Nrf1 is an indispensable redox-determining factor for mitochondrial homeostasis by integrating multi-hierarchical regulatory networks

Abstract

To defend against a vast variety of challenges in oxygenated environments, all life forms have evolutionally established a set of antioxidants, detoxification, and cytoprotective systems during natural selection and adaptive survival, to maintain cell redox homeostasis and organ integrity in the healthy development and growth. Such antioxidant defense systems are predominantly regulated by two key transcription factors Nrf1 and Nrf2, but the underlying mechanism(s) for their coordinated redox control remains elusive. Here, we found that loss of full-length Nrf1 led to a dramatic increase in reactive oxygen species (ROS) and oxidative damages in Nrf1α^–∕– cells, and this increase was not eliminated by drastic elevation of Nrf2, even though the antioxidant systems were also substantially enhanced by hyperactive Nrf2. Further studies revealed that the increased ROS production in Nrf1α^–∕– resulted from a striking impairment in the mitochondrial oxidative respiratory chain and its gene expression regulated by nuclear respiratory factors, called αPal^NRF1 and GABP^NRF2. In addition to the antioxidant capacity of cells, glycolysis was greatly augmented by aberrantly-elevated Nrf2, so to partially relieve the cellular energy demands, but aggravate its mitochondrial stress. The generation of ROS was also differentially regulated by Nrf1 and Nrf2 through miR-195 and/or mIR-497-mediated UCP2 pathway. Consequently, the epithelial-mesenchymal transformation (EMT) of Nrf1α^–∕– cells was activated by putative ROS-stimulated signaling via MAPK, HIF1α, NF-ƙB, PI3K and AKT, all players involved in cancer development and progression. Taken together, it is inferable that Nrf1 acts as a potent integrator of redox regulation by multi-hierarchical networks.

Graphical abstract

Highlights

• •Knockout of Nrf1α rather than Nrf2 leads to severe oxidative stress.
• •Nrf1α is indispensable for maintaining mitochondrial homeostasis.
• •Glycolysis enhanced by hyper-active Nrf2 promotes ROS yield in Nrf1α-deficient cells.
• •Negative regulation of UCP2 by Nrf1 and especially Nrf2 actually elevates ROS levels.
• •Adaptive activation of mitochondrial stress by FCCP is blocked upon loss of Nrf1α.

1. Introduction

Under the oxygenated environments, almost all cellular life forms have been successively generating, transforming and eliminating those reactive oxygen species (ROS, also including reactive nitrogen and sulphur species) in a variety of cell processes (e.g., metabolism, proliferation, differentiation), immune regulation and vascular remodeling [1]. Approximately 90% of ROS in eukaryotic cells are generated from the mitochondria, which are hence accepted as a major source of intracellularly-producing ROS. Approximately 2% of oxygens are consumed by mitochondria to produce ROS, and this number could be incremented when their electron transport chains (ETCs) are damaged and/or uncoupled with oxidative phosphorylation (OXPHOS, for ATP production) [2]. Besides, NADPH oxidases, cytochrome p450 isoenzymes and the other oxidases existing particularly in the peroxisome and endoplasmic reticulum (ER) are considered as the secondary sources of ROS production [3]. ROS are endowed with a cytotoxic capability to beget oxidative damage and even lead to cell death if excessive ROS is yielded. To defend against potential challenges from oxidative stress, all life forms have also evolutionally established a set of antioxidant, detoxification and cytoprotective systems to maintain cellular redox homeostasis and organ integrity during their adaptive survival. Such a homeodynamic balance between oxidative stress and antioxidant defense system has been established and also perpetuated at a steady state of redox, just at which all living organisms are allowed for healthy survival under such robust redox homeostasis.

Collectively, the levels of ROS (albeit with higher cytotoxicity) under the normal conditions are finely tuned and also restricted to considerably lower extents, such that they can also serve as potent eustress that triggers certain hormestic effects on the physiology of living organisms insomuch as to preserve the healthy life processes [4,5]. Thereby, a redox threshold is set at a certain steady-state of physiology. Once ROS are produced to exceed the preformed redox threshold, they exert oxidative distress to stimulate a vast variety of pathophysiological responses and even pathological effects to different extents [6]. The proper ROS levels are required to induce many biological activities, including development, growth, cell division, proliferation, differentiation, survival and apoptosis, as well as immune responses [1,3]. Rather, excessive ROS has manifested with some pathogenic relevance to many human diseases, such as neurodegenerative, cardiovascular, inflammatory and autoimmune diseases, diabetes, aging and cancer [7].

In mammalians, Nrf1 and Nrf2 are two principal regulators of the intracellular redox homeostasis by coordinately governing transcriptional expression of distinct subsets of cognate genes through their functional heterodimers with a partner of small Maf (sMaf) or other bZIP proteins (e.g., AP-1 and ATF4), that are recruited for directly binding to those consensus sites, called antioxidant and/or electrophile response elements (AREs/EpREs), within their target promoter regions [8,9]. Both Nrf1 and Nrf2 are differentially activated in distinctive tempo-spatial responses to different extents of ROS-led oxidative stress (Figs. 1A and S1A). Under normal conditions, Nrf1 is anchored to the ER and most degraded via the ubiquitin-mediated proteasome pathway. When Nrf1 is required for biological responses to stimulation, this CNC-bZIP protein is allowed for dynamic dislocation from the ER lumen across membranes into extra-ER compartments, in which it is subjected to its selective proteolytic cleavage by DDI1/2 or other cytosolic proteases to become a mature transcription factor, before entering the nucleus and transactivating target genes [10,11]. By contrast, most of Nrf2 is inactive under normal conditions, because it is segregated in the cytoplasm by physic interaction with Keap1, a redox-sensitive adaptor for Cullin3, to target the second CNC-bZIP protein to the ubiquitin-mediated proteasomal degradation [12]. Once Keep is activated by increased ROS, Nrf2 is allowed for disassociation from its interactor Keap1 to enter the nucleus, where it enables to heterodimerize with sMaf and transactivate target genes.

Fig. 1.

Distinctions in the redox status and expression of antioxidant genes in WT, Nrf1α^–∕– and Nrf2^−/− cell lines.

(A) A schematic representation of signaling mechanisms that Nrf1 and Nrf2 are activated by ROS to play cytoprotective roles.

(B) Fluorescence images representing the ROS levels in WT, Nrf1α^–∕– and Nrf2^−/− cells, which were stained by DCFH-DA (excited in 488 nm). The fluorescence intensity was statistically calculated as fold changes (mean ± SEM, n = 3 × 3, $, p < 0.05 and $$, p < 0.01).

(C, D, E) Distinctive ROS levels were detected by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cell lines that had been stained by DCFH-DA (C), HPF (D), or MitoSOX (E), respectively.

(F) Different GSH levels were determined by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cell lines that had all been stained by monochlorobimane (mBCI. Ex/Em = 394/490 nm).

(G) Relative malondialdehyde (MDA) levels were measured in WT, Nrf1α^–∕– and Nrf2^−/− cells. The data are shown as fold changes (mean ± SEM, n = 3 × 3; $, p < 0.05 and $$, p < 0.01).

(H) The mRNA levels of those antioxidant genes CAT, PRDX1, PRDX3, PRDX4, PRDX5, PRDX6, GPX1, GPX2, SOD1, and SOD2 were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The data are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(I) The mRNA levels of GSH metabolism genes GCLC, GCLM, GLS1, GSR, SLC6A9, and SLA7C11 were examined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The data are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(J) The mRNA levels of TXN1, TXN2, TXNRD1, TXNRD2 and SRXN1 determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(K) The mRNA levels of G6PD, IDH1, IDH2, ME1, PGD and TIGAR determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(L) The mRNA levels of FTH1, FTL and HO1 determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(M) The mRNA levels of SESN1, SESN2, and SESN3 were determined by qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The resulting data are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(N) Distinct protein levels of Nrf1, Nrf2, CAT, PRDX1, GPX1, SOD1, SOD2, TXN1, TXN2, G6PD, IDH1, IDH2, GSR, GCLM, and HO1 in WT, Nrf1α^–∕– and Nrf2^−/− cells were visualized by Western blotting. The intensity of all immunoblots was also calculated as shown in Fig. S3.

Those antioxidant genes mediated by Nrf1 and/or Nrf2 are mainly involved in: i) glutathione production and regeneration, which are regulated by glutamate-cysteine ligase modifier (GCLM) and catalytic subunits (GCLC), glutathione reductase (GSR) [13]; ii) glutathione utilization, which is conducted by glutathione S-transferases (GSTs) and glutathione peroxidase 2 (GPXs); iii) thioredoxin production and utilization (e.g., TXNs, PRDXs); iv) NADPH production, controlled by glucose-6-phosphate dehydrogenase (G6PD), malic enzyme 1 (ME1) and isocitrate dehydrogenase 1 (IDH1) [14]; and v) the other responsible genes for antioxidant and detoxification, e.g., both encoding NAD(P)H:quinone oxidoreductase 1 (NQO1) [15] and heme oxygenase 1 (HO1) [16]. Furthermore, Nrf1 acts as a vital player in maintaining protein homeostasis (proteostasis), lipid metabolism, glucose metabolism, inflammation, cell differentiation, and embryonic development by controlling transcriptional expression of related target genes [9,[17], [18], [19], [20], [21], [22], [23]]. Of stinking note, Nrf2 can also exert a protective effect on normal cells against chemical carcinogens, but rather promotes cancer development and progression, particularly when it is hyper-activated for a long term [24,25]. By contrast, Nrf1α (and long isoform TCF11) is identified as a potent tumor-repressor to inhibit cancer malgrowth in xenograft model mice [26]. This is supported by liver-specific knockout of Nrf1 in mice with obvious pathological phenotypes, resembling human non-alcoholic steatohepatitis (NASH) and hepatoma [27,28]. Our previous evidence also revealed that knockout of Nrf1 from human hepatoma cells results in the exacerbation of tumor growth and migration, but such malignant growth appears to be strikingly prevented upon overexpression of Nrf1α or TCF11 [26,29].

In this study, we found that knockout of Nrf1α (a major full-length isoform of prototypic Nrf1) caused a substantial increase of the intracellular ROS determined in Nrf1α^–∕– cells, when compared with wild-type (WT) and Nrf2^–∕–cells. Such increased ROS levels were accompanied by up-regulation of most genes responsible for antioxidant and detoxification; this is attributable to aberrant accumulation of Nrf2 upon loss of Nrf1α. Interestingly, abnormal changes in the mitochondrial morphology of Nrf1α^–∕– rather than Nrf2^–∕– cells were observed, as accompanied by impaired ETC complexes. Further examinations also unraveled that Nrf1α^–∕–-derived mitochondrial dysfunction was owing to deregulation of two nuclear respiratory factors αPal^NRF1 (nuclear respiratory factor 1) and GABPα^NRF2 (nuclear respiratory factor 2), along with the co-factor PGC1α, resulting in strikingly increased ROS levels in Nrf1α-deficient cells. Besides, glycolysis of Nrf1α^–∕– cells was greatly augmented by hyperactive Nrf2, leading to an aggravated pressure of mitochondrial stress. The mitochondrial UCP2 was markedly suppressed by Nrf2 through AREs-driven miR-195 and miR-497. These collectively contribute to a drastic increase of ROS in Nrf1α^–∕– cells; this is concomitant with the activation of multiple signaling pathways via MAPK, PI3K-AKT, HIF1α, and NF-κB to the EMT process involved in cancer development and progression. Overall, Nrf1 acts as a vital player in determining and maintaining robust redox homeostasis by governing mitochondrial homeostasis integrated from multi-hierarchical signaling pathways towards cognate gene regulatory networks. Therefore, this work provides a novel understanding of the distinct roles of Nrf1 and Nrf2 in cancer development and prevention.

2. Materials and methods

2.1. Cell lines, culture and transfection

Nrf1α^−/− cells were established by TALENs-led genome editing with HepG2 cells [26], whereas Nrf2^−/− cells were constructed by CRISPR/Cas9-editing system with HepG2 cells [29]. These cell lines expressing Nrf1α and Nrf2, as well as an empty control, were established by using the Flp-In™ TREx™-293 system with HepG2 cells (Invitrogen) [30]. These cell lines were maintained in DMEM supplemented with 5 mM glutamine, 10% (v/v) fetal bovine serum (FBS), and 100 units/mL of either penicillin or streptomycin, in the 37 °C incubators with 5% CO₂. In addition, some cell lines were transfected for 8 h with the indicated constructs mixed with the Lipofectamine®3000 agent in the Opti-MEM (gibca, Waltham, MA, USA). The cells were then allowed for recovery from transfection in a fresh complete medium for 24 h before the other experiments were conducted.

2.2. Plasmid construction

The expression constructs for human Nrf1 and Nrf2 were created by inserting their full-length cDNA sequences into the KpnI/XbaI site of pcDNA3.1/V5His B. The six reporter gene plasmids for PGC1α-luc, αPal^NRF1-luc, GABPα-luc, UCP2-luc, miR-195-luc and miR-497-luc were created by inserting their promoter sequences into the PGL3-basic vector, distinct lengths of each gene promoter are indicated in Fig. 3, Fig. 5H. In addition, the ARE-luc plasmids were created by inserting the consensus ARE-adjoining sequences (Fig. 3, Fig. 5J) from indicated gene promoters into the PGL3-promoter vector. The 3’-UTR of UCP2 was cloned into the psiCHECK-2 plasmid.

Fig. 3.

Distinct effects of Nrf1 and Nrf2 on nuclearly-controlled mitochondrial factors PGC1α, PGC1β, αPal^NRF1, GABPα

(A) The mRNA expression levels of PGC1α, PGC1β, αPal^NRF1, GABPα, TFAM, TFB1M, and TFB2M were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The data are shown as fold changes (mean ± SEM, n = 3 × 3; *, p < 0.05 and **, p < 0.01).

(B) The protein levels of PGC1α, PGC1β, αPal^NRF1, GABPα, and TFAM in WT, Nrf1α^–∕– and Nrf2^−/− cells were visualized by Western blotting. The intensity of all the immunoblots was calculated and shown in Fig. S6A.

(C) The mRNA levels of Nrf1 and Nrf2 were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated with tBHQ for 0 h, 12 h, or 24 h. The data are shown as mean ± SEM (n = 3 × 3) with significant increases ($, p < 0.05 and $$, p < 0.01) as compared to the cells without this chemical treatment.

(D) The mRNA levels of αpal^NRF1, GABPα, PGC1α, and TFAM were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated with tBHQ for 0 h, 12 h, or 24 h. The resulting data are shown above.

(E) Schematic representation of the PGC1α-luc, αPal^NRF1-luc, and GABPα-luc, which were constructed into the PGL3-Promoter plasmid. Their promoter regions were also indicated.

(F) Relative luciferase activity of PGC1α-luc, αPal^NRF1-luc, and GABPα-luc were determined in COS-1 cells that had been co-expressed with each reporter gene and pRL-TK (as an internal reference), plus an expression construct for Nrf1 or Nrf2, or empty pcDNA3 vector. The data are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01).

(G) The consensus ARE sites within the promoter of PGC1α, αPal^NRF1, and GABPα were listed herein.

(H) Fold ARE-luc activity mediated by Nrf1 or Nrf2 was determined. The indicated ARE-adjoining sequences were cloned into the PGL3-promoter vector, and were co-expressed with pRL-TK, plus expression constructs for Nrf1 or Nrf2, or empty pcDNA3 plasmid, then operated as above.

(I) A proposed model that Nrf1 and Nrf2 regulate mitochondrial function by αpal^NRF1, GABPα, and PGC1α.

Fig. 5.

Distinct effects of Nrf1 and Nrf2 on ROS production by inhibiting UCP2 via miRNA-195 and miRNA-497

(A) The mRNA (upper column) and protein (lower panel) levels of UCP2 were determined by RT-qPCR and Western blotting of WT, Nrf1α^–∕– and Nrf2^−/− cells. The qPCR data are shown as mean ± SEM (n = 3 × 3; $$, p < 0.01; **, p < 0.01). The intensity of all the immunoblots was calculated as shown on the bottom.

(B) WT, Nrf1α^–∕– and Nrf2^−/− cell lines were allowed for the forced expression of UCP2 or empty vector, and then for 24-h recovery from transfection, before ROS levels of cells are detected by flow cytometry.

(C) Relative mRNA level of miR-15a-5P, miR-15b-5P, miR-16b-5P, miR-24-3P, miR-195-5P, miR-424-5P, miR-491-5P, and miR-497-5P in WT, Nrf1α^–∕– and Nrf2^−/− cells were determined by RT-qPCR. The data are shown as mean ± SEM (n = 3 × 3; $$, p < 0.01; *, p < 0.05).

(D) Schematic representation of miR-195 and miR-497 targeting the 3’-UTR of UCP2.

(E) WT cells were transfected with miR-24, miR-195 or miR-497 mimics in indicated concentration, and then allowed for 24-h recovery from transfection, because the mRNA levels of UCP2 were determined by RT-qPCR. The data are shown as fold changes (mean ± SEM, n = 3 × 3; **, p < 0.01).

(F) After Nrf1α^–∕– cells were transfected with the inhibitors of miR-195 and miR-497, the mRNA levels of UCP2 were determined by RT-qPCR. The data are shown as mean ± SEM (n = 3 × 3; $$, p < 0.01).

(G) The 3’-UTR of UCP2 were cloned into a psiCHECK-2 plasmid, and co-transfected with miR-195, miR-497, or negative controls. After 24-h recovery from transfection, the miRNA activity of binding to 3’-UTR of UCP2 was assayed by indicated Ranilla luciferase reporters, luciferase gene in psiCHECK-2 plasmid was used as an internal reference. The data are shown as fold changes (mean ± SEM, n = 3 × 3; **, p < 0.01).

(H) Schematic representation of the UCP2-luc, miR-195-luc, and miR-497-luc reporters, which were constructed into the PGL3-Promoter plasmid. These gene promoter regions were also indicated.

(I) Relative activity of UCP2-luc was determined in the cells that were co-expressed with each of indicated luciferase reporters, and pRL-TK, plus an expression construct for Nrf1 or Nrf2, or empty pcDNA3 plasmid and then allowed for 24-h recovery from transfection. The resulting data are shown as mean ± SEM (n = 3 × 3).

(J) Relative activity of miR-195-luc and miR-497-luc were determined as described above (mean ± SEM, n = 3 × 3; $$, p < 0.01).

(K) Fold activity of ARE-luc activated by Nrf1 or Nrf2 was determined. These ARE-adjoining sequences listed in (L) were cloned into PGL3-promoter plasmid and were co-transfected with pRL-TK, plus an expression construct for Nrf1 or Nrf2, or empty pcDNA3 plasmid.

(L) Distinct locations of ARE sites from the promoters of miR-195 and miR-497 were listed herein.

(M) A proposed model to explain the distinct effects of Nrf1 and Nrf2 on ROS production by inhibiting UCP2 via miR195 and miR497.

2.3. Real-time qPCR analysis of mRNA expression

Their total RNAs were extracted by using an RNA extraction kit (TIANGEN, Beijing, China), then approximately 2.0 μg of total RNAs were added in a reverse-transcriptase reaction to generate the first strand of cDNA (by using the Revert Aid First Strand Synthesis Kit, Thermo, Waltham, MA, USA). The synthesized cDNA served as the template for qPCR, in the GoTaq®qPCR Master Mix (Promega, Madison, WI, USA). Subsequently, relative mRNA expression levels were measured by qRT-PCR with indicated pairs of primers. Of note, the mRNA expression level of β-actin served as an optimal internal standard control, relative to other mRNA expression levels presented as fold changes. All the forward and reverse primers of those indicated genes were shown in Table S1.

2.4. Western blotting (WB)

Experimental cells were harvested in a denatured lysis buffer (0.5% SDS, 0.04 mol/L DTT, pH 7.5, containing 1 tablet of cOmplete protease inhibitor EASYpacks in 10 ml of this buffer). The total lysates were further denatured by boiling at 100 °C for 10 min, sonicated sufficiently, and diluted with 3 × loading buffer (187.5 mmol/L Tris-HCl, pH 6.8, 6% SDS, 30% Glycerol, 150 mmol/L DTT, 0.3% Bromphenol Blue), before being re-boiled at 100 °C for 5 min. Thereafter, equal amounts of protein extracts were subjected to separation SDS-PAGE before being transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA), and subsequent visualization by Western blotting with distinct antibodies as indicated. β-actin served as an internal control to verify equal loading of proteins in each of the electrophoretic wells.

2.5. Intracellular ROS and GSH detection

Experimental cells were incubated in a serum-free medium containing 20,70-Dichlorodihydrofluorescein diacetate (DCFH-DA, S0033, Beyotime, Shanghai, China), 3ʹ-(p-hydroxyphenyl) fluorescein (HPA, H36004, ThermoFisher), MitoSOX Red (40778ES50, Yeasen Biotechnology, Shanghai, China) or Monochlorobimane (mBCL, HY-101899, MedChemExpress) at 37 °C for 20 min. After the cells were washed three times with a fresh serum-free medium, the relevant fluorescence intensity is measured by a flow cytometer or fluorescent inverted microscope in specific excitation (Ex) and emission (Em) lights. Among them, Ex/Em of DCFH-DA is 488/525 nm, Ex/Em of HPF is 490/515 nm, and Ex/Em of MitoSOX is 510/580 nm, Ex/Em of Monochlorobimane is 380/470 nm.

2.6. MDA, pyruvate, triglyceride and ATP content detection

Both MDA and ATP levels were determined, respectively, according to the instruction of an enhanced ATP assay kit and another kit of lipid oxidation (MDA) detection (both kits from Beyotime, Shanghai, China). The pyruvate and triglyceride contents were measured according to their respective assay kits (from Njjcbio, Nanjing, china).

2.7. CHIP-sequencing analysis

The CHIP-sequencing data was analyzed by Encode database (https://www.encodeproject.org). The project numbers in Encode were ENCSR543SBE (Nfe2l1 in HepG2 cells) and ENCSR488EES (Nfe2l2 in HepG2 cells).

2.8. Luciferase reporter assay

Equal numbers (1.0 × 10⁵) of COS1 cells were allowed for growth in each well of 12-well plates. After reaching 70–80% confluence, the cells were co-transfected for 8 h with an indicated luciferase plasmid alone or together with one of the expression constructs with the Lipofectamine®3000 agent in the Opti-MEM (gibca, Waltham, MA, USA) the Renilla-expressing pRL-TK plasmid served as an internal control for transfection efficiency. After being cultured in a fresh complete medium for 24 h, the cells were lysed and then the luciferase activity was measured by the dual-reporter assay (Promega, Madison, WI, USA).

2.9. Transmission electron microscopy

Experimental cells were pelleted by centrifuging (1000 rpm) before being fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer pH 7.4 for 30 min at 37 °C. The fixed cells stored in 4 °C in 0.1 M sodium cacodylate buffer were performed with a mixture of 1% OsO4, 1.5% K4Fe(CN)6 in 0.1 M sodium cacodylate pH 7.4 for 1 h at 4 °C and then overnight incubated in 0.25% uranyl acetate at 4 °C. After three water washes, samples were dehydrated in series of 15 min steps in 25%, 50%, 75%, 95% and 100% (v/v) ethanol and embedded in an epoxy resin (Sigma-Aldrich). The ultrathin sections (of 60–70 nm in thickness) of cells were prepared with a Leica EM UC7 ultramicrotome and counterstained with 1% uranyl acetate for 15 min and 1% lead citrate for 6 min. Thin sections were imaged using a HITACHI HT 7800 120kv transmission electron microscope operating at 80 kV.

2.10. Flow cytometry analysis of apoptosis

Experimental cells (3 × 10⁵) were allowed for growth in each well of 6-well plates. After transfection or drug treatment, the cells were pelleted by centrifuging at 1000×g for 5 min and washed with PBS for three times, before being incubated for 15 min with 5 μL of Annexin V-FITC and 10 μL of propidium iodide (PI) in 195 μL of the binding buffer. The results were analyzed by the FlowJo 7.6.1 software (FlowJo, Ashland, OR, USA) before being presented.

2.11. The transwell-based migration and invasion assays

The Transwell-based migration and invasion assays were conducted in the modified Boyden chambers (Transwell; Corning Inc. Lowell, MA, USA). Equal numbers of cells were allowed for growth in each well of 12-well plates. After reaching 70–80% confluence, they were starved for 12 h in a serum-free medium. The experimental cells (1 × 10⁴) were suspended in a 0.5 ml medium containing 5% FBS and seeded in the upper chamber of a Transwell. The cell-seeded Transwells were placed in each well of 24-well plates containing 1 ml of complete medium (i.e., the lower chamber), and then cultured for 24 h in the incubator at 37 °C with 5% CO2. Of note, the bottom of the upper Transwell was pre-coated by matrigel basement matrix (BD, Biosciences, USA), before the cells were placed in the invasion assay. The remaining cells in the upper chamber were removed, and the cells attached to the lower surface of the Transwell membranes were fixed with 4% paraformaldehyde (AR10669, BOSTER) and stained with 1% crystal violet reagent (Sigma) before being counted.

2.12. Statistical analysis

Statistical significance of changes in the reporter gene activity or other gene expression was determined using either the Student’s t-test or Multiple Analysis of Variations (MANOVA). The resulting data are shown as a fold change (mean ± S.D), each of which represents at least 3 independent experiments that were each performed triplicate.

3. Results

3.1. A dramatic increase of ROS results from loss of Nrf1α, but is not eliminated by hyperactive Nrf2 in Nrf1α^–∕– cells

Albeit it was previously reported that the intracellular ROS levels were increased in mouse Nrf1-and Nrf2-deficient cells [31], the underlying mechanism remains elusive, to date. Contrarily, Nrf2 was also shown to amplify oxidative stress by induction of Kruppel-like factor 9 (KLF9) in response to elevated ROS over the presetting threshold [32]. To gain an insight into the distinct roles of Nrf1 and Nrf2 in determining and maintaining redox homeostasis, we reexamined human Nrf1α^–∕– and Nrf2^–∕–cell lines; both were established by the genomic editing of HepG2 cells [26,29]. Here, we found that knockout of either Nrf1α^–∕– or Nrf2^–∕– caused an obvious increase in their intracellular ROS levels (detected by DCFH-DA, as the most commonly used probe with a relatively large range from detecting H₂O₂, O₂^.−,·OH to ONOO⁻), but the increased ROS in Nrf1α^–∕– cells were significantly higher than those in Nrf2^−/− cells (Figs. 1B and 1C). To confirm this result, two additional fluorescent probes HPF (hydroxyphenyl fluorescein, sensing particularly to intracellular O₂^.− and ONOO⁻ changes) and MitoSOX (specifically sensing to mitochondrial O₂^.− changes) were also employed (Fig. S1B). The resulting data revealed that distinct extents of the increased ROS in between Nrf1α^–∕– and Nrf2^–∕– cell lines were amplified by HPF (Fig. 1D). However, it is, to our surprise, found that almost no changes in the mitochondrial O₂^.—-sensing fluorescence were detected by MitoSOX in Nrf2^–∕–cells, but Nrf1α^–∕– cells were manifested with a significant magnitude of its mitochondrial ROS (Fig. 1E) when compared to WT controls. A concomitant decrease in the reduced glutathione (GSH, as a free radical scavenger and potent detoxifying agent) was also determined in WT, Nrf1α^–∕– and Nrf2^–∕– cell lines (Figs. 1F, S1C); this is negatively correlated with the intracellular ROS levels, as described [33]. As such, further examination of malondialdehyde (MDA, a marker of polyunsaturated fatty acids peroxidation in the cell) unraveled a substantially increased production in Nrf1α^–∕– cells, to 8 times higher than that obtained from WT cells, whereas Nrf2^–∕– cells only displayed ∼2-fold changes (Fig. 1G). Collectively, these demonstrate that loss of Nrf1α has led to severe endogenous oxidative distress which may be caused primarily by increased ROS from mitochondria, and, by contrast, loss of Nrf2 only causes modest oxidative stress in the presence of Nrf1.

To explore the reasons underlying such distinct increases of ROS caused by deficiency of Nrf1α from Nrf2, it was surprisingly unveiled by transcriptome sequencing that most of those differential expression genes responsible for antioxidant and detoxification were significantly up-regulated in Nrf1α^–∕–cells, but also markedly down-regulated in Nrf2^–∕– cells (Fig. S2), when compared to WT cells. Further quantitative PCR of critical genes for ROS elimination revealed that Nrf1α^–∕–cells gave rise to obvious increases in basal expression of SOD2 (superoxide dismutase 2, converting the mitochondrial O₂^.— into H₂O₂), PRDX1, PRDX4, PRDX5 (peroxiredoxins, all three scavenging H₂O₂ by consuming reduced thioredoxin) and GPX1 (glutathione peroxidase 1), but extra-mitochondrial SOD1 expression was decreased (Fig. 1H). Conversely, Nrf2^–∕– cells manifested significantly decreased expression of SOD2, PRDX5 and GPX1, but PRDX4 was highly expressed, while SOD1 was modestly increased. Also, it was, much to our surprise, found that basal expression of those examined genes GCLC, GCLM, GLS (glutaminase), GSR, SLC6A9 and particularly SLC7A11 (all involved in GSH biosynthesis and regeneration) was significantly up-regulated in Nrf1α^–∕–cells, but down-regulated in Nrf2^–∕– cells, when compared with WT controls (Fig. 1I). Additional two reducing powers TXN (thioredoxin) and TXN2 (thioredoxin-2) were up-expressed in Nrf1α^–∕–cells, but down-expressed in Nrf2^–∕– cells, whereas basal expression of TXNRD1 (thioredoxin reductase-1) and TXNRD2 (thioredoxin reductase-2) was almost not changed in both cell lines (Fig. 1J). Moreover, basal expression of SRXN1 (sulfiredoxin 1) was reduced only in Nrf2^–∕–cells, but largely unaltered in Nrf1α^–∕–cells (Fig. 1J). Collectively, these findings demonstrate that loss of Nrf2 results in obvious defects in de novo biosynthesis of GSH and TXN and their recycling, but rather unusual increases of them have strikingly emerged upon loss of Nrf1α.

Further examination revealed that basal expression of G6PD (glucose-6-phosphate dehydrogenase, a key enzyme in the pentose phosphate pathway to yield NADPH) and PGD (6-phosphogluconate dehydrogenase, as the second dehydrogenase in the pentose phosphate shunt) was slightly increased or unaltered in Nrf1α^–∕–cells respectively, but decreased in Nrf2^−/− cells (Fig. 1K). Conversely, significant increases in basal TIGAR (TP53 induced glycolysis regulatory phosphatase, blocking glycolysis and directing the pathway into the pentose phosphate shunt to protect cells from DNA-damaging ROS) expression were observed in both Nrf1α^–∕– and Nrf2^−/− cell lines (Fig. 1K). Besides, basal expression of ME1 (malic enzyme, as another NADPH supplier) was also substantially up-regulated in Nrf1α^–∕–cells to ∼4-fold changes higher than its WT control, but almost unchanged in Nrf2^−/− cells. By contrast, IDH1 (isocitrate dehydrogenase, a key enzyme catalyzing the oxidative decarboxylation of isocitrate to alpha-ketoglutarate to generate CO₂ and NADPH in the citric acid cycle), rather than IDH2, was significantly lowered in Nrf1α^–∕–cells, but highly expressed in Nrf2^−/− cells (Fig. 1K). Altogether, these indicate knockout of Nrf1α leads to a redox metabolism reprogramming, that is distinctive from the knockout of Nrf2.

Next, by quantitative investigation of another antioxidant capability to prevent free heme from forming free radicals during oxidative stress [16], it was unraveled that basal expression levels of HO1 and FTH1 (ferritin heavy chain) were up-regulated in Nrf1α^–∕–cells, but down-regulated in Nrf2^−/− cells (Fig 1L), while no changes in the expression of FTL (ferritin light chain) were observed in both cell lines when compared to WT controls. In addition, another potent antioxidant family of highly conserved sestrin (SESN, involved in the reduction of PRDXs, so to negatively regulate mTORC signaling pathways was also investigated herein) [34]. Among them, SESN1 was up-expressed in both Nrf1α^–∕–and Nrf2^−/− cell lines, whereas SESN2 was only up-expressed in Nrf1α^–∕– cells, but rather down-expressed in Nrf2^−/− cells (Fig 1M). By sharp contrast, basal expression of SESN3 was significantly down-regulated in Nrf1α^–∕– cells and also almost abolished in Nrf2^−/− cells when compared with WT controls. These imply distinct roles for Nrf1α and Nrf2 in governing antioxidant potentials and also redox signaling to mTORC-regulated networks to meet the needs of cell survival.

Further insights into the protein expression of critical genes by Western blotting revealed that, albeit the processed Nrf1-C/D isoform is only slightly decreased in Nrf2^−/− cells, basal Nrf2 protein was aberrantly, substantially accumulated in Nrf1α^–∕– cells (Fig 1N and S3), as consistent with our previous results [29,35]. Such hyperactive Nrf2 should be interpreted as a compensation for loss of Nrf1, and consequently attributable to the constant up-regulation of most antioxidant and detoxifying genes in Nrf1α^–∕– cells. As anticipated, most of the examined proteins including CAT (catalase), PRDX1, GPX1, SOD2, TXN1, TXN2, G6PD, GSR, GCLM and HO1 were highly expressed as accompanied by hyperactive Nrf2 accumulation in Nrf1α^–∕– cells (Figs 1N and S3). However, basal abundances of SOD1 and IDH1, rather than IDH2, were down-regulated in Nrf1α^–∕– cells, but almost unaffected in Nrf2^−/− cells, implying they may serve as Nrf1-specific targets.

3.2. Loss of Nrf1α results in mitochondrial dysfunction and oxidative damage in Nrf1α^–∕– cells

Intriguingly, the above-described data indicate that even though hyperactive Nrf2 accumulated with up-regulation of antioxidant and detoxification genes, Nrf1α^–∕– cells are still manifested with severe endogenous oxidative stress, caused primarily by ROS yielded from its mitochondria, but rather not similar distress has been suffered from Nrf2^−/− cells. To address this, we herein examine distinct subunits of mitochondrial ETC complexes I to IV in Nrf1α^–∕– cells, by comparison with Nrf2^−/− cells and WT controls. As anticipated, most subunits of the mitochondrial respiratory chain were, to different extents, down-regulated in Nrf1α^–∕– cells (Fig. 2A). By contrast, their mRNA expression in Nrf2^−/− cells appeared to be unaffected or slightly enhanced, with an exception of SDHA (succinate dehydrogenase complex flavoprotein subunit A). Of note, ND1, MTCYB, MTCO1 and MTCO2 are encoded by the mitochondrial genome per se, all other examined subunits are encoded by the nuclearly-located genes. Western blotting also revealed that protein abundances of NDUFC2, SDHA, SDHB, NQCRFS1, MTCO1, and MTCO2 were decreased in Nrf1α^–∕–, but not Nrf2^−/−, cell lines (Fig. 2B). These demonstrate that loss of Nrf1α, rather than Nrf2, leads to a certain dysfunction of the mitochondrial respiratory chain.

Fig. 2.

Changes in mitochondrial genes and morphology of Nrf1α^−/− distinctive from Nrf2^−/− cells

(A) The mRNA levels of NDUFS1, NDUFV1, NDUFV2, ND1, NDUFC2, SDHA, SDHB, SDHC, SDHD, UQCRFS1, CYC1, MTCYB, COX5A, COX6A, COX7A, COX8A, MTCO1, and MTCO2 were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The resulting data are shown as mean ± SEM (n = 3 × 3; *, p < 0.05 and **, p < 0.01).

(B) The protein expression levels of NDUFC2, SDHA, SDHB, NQCRFS1, MTCO1, and MTCO2 in WT, Nrf1α^–∕– and Nrf2^−/− cells were visualized by Western blotting. The intensity of all the immunoblots was calculated and shown on the bottom.

(C) Distinct ROS levels and detected by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cells that had been treated with Rotenone for 0 h, 12 h, or 24 h and then stained by DCFH for 30 min.

(D) The electronic micrographs representative of WT, Nrf1α^–∕– and Nrf2^−/− cells. The scale bar = 2 μm in × 2 K pictures, or = 1 μm in × 7 K pictures, or = 0.25 μm in × 20 K pictures.

(E) Differential mRNA expression levels of Drp1, OPA1, MFN1, MFN2 and MTFR1 determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(F) The protein levels of OPA1, MFN1, MFN2, and Drp1 in WT, Nrf1α^–∕– and Nrf2^−/− cells were visualized by Western blotting. The intensity of all the immunoblots was calculated and shown on the bottom.

Next, treatment of WT cells with rotenone (a specific inhibitor of mitochondrial ETC complex I [36]) unraveled that the intracellular oxidative stress was aggravated by incrementing the mitochondrial ROS production in a time-dependent manner (Fig. 2C, left panel), albeit Nrf2 protein expression was enhanced by this chemical (Fig. S4A). Rather, similar rotenone treatment of Nrf1α^–∕– cells rendered its severe endogenous oxidative stress to be significantly mitigated to a certain extent (Fig. 2C, middle panel), but roughly no obvious changes were determined after treatment of Nrf2^−/− cells (Fig. 2C, right panel). Also, no significant differences in rotenone-induced apoptosis were observed in these three cell lines (Figs S4B and S4C). These indicate that Nrf1α^–∕–-derived dysfunction of its mitochondrial respiratory chain to give rise to considerable higher ROS levels is not further worsened but conversely relieved by rotenone, while not a similar event appears to take place in Nrf2^−/− cells. In addition, rotenone could only reduce a small amount of ATP production while glycolysis inhibitors 2-DG could reduce most ATP production in Nrf1α^–∕– cells (Figs. S5A and 5B); this indicates that ATP is mainly supplied by glycolysis after Nrf1 knockout.

Electron micrographic imaging showed that morphological changes of Nrf1α^–∕– cells were distinctive from Nrf2^−/− cells (Figs. 2D and S5C). The mitochondria of Nrf1α^–∕– cells seem to be shrunk, with significant decreases in their mitochondrial number and their cristae in each mitochondrion (middle panels), whereas the mitochondria of Nrf2^−/− cells appeared to be modestly reinforced (right panels) when compared with WT cells (left panels). Further examination of critical players in the mitochondrial fission and fusion (for maintaining a steady state of mitochondrial function under cellular metabolic or environmental stress [37])by RT-PCR and Western blotting revealed that MFN1 (mitofusin 1) and OPA1 (optic atrophy 1, serving mitochondrial dynamin-like GTPase) were down-regulated in Nrf1α^–∕– cells (Figs. 2E and 2F). By contrast, DRP1 (dynamin-related protein 1, a key regulator of mitochondrial division) was substantially up-regulated in Nrf1α^–∕– cells, even though MTFR1 (mitochondrial fission regulator 1) was decreased. However, slight changes in these players were examined in Nrf2^−/− cells. Altogether, loss of Nrf1α results in dysfunctional mitochondria, thereby producing considerably higher levels of ROS along with oxidative damages.

3.3. Nrf1 directly regulates two nuclear respiratory factors αPal^NRF1 and GABPα^NRF2, and co-factor PGC1α

To gain an insight into the underlying mechanism for mitochondrial dysfunction caused by loss of Nrf1α^–∕–, we here examined the constitutive expression of two nuclear respiratory factors αPal^NRF1 and GABPα^NRF2 [38,39], cofactor PGC1α, and critical target genes (responsible for mitochondrial DNA replication and transcription, ETC/OXPHOS gene expression, and mitochondrial biogenesis [40]). The results showed that basal mRNA expression levels of αPal^NRF1, GABPα^NRF2 and co-targeting mitochondrial transcription factors TFAM (transcription factor A, mitochondrial), TFB1M (transcription factor B1, mitochondrial) and TFB2M (transcription factor B2, mitochondrial), together with PGC1α, but not PGC1ß, were down-regulated, though to different extents, in Nrf1α^–∕– cells (Fig. 3A). However, such key factors αPal^NRF1 and GABPα^NRF2 were not significantly affected in Nrf2^−/− cells, which only gave rise to marginally decreased expression of PGC1α, TFAM, TFB1M, but not TFB2M. Further examinations also revealed that basal abundances of αPal^NRF1, GABPα^NRF2, PGC1α and TFAM proteins were down-regulated in Nrf1α^–∕– cells, while in Nrf2^−/− cells only TFAM was down-expressed (Figs. 3B and S6A). Collectively, these indicate that Nrf1α^–∕–-caused mitochondrial dysfunction is attributable to deregulation of nuclear respiratory controls towards the mitochondrial gene transcriptional networks. This is also further supported by the restoration of ectopic Nrf1 factor into Nrf1α^–∕– cells (Fig. S6B), showing greater or lesser extents of recovery of those four key factors (αPal^NRF1, GABPα^NRF2, PGC1α and TFAM), together with other mitochondrial proteins NDUFC2, NQCRFS1, MTCO1, MTCO2, and SDHB, but not SDHA. Additional supportive evidence was obtained from the tetracycline-inducible Nrf1-expressing cell lines (Figs. S6C and S6D), in which almost all those examined mRNAs and proteins were up-regulated. By contrast, Nrf2 induction by tetracycline led to significantly increased expression of PGC1α, GABPα^NRF2 and TFAM, besides HO1 and GCLM. This implies at least a partial involvement of Nrf2 in regulating the mitochondrial function, particularly when it is required for certain inducible cues.

To further investigate distinct roles for Nrf1 and Nrf2 in the nuclear-to-mitochondrial regulation, the experimental Nrf1α^–∕–, Nrf2^−/−, and WT cell lines were treated with tert-butylhydroquinone (tBHQ, as a pro-oxidative stressor [41]). The results revealed that Nrf1, Nrf2, PGC1α, αPal^NRF1, GABPα^NRF2, and TFAM (Figs. 3C & 3D and S7A & S7B) along with HO1 and GCLM (Fig. S7C) were significantly induced by tBHQ stimulation in WT cells. Such induction of PGC1α, αPal^NRF1, GABPα^NRF2 and TFAM (controlling the mitochondrial function) was almost completely abolished in Nrf1α^–∕– cells (Fig. 3C), even albeit hyperactive Nrf2 and its targeting HO1 and GCLM were further enhanced by tBHQ (Figs. S7A–C). Conversely, although induction of HO1 and GCLM by tBHQ was abolished in Nrf2^−/− cells, different inducible extents of αPal^NRF1, GABPα^NRF2, TFAM and PGC1α were still stimulated by this chemical, and also accompanied by induction of Nrf1 in Nrf2-deficient cells. Taken altogether, these demonstrate that Nrf1 rather than Nrf2 is essential for maintaining mitochondrial functional homeostasis by governing the nuclear respiratory factors.

To clarify whether Nrf1 or Nrf2 directly regulates such key genes αPal^NRF1, GABPα^NRF2 and PGC1α, their promoter regions were constructed into relevant luciferase reporters (as shown in Fig. 3E). The results revealed that these reporter genes αPal^NRF1-Luc, GABPα^NRF2-Luc and PGC1α-Luc were transcriptionally activated by ectopic expression of Nrf1 (Fig. 3F). Of note, transactivation activity of GABPα^NRF2-Luc mediated by Nrf1 was substantially increased to ∼2.5 times over its control values. By contrast, Nrf2 only gave rise to marginal transactivation of these reporter genes at considerably lower levels. Further sequence analysis unraveled that several ARE consensus sites are present in the promoter regions of αPal^NRF1, GABPα^NRF2 and PGC1α, respectively (Fig. 3G). Some of such ARE-driven luciferase reporter genes (with ARE site of HO1 acting as a positive control) were also trans-activated by Nrf1 and/or Nrf2 (Figs. 3H and S7D). In search of the Encode database, the ChIP-sequencing data for binding Nrf1 and Nrf2 in HepG2 cells failed to show significant signals in the promoter regions of αPal^NRF1, GABPα^NRF2, and PGC1α, because signals are considerably weak (Fig. S8), while further validation is needed. Overall, these results suggest that Nrf1 (alongside with Nrf2) in a combination with putative DNA-binding partners or other cofactors regulates αPal^NRF1, GABPα^NRF2 and PGC1α for determining mitochondrial homeostasis (Fig. 3I). But, conversely loss of Nrf1α^–∕– leads to a fatal defect in this critical gene expression, resulting in mitochondrial dysfunction and oxidative damages.

3.4. Glycolysis of Nrf1α^–∕– cells is enhanced to aggravate its mitochondrial pressure with increased ROS production

Since as by-products of mitochondrial oxidative metabolism, ROS are allowed to connect with controls of metabolic process, such that the yield of ROS in this organelle depends on distinct types of fuel loading (carbohydrate, lipid, protein), besides mitochondrial functional homeostasis per se [42]. Here, transcriptomic sequencing analysis revealed significant differential expression of those genes responsible for energy metabolism, as well as carbohydrate and lipid metabolisms of Nrf1α^–∕– cells (Fig. S9), but these metabolic gene changes were strikingly diminished, abolished, or even reversed in Nrf2^−/− cells. Thereafter, Nrf1α^–∕–, Nrf2^−/− and WT cell lines had been allowed for growth in a complete medium containing 25 mM versus 5 mM glucose for 24 h, before the putative effect of glucose, as a fuel load, on the amount of intracellular ROS production was determined. As shown in Fig. 4A, the ROS levels detected by flow cytometry of WT cells fed with 25 mM glucose were augmented when compared with those arising from 5 mM glucose conditions. Such a supply of 25 mM glucose deteriorated endogenous oxidative stress in Nrf1α^–∕– cells (middle panel), whereas Nrf2^−/− cells appeared to be unaffected (right panel). Conversely, glycolytic inhibition by 2-deoxy-d-glucose (2-DG in 25 mM glucose media) led to distinct extents of decreased ROS yield in Nrf1α^–∕– or Nrf2^−/− cell lines (Fig. S10A), but an apparent increase of ROS in WT cells occurred after 2-DG treatment. However, treatment of Nrf1α^–∕–cells with insulin (to activate glycolysis) caused a modest increase in ROS, while no changes in ROS yielded from insulin-treated WT or Nrf2^−/− cell lines were determined (Figs. S10B and S10C). Collectively, these data suggest that glycolysis contributes to ROS production, especially in Nrf1α^–∕–cells, and decreased metabolism in Nrf2^−/− cells could impede the yield of ROS by feeding 25 mM glucose to oxidative metabolism (e.g., glycolysis).

Fig. 4.

Changes in ROS levels arising from differential expression of redox metabolic genes in Nrf1α^–∕– and Nrf2^−/− cells lines.

(A) Changes in ROS levels were detected by flow cytometry of WT, Nrf1α^–∕– and Nrf2^−/− cell lines that had been cultured in 25 mM or 5 mM glucose media for 24 h, and then stained by DCFH for 30 min.

(B) A schematic representation of the redox metabolic process after glucose enters the cell.

(C) Differential mRNA expression levels of HK1, HK2, GLUT1, GLUT4, LDHA, PDH, PDK1, SREBP1, SREBP2, ACCα, ACLY, FASN, SCD1, and CPT1A were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells. The data are shown as mean ± SEM (n = 3 × 3; $, p < 0.05 and $$, p < 0.01; *, p < 0.05 and **, p < 0.01).

(D) Changed abundances of HK1, HK2, Glut1, Glut4, LDHA, PDH, SREBP1, SREBP2, FASN, SCD1, and CPT1A in WT, Nrf1α^–∕– and Nrf2^−/− cells were visualized by Western blotting. The immunoblotting intensity was calculated as shown on the bottom.

(E) The mRNA level of Nrf2, GCLM, HO1, HK1, HK2, GLUT1, GLUT4, G6PD, LDHA, PDH, PDK1, SREBP1, ACCα, ACLY, FASN, SCD1, and CPT1A were determined by RT-qPCR in Nrf1α^–∕– cells, which had been interfered by siRNA specific targeting Nrf2 (siNrf2) or scrambled siRNA (siNC). The resulting data are shown as mean ± SEM (n = 3 × 3; *, p < 0.05 and **, p < 0.01).

(F) The protein levels of Nrf2, GCLM, HO1, HK1, HK2, Glut1, Glut4, G6PD, LDHA, PDH, SREBP1, ACCα FASN, and SCD1 were visualized by Western blotting in Nrf1α^–∕– cells that were interfered by siNrf2 or siNC. The intensity of all the immunoblots was calculated as shown on the bottom.

(G) Changes in ROS levels were detected by flow cytometry of Nrf1α^–∕– cells that had been interfered with by siNrf2 or siNC and cultured in 25 mM or 5 mM glucose media for 24 h and then stained by DCFH for 30 min.

Those key genes involved in glucose and lipid metabolic pathways (as illustrated in Fig. 4B) were further examined herein. Both RT-qPCR and WB results showed that hexokinases (HK1 and HK2, the first rate-limiting enzymes of glycolysis to yield glucose 6 phosphate), along with glucose transporters (Glut1, and Glut4) are all highly expressed in Nrf1α^–∕– cells, of which HK1 and Glut4 are also obviously up-regulated in Nrf2^−/− cells to considerably higher levels, but that were rather lower than those obtained in Nrf1α^–∕– cells (Figs. 4C and 4D). In addition, the pyruvate content in Nrf1α^–∕– cells was increased, while decreased in Nrf2^−/− cells (Fig. S10D), this implies an enhancement of glycolysis in Nrf1α^–∕–cells, as consistent with a previous report [20,43]. This is also evidenced by further examinations, revealing that pyruvate dehydrogenase (PDH) was modestly up-regulated in Nrf1α^–∕–cells (with hyperactive Nrf2 accumulated), and thus down-regulated in Nrf2^−/− cells (Figs. 4C and 4D). Rather, no changes in basal expression of pyruvate dehydrogenase kinase 1 (PDK1, as a specific inhibitor of PDH) and lactate dehydrogenase A (LDHA) were observed in these experimental cells (Figs. 4C and 4D). Such being the conditions, this facilitates the fuel loading by glycolysis insomuch as to enter the mitochondria, leading to an increased pressure on the oxidative respiratory chain to produce the excessive ROS in this organelles of Nrf1α^–∕–cells. Furthermore, it was also found that other critical genes SREBP1, SREBP2, ACLY (ATP citrate lyase), ACCα (Acetyl-CoA Carboxylase Alpha), FASN (fatty acid synthase) and SCD1 (stearoyl-CoA desaturase 1) (all responsible for fatty acid synthesis) were significantly up-regulated in Nrf1α^–∕–cells, but rather down-regulated in Nrf2^−/− cells (Figs. 4C and 4D). In addition, the triglyceride content in Nrf1α^–∕– cells was increased (Fig. S10E), this implies that loss of Nrf1α rather than Nrf2 results in an increase in fatty acid synthesis, providing a material basis for the proliferation of Nrf1α^–∕–cells. Also, such increased synthesis of fatty acids has to consume a certain amount of NADPH generated by the pentose phosphate pathway and hence leads to a decrease in the antioxidant capability of the cell. Conversely, this appears to be supported by Scheffler’s group, demonstrating that a genetic respiratory chain deficiency could block the TCA cycle [44], serving as a vital hub towards de novo lipid synthesis (as illustrated in Fig. 4B). Thereby, it is inferable that the elevated lipid synthesis is considered an important outlet for glycolytic flux in Nrf1α^–∕–cells, as accompanied by its mitochondrial oxidative damages.

Since Nrf2 is aberrantly accumulated in Nrf1α^–∕–cells as aforementioned, the role of Nrf2 in this metabolic process was investigated by specific siRNAs interfering with Nrf2. As shown in Figs. 4E & 4F, significant decreases in mRNA and protein expression of Nrf2 and target genes HO1 and GCLM were caused by the silencing of Nrf2 in Nrf1α^–∕–cells. In the meanwhile, all other examined genes except PDH and PDK (involved in glucose and lipid metabolism) are also down-regulated by knockdown of Nrf2 to varying degrees (Figs. 4E and 4F). Among them, HK2, Glut1, Glut4, G6PD, ACCα, ACLY, FASN and SCD1 were substantially inhibited by interfering Nrf2 (Figs. 4E and 4F). Further studies unraveled that such knockdown of Nrf2 in Nrf1α^–∕–cells did not reverse but shortened the reduced ROS led by 5 mM glucose (Fig. 4G), with the apoptosis unchanged (Fig. S11A). In addition, knockdown of Nrf2 in Nrf1α^–∕–cells led to an increase in the intracellular ROS and apoptosis, when compared to equivalent controls (Fig. S11B), implying an antioxidant cytoprotective role of Nrf2 against oxidative damages. Altogether, loss of Nrf1α may stimulate a surge of Nrf2 accumulated to enhance glycolysis and lipid synthesis; this is also accompanied by increased pressure of the Nrf1α^–∕–-damaged mitochondria to generate the excessive ROS. In turn, the elimination of ROS by the Nrf2-mediated antioxidant cytoprotective mechanism is also reinforced. Therefore, it is postulated that Nrf2 may provide a decisive interplay to balance between ROS arising from cellular metabolism and antioxidant response.

3.5. Mitochondrial UCP2 is negatively regulated by Nrf2 via miR-195 and miR-497 to augment ROS, even in Nrf1α^–∕– cells

Besides the glycolytic flux entering the mitochondria, its uncoupling proteins (UCPs) can also monitor the production of ROS in these organelles [45]. Among them, UCP1 was reported to be positively correlated with Nrf1, but it is dominantly expressed in white adipose tissue [46], while UCP2 is widely expressed in various tissues, though they can reduce the mitochondrial Δψ_m by ‘mild uncoupling’ insomuch as to negatively regulate the yield of ROS [45,47]. Here, our evidence revealed that basal mRNA and protein levels of UCP2 were significantly decreased in Nrf1α^–∕–cells, but up-expressed to considerably higher extents in Nrf2^−/−cells when compared to WT controls (Fig. 5A). Next, we examined the effect of UCP2 on the production of ROS by forced expression of UCP2 in distinct genotypic cell lines (Figs. 5B and S12A). The results unraveled that the ROS levels in Nrf1α^–∕–cells (with hyperactive Nrf2) were significantly decreased by over-expression of UCP2, but almost unaffected in Nrf2^−/− and WT cell lines (Fig. 5B). Collectively, these indicate that Nrf2 may exert a dominant inhibitory effect on UCP2 to promote the yield of ROS in Nrf1α^–∕–cells.

To gain insight into the putative inhibitory effect of Nrf2 on UCP2, we further investigated this gene promotor and its mRNA regulatory mechanisms, As shown in Fig. 5C, among eight candidate miRNAs for targeting 3’-UTR of UCP2’s mRNA predicated by online databases (http://www.targetscan.org/ and http://ophid.utoronto.ca/mirDIP/index.jsp#r), only miR-195 and miR-497 were highly expressed in Nrf1α^–∕–cells, but unaltered or even down-regulated, respectively, in Nrf2^−/−cells, just as opposed to basal expression of UCP2 (Fig. 5A). The sequence analysis of miR-195 and miR-497 deciphered a conserved homology targeting their complementary 7-nucleotide motif within 3’-UTR of UCP2 (Fig. 5D). As anticipated, the results from RT-qPCR revealed that the mRNA expression levels of UCP2 were effectively down-regulated by transfecting different concentrations of miR-195 or miR-497, but not miR-24, into WT cells (Fig. 5E). In the parallel experiments, the decreased expression of UCP2 in Nrf1α^–∕–cells was restored by a co-inhibitor of both miR-195 and miR-497 (Fig. 5F).

To validate the putative interaction of either miR-195 or miR-497 with UCP2 mRNA, we cloned their 3’-UTR or mutants into a luciferase reporter vector before being co-transfected into WT cells. As expected, the results demonstrated that the 3’-UTR-driven reporter activity was inhibited by miR-195 or miR-497, but such inhibitory effects were sufficiently abrogated by its mutants from 3’-UTR of UCP2 mRNA (Fig. 5G). Next, we constructed three reporter genes by cloning distinct lengths of the examined promoter regions of UCP2, miR-195 and miR-497 (i.e., UCP2-Luc, miR-195-Luc, and miR-497-Luc, Fig. 5H), to verify whether they are regulated by Nrf1 and/or Nrf2. The luciferase assays showed that the UCP2-Luc transcriptional activity was almost unaffected by Nrf1 or Nrf2 (Fig. 5I). However, transactivation activity of miR-195-Luc or miR-497-Luc reporters was significantly induced by co-transfection with Nrf1 or Nrf2 (Fig. 5J). Further analysis of the promoter regions of miR-195 and miR-497 uncovered that their promoters are almost completely overlapped and comprise five common ARE consensus sites (Fig. 5K). Among these ARE-driven luciferase reporters made by ligating the indicated sequences into the PGL3-promoter vector, only #1 transactivation activity was modestly mediated by Nrf1, while Nrf2 enabled to mediate significant induction of #1, #2 and #4 reporters (Fig. 5L). However, ChIP-sequencing data obtained from the Encode database failed to show significant peaks of both Nrf1 and Nrf2 activities to binding multiple ARE sites within the promoter regions of miR-195 and miR-497 (Fig. S12B). As such, taking all available data into consideration, we propose that Nrf2 (and Nrf1) possess a capability to down-regulate the UCP2 gene activity indirectly by cognate ARE-driven expression of miR-195 and/or miR-497 (Fig 5M). This notion is also further supported by additional evidence that a remarkable decrease of UCP2 at its mRNA and protein expression levels occurred upon tetracycline-inducible expression of either Nrf1 or Nrf2 in HEK293 cells (Fig. S12C).

3.6. Loss of Nrf1α leads to inactivation of the mitochondrial stress response, albeit Nrf2 is hyperactive, in Nrf1α^–∕– cells

To maintain mitochondrial homeostasis, all eukaryotic cells have evolutionarily developed a nuclearly-controlled transcriptional responsive programme, called mitochondrial unfolded protein response (UPR^mt) [48]. If it is required for mitochondrial functioning, UPR^mt is triggered to actively promote the repair and recovery of mitochondrial function and integrity. In Caenorhabditis elegans, the UPR^mt is monitored primarily by the stress-activated transcription factor ATFS-1 (of the basic-region leucine zipper family), as well by SKN-1 (skinhead-1, sharing an orthologous homology with Nrf1 and Nrf2) [49,50]. The mammalian UPR^mt is regulated by ATF5 (activating transcription factor 5)-mediated expression of several mitochondrial chaperones and protease genes to promote its OXPHOS and cell growth during mitochondrial dysfunction [51]. Besides, ATF4 (activating transcription factor 4) and CHOP are also involved in the activation of UPR^mt [52,53]. Our previous study has found that loss of Nrf1 results in down-regulation of ER-stress-related genes (Figs. S13A and S13B) [54]. Herein, our evidence revealed that basal expression of ATF5, ATF4 and CHOP, along with their targets HSP60 (60 KDa heat shock protein, mitochondrial), GRP75 (i.e. mtHSP70, stress-70 protein, mitochondrial), and FGF21 (fibroblast growth factor 21) was, to different degrees, lowered in Nrf1α^–∕–cells (Fig. S13B). Of note, these two key chaperones HSP60 and GRP75 are controlled by ATF5 in folding the denatured and nascent polypeptides in the mitochondria, while FGF21 is regulated by ATF4 in metabolic adaptation to fasting [55]. By contrast with Nrf1α^–∕–cells, putative inactivation of UPR^mt did not appear to occur in Nrf2^−/−cells when compared with a similar status of WT cells (Fig. S13B).

To determine the effect of Nrf1 or Nrf2 deficiency on UPR^mt, Nrf1α^–∕– and Nrf2^−/−cell lines were treated with carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP, an uncoupling reagent to disrupt ATP synthesis by transporting H⁺ ion through the mitochondrial membrane before being used for oxidative phosphorylation, thus to activate UPR^mt as reported by Refs. [53,[56], [57], [58]]). The results unraveled that Nrf1 and Nrf2 at the mRNA and protein levels were induced by FCCP in WT cells (Figs. 6A and 6B). Albeit transcriptional expression of Nrf1 was reportedly regulated by Nrf2 [29], it was still stimulated by FCCP in Nrf2^−/−cells. Conversely, mRNA expression levels of Nrf2 were unaffected by FCCP in Nrf1α^–∕– cells, even though its protein appeared to be further accumulated (Figs. 6A and 6B). Further examinations revealed that FCCP-inducible expression levels of ATF4, ATF5, CHOP, HSP60, GRP75 and FGF21 were almost completely abolished in Nrf1α^–∕– cells, but not or less altered in Nrf2^−/−cells than those obtained from WT cells (Figs. 6A & 6C). Collectively, these demonstrate that Nrf1, rather than Nrf2, is a dominant activator involved essentially in the FCCP-induced UPR^mt.

Fig. 6.

Nrf1 is essentially required for activation of mitochondrial stress-related responsive genes

(A) The mRNA levels of Nrf1, Nrf2, ATF5, ATF4, CHOP, HSP60, GRP75, and FGF21 were determined by RT-qPCR in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated with FCCP for 0 h, 12 h or 24 h. The data are shown as mean ± SEM (n = 3 × 3) with significant increases ($, p < 0.05 and $$, p < 0.01) as compared to the untreated cases.

(B) The protein levels of indicated genes above were visualized by Western blotting in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated with FCCP for 0 h, 12 h, or 24 h. The intensity of all the immunoblots was calculated as shown on the bottom.

(C) Changes in ROS levels were detected by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cells that had been treated by FCCP for 0 h, 12 h, or 24 h, and then stained by DCFH for 30 min.

(D) The apoptosis was analyzed by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cells, after they had been treated by FCCP for 0 h, 12 h, or 24 h and then incubated with Annexin V-FITC and PI.

(E) A proposed model to explain an essential role of Nrf1 in mediating mitochondria unfolded protein response.

Further treatment of WT cells with FCCP caused a significant increment in the production of ROS in a time-dependent manner (Fig. 6D); this was also accompanied by increased apoptosis (Fig. 6E). In sharp contrast, FCCP-caused ROS and apoptosis were substantially augmented in Nrf2^−/−cells (Fig. 6D and 6E). However, it is, much to our surprise, found that almost no changes in the intracellular ROS levels and apoptosis of Nrf1α^–∕– cells occurred after treatment of FCCP (Figs. 6D and 6E). This implies that no proper UPR^mt to FCCP is instigated in the dysfunctional mitochondria caused by loss of Nrf1α, although hyperactive Nrf2 is retained in Nrf1α^–∕– cells (Fig. 6F). Conversely, loss of Nrf2 may render its deficient cells to rely much on the energy supply of mitochondria.

3.7. The EMT-relevant signaling pathways are constitutively activated in Nrf1α^–∕– cells

The above-described evidence revealed that mildly increased ROS in Nrf2^−/−cells is attributable to the diminishment of the antioxidant capability to scavenge ROS, while Nrf1α^–∕– cells give rise to a severe amount of ROS in impaired mitochondria, possibly leading to malignant cell proliferation as reported previously [29]. Herein, to further determine the underlying mechanisms by which loss-of-function of Nrf1α enables to serve an original impact on tumorigenesis and development, we examined several discrete signaling pathways provoked by ROS, which are all converged on the EMT-relevant process involved in carcinogenesis and malignance (including invasion and migration). As illustrated in Fig. 7A, the migrating ability of Nrf1α^–∕– cells through transwells was greatly enhanced, while the migration of Nrf2^−/−cells became rather weakened, when compared to that of WT cells. Further examinations revealed that the epithelial marker proteins CDH1 (E-cadherin) were down-regulated in Nrf1α^–∕– cells, but the mesenchymal marker proteins CDH2 (N-cadherin), vimentin and FN1 (fibronectin 1) were up-regulated in this deficient cells (Fig. 7B, b1 to b4). Conversely, in Nrf2^−/−cells, CDH1 was substantially incremented, while CDH2 and FN1 were roughly unaltered. Interestingly, ITGB4 (integrin subunit beta 4, a key node making the cell-cell interaction and communication with the extracellular matrix) was markedly increased in Nrf1α^–∕– cells but rather diminished in Nrf2^−/−cells (Fig. 7B, b6). Aside from SNAI1 (snail family transcriptional repressor 1) which was unaffected in these two deficient cell lines, SNAIL2 (snail family transcriptional repressor 2) was significantly augmented in Nrf1α^–∕– cells, but was unchanged in Nrf2^−/−cells (Fig. 7B, b7 & b8), albeit both factors were shown to activate the EMT-relevant transcriptional programme during liver fibrosis and cancer development, but also repress the epithelial genes by binding to their E-box DNA sequences [59]. In addition, MMP9 (matrix metallopeptidase 9, which can enhance the extracellular matrix protein degradation and hence enable invasion) was up-regulated in Nrf1α^–∕– cells, but down-regulated in Nrf2^−/−cells (Fig. 7B, b9). Taken altogether, these demonstrate that loss of Nrf1α leads to promotion of the EMT and malignant behavior, but such effects appear to be suppressed by loss of Nrf2, although its ROS levels were also increased.

Fig. 7.

Distinctive effects of Nrf1α^–∕– and Nrf2^−/− on EMT-related genes and ROS-triggered signaling pathways

(A) WT, Nrf1α^–∕– and Nrf2^−/− cells were starved for 12 h in a serum-free medium and subjected to transwell migration before being Captured by microscope. Counts of migrated cells are shown as fold changes in right panel (mean ± SEM, n = 3 × 3; $$, p < 0.01; **, p < 0.01).

(B) The protein levels of CDH1, CDH2, Vimentin, FN1, ITGB4, SNAI1, SNAIL2, MMP9, and P53 were visualized by Western blotting in WT, Nrf1α^–∕– and Nrf2^−/− cells. The intensity of all the immunoblots was calculated as shown on the bottom.

(C) A schematic representation of EMT and ROS-triggered signaling pathways.

(D) Abundances of p-JNK, JNK, p-ERK1/2, ERK1/2, p-MEK1, MEK, p-p38, p38, HIF1α, PTEN, PI3K, p-AKT, NF-κB 2, NF-κB 1 (p105/p50), P65, and c-FOS, c-JUN proteins were visualized by Western blotting in WT, Nrf1α^–∕– and Nrf2^−/− cells. The intensity of all the immunoblots was calculated and shown in Fig. S14.

(E) The protein levels of p-JNK, Nrf1, c-JUN, c-FOS, Nrf2, HO1, CDH, and SNAIL were visualized by Western blotting in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated with SP600125 for 0 h, 12 h or 24 h.

(F) ROS levels were detected by flow cytometry in WT, Nrf1α^–∕– and Nrf2^−/− cells, which had been treated by SP600125 for 0 h, 12 h, or 24 h, and then stained by DCFH for 30 min.

Such discrepant consequences in the migration of between Nrf1α^–∕– and Nrf2^−/−cell lines, although both have given rise to evident increases in their intracellular ROS levels, suggest that distinct signaling pathways are likely to be activated by different extents of ROS-caused redox stress in the two distinctive genotypic cell lines (Fig. 3C). Firstly, examinations of the expression of mitogen-activated protein kinases (MAPKs, including ERKs, JNK and p38 kinase, that play key roles in tissue homeostasis, cell proliferation, differentiation survival and migration, as well in inflammation and carcinogenesis [60]) unraveled that phosphorylated abundance of JNK was substantially augmented in Nrf1α^–∕– cells, but only modestly increased in Nrf2^−/−cells, albeit its total protein expression levels were slightly enhanced in the two cell lines (Fig. 7D, d1 & d2). Conversely, the phosphorylated ERKs and p38, as well as their upstream kinase MEK were strikingly elevated in Nrf2^−/−cells, but almost unchanged in Nrf1α^–∕– cells (Fig. 7D, d3 to d8). Their downstream c-JUN and c-FOS (two key constituents of AP1 involved in tumorigenesis) were markedly up-regulated in both Nrf1α^–∕– and Nrf2^−/−cell lines (Fig. 7D, d9 & d10). Such increased c-JUN abundance was nearly 6 folds, while the increased c-FOS was also nearly 4 folds in Nrf1α^–∕– cells, but both proteins were about 2 folds in Nrf2^−/−cells, as compared to WT controls (Fig. S14). Thereafter, it was found that HIF1α (a critical factor for hypoxic stress response) was increased in Nrf1α^–∕– cells, but almost unaffected in Nrf2^−/−cells (Fig. 7D, d11). However, the versatile p53 expression was almost abrogated in Nrf1α^–∕– cells but significantly up-regulated in Nrf2^−/−cells (Fig. 7D, d12).

Next, basal expression of key signal molecules involving the PTEN-PI3K-AKT and NF-κB pathways were examined, albeit a previous study showed that PTEN (a potent tumor suppressor, that negatively regulates the PI3K-AKT signaling to inhibit cell proliferation and survival [61]) is blocked by ROS-induced Nrf2 through miR-22 [29]. Herein, basal PTEN expression was almost completely abolished in Nrf1α^–∕– cells, but conversely remarkably increased in Nrf2^−/−cells (Fig. 7D, d13). Accordingly, both phosphorylated proteins of PI3K (p110) and AKT were significantly up-regulated in Nrf1α^–∕– cells, but rather down-regulated in Nrf2^−/−cells (Fig. 7D, d14 to d15). This finding provides a better understanding of the malignant proliferation of tumors occurring upon loss of Nrf1α. Furthermore, only the p65 subunit of NF-κB (as a key factor responsible for the inflammatory response to ROS) was only modestly up-regulated in both Nrf1α^–∕– and Nrf2^−/−cell lines (Fig. 7D, d19), but the other relevant NF-κB1 and NF-κB2 were unchanged (Fig. 7D, d17 & d18).

Lastly, treatments of WT cells with a JNK-specific inhibitor SP600125 revealed that Nrf1, Nrf2 and all relevant target genes were suppressed significantly (Fig. 7E). Similar inhibitory effects were also exerted in Nrf2^−/−cells. By contrast, although auto-phosphorylated JNK, along with c-JUN and c-FOS in Nrf1α^–∕– cells were strikingly prevented by SP600125 (Fig. 7E), Nrf2, CDH1 and CDH2 were largely unaffected by SP600125, while HO1, SNAIL1 and SNAIL1 were partially altered (e5 to e10). This implies they may be also modulated independently of the JNK-AP1 signaling, particularly in Nrf1α^–∕– cells. Intriguingly, inhibition of JNK caused a sharp increment in intracellular ROS in WT and Nrf2^−/−cells, but nearly unchanged in Nrf1α^–∕– cells (Fig. 7F), reflecting a crucial cytoprotective role of JNK and AP1 in the response to oxidative stress. Overall, loss of Nrf1α^–∕– or Nrf2^−/− leads to differential dysregulation of multi-hierarchical signaling pathways to cognate target gene networks at distinct layers, by varying extents of their endogenous oxidative stress during different cell processes (Fig. 8).

Fig. 8.

Distinct roles of Nrf1 and Nrf2 in nuclearly-controlled mitochondrial functional networks.

(A) A multi-hierarchical regulatory network monitored by Nrf1 and Nrf2 alone or together. Those protein-protein associations are determined in various ways, and thus represented by different colored edges as indicated.

(B) A comprehensive regulatory model is proposed to explain an indispensable function of Nrf1 as a redox-determining factor for mitochondrial homeostasis.

4. Discussion

Just because ROS has a ‘double-edged sword’ effect with dual characteristics of beneficial hormesis and harmful cytotoxicity to all living cells, a steady state of redox homeostasis is robustly maintained by balancing ROS production (primarily in mitochondria) and elimination by a set of antioxidant responses and detoxification protective systems. Such a certain homeodynamic range of redox threshold should be tightly controlled by redox signaling to gene regulatory networks predominantly mediated by Nrf1 and Nrf2. Despite earlier discovery of antioxidant Nrf1 than Nrf2, less is known about Nrf1 relatively to Nrf2 albeit the highly conserved Nrf1 rather than Nrf2 is essential for redox homeostasis. This is owing to the fact that abundance of Nrf2 is monitored by Keap1 as a redox sensor, enabling Nrf2 to sensitively respond to changes in intracellular redox state, whereas Nrf1 as a membrane protein seems to relatively hardly respond to the changing redox levels. However, in addition to regulation of antioxidant genes by Nrf1 that has been confirmed, the sensitivity of Nrf1 to proteasome inhibition and cholesterol changes enables it to play a pivotal role in connecting basic antioxidants to other biological processes. In this study, we have provided a brand-new view of distinct roles for Nrf1 and Nrf2 played in redox regulation by distinctive mechanisms.

4.1. Distinct roles of Nrf1 and Nrf2 in redox regulation

Herein, we found that loss of either Nrf1 or Nrf2 leads to a redox imbalance to increased levels of intracellular ROS, which occurs by distinct rational mechanisms. This is because knockout of Nrf2 results in a reduction of its cellular capability to eliminate ROS insomuch as to make a redox bias towards yield of ROS. In sharp contrast, a similar capability to eliminate ROS appears to be undiminished by the knockout of Nrf1, but conversely, the production of ROS in the mitochondria is increased strikingly in Nrf1α^–∕– cells. Consequently, Nrf2 is accumulated in Nrf1α^–∕–-deficient cells by its proteasomal dysfunction and further activated by increased ROS, leading to an evident enhancement of antioxidant, detoxifying and cytoprotective systems. In turn, the activity of Nrf1 appears to be not further reinforced by elevated ROS in Nrf2^−/− cells. This implies that Nrf1 acts as a key constitutive transcription factor of redox regulation, but it is activated predominantly by another main way rather than ROS, or the ROS activation of Nrf1 is dependent on Nrf2. Coincidently, this is also fully consistent with our previous finding that Nrf2 serves as an upstream regulator to activate the transcriptional expression of Nrf1 [29]. Of note, as an ER-anchored membrane protein, Nrf1 is de facto regulated by ER-derived unfolded proteins and/or other metabolic stressors [9], to coordinate the ER homeostasis with redox homeostasis (Fig. 8), but this is required for further experiments to be elucidated.

With an aberrant accumulation of hyperactive Nrf2 in Nrf1α^–∕– cells, most of the antioxidant and detoxifying enzymes, as well as relevant genes responsible for the elimination of ROS are markedly augmented. Many of these up-expressed genes (e.g., encoding PRDX5, PRDX6, GPX1, SOD2, SESN2, G6PD, GCLC, GCLM, GLS1, GSR, SLC6A9, FTH1, HO1, TXN1, and TXN2) are diminished or abolished by knockout of Nrf2^−/−, indicating they are mainly regulated by Nrf2. Also, it cannot be ruled out that some genes are, to certain lesser extents, regulated by Nrf1, but its overlapping effects are much more likely to be concealed by hyperactive Nrf2. For example, SRXN1 and PGD were reportedly regulated by Nrf2 [62,63], and thus down-regulated in Nrf2^−/− cells, but largely unchanged in Nrf1α^–∕– cells, implying both are co-regulated by Nrf1 and Nrf2, but the regulatory effect of Nrf1 is postulated to be offset by hyperactive Nrf2. Conversely, SESN3 was down-regulated in Nrf1α^–∕– cells, but almost completely abrogated in Nrf2^−/− cells, implying that this putative effect of Nrf1 on SESN3 was only partially counteracted by accumulated Nrf2. In addition, PRDX4, SESN1 and TIGAR are elevated in both Nrf1α^–∕– and Nrf2^−/− cell lines, suggesting they may be over-stimulated by elevated ROS.

4.2. Nrf1 serves as an indispensable redox-determining factor in the mitochondrial homeostasis

Albeit two independent groups had presumed that the loss of Nrf1 is likely to cause an impairment of mitochondrial functioning [23,46], no further experimental evidence has been provided in the current literature. Another group led by Yoon reported that Nrf1 acts as one of the most responsive genes when the mitochondrial respiratory chain is damaged [64,65]. In addition, a recent study also showed that interfering with Nrf1 in HepG2 cells caused mitochondrial changes and affected metabolic reprogramming [43]. These putative effects of Nrf1 prompt us to gain insights into its redox regulation for mitochondrial homeostasis because this organelle is the main resource of intracellular ROS production. The experimental evidence has been presented here, revealing that a dramatic increase of the mitochondrial ROS levels was manifested in Nrf1α^–∕– cells, simultaneously accompanied by significant down-regulation of key mitochondrial complex subunits upon loss of its function. However, no further increases or even modest decreases in the yield of ROS were observed after Nrf1α^–∕– cells had been treated with either the mitochondrial respiratory chain inhibitor rotenone or another uncoupling agent FCCP. Such dysfunctional mitochondria in Nrf1α^–∕– cells were damaged as evidenced by further electroscopic observations. We also observed that ROS levels in Nrf1α^–∕– cells decreased after rotenone treatment, presumably because rotenone blocks electrons being transferred to complex I from complex II via coenzyme Q, thereby reducing generation of ROS, which also indicates high electron transport stress on mitochondrial complex I in Nrf1α^–∕– cells. As such, Nrf2 is also inferred to be partially involved in mitochondrial redox stress response, because rotenone treatment of Nrf2^−/− cells only modestly increased ROS generation to certain extents that are not higher than WT controls, meanwhile the reduction of glycolysis caused by Nrf2 knockout also reduces mitochondrial fuel loading. But, remarkable changes in its ROS occurred after FCCP treatment of Nrf2^−/− cells, implying that Nrf1 remains to be expressed to a considerably high degree in Nrf2^−/− cells so that deregulated mitochondria were further stressed by FCCP insomuch as to trigger excessive ROS products. In addition, it should also be noted that the basal metabolism of Nrf2^−/− cells is likely placed at a rather lower level, so to enable these deficient lines to make more resistance to glucose starvation, as reported previously [35].

It is of crucial importance to discover that loss of Nrf1α results in a defective down-regulation or even abolishment of the nuclearly-controlled mitochondrial respiratory factors involving the PGC1α-αPal^NRF1 and PGC1α-GABPα^NRF2 pathways in Nrf1α^–∕– cells. Of note, PGC1α is a key transcriptional coactivator enabling for activation of estrogen-related receptors, peroxisome proliferator-activated receptors (PPARs), and other specific partners, in addition to two nuclear respiratory factors αPal^NRF1and GABPα^NRF2 (Fig. 3, Fig. 8B). Co-activation of nuclear respiratory factors αPal^NRF1and GABPα^NRF2 by PGC1α further induces the expression of mitochondrially-specific transcriptional factors TFAM and TFBM, which can bind directly to both strands of mtDNA and thus play key roles in the mtDNA replication, transcription and maintenance [66]. In addition to regulating the expression of these mitochondrially-located genes through TFAM and/or TFBM, two nuclear respiratory factors αPal^NRF1and GABPα^NRF2 also directly activate those nuclearly-located genes involved in the oxidative respiratory chain and mitochondrial biogenesis. Herein, our experimental evidence has unraveled that overexpression of Nrf1 and Nrf2 activates the transcriptional expression of αPal^NRF1, GABPα^NRF2 and PGC1α, as well as respiratory chain subunits and that they also exert transcriptional activation effects on each of αPal^NRF1, GABPα^NRF2 and PGC1α promoter-driven reporter genes. Rather, the CHIP-sequencing data for binging Nrf1 and Nrf2 obtained from the ENCODE database showed considerably weak signals on the promoters of the above-described genes (Fig. S8). This is likely owing to less amounts of the nuclearly-imported Nrf1 and Nrf2, particularly in the normal unstressed conditions or requirements of both factors for the functional heterodimerization with their putative DNA-binding partners, as well as other cofactors. As such, our further experimental evidence obtained from Nrf1α^–∕– cells has demonstrated that Nrf1, rather than Nrf2, can serve as a vital dominant-positive determinant of mitochondrial functioning by governing transcriptional expression of αPal^NRF1, GABPα^NRF2 and PGC1α because even though hyperactive Nrf2 was accumulated, it cannot compensate for a fatal defect in basal expression of αPal^NRF1, GABPα^NRF2 and PGC1α in Nrf1α^–∕– cells. But, no decrease of PGC1β was determined in Nrf1α^–∕– cells, although it was reportedly regulated by Nrf1 in mouse cells [23]; this difference may be attributable to distinct cell types in between the human and mouse species. Moreover, when cells are required for impaired mitochondrial function, a highly-conserved mechanism called UPR^mt is activated to trigger cell repair of the mitochondrial function. Such UPR^mt-related genes were still activated by FCCP-induced mitochondrial stress in Nrf2^−/− cells but were almost unaffected in Nrf1α^–∕– cells. Thereby, it is inferable that the mitochondrial UPR^mt is prevented or even completely abolished in Nrf1α^–∕– cells, due to its fetal mitochondrial damage arising from the loss of Nrf1α’s function. This finding further indicates that Nrf1, but not Nrf2, is an indispensable dominantly-determining factor in maintaining mitochondrial homeostasis and its integrity, albeit Nrf2 is also involved in this process.

4.3. Nrf2 exerts a ‘double-edged sword’s role in the redox regulation, particularly in Nrf1α^–∕– cells

Mitochondria produce ROS mainly from escaped electrons in the transport respiratory chain, and these electrons [e^–, together with hydrogen ion (H⁺)] originated primarily from aerobic glycolysis. Herein, it is, to our surprise, found that the apparent rise and fall of the intracellular ROS levels in Nrf1α^–∕– cells are accompanied by an increase or decrease of glycolysis, respectively when they had been treated with its activator insulin or inhibitor 2-DG. This is also supported by additional observations showing that the yield of ROS in 5 mM-glucose cultured Nrf1α^–∕– cells is at a rather lower level than that arising from the 25 mM glucose cultured conditions. Similar glucose effects on the yield of ROS were also manifested in wild-type cells, but Nrf2^−/− cells appeared to make little difference. Further experiments revealed that basal expression levels of glycolysis and lipid synthesis-related genes were up-regulated in Nrf1α^–∕– cells, but also enabled to be interfered to considerably lower extents by silencing of Nrf2. Collectively, these data suggest that hyperactive Nrf2 accumulated in Nrf1α^–∕– cells has a strong ability to promote glycolysis and lipid biosynthesis insomuch as to give rise to excessive ROS from its mitochondria, particularly under the pressure of a high glycolysis flow. In addition, we found that insulin can increase the expression of Nrf1, and 5 mM glucose cultured conditions (glucose restricted medium) reduce the expression of Nrf1 (Figs. S10C and S10F). When cells are treated with high concentrations of glycolysis inhibitors (2-DG) or starved in a non-glucose medium, glycosylated Nrf1 decrease and the active isoform of Nrf1 significant increase (Fig. S10G) [54]. This also implies the relevance between Nrf1 and glucose metabolism. Perhaps under lower metabolic conditions Nrf1 is down-regulated to weaken mitochondrial-related metabolism, and under extreme glucose deficiency conditions Nrf1 is transitorily activated to enhance mitochondrial function to meet the needs of energy and promote the metabolism of non-sugar substances such as lipids, but that requires more experiments to prove.

Among those genes possibly mediated by Nrf2, PDH plays a role in controlling the entry of pyruvate produced by glycolysis into mitochondria, particularly when PDH is elevated concomitantly with unchanged LDHA, in Nrf1α^–∕– cells, making the flue flow produced by glycolysis more easily enter the mitochondria. However, fatal damage to the electron transport chain slows down the TCA cycle [67], it is hence difficult for the impaired TCA cycle to consume acetyl-CoA that enters the mitochondria in Nrf1α^–∕– cells. The resulting increase in fatty acid biosynthesis provoked by hyperactive Nrf2 provides an alternative way to consume the metabolites from the increased glycolysis flow. This enables to allow the citric acid produced by oxaloacetate and acetyl-CoA to skip most of the damaged TCA cycle and then convert it into apple acids. Overall, deficiency of Nrf1α cannot only lead to a marked increase in intracellular ROS yield from its defective mitochondria, and also augment glucose consumption and lipid biosynthesis by increasing Nrf2 to aggravate ROS production. Such incremented lipid biosynthesis may be explained as a reasonable cause of lipid deposition in Nrf1^–∕–cells, ultimately resulting in the spontaneous development of non-alcoholic steatohepatitis and ensuing hepatoma in liver-specific Nrf1^–∕– mice [27,29].

Notably, a direct main source of ROS arises from the electron escape of NADH produced by cellular metabolism in the respiratory chain. Such escaped electrons enable exacerbation of the mitochondrial membrane potential (Δψm) to a rather higher degree due to the increasingly sluggish electron transport within a longer half-life of relevant respiratory chain intermediates [68]. This is distinctive from destructive uncoupling, because a modest decrease in Δψm, called ‘mild uncoupling’, has been shown to exert a cytoprotective effect [69]. Thereby, we also found that UCP2 is negatively regulated by hyperactive Nrf2 in Nrf1α^–∕– cells to increase the intracellular ROS production, even though Nrf2 has been generally accepted as a master regulator of the antioxidant and detoxifying system and also as a key player in metabolic regulation. When required for cellular demands, UCP2 is widely expressed, so that it cannot only relieve the pressure of the electron transport chain and also reduce byproducts of ROS from mitochondria, but the resulting production of ATP is decreased concomitantly by reducing Δψm. Thus, it is inferable that Nrf2 is selectively allowed for inhibition of UCP2 to facilitate the energetic production of ATP, albeit with an accompanying increase in the mitochondrial ROS byproducts, particularly in Nrf1α^–∕– cells. Conversely, when UCP2 was over-expressed, the yield of ROS in Nrf1α^–∕– cells was greatly reduced, but in Nrf2^−/− cells was unchanged similarly to WT controls, implying that UCP2 is also selectively contributable to antioxidant cytoprotection. Further experiments unraveled that UCP2 is negatively regulated by Nrf2 through ARE-driven miR-195 and/or miR-497. The expression levels of miR-195 and miR-497 were up-regulated by hyperactive Nrf2 in Nrf1α^–∕– cells, even albeit both can also be regulated directly by Nrf1. In addition, Kuosmanen et al. identified 116 novel Nrf2-targeted miRNAs by CHIP-sequencing, including miR195 and miR497 [70], as fully consistent with our results. Altogether, these demonstrate that Nrf2 cannot only exert its intrinsic antioxidant effect but also promote the generation of ROS within proper tempo-spatial contexts. And it also indicates that Nrf2 does not only play a role in the antioxidant, but also, more importantly, regulates the balance between production of ROS and their elimination, and even another balance existing between ROS elimination and energy production, albeit this requires further investigation.

4.4. Inter-regulatory effects of Nrf1 and Nrf2 are integrated by multiple signaling to gene expression networks

Several studies have uncovered that Nrf2 acts as a versatile chameleon-like player in both cancer prevention and promotion [71], but its oncogenic activity to promote cancer development is tightly confined by Nrf1 (along with its long isoform TCF11), which is endowed as a potent tumor-repressor [26,29,30]. Such oppositely inter-regulatory effects of Nrf1 and Nrf2 are unified integrally by multiple redox signaling pathways to cognate gene expression networks at distinct levels (Fig. 8A). These multi-hierarchical signaling networks gradually converged on the base of redox regulation by Nrf1 and Nrf2 (Fig. 8B). Of note, the primary ROS-caused stress and secondary oxidative damages are well known to serve as an initiator of cancer development and also as a promotor of cancer progression and malignance [72]. However, our evidence revealed that significantly elevated ROS in Nrf2^−/− cells does not render its migration ability and growth rate to be increased, but conversely becomes weaker than WT controls. Further xenograft animal experiments by inoculating Nrf2^−/− cells unraveled that loss of Nrf2’s function leads to a striking reduction in vivo malgrowth of hepatoma and its metastasis [29]. These indicate that loss of Nrf2-mediated antioxidant and detoxifying cytoprotection results in certain ‘mild extents’ of ROS that are not enough to sufficiently initiate cancer development or promote malignant progression, but conversely can serve as eustress to stimulate beneficial hormesis effects.

By contrast with Nrf2^−/− cells, severe extents of ROS-caused distress and damages are determined in Nrf1α^–∕– cells, although accompanied by hyperactive Nrf2 accumulation and reinforced antioxidant response. Such oxidative distress is attributable primarily to Nrf1α^–∕–-leading mitochondrial dysfunction and its UPR^mt failure, and secondarily to Nrf2-augmenting the yield of ROS by strengthening glycolysis (resembling the Warburg’s effect) to aggravate the pressure of electron respiratory chain and also by inhibiting UCP2 to enable for electron escape from the transport respiratory chain. Consequently, loss of Nrf1α^–∕–results in marked increases in its malgrowth, migration and metastasis, as well in the EMT process [29,73]; such increased ability was also greatly prevented by silencing of Nrf2, in addition to the restored expression of Nrf1α or TCF11 [29,30]. These indicate that Nrf2 acts as a tumor-promoter to exert its oncogenic activity, only when Nrf1α/TCF11 is disrupted, whereas the latter Nrf1α/TCF11 is likely to possess an intrinsic capability to repress initiation of cancer development. This cancer-repressing effect may also be executed by two additional ‘star’ tumor-repressors PTEN and p53 because both were diminished in Nrf1α^–∕– cells. Besides, the EMT-relevant signaling, together with the MEK-MAPKs (i.e., JNK, ERKs, p38 kinase)-AP1 (JUN + FOS) and PI3K-AKT, as well as HIF1α and NF-κB, signaling networks are constitutively activated in Nrf1α^–∕– cells. In addition, the redox metabolic reprogramming of Nrf1α^–∕– cells (owing to dysfunctional mitochondria) may be contributable to cancer development and progression [35]. Contrarily, most cell killing results from the further elevated yield of ROS after Nrf1α^–∕– cells were stimulated by glucose deprivation. Overall, Nrf1 acts as a potent integrator of the cellular redox regulation by multi-hierarchical signaling to gene expression networks to maintain cell homeostasis and organ integrity.

In summary, systematic examinations of distinctive roles for Nrf1 and Nrf2 in redox regulation are carried out from a holistic view in a combination of reductionist approaches. The results unveiled that Nrf1 functions as an indispensable redox-determining factor for mitochondrial homeostasis because the loss of its function leads to a potentially fatal defect in dysfunctional mitochondria to give rise to severe oxidative stress along with the failure of UPR^mt, and such detriment effects cannot be counteracted by hyperactive Nrf2 accumulated in Nrf1α^–∕– cells, but conversely malgrowth of Nrf1α^–∕–-derived tumor appears to be protected by Nrf2-strengthening antioxidant response and glycolysis pathways. From this, it is inferable that Nrf2 plays a critical role in Warburg’s effect in Nrf1α^–∕– cells. The reinforced glycolysis aggravates the mitochondrial pressure to allow for the electron escape to yield excessive ROS, particularly when UCP2 is suppressed by hyperactive Nrf2. Therefore, the versatile Nrf2 exerts a ‘double-edged sword’s role in redox regulation, particularly in Nrf1α^–∕– cells. Such inter-regulatory effects of Nrf1 and Nrf2 are integrated by multi-hierarchical signaling towards gene expression networks to perpetuate cell homeostasis and organ integrity during development and growth.

Author contributions

S.H. performed most of the experiments except indicated elsewhere, made all figures, and wrote the manuscript draft. J.F. constructed relevant plasmids used in this study. M.W. performed a statistical analysis of transcriptome data. R.W. and K.L. participated in the detection of intracellular ROS and transmission electron microscopy. Z.Z. participated in drawing some pictures and edited the manuscript draft in English. Y.Z. designed and supervised this study, analyzed all the data, helped to prepare all figures with cartoons, and wrote and revised the paper.

Declaration of competing interest

The authors declare no conflict of interest. Besides, it should be noted that the preprinted version of this paper had been initially posted at doi: https://doi.org/10.1101/2022.05.04.490622.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

No data was used for the research described in the article.