When NRF2 Talks, Who's Listening?

Introduction

NRF2 (NF-E2-related factor 2, NFE2L2) is a transcription factor that mediates a broad-based set of adaptive responses to intrinsic and extrinsic cellular stresses (73). As such, NRF2 influences sensitivity to physiologic and pathologic processes affected by oxidative and electrophilic stresses, such as those imposed by inflammation and exposures to environmental toxicants. Carcinogenesis, chronic obstructive pulmonary disease, obesogenesis, and neurodegeneration are known to be affected by the Nrf2 genotype in murine models (73).

Molecular details of the NRF2 signaling pathway have emerged over the past decade, as reviewed elsewhere in this issue. Notably, definition of the interacting partners of NRF2, such as KEAP1 (Kelch-like ECH-associated protein 1), and the means by which they sense and transduce chemical signals of stress are under intense investigation in many laboratories (48). Characterization of the immediate downstream target genes of NRF2 has also been well enumerated, although features that define the tissue and cell-type specificity in responses are poorly understood. The gene-expression responses to activation of NRF2 signaling, mediated through interactions of NRF2 and the antioxidant response element(s) (ARE: 5′-NTGAG/CNNNGC-3′) in the regulatory domains of its target genes, result in a distinctive cytoprotective response. Beyond the now-seminal response of catalyzing the detoxification of carcinogens and other xenobiotics through conjugation and trapping processes, genomic analyses indicated that gene families affected by Nrf2 (a) provide direct antioxidants; (b) encode enzymes that directly inactivate oxidants; (c) increase levels of glutathione synthesis and regeneration; (d) stimulate NADPH synthesis; (e) enhance toxin export through the multidrug-response transporters; (f ) enhance the recognition, repair, and removal of damaged proteins; (g) elevate nucleotide excision repair; (h) regulate expression of other transcription factors, growth factors and receptors, and molecular chaperones; and (i) inhibit cytokine-mediated inflammation (48, 73).

It also is clear, though, that these direct genomic responses do not account for all of the actions of pharmacologic agents that activate NRF2 signaling, nor do they explain the range of stress-response phenotypes observed when components of the pathway, notably, Nrf2 or Keap1, have been genetically disrupted in mice (162). Analyses of global transcriptional responses have hinted toward interactions between NRF2 signaling and other prominent signaling pathways. These interactions can occur in multiple forms. Post-translational modifications such as phosphorylation play a major role in the regulation of gene expression and function. These covalent modifications control intracellular distribution, transcriptional activity, and stability of transcription factors, including NRF2 (81, 143). Some transcription factors can antagonize NRF2 either by competing for binding to AREs or by inhibiting NRF2 through a physical association. Small MAF proteins, BACH1, and the immediate early proteins c-FOS and FRA1 can compete with NRF2 for binding to AREs (103, 151). Furthermore, several transcription factors, including activating transcription factor 3, proliferator-activated receptor (PPAR)γ, and retinoic acid receptor α have been reported to form inhibitory complexes with NRF2 (6, 13, 59, 155, 168). In this review, we focus on recent evidence for transcriptional cross-talk between NRF2 and the arylhydrocarbon receptor (AhR), NF-κB, p53, and Notch pathways, as summarized schematically in Fig. 1.

NRF2 Interactions with AhR Signaling

The aryl hydrocarbon receptor is a ligand-activated member of the bHLH/PAS (basic helix-loop-helix/Per-Arnt-Sim) family of transcription factors that mediates the biologic and toxic effects of its xenobiotic ligands (44). When bound by polycyclic aromatic hydrocarbons such as dioxins (TCDD), AhR translocates from the cytoplasm to the nucleus, heterodimerizes with AhR nuclear translocator (ARNT), and activates transcription through the xenobiotic-responsive element (XRE). The XRE consists of a canonic motif of 5′-TNGCGTG-3′. After nuclear export, AhR is degraded through the 26S proteasome pathway. It is well recognized that both AhR and NRF2 signaling regulate the expression of genes affecting the metabolism of xenobiotics. In some instances, the response elements recognizing these transcription factors can be found in the regulatory domains of the same target genes, such as Nqo1 (104). Ma et al. (95) reported that the inducible expression of NQO1 by the potent AhR ligand TCDD depends on both AhR and NRF2. Genetic experiments using Ahr-, Arnt-, or Nrf2-deficient cells revealed that induction of NQO1 by TCDD depended on the presence of AhR and ARNT, whereas the basal and inducible expression required a functional NRF2 pathway. Yaeger et al. (163) extended these findings by using Nrf2-deficient mice to demonstrate the requirement of NRF2 for the induction of Nqo1, Ugt1a6, and Gsta1 transcripts, among other genes, by TCDD in mouse liver. AhR–Nrf2 compound null mutant mice respond to neither AhR ligands nor prototypical NRF2 activators with target genes, such as Nqo1 (105). A functional collaboration between these pathways is also highlighted by the earlier findings of Talalay and colleagues (115), in which they described the roles of “monofunctional” and “bifunctional” activators of NRF2 signaling. “Monofunctional” inducers affect only NRF2 signaling. “Bifunctional inducers,” such as β-napthoflavone, may interact with the AhR directly to activate AhR-regulated genes, such as Cyp1a1, Cyp1a2, and Cyp1b1, and then undergo transformation by these enzymes to reactive intermediates that then trigger NRF2 signaling (see Fig. 2). In this way, a comprehensive set of responses to xenobiotic challenges can be mounted (79).

The mechanisms underlying the linkages between AhR and NRF2 signaling need further investigation. Miao et al. (98) demonstrated that Nrf2 gene transcription is directly modulated by AhR activation. DNA sequence analyses of the murine Nrf2 promoter revealed an XRE located at −712 bp and two additional XRE-like motifs located further upstream. Direct binding of AhR to the Nrf2 promoter was observed. Moreover, silencing of AhR expression with siRNA obviated induction of Nrf2 mRNA by TCDD. NRF2 also autoregulates its own expression through an ARE-like element located in the proximal region of its promoter, leading to persistent nuclear accumulation of NRF2 and protracted induction of its target genes (83).

NRF2 also directly modulates AhR signaling, highlighting bidirectional interactions of these pathways (138). Constitutive expression of AhR was affected by Nrf2 genotype. Moreover, a pharmacologic activator of NRF2 signaling, CDDO-Im {1-[2-cyano-3,12-dioxooleana-1,9(11)-dien-28-oyl]imidazole}, induced Ahr, Cyp1a1, and Cyp1b1 transcription in Nrf2^+/+ mouse embryonic fibroblasts (MEFs) but not in Nrf2^-/- MEFs. Reporter analysis and chromatin immunoprecipitation assays revealed that NRF2 directly binds to one ARE found in the −230 bp region of the promoter of Ahr. Better understanding of this bidirectional regulation of AhR and NRF2 pathways should provide clues on the roles of NRF2 in complex diseases and in uncharacterized phenotypes.

Cell-culture studies and knockout mouse models have shown that, in addition to controlling the xenobiotic detoxification response, AhR activation leads to G₀/G₁ arrest, diminished capacity for DNA replication, inhibition of cell proliferation, and impaired cell differentiation. AhR may function also as a tumor-suppressor gene that becomes silenced during the process of tumor formation (32). Transcriptional responses to ligand-dependent activation of AhR signaling vary across cell types and by species, but consistently influence cytochrome P450 and other monooxygenase activities (17). Ligand-dependent and -independent AhR signaling influences (a) positive regulation of transcription, vascular development, organ morphogenesis, and metabolic processes; (b) processes related to mRNA processing and transport; (c) processes related to cell death, proliferation, and regulation of apoptosis; (d) mitotic cell cycle; and (e) embryonic and other developmental processes (132).

Ahr-null mice and Nrf2-null mice share some common phenotypes. These transcription factors appear to function as cell-survival factors; both types of knockout mice exhibit increased sensitivity to infection (75, 147), carcinogenesis (32, 72, 117), altered responses to wounding (12, 18), and metabolic syndrome (137, 144). Studies on responses to adipogenic stimuli in vitro and high-fat stress diets in vivo have provided some opportunities for evaluating the physiological interactions between these pathways.

Shimba et al. (136) observed that TCDD treatment suppresses the conversion of 3T3-L1 cells into adipocytes. By using MEF cell lines derived from Ahr-knockout mice, Alexander et al. (5) reported that the AhR is a constitutive inhibitor of triglyceride synthesis and an early regulator of adipocyte differentiation. In addition, one of the phenotypes of Ahr-null mice is transient fatty liver (133), implying an in vivo regulatory role of AhR in the adipogenic process. Shin et al. (138) postulated that NRF2 would inhibit adipogenesis through the interaction with the AhR pathway. Nrf2^-/- MEFs showed markedly accelerated adipogenesis on stimulation, whereas Keap1^-/- MEFs (which exhibit higher NRF2 signaling) differentiated slowly compared with their congenic wild-type MEFs. Ectopic expression of AhR and dominant-positive Nrf2 in Nrf2^-/- MEFs also substantially delayed differentiation. To evaluate how NRF2 regulates adipogenesis, differentiation markers involved at multiple stages of adipogenesis were analyzed. CEBPs (CCAAT-enhancer–binding proteins) and PPARs are the two families of transcription factors that play critical roles in adipogenesis (159). CEBPβ and CEBPδ function at an early phase of the differentiation process by sensing adipogenic stimuli and initiating expression of CEBPα and PPARγ (122). CEBPα and PPARγ play roles at a later stage by inducing and maintaining expression of adipocyte-specific genes, such as Fabp4. The elevation of CEBPβ on adipogenic stimulation is transient, whereas CEBPα and PPARγ remain upregulated for the duration of adipogenesis. Although the mechanism by which AhR regulates adipogenesis has not been fully characterized, recent work suggests that AhR affects differentiation stages that follow CEBPβ activation (i.e., CEBPα or PPARγ upregulation. In 3T3-L1 preadipocytes, forced expression of AhR resulted in lower induction of Fabp4 and Cebpα upon differentiation than did that in control cells, whereas the induction of Pparγ was not affected (136). Pparγ expression could be induced by differentiation in Ahr^-/- MEFs, but levels were lower than those in Ahr^+/+ MEFs (5). Similarly, mRNA levels of Cebpα and Fabp4 were higher in Nrf2^-/- MEFs than in Nrf2^+/+ MEFs, both before and after differentiation, whereas induction of Pparγ2 was not affected by the Nrf2 genotype. mRNA levels of Cebpβ were lower in Nrf2^-/- MEFs than in Nrf2^+/+ MEFs. In primary Keap1^-/- MEFs, disruption of Keap1 resulted in minimal induction of Cebpα, Fabp4, and Pparγ2 upon differentiation. Cebpβ mRNA levels remained higher in Keap1^-/- MEFs than in Keap1^+/+ MEFs, both before and after induction of adipogenesis. Thus, NRF2 negatively modulates expression of Cebpα and Pparγ2 but not Cebpβ during the course of adipogenesis. The stages of differentiation that are affected by NRF2 directly overlap with those affected by AhR, thereby supporting the hypothesis that NRF2 inhibits adipogenesis through cross talk with AhR signaling (see Fig. 3).

Treatment with CDDO-Im effectively prevented high-fat diet–induced increases in body weight, adipose mass, and hepatic lipid accumulation in Nrf2 wild-type mice but not in Nrf2-disrupted mice (137). Wild-type mice on a high-fat diet and treated with CDDO-Im exhibited higher oxygen consumption and energy expenditure than did vehicle-treated mice, whereas food intake was lower in CDDO-Im–treated than in vehicle-treated mice. Levels of gene transcripts for fatty acid synthesis enzymes were downregulated after CDDO-Im treatment in the liver of wild-type mice. This inhibitory effect of CDDO-Im on lipogenic gene expression was significantly reduced in Nrf2-disrupted mice. However, the extent to which this pharmacologic activation of Nrf2 to produce an anti-obesogenic effect is mediated through AhR, if at all, is unclear.

As with many transcription factors, the actions of the direct target genes of AhR alone do not fully explain its toxicologic and physiologic effects. In addition to interactions with NRF2 signaling, AhR extends its regulatory sweep by modulating the function of other transcription factors, including estrogen receptor (ERα and ERβ) and androgen receptor (AR). These cross-talk pathways are important mediators of the functions of endogenous and exogenous AhR ligands. The liganded AhR recently was shown to promote the ubiquitination and proteasomal degradation of ERs and AR by assembling a ubiquitin ligase complex, CUL4B^AHR (108). Thus, complexes of the AhR with ERs or AR appear to regulate transcription as functional units by multiple mechanisms. As depicted in Fig. 4, KEAP1 also serves as an adaptor protein for an E3 ubiquitin ligase complex (78). Thus, broad similarities exist in the ways these pathways sense and respond to stresses. Indeed, other stress response–signaling pathways such as HIF-1α and NF-κB also interact with ubiquitin ligase systems to facilitate proteolytic turnover of signaling components (109).

Direct and Indirect NRF2–NF-κB Interactions

Cross talk between NRF2 and nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB) is an area of extensive interest. NF-κB proteins are a family of transcription factors involved in several processes such as inflammation, immune response, apoptosis, development, and cell growth. Targets of NF-κB include genes classified as chemokines, cytokines, immunoreceptors, cell-adhesion molecules, stress-response genes, regulators of apoptosis, growth factors, and transcription factors, among many others. Tumor necrosis factor α (TNF-α), inducible NOS (iNOS), interleukin-1 (IL-1), intracellular adhesion molecule-1, and cyclooxygenase (COX-2) have all been shown to be induced by NF-κB (112). Pathologies related to alterations of the NF-κB pathway include allergies (25), Alzheimer disease (97), autoimmunity (46), obesity (40), atherosclerosis (125), arthritis (123), cancer (41), Crohn's disease (114), diabetes (164), and stroke (51).

All NF-κB proteins contain a conserved Rel homology domain responsible for dimerization and DNA binding to the consensus sequence GGGRNNYYCC, (R = purine, N = any base, Y = pyrimidine). The NF-κB family of proteins can be divided into two distinct groups based on the presence of a transactivation domain. RelA (p65), RelB, and c-Rel all contain transactivation domains, whereas p50 and p52 do not, and they require heterodimerization with the Rel proteins to activate transcription. Inactive NF-κB is located in the cytoplasm associated with the negative regulator IκB, which can conceal the nuclear-localization sequence or DNA-binding domain of NF-κB to prevent transcription (148, 157, 165). Activation of NF-κB is accomplished through phosphorylation of IκB by IκB kinases (IKKs), which leads to the release and nuclear translocation of NF-κB (22, 28, 150). A wide range of agents have the ability to stimulate IKK to activate NF-κB, including H₂O₂, TNF-α, IL-1, phorbol esters, hypoxia followed by reoxygenation, ultraviolet radiation, or microbial infection (112).

The NRF2 and NF-κB signaling pathways interface at several points to control the transcription or function of downstream target proteins. Many examples exist in which activation and repression occur between members of the two pathways through mechanisms of regulation ranging from direct effects on the transcription factors themselves to protein–protein interactions and second-messenger effects on target genes. Several cancer chemopreventive agents trigger NRF2 signaling with a concomitant repression of NF-κB and its target genes. As examples, sulforaphane [(-)-1-isothiocyanato-(4R)-methylsulfinyl)butane] reduces DNA binding of NF-κB and decreases generation of NO, PGE₂, and TNF-α in Raw 264.7 macrophages without affecting IκB or nuclear translocation of the transcription factor (50). 3H-1,2-Dithiole-3-thione reduces the nuclear translocation and DNA binding of NF-κB, along with changes in phosphorylation of IκB in the hepatocytes of lipopolysaccharide (LPS)-treated rats (69). Epigallocatechin-3-gallate induces NRF2 and reduces NF-κB, TNF-α and IL-1β in the lungs of bleomycin-treated rats (140). Chalcone has been shown to induce NRF2 and inhibit the activation of NF-κB in endothelial cells (94). The synthetic triterpenoid CDDO-Me can inhibit NF-κB activity through repression of IKKβ (1). Curcumin is able to inhibit NF-κB by blocking phosphorylation and degradation of IκB in macrophages challenged with LPS (113).

NRF2 and NF-κB can interact to have direct effects on gene expression. NRF2 and NF-κB mediate the activation of IκB and ARE sites, respectively, through antagonism of transcription-factor binding to DNA. NF-κB was recently shown to prevent the transcription of NRF2-dependent genes by reducing available co-activator levels and promoting recruitment of a co-repressor. p65 and Nrf2 both bind to the CH1-KIX domain of CREB-binding protein (CBP), and after phosphorylation of p65 at Ser276, NF-κB is able to suppress transcription of ARE-dependent genes by preventing CBP from binding to Nrf2 (93). A second mechanism of p65 transcriptional repression of the ARE involving histone deacetylase 3 (HDAC3) has also been described. Overexpression of p65 causes the recruitment of HDAC3 to the ARE by binding to CBP or MafK. HDAC3 was shown to bind MafK in the NRF2 dimerization region and to prevent the acetylation of MafK by CBP (93). Further experiments showed that p65 is able to affect NRF2-dependent transcription through the recruitment of HDAC3 to the ARE by binding to CBP. HDAC3 was previously shown to bind and deacetylate CBP, resulting in reduced CBP-mediated transcriptional coactivation (24). A model for effects of reduced coactivator and recruitment of a co-repressor is depicted in Fig. 5. Phosphorylation of p65 on Ser276 reduces the available coactivator levels of CBP from NRF2, which causes decreased transcription through the ARE and also allows HDAC3 to bind CBP. The loss of CBP from the MafK-NRF2 heterodimer may then cause destabilization and allow HDAC3 to bind and deacetylate MafK. The HDAC could be then be recruited to the ARE through the action of the CBP-bound HDAC or a MafK homodimer–bound HDAC3 (93). Interestingly, this mechanism may be countered by induction of NRF2, which leads to increased cytoplasmic-to-nuclear translocation of the transcription factor. Experiments have shown that repression of NRF2-dependent transcription was eliminated by transfection of a dominant negative NRF2 that does not undergo cytoplasmic sequestration (93).

Several experiments have shown increased NF-κB activation in Nrf2^-/- mice when compared with wild-type after stimuli such as traumatic brain injury (65), LPS (147), TNF-α (147), ovalbumin (119), and respiratory syncytial virus (23). NRF2 has been implicated in NF-κB control through attenuation of phosphorylated IκB, which causes NF-κB degradation. Nrf2 ^-/- MEFs have higher levels of phosphorylated IκB and greater IKK kinase activity when compared with Nrf2 ^+/+ MEFs after challenge with LPS or TNF-α (147). Interestingly, overexpression of NRF2 seems to change only downstream NF-κB targets, not NF-κB or IκB (21, 68, 90). Overexpression of NRF2 in human aortic endothelial cells by using an adenoviral vector did not inhibit TNF-α–induced NF-κB activity or IκB degradation, but induction of monocyte chemoattractant protein-1 and VCAM-1 with TNF-α were inhibited (20). These inconsistencies could be due to the loss of NRF2, creating an environment with a diminished capacity to scavenge reactive oxygen species (ROS). The impaired expression of antioxidative genes could potentially lead to increased activation of the NF-κB pathway. Alternatively, enhancement of NF-κB due to direct transcriptional action of NRF2 may occur. In genetic studies, Nrf2^-/- fibroblasts have shown reduced steady-state protein levels of p50 and p65 when compared with wild-type, whereas in vivo studies revealed that livers of Nrf2^-/- mice exhibited greatly reduced transcript levels of p65 (161). By contrast, TNF-α or LPS stimulation of Nrf2^-/- fibroblasts caused increased activation of NF-κB (147). Enhancement of NF-κB by NRF2 may be coordinated in a subset of the NF-κB family members to drive expression of specific genes. In Nrf2^-/- fibroblasts p50 and p65 show reduced steady-state protein levels to 30% of wild type, whereas c-Rel protein levels were increased 400% (161). It is unknown whether NRF2 is able specifically to control expression of NF-κB proteins that lack a transactivation domain to repress transcription of NF-κB target proteins (88, 91). The enhancement of NRF2 due to direct transcriptional action of NF-κB may also be possible. In silico transcription factor–binding site analysis has identified an NF-κB consensus binding site in the murine promoter of NRF2 (102), which may increase transcription of NRF2 to blunt prolonged NF-κB activity. As multiple agents and stimuli such as ROS, LPS, flow shear stress, oxidized low-density lipoprotein, and cigarette smoke have been shown to induce both NRF2 and NF-κB activity (2, 7, 16, 42, 56, 77, 127), a completely antagonistic effect of the two transcription factors against one another in all situations is unlikely.

Interactions between downstream targets of NRF2 and NF-κB that ultimately lead to modulation of transcription factor activity have been discovered, which adds complexity to the two pathways. Three examples of NRF2 target genes that are able to influence NF-κB activity are heme oxygenase-1 (HO-1), NAD(P)H:quinone oxidoreductase (NQO1), and thioredoxin (TRX). HO-1, which breaks down heme to produce biliverdin, CO, and free iron, is able to modulate activation of NF-κB through the action of bilirubin and free iron (3, 66, 111, 135, 139, 145) (see Fig. 6). HO-1 has been shown specifically to prevent phosphorylation of RelA through the action of free iron at a site required for TNF-α–dependent NF-κB activation (135). Induction of HO-1 can inhibit IκB degradation, whereas inhibition of HO-1 increases p65 activity in HT-29 cells and colonic mucosa after treatment with TNF-α or IL-1ß (66). NQO1 has both positive and negative effects on NF-κB. NQO1 overexpression reduces LPS-induced expression of the downstream inflammatory genes TNF-α and IL-1 in THP-1 cells independent of NF-κB (128). Mice deficient of NQO1 in bone marrow, spleen, and thymus showed reduced NF-κB DNA binding under basal or LPS-treated conditions (61). TRX also has both positive and negative effects on NF-κB. TRX can modulate NF-κB activity through the reduction of cysteine residues (26, 36, 47, 74, 96, 116). Nuclear TRX is required for the reduction of cysteine residues on p50 to allow binding to DNA, whereas cytoplasmic TRX is able to block the degradation of IκB (54).

Target proteins of the NRF2 and NF-κB pathways and their products can also interact to regulate enzyme activity. The NF-κB target IL-10 (15) has been shown to induce HO-1 in murine macrophages and in vivo (87). A feedback loop is present between HO-1 and iNOS in macrophages to protect the cell against injury due to excess NO. Expression of iNOS, which leads to overproduction of NO, causes induction of HO-1 (34, 60). HO-1 is then able to inhibit iNOS through the action of free iron (156) and CO (131), as well as through the overall reduction of heme (4) to prevent further production of NO (141). Interestingly, NO can be scavenged by biliverdin and bilirubin to further entangle the signaling pathways (71) (Fig. 6). HO-1 has also been shown negatively to affect VCAM-1 expression (9). Additional interactions between HO-1 and the NF-κB pathway can be seen with the HO-1 product CO. CO can repress the formation of TNF-α, IL-1β, and macrophage inflammatory protein-1β and increase IL-10 after LPS stimulation (110).

An NF-κB target that has been shown to influence NRF2 activity is COX-2 and its downstream product 15-deoxy Δ(12,14) prostaglandin J2 (15d-PGJ2). Induction of COX-2 by shear stress has been shown to inhibit phosphatidylinositol 3-kinase activity, causing reduction of the transcription of NRF2, NQO1, HO-1, GCLR, and GSTM1 in human chondrocytes. Restoration of NRF2-dependent induction of the genes was seen on treatment with the COX-2–specific inhibitors CAY10404 and NS398 (49). 15d-PGJ2 has been shown to play a role in NRF2 modulation by binding to KEAP1 and causing nuclear translocation of NRF2 in macrophages (63). Further evidence of a direct effect of 15d-PGJ2 on NRF2 activation was found in vivo by using Nrf2^-/- mice and the COX-2–selective inhibitor NS-398 (99). 15d-PGJ2 has also been shown to provide negative feedback through its own pathway by binding IKK to prevent the activation of NF-κB in HeLa cells and appears to do the same in human peripheral blood monocytes (126).

NRF2 and NF-κB have a subset of target genes that are regulated through the actions of both transcription factors on ARE and κB sites. HO-1 has been shown to have a functional ARE that can be activated by NRF2 (3), as well as a functional NF-κB site, by using EMSA and DNase I footprinting analysis (86). Glutamate-cysteine ligase catalytic subunit (GCLC) can be activated by NRF2 through an ARE (101) or by NF-κB through a κB site after ethanol exposure (76). The inhibitory G protein, Gαi2, is another example of a gene activated by both NF-κB and NRF2 (8). Expression of the chemokine IL-8 can be increased by both NRF2 and NF-κB, but not through the usual mechanisms. NF-κB activates transcription of IL-8 by binding to the κB site (82), whereas NRF2 instead uses a mechanism that stabilizes IL-8 mRNA (166).

p53 and NRF2: Cooperation and Antagonism in Protection Against Cancer

The examination of p53 after its discovery in 1979 (92) has been extensive, with approximately 52,000 articles accessioned when “p53” is searched on PubMed. This investigative interest is not without reason, inasmuch as p53 is mutated in upward of 50% of human cancers through single, compound, or deletion mutations (55). This common frequency of mutations makes p53 an attractive target for chemotherapeutics. If p53 is to be targeted, however, the body of knowledge surrounding the protein should be expanded to provide full understanding of the interactions of p53 with other signaling molecules and networks.

p53 regulates a host of processes within the cell, both directly, including the induction of apoptosis through the permeabilization of the outer mitochondrial membrane through interaction with proapoptotic factors (43, 89, 100), and indirectly, through the transcription of genes that effect changes on the cell. Examples include both pro- and antiapoptotic responses, the induction of cellular senescence, and the repair of genotoxic damage (39). These various activities of p53 make it an integral part of the cellular defense against transformation: p53 controls cell fates by regulating the activation of apoptosis to remove a heavily damaged cell, induces terminal differentiation to remove the threat of a moderately damaged cell, or induces the transcription of cellular scavenging proteins to reestablish cellular homeostasis.

Iida et al. (58) showed that Nrf2 and p53 collaborate to protect against carcinogenesis (58), with the demonstrations that, compared with wild-type mice, either Nrf2 knockout mice or p53 heterozygotic mice are more susceptible to nitrosamine-induced bladder carcinogenesis, and that the Nrf2^-/-::p53^+/- compound knockout mice are even more susceptible. This interaction is not unexpected, as when either the cytoprotective response system of NRF2, or p53, a cell-fate decision molecule is removed, the incidence of cancer increases after carcinogen challenge (117, 146). Although broad explanations based on general understandings of pathways can be posited, the actual interactions between NRF2 and p53 in this setting and the collective effects on the expression and function of their downstream target genes remain to be established.

In a more-direct examination of the interaction between p53 and NRF2, p53 has been shown to influence NRF2-based transcription by Cimino and colleagues (33). Specifically, p53 suppresses the transcription of three NRF2 target genes driven through the ARE. These genes, x-ct, Nqo1, and Gst1α1, are all involved in mounting an antioxidant response within the cell, specifically neutralizing ROS. In the study, various cell lines were transfected with expression vectors containing Nrf2 or p53. In all circumstances, the transformation with both Nrf2 and p53 results in a reduction of transcript levels of the target genes when compared with cells transformed with Nrf2 alone. This, in combination with experiments in which cells were exposed to etoposide (a topoisomerase II–inhibiting chemotherapeutic drug) also shows that p53 has the ability to regulate negatively the NRF2 target gene transcripts. When p53 is highly upregulated, through either overexpression or a strong DNA-damaging agent, antioxidant response genes are downregulated through p53-mediated disruption. Whether or to what extent NRF2 regulates p53 signaling is not clear. Transcriptome analyses, along with unpublished observations (M.K. Kwak and N. Wakabayashi), suggest that NRF2 contributes to the basal expression of MDM2, an inhibitor of p53 (85). The potential additional level of p53 regulation through NRF2 allows the dampening of MDM2-based proteasomal degradation of p53 in circumstances in which p53 may be modulating the expression of NRF2 target genes. Dampening of Mdm2 would allow an enhanced p53 signal. This effect makes intuitive sense, as if p53 were mediating an apoptotic response through ROS production, NRF2-driven gene expression of anti-ROS enzymes would be counterproductive. This observation, however, is confounded by a recent article examining the relation between p21, a direct downstream target of p53, and NRF2 (19).

p21 is involved with many different cellular processes, including cell-cycle arrest, cell differentiation, senescence, apoptosis and DNA replication and repair, and oxidative stress (29, 31, 38, 86, 107). Chen et al. (19) recently showed a direct interaction between p21 and NRF2, which may explain the cytoprotective properties of p21 when faced with oxidative stress. These investigators demonstrated in vitro that p21 is able to interact with the DLG motif within NRF2, thereby attenuating KEAP1-based ubiquitination and subsequent proteasomal degradation. The antioxidative properties of p21 rely on the presence of NRF2, as an increase in p21 concentration leads to increased transcript levels of ARE-regulated genes, and the presence of p21 within cells yields an increase in the half-life of NRF2. Additionally, it was observed in vivo that both basal and induced levels of NRF2 protein are higher in p21 wild-type mice than in p21-null mice, with induced levels of NRF2 target gene proteins NQO1 and HO-1 being higher in p21 wild-type mice. This discovery contributes significantly to appreciation of the emerging network of signaling between p53 and NRF2 by demonstrating that a direct target of p53 upregulates NRF2 levels and activity.

The three puzzle pieces presented here come together in an intricate fashion, demonstrating a tunable response by p53 to induction by any number of stimuli, as shown in Fig. 7. When p53 is strongly induced, the induction of apoptosis through the expression of proapoptotic proteins through the generation of intracellular ROS is secured by the suppression of antioxidant gene expression regulated by Nrf2, as well as the suppression of basal levels of Mdm2, to ensure the strongest possible cell-death signal both through direct levels of p53 and by its direct activities remaining uncompromised. The upregulation of proapoptotic factors by p53 may rely on increased levels of intracellular ROS production for signaling. In this setting, ROS is not neutralized by NRF2 target genes such as Nqo1 and Gst1α1 because of transcriptional interference by p53 (33), and the proteasomal degradation of p53 is hampered because of the reduction in MDM2 levels.

On the other extreme, in the face of weak p53 induction, p21 is activated, stalling the cell cycle at the G₁/S-phase checkpoint (30, 45, 160) This stalling allows the induction of DNA damage-repair functions by p53, the expression of proteins in an attempt to lower intracellular ROS (10, 129), and the stabilization of NRF2 by p21, thereby disrupting proteasomal degradation of NRF2 by KEAP1. This disruption of NRF2 turnover allows the nuclear translocation of NRF2, and the subsequent transcription of ARE-driven genes, leading to a general cytoprotective response. The collective actions of p53 and NRF2 lead to the repair of the cell and evasion of apoptosis. Links between NRF2 and the oxidative-stress–induced inhibition of proliferation have been well described (120, 121); in some instances, they may operate independent of p53.

The circumstance of a moderate induction of p53 allows some speculation that a middle ground exists between the promotion of programmed cell death and full cell repair and restoration of homeostasis, namely, that of cellular senescence through terminal differentiation. This balance is achieved through p21 activation after p53 induction through signaling. p21 blocks the cell cycle by stopping progression through the G₁/S checkpoint (30, 45, 160), and allows the stabilization of NRF2 and an increased level of detoxification enzymes (19). p53 almost certainly plays a role in activating additional proteins that contribute to the process of rescuing a damaged cell that may be useful in a state of terminal differentiation but has experienced some insult that renders the cell a liability when used for mitotic purposes.

This system of dynamic response to cellular stresses by p53, p21, and NRF2 demonstrates a powerful mechanism for cellular and organismal maintenance, through the repair, differentiation, or destruction of damaged cells. In the case of mild damage, NRF2-mediated detoxification of carcinogens combined with a p21-mediated pause in the cell cycle and a p53/p21-mediated induction of DNA damage-repair enzymes salvages cells from damage, whereas in the case of extensive damage to a cell, p53 mediates the disruption of NRF2-mediated antioxidant response proteins, allowing p53-driven apoptosis, in toto reflecting a tunable response to environmental stresses.

NRF2 and NOTCH Signaling: Cell-Survival Factor Meets Cell-Fate Factor

Several lines of differential microarray analyses using RNAs isolated from MEFs prepared from Nrf2^-/- and wild-type mice have shown that Notch1 and its direct downstream genes (Hes, Herp) (62) displayed decreased expression in Nrf2^-/- MEF. Studies from our laboratory demonstrate that Notch1 signaling triggered through its activating ligands, DLL1 and JAG1, is dependent on Nrf2 genotype. Moreover, one or more functional ARE sequences exist in the promoter of Notch1. Therefore, it appears that NRF2 directly regulates Notch1 gene expression (153). It is now well established that signals through the NOTCH receptor are involved in the development of several cell types and that the modulation of these signals can markedly affect differentiation, proliferation, and apoptotic events. Genetic ablation studies indicate that Notch1 is crucial for early development and regrowth in a variety of tissues (57, 154). Activation of the pathway has been shown to be a potent inhibitor of differentiation in different developmental contexts and has been associated with the amplification of certain somatic stem cells (27, 35, 64, 80, 134, 142). Considering the significance of the NOTCH1 signal cascade in developmental biology, this indicated the possibility that NRF2 could be a key molecule affecting both embryonic and adult tissue stem cell renewal, as well as cell fate. Of course, this cross talk might not only be mediated by transcriptional machinery, as NOTCH signaling is regulated by various dynamic mechanisms, including posttranslational modifications (149). Critical to the evaluation of cross talk between transcription factors are assessments of their functional consequences. Several physiological lines of evidence support functional interactions between NRF2 and Notch1: liver regeneration, keratinogenesis, osteoblastogenesis, and adipogenesis (Fig. 8).

Nrf2^-/- mice display an interesting phenotype in the process of liver regeneration (11). Nrf2^-/- mice have been challenged with a two-thirds partial hepatectomy to view the impact of genotype on the kinetics of liver regeneration. These knockout mice showed a significant delay in the recovery of liver mass after partial hepatectomy. We have observed that this regeneration-defective phenotype is rescued by hepatocyte-specific expression of the NOTCH1 intracellular domain (NICD) in Nrf2^-/- mice. NICD functions as a transcription co-factor of RBPJ. This finding provides clear evidence for the existence of NRF2-Notch1 cross talk within the damaged liver, and within the process of liver regeneration (153). In a rat model in which siRNAs against NOTCH1 and JAG1 (a NOTCH1 ligand) were injected into the superior mesenteric vein before the two-thirds partial hepatectomy, significantly suppressed proliferation of hepatocytes was reported at days 2 to 4 of the regenerative response (80). Apparently, the NRF2-NOTCH1 signaling pathway may be activated during liver regeneration and is potentially contributing to signals affecting cell growth and differentiation.

Keap1^-/- mice, which exhibit hyperactivated NRF2 signaling, have a thickened cornified layer on the suprabasal layer of the esophagus (152), with a huge keratin mass derived from the cornified layer around the limiting ridge and cardia. All Keap1-null mice die within 3 weeks of malnutrition caused from esophageal and forestomach obstruction. Through microscopic observation, a clear difference is noted between the processes of cornified layer thickening in the esophagus versus the foregut. In the case of the esophagus, cell numbers in the basal layer seem to be unchanged, with the cornified layer exhibiting thickening. This appears to be a prodifferentiation mode of hyperkeratosis. Within the limiting ridge of the greater curvature and the cardiac portion of the forestomach, however, a proliferative form of hyperkeratosis is observed, with proliferation of undifferentiated cells. This phenotype might be explained by the excess Notch signaling produced by excessive NRF2 activation. NOTCH signaling is expressed in the gastrointestinal tract. Specifically, in both the esophagus and forestomach, NOTCH1 and JAG2 display the highest levels of expression within their families in the basal layer (70), where stem and progenitor cells are located (130). Progenitor cells are differentiated to keratinocytes in the suprabasal layer and denucleated in the cornified layer, with Notch signaling being weakly expressed in the suprabasal layer. Interestingly, NOTCH signaling induces terminal differentiation to activate p21 gene expression directly in keratinocytes (118, 158). This function of the NOTCH signal might be stronger in the esophagus and forestomach of Keap1-null mice.

Hinoi et al. (52) demonstrated that NRF2 negatively regulates osteoblast differentiation in the MC3T3-E1 cell system. They also showed that NRF2 might inhibit cellular differentiation and maturation in chondrocytes by using the pre-chondrogenic cell line ATDC5, derived from the mouse teratocarcinoma AT805 (53). As a proposed mechanism of these phenomena, they conjectured that the inhibition of RUNX2 (one of the essential factors for bone formation and chondral maturation) was caused by either direct or indirect interaction with NRF2. Conversely, Notch-signaling directly influences osteogenesis and chondrogenesis (14). Interestingly, these phenomena are all related to bHLH-transcription factors, especially HES and HERP, which are primary target genes of NOTCH signaling. These proteins function as negative regulators of target gene expression through direct cis-element (E-box, N-box)–binding machinery. They also facilitate proteasomal degradation through heterodimerization with other bHLH transcription factors. The heterodimer partners of HES or HERP primarily promote differentiation. Thus, they can lead to maintenance of current cell status. HES1 directly binds to the osteocalcin promoter and represses osteocalcin gene expression (167). Additionally, HES1/HERP2 physically interacts with RUNX2 in osteoblasts (37). It is also possible that NOTCH1 expression in osteoblasts is influenced by the NRF2 transcription factor, much as NOTCH1 is in MEF.

As discussed earlier, AhR and its downstream gene expression is reduced in Nrf2^-/- MEF (138). This dampened gene expression impairs adipocyte differentiation from MEF. Nrf ^-/- MEF show markedly accelerated adipogenesis on stimulation, whereas Keap1^-/- MEF, which demonstrate higher levels of NRF2 signaling, differentiate slowly compared with their congenic wild-type MEF. Within this system, ectopic expression of AhR rescues adipogenesis seen in Nrf2^-/- MEF. However, recently Ross et al. (124) reported that HES1, a primary downstream gene product of NOTCH1 signaling, inhibits the differentiation of adipocytes from 3T3-L1 cells. They also demonstrated that both the bHLH and the WRPW (67) [which is a co-repressor (GROUCHO/TLE) binding motif ] domains in HES1 are required for its inhibitory effect. This study raised the possibility that the enhanced adipogenesis shown in Nrf2^-/- MEF is caused by a weaker output of both NOTCH1 and AhR signaling than within wild-type MEF.

Conclusions

The NRF2 pathway has been studied in diverse contexts because of potential roles in stroke recovery, neurodegeneration, chronic obstructive pulmonary disease, and cancer prevention in multiple organs. The profiling of gene-expression changes mediated by the NRF2 pathway has been well documented. The interpretation of such studies, however, has been hampered by the inability to distinguish genes modulated directly by NRF2 from genes induced secondarily by NRF2-sensitive transcription factors and the mitigating impacts of tissue-specific responses. Nonetheless, it is clear that the protective effects of upregulation of NRF2 signaling can take several forms. Protection can be immediate, reflecting induction of genes directly regulated through NRF2 binding to AREs in target genes (e.g., the innate immune response and elevated cytoprotective responses to blunt cytokine surges or detoxify reactive intermediates, respectively) (73). The protective effects can be secondary through induction of macromolecular damage repair/removal systems (proteasome, DNA repair) (84, 106). Last, the protective effects can be tertiary through activation of tissue repair/regeneration pathways. In these latter cases, involvement in cross talk with additional pathways affecting cell survival and other aspects of cell fate most certainly play important collaborating roles. This review has highlighted several such collaborators (and sometime competitors): AhR, NF-κB, p53, and Notch1. Future studies evaluating the concordance between global mRNA expression profiles at the transcriptomic levels (i.e., oligonucleotide microarray analyses) and genome-wide NRF2-DNA binding analysis using high-throughput methods (i.e., chromatin-immunoprecipitation with parallel sequencing) should better establish a systems-level view of NRF2-dependent processes occurring within cells. Functional importance will require well-crafted molecular genetic studies in cell culture, and more important, in vivo, to define the critical cross-talk interactions between NRF2 and other transcription factors.

Acknowledgments

This work was supported by NIH grants CA39416 and CA94076 (TWK) and the Maryland Cigarette Restitution Fund. SLS was supported by T32 ES07141 and T32 CA009110; JJS by T32 GM08763; and SS was the recipient of a Samsung Scholarship (Samsung Foundation of Culture).