease in optical density (600 nm) was immediately recorded for as long as 5 min with a 30-s interval by using a SPECTRO star^Nano (BMG LABTECH GmbH, Ortenberg, Germany). The complex I activity was calculated in mU/μg as Δ [reduced complex I dye concentration]/[(t₂ − t₁) × mitochondria protein (μg)] × dilution factor, where Δ [reduced complex I dye concentration] is obtained by inserting OD values to a standard curve in the time point.

Measurement of cytochrome c oxidase (complex IV) activity

Complex IV activity was measured by using a cytochrome c oxidase assay kit (BioVision). Briefly, cytochrome c mixture was added to cell lysates, and the decrease in optical density (550 nm) was immediately recorded for 30 min with a 30-s interval by using a SPECTRO star^Nano. The complex IV activity was calculated in U/mg as [(OD₁ − OD₂)/(t₁ − t₂)]/(ɛ × mg protein), where ɛ is the molar extinction coefficient of cytochrome c at 550 nm (7.04 mM ⁻¹cm⁻¹).

ATP content measurement

Cells were incubated in 24-well plates with trichloroacetic acid (0.5%, 200 μl) for 30 min on ice, centrifuged, and neutralized by using 250 mM Tris-acetate (pH 7.8) buffer. ATP levels in the obtained supernatants were determined with ENLITEN ATP assay kits (Promega Corporation) with a luminometer and normalized to the protein content.

Cell number count

Viable cell numbers were determined with a TC10 Automated Cell Counter (Bio-Rad Laboratories, Inc., Hercules, CA) after the addition of trypan blue.

Immunocytochemical analysis

Cells on cover glass were fixed in cold methanol for 10 min and then incubated with the anti-NFE2L2/NRF2 (1:200) or anti-MT-CO1 (1:200) antibody at 4°C overnight. Cells were then incubated with the Alexa Fluor 488 or 549 goat anti-rabbit antibodies (1:500) for 1 h. For nuclear staining, cells were incubated with PI (1:500) or H342 (2 μg/ml) for 10 min. Fluorescent images were obtained with an LSM 710 confocal microscope, and fluorescence intensities were quantified with ZEN2011 software (27).

Western blotting

Cells were lysed as previously described (31), and protein concentrations were determined with bicinchoninic acid kits (Thermo Fisher Scientific, Inc.). Electrophoresed protein samples were transferred to nitrocellulose membranes (GE Healthcare Bio-Sciences, Pittsburgh, PA), and membranes were incubated with primary and secondary antibodies. Chemoluminescent images were obtained with an ImageQuant LAS 4000 Mini (GE Healthcare Bio-Sciences). Mitochondria were isolated with a mitochondria isolation kit (Thermo Fisher Scientific, Inc.). Briefly, cells were lysed after the serial addition of reagents A, B, and C, and supernatants were obtained by centrifugation at 700 g for 10 min. The supernatants were then centrifuged at 12,000 g for 15 min, and mitochondria-containing pellets were prepared.

Luciferase reporter assay for NF-κB and 3′UTR of MT-CO1

The cells were incubated with a mixture of NF-κB-luciferase plasmid (Affymetrix, Inc.), pRLtk control plasmid (Promega Corporation, Madison, WI), and Lipofectamine 2000 for 18 h. After a 24 h recovery, cells were lysed and luciferase activities were measured with a 20/20ⁿ luminometer (Promega Corporation). For, 3′UTR-luciferase activity, the cells were incubated with a mixture of the 3′UTR-MT-CO1-luciferase plasmid (Genecopoeia, Inc., Rockville, MD) and Lipofectamine 2000 for 18 h.

Overexpression of p50 and p65

For NF-κB overexpression, pCMV4-p50 and pCMV4-p65 (Addgene, Cambridge, MA) were transfected in cells by using Lipofectamine 2000.

Transfection with miR-181c mimic, miR-181c inhibitor, or AMPKα siRNA

AccuTarget human miRNA mimics for miR-181c, miR-181a-1, miR-181a-2, the miR-181c inhibitor, and negative controls were obtained from the Bioneer Corporation. To silence AMPKα, siRNAs for AMPKα₁ and AMPKα₂ (Bioneer Corporation) were transfected into cells. Transfections were performed with Lipofectamine 2000.

Tumor xenografts in mice

A suspension of 3 × 10⁶ cells in 0.2 ml serum-free medium was injected subcutaneously into the flanks of 5-week-old BABLc (nu/nu) mice (Orient Bio, Inc., Gyeonggi-do, Republic of Korea). Mice were maintained in a specific pathogen-free environment, and the animal experiment was approved by the Institutional Ethics Committee on Animal Care and Experimentation at The Catholic University of Korea (Approval No. 2014-032-01). When tumor formation was observed, compound C (5 mg/kg) or saline was injected intraperitoneally into the mice 2–3 times a week for a total of 10 times. The tumor growth was monitored twice a week by measuring two diameters of tumors with calipers. The tumor volume was calculated by the formula V = (a ² × b)/2, where a and b are the width and the length of tumors in mm, respectively. Thirty-one days after inoculation, the mice were euthanized with carbon dioxide (CO₂) in an appropriate chamber, and the collected tumors were weighed and frozen. Tumor tissues in the lysis buffer were homogenized on ice with an IKA homogenizer (IKA-Werke GmbH & Co. KG, Staufen, Germany). The homogenates were centrifuged twice at 12,000 g for 20 min at 4°C, and supernatants were obtained.

Statistical analysis

Statistical significance was determined with Student's unpaired t-tests or one-way analyses of variance, followed by Student-Newman-Keuls tests for multiple comparisons. These analyses were conducted with GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA).