Molecular Basis Distinguishing the DNA Binding Profile of Nrf2-Maf Heterodimer from That of Maf Homodimer*

Nrf2-small Maf heterodimer activates the transcription of many cytoprotective genes through the antioxidant response element and serves as a key factor in xenobiotic and oxidative stress responses. Our surface plasmon resonance-microarray binding analysis revealed that both Nrf2-MafG heterodimer and MafG homodimer bind to the consensus Maf recognition element with high affinity but bind differentially to the suboptimal binding sequences degenerated from the consensus. We examined the molecular basis distinguishing the binding profile of Nrf2-MafG heterodimer from that of MafG homodimer and found that the Ala-502 residue in the basic region of Nrf2 is a critical determinant of its binding specificity. In Maf proteins, a tyrosine resides in the position corresponding to Ala-502 in Nrf2. We prepared a mutant Nrf2 molecule in which Ala-502 was replaced with tyrosine. In surface plasmon resonance-microarray analysis, heterodimer of Nrf2(A502Y) and MafG displayed a binding specificity similar to that of MafG homodimer. The target genes activated by mutant Nrf2(A502Y)-small Maf heterodimer were largely different, albeit with some overlap, from those activated by wild-type Nrf2-small Maf, indicating that the array of target genes regulated by Nrf2-small Maf heterodimer differs substantially from that regulated by Maf homodimer in vivo. These results suggest that the distinct DNA binding profile of Nrf2-Maf heterodimer is biologically significant for Nrf2 to function as a key regulator of cytoprotective genes. Our contention is supported that the differential DNA binding specificity between Maf homodimers and Nrf2-Maf heterodimers establishes the differential gene regulation by these dimer-forming transcription factors.


Nrf2-small Maf heterodimer activates the transcription of many cytoprotective genes through the antioxidant response element and serves as a key factor in xenobiotic and oxidative stress responses. Our surface plasmon resonance-microarray binding analysis revealed that both Nrf2-MafG heterodimer and
MafG homodimer bind to the consensus Maf recognition element with high affinity but bind differentially to the suboptimal binding sequences degenerated from the consensus. We examined the molecular basis distinguishing the binding profile of Nrf2-MafG heterodimer from that of MafG homodimer and found that the Ala-502 residue in the basic region of Nrf2 is a critical determinant of its binding specificity. In Maf proteins, a tyrosine resides in the position corresponding to Ala-502 in Nrf2. We prepared a mutant Nrf2 molecule in which Ala-502 was replaced with tyrosine. In surface plasmon resonance-microarray analysis, heterodimer of Nrf2(A502Y) and MafG displayed a binding specificity similar to that of MafG homodimer. The target genes activated by mutant Nrf2(A502Y)-small Maf heterodimer were largely different, albeit with some overlap, from those activated by wild-type Nrf2-small Maf, indicating that the array of target genes regulated by Nrf2-small Maf heterodimer differs substantially from that regulated by Maf homodimer in vivo. These results suggest that the distinct DNA binding profile of Nrf2-Maf heterodimer is biologically significant for Nrf2 to function as a key regulator of cytoprotective genes. Our contention is supported that the differential DNA binding specificity between Maf homodimers and Nrf2-Maf heterodimers establishes the differential gene regulation by these dimer-forming transcription factors.
Cellular detoxification is critical for the maintenance of health by providing protection against various chemicals of either endogenous or exogenous origin. Many drug-metabolizing enzymes and antioxidant proteins are coordinately induced at the transcriptional level in response to electrophiles and oxidative stress. The antioxidant response element (ARE) 2 was identified as a critical cis-regulatory element for the inducible expression of these cytoprotective genes (1). A sequence common to individual regulatory elements was extracted and TGA(G/C)NNNGC was defined as the ARE consensus sequence.
Maf family proteins share a conserved basic region and leucine zipper (bZip) motif and are divided into two groups, large Maf (c-Maf, MafA, MafB, and Nrl) with an activation domain and small Maf (MafG, MafK, and MafF) without an activation domain (2). The Maf recognition element (MARE) was defined as a 13-bp (TGCTGA(G/C)TCAGCA) or 14-bp (TGCTGA(GC/CG)TCAGCA) palindromic sequence through the systematic evolution of ligands by exponential enrichment (3,4) that identifies a sequence with the highest affinity to the transcription factor as its consensus sequence. The MARE core contains a 12-O-tetradecanoylphorbol-13-acetate-responsive element (TGA(G/C)TCA) or a cAMP-responsive element (TGA(GC/CG)TCA), which is the binding sequence of Jun/Fos family proteins or CREB (cAMP-response element-binding protein)/ATF family proteins, respectively. In addition, the MARE core is flanked on both sides by extended elements (5Ј-TGC-(12-O-tetradecanoylphorbol-13-acetate-responsive element/cAMP-responsive element)-GCA-3Ј), a unique feature of the binding site of Maf family proteins. A core sequence on one side and a flanking sequence on the other side of the MARE were found to comprise the ARE consensus sequence. This striking similarity suggested the involvement of Maf proteins in ARE-dependent gene regulation (2,5).
Small Maf and Nrf2, a member of the Cap'n'Collar (CNC) family of transcription factors, form a heterodimer that serves as a key regulator for the inducible expression of cytoprotective genes (6 -8). Other Maf-containing dimers also interact with the MARE and its related sequences, including the ARE. Large Maf homodimers and heterodimers of small Maf and a CNC family member (NF-E2 p45, Nrf1, Nrf2, and Nrf3) are MAREdependent activators, whereas small Maf homodimers and heterodimers of small Maf with a Bach family member (Bach1 and Bach2) are MARE-dependent repressors (2, 9 -12). Considering that all these dimers bind to similar target sites, the question is how each Maf-containing dimer elicits its specific biological function.
The individual MARE-related sequences in the regulatory regions of endogenous genes have often diverged from the consensus MARE. We surmised that such divergence has enabled the differential binding of distinct Maf-containing dimers and generated the diversity in MARE-dependent gene regulation. To test our hypothesis, we performed a comprehensive analysis on the binding affinities of Maf homodimer and Nrf2-Maf heterodimer to MARE-related sequences generated by systematic base alterations using the surface plasmon resonance (SPR) microarray imaging technique (13,14). We found that Maf homodimer and Nrf2-Maf heterodimer bind similarly to the consensus MARE with high affinity but bind differentially to the suboptimal binding sequences degenerated from the MARE consensus. The biological significance of the latter has not been clarified.
Because Nrf2-MafG heterodimer was more sensitive to a core region mutation than MafG homodimer, we suspected that Nrf2 mainly recognizes the core region of the MARE, which is a characteristic common to other bZip transcription factors, including the Jun and Fos family. On the contrary, Maf family DNA recognition has been considered unique due to the requirement of flanking regions within the MARE (3,15). The structural characteristics accounting for the unique DNA recognition by Maf family proteins reside in the basic region and the extended homology region (EHR), a unique stretch of 24 amino acids just in front of the basic region (Fig. 1A). We showed that the EHR partially supports the Maf-specific binding mode (16), and the critical contribution of the basic region was also suggested (17).
In this study, we found that Ala-502 in the basic region of Nrf2 is critical in determining the binding specificity of Nrf2small Maf heterodimer. Ala-502 of Nrf2 corresponds to the well conserved alanine residue in the basic region of all bZip transcription factors, except for Maf family proteins that possess a tyrosine residue instead. We generated a mutant Nrf2(A502Y) molecule by replacing the Ala-502 residue with tyrosine. Nrf2(A502Y)-MafG heterodimer and MafG homodimer displayed similar binding specificities in an SPR microarray anal-ysis. Using the Nrf2(A502Y) mutant with its altered specificity, we examined whether the distinct profiles of suboptimal binding of Nrf2-MafG heterodimer and MafG homodimer prove significant in the induction of Nrf2 target genes in vivo. Compared with those of wild type, the target genes activated by Nrf2(A502Y)-small Maf heterodimer were fundamentally different, although somewhat overlapping. This advocates that a significantly different set of target genes is regulated by Nrf2small Maf heterodimer compared with those associated with Maf homodimer. These results suggest that the distinct DNA binding profile of Nrf2-small Maf heterodimer is essential for Nrf2 to function as an activator of cytoprotective genes. Our results also support our contention that the differential DNA binding specificity between Maf homodimers and CNC-Maf heterodimers forms a strong foundation for differential gene regulation by dimer-forming bZip transcription factors.
Cell Culture and Reporter Assay-293T cells were maintained in DMEM (Sigma) with 10% fetal bovine serum and 1% penicillin-streptomycin. Reporter plasmids were introduced into 293T cells along with Nrf2 and Nrf2(A502Y) expression vectors using FuGENE6 transfection reagent (Roche). Luciferase activity was measured 24 h after transfection using a Dual Reporter Assay System (Promega) and a luminometer (Berthold Japan, Tokyo). All samples were prepared in triplicate and means and standard deviations were calculated.
Establishment of Stable Cell Lines-Stable cell lines were established as instructed earlier (21). pcDNA5/FRT/TO-FLAG-Nrf2 and pcDNA5/FRT/TO-FLAG-Nrf2(A502Y) were transfected into a parental FlpIn TREx 293 cell line (Invitrogen) with pOG44 (Invitrogen) that encodes Flp recombinase. The parental cells contain a single Flp recombinase target (FRT) site located at a transcriptionally active genomic locus to which the gene of interest can be targeted. Transfectants were selected and maintained in Dulbecco's modified Eagle's medium containing 10% tetracycline-reduced fetal bovine serum (BD Biosciences), 15 g/ml blasticidin, and 150 g/ml hygromycin B.
Chromatin Immunoprecipitation (ChIP) Assays-ChIP assays were performed before and 3 and 6 h after treatment with 1 g/ml tetracycline. Immunoprecipitation was performed using control rabbit IgG, anti-Nrf2 (Santa Cruz, sc-13032), and anti-FLAG (Sigma, A2220) antibodies. PCR primer sets HO-1 E1, NQO1 pr, and HO-1 exon 3 were as described previously (22). Chromatin DNA was subjected to PCR analysis and is indicated as input.
Microarray Analysis-Before and 6 h after tetracycline treatment, total RNA samples were extracted from 293/FLAG-Nrf2 and 293/FLAG-Nrf2 A502Y cells using Isogen (Nippon Gene). Isolated RNA was purified using an RNeasy Mini kit (Qiagen), then processed and hybridized to a human expression array Affymetrix Human Genome U133 Plus 2.0 Array (Affymetrix). Experimental procedures for Gene Chip were performed according to the Affymetrix Technical Manual.
Quantitative Real Time Reverse Transcription-PCR-cDNAs were synthesized from total RNAs prepared from 293/FLAG-Nrf2 and 293/FLAG-Nrf2 A502Y cells using random hexamers. Real time PCR was performed using an ABI7700 or ABI7300 sequence detection system. The reaction was carried out for 40 cycles of 30 s at 95°C and 1 min at 60°C (HO-1, NQO1, and ribosomal RNA) or 55°C (GLIPR1 and TPM4) using qPCR Mastermix (Eurogentec). Ribosomal RNA control reagents (Applied Biosystems) were used as an internal control. The probes and primers used for the quantitative real time PCR are described in Table 1.

Replacing an Alanine Residue in the Basic Region of Nrf2 with the Corresponding Tyrosine Residue of MafG Converts the Specificity of Heterodimeric Binding to That of Homodimeric
Binding-Our previous SPR-microarray analysis demonstrated the differential DNA binding profiles of Nrf2-MafG heterodimer and MafG homodimer, which seemed to be attributed to the distinct DNA recognition modes of Nrf2 and MafG, mainly recognizing the core and the flanking regions of the MARE, respectively (14). Chief recognition of the core is the DNA binding feature typical of dimer-forming bZip transcription factors, whereas Maf proteins uniquely recognize the flanking regions. One of the structural features distinguishing between the Maf family and the rest of the bZip factors including Nrf2 is a well conserved amino acid in the basic region (Fig.  1A). Maf family proteins possess a tyrosine residue in their basic regions, whereas the other bZip family members possess an alanine residue at the corresponding position. The importance of tyrosine for Maf-specific DNA recognition was suggested by analysis of Maf proteins with tyrosine mutations (17). Replacement with alanine converted sequence recognition from the flanking regions to the core, whereas replacement with pheny-

Name
Sequence lalanine generally reduced the binding affinity. From this, we suspected that an alanine residue might be essential for recognition of the core. To test our hypothesis, we replaced alanine with tyrosine in the basic region of Nrf2 and generated an Nrf2(A502Y) mutant molecule. We first performed comprehensive binding analysis using the SPR microarray imaging technique (13). Protein fragments of MafG, Nrf2, and Nrf2(A502Y) containing bZip domains were expressed in bacteria and purified. The DNA microarrays were assembled using double-stranded oligonucleotides containing MARE variants generated by systematic base alterations as described previously (Fig. 1B, Ref. 14). MafG homodimer, Nrf2-MafG heterodimer, and Nrf2(A502Y)-MafG heterodimer were applied to the DNA microarray, and SPR signals were obtained.
MafG homodimer bound to the consensus MARE (CEN-C and CEN-G, see The Binding Specificity Observed in Vitro Is Reflected in the Differential Transcriptional Activation through MARE Variants in Vivo-We next asked whether the difference in suboptimal binding to MARE-related sequences is important to the function of Nrf2-MafG heterodimer in vivo. We performed a reporter assay by transient expression in 293T cells using Nrf2(A502Y) for evaluating the in vivo significance of the altered binding specificity.
Increasing amounts of Nrf2 and Nrf2(A502Y) expression vectors were transiently introduced into 293T cells with a lucif-

TABLE 2 MARE-related sequences for EMSA and reporter assays
Each MARE-related sequence is underlined and flanked by fixed 6-bp regions. Lowercase letters indicate the mutated bases in the consensus MARE.  NOVEMBER 16, 2007 • VOLUME 282 • NUMBER 46 erase reporter gene. The assay was not supplemented with MafG, as endogenous small Maf was expected to be sufficiently abundant for heterodimerization with Nrf2 and its mutant molecule. Luciferase activities were measured 24 h after transfection. When the luciferase gene was driven by the consensus MARE (Fig. 3A), reporter gene activation was comparable between Nrf2 and Nrf2(A502Y). The nuclear accumulation of Nrf2 and Nrf2(A502Y) was also analogous (Fig. 3F). These results and the comparable binding affinities of Nrf2 and Nrf2(A502Y) to the consensus MARE (see Fig. 2A) suggested that the Ala to Tyr mutation in the basic region did not affect the transcriptional activity of Nrf2. When the reporter gene was driven by heterodimer MARE (Fig. 3, B and C), Nrf2 elicited a stronger activation than Nrf2(A502Y). On the contrary, Nrf2(A502Y) elicited a stronger activation than Nrf2 when the reporter gene was driven by homodimer MARE (Fig. 3, D and E). These findings show that the difference between the binding profiles of Nrf2-MafG and Nrf2(A502Y)-MafG heterodimers was faithfully reflected in the differential activation of reporter genes driven by different subtypes of MARE variants.

Differential DNA Recognition by Maf-containing Dimers
Differential DNA Binding of Nrf2 and Nrf2(A502Y) to the Endogenous Regulatory Region-To verify whether the binding specificity alteration of Nrf2(A502Y) mutant is actually operating in vivo, we examined the binding of Nrf2 and Nrf2(A502Y) in vivo by a ChIP analysis through establishing stable cell lines expressing FLAG-tagged Nrf2 or Nrf2(A502Y), which is inducible by tetracycline (293/FLAG-Nrf2 and 293/FLAG-Nrf2(A502Y) cells). The nuclear accumulation of Nrf2 and Nrf2(A502Y) after the tetracycline induction was comparable between the two cell lines (Fig. 4A).
In this analysis we chose two Nrf2 target genes with well characterized regulatory regions, hemeoxygenase-1 (HO-1) and NADP(H):quinone oxidoreductase 1 (NQO1). Two tandem MAREs, referred to as the E1 distal enhancer (21), are located 4 kbp upstream from the transcription initiation site of the human HO-1 gene (Fig. 4B). HO-1 MARE-1 and HO-1 MARE-2 have the characteristics of heterodimer MARE and homodimer MARE, respectively. A single MARE with characteristics of heterodimer MARE is located 400 bp upstream from the transcription initiation site of the human NQO1 gene (Fig. 4C). A ChIP assay was performed with the cells before and after induction by tetracycline using anti-FLAG antibody and enrichment of DNA fragments containing HO-1 E1 enhancer (containing both HO-1 MARE-1 and -2), or NQO1 MARE was examined by PCR (Fig. 4, D-F). Nrf2 and Nrf2(A502Y) bound similarly to the HO-1 E1 enhancer (Fig. 4D), whereas Nrf2(A502Y) bound to the NQO1 MARE less efficiently than Nrf2 (Fig. 4E). Anti-Nrf2 antibody gave a similar result (data not shown). These outcomes suggested that Nrf2(A502Y) could interact with the E1 enhancer, since HO-1 MARE-2 should allow its binding and NQO-1 MARE, with its preference for the heterodimer, did not strongly recruit Nrf2(A502Y). The third exon fragment of the HO-1 gene was detected as a negative control (Fig. 4F). Thus, the different binding specificities between Nrf2 and Nrf2(A502Y) were significant in vivo.
Endogenous Genes Respond Differentially to Nrf2 and Nrf2(A502Y)-The effects of Nrf2 and Nrf2(A502Y) on the expression levels of endogenous genes using 293/FLAG-Nrf2 and 293/FLAG-Nrf2(A502Y) cells were examined. Total RNAs were prepared from each cell line before and after induction and subjected to transcriptome analysis. 255 genes and 126 genes were induced by over 1.4-fold after tetracycline  Table 2. Black and gray bars show the average relative luciferase activities of triplicate samples of Nrf2 and Nrf2(A502Y), respectively. Error bars indicate the S.D. of the triplicate samples. Data representative of three independent experiments are presented. F, the nuclear accumulation of Nrf2 and Nrf2(A502Y) was examined by immunoblot analysis using anti-Nrf2 antibody.  (Fig. 5A). Among them, 32 genes were induced by both Nrf2 and Nrf2(A502Y), whereas 223 genes and 94 genes were induced exclusively by Nrf2 and Nrf2(A502Y), respectively. Genes induced by Ͼ2-fold are listed in Table 3. The expression levels of representative genes were examined by quantitative real time-PCR (Fig. 5B).

Differential DNA Recognition by Maf-containing Dimers
Detoxifying enzymes and antioxidant-responsive genes, which are well characterized Nrf2 target genes, were well induced by Nrf2 but not by Nrf2(A502Y). NQO1 was highly induced by Nrf2, but not by Nrf2(A502Y) (Fig. 5B, upper left), consistent with the observation that Nrf2, but not Nrf2(A502Y), associated with the promoter region of the NQO-1 gene (Fig. 4E). The genes belonging to this first group (group A in Table 3) were mainly regulated by Nrf2-small Maf heterodimer.
HO-1, GLI pathogenesis-related 1 (glioma) (GLIPR1), and tropomyosin 4 (TPM4) were induced by both Nrf2 and Nrf2(A502Y) (Fig. 5B, lower panels). Nrf2(A502Y) contributed to HO-1 gene activation, probably because the mutant could still associate with the E1 distal enhancer of the HO-1 gene. Although the MAREs responsible were not identified for GLIPR1 and TPM4, we surmised that they possess heterodimer MAREs and homodimer MAREs simultaneously or dual MAREs in their regulatory regions. Because this second group of genes was sensitive to both types of heterodimer, group B in Table 3 included genes under the dual regulation of Nrf2-small Maf heterodimer and Maf homodimer.
The third group (group C in Table 3) represents a new battery of genes directly or indirectly affected by Nrf2(A502Y). Protein kinase, cAMP-dependent, regulatory, type II, ␣ (PRKAR2A) was induced by Nrf2(A502Y), but not by Nrf2 (Fig.  5B, upper right). The induction levels of the genes belonging to this group were generally modest, probably because the Nrf2 activation domain might be disadvantageous in cooperating with transcriptional cofactors at heterologous loci different from the natural Nrf2 targets.
These results clearly indicate that the profile of genes activated by Nrf2-small Maf heterodimer differs from that activated by Nrf2(A502Y)-small Maf heterodimer, suggesting that the difference in suboptimal binding to MARE-related sequences between Maf homodimer and Nrf2-small Maf heterodimer generates a significant difference in transcriptional gene regulation in vivo.

DISCUSSION
In this study we found that a heterodimer of MafG and Nrf2(A502Y), the basic region mutant of Nrf2, displayed a DNA   NOVEMBER 16, 2007 • VOLUME 282 • NUMBER 46 binding specificity similar to that of MafG homodimer. Our microarray analysis utilizing the cell line stably expressing the Nrf2(A502Y) mutant molecule revealed the substantial difference between the gene expression profiles activated by Nrf2small Maf and Nrf2(A502Y)-small Maf heterodimers. These results strongly argue that subtle changes accumulated in the binding sequences for the Maf and CNC family of transcription factors during molecular evolution give rise to the functionally discrete cis-acting elements categorized as homodimer MAREs or heterodimer MAREs (e.g. ARE), which prefers Maf homodimer binding and Maf-CNC heterodimer binding, respectively. This study further clarified that one amino acid substitution in the basic region of the CNC and Maf family transcription factors serves as part of the molecular basis for this specific matching of cis-elements and transcription factors. Thus, the biological significance of differential binding by Maf homodimer and CNC-Maf heterodimer should be interpreted based on the accumulation of base changes in the cis-acting elements and evolution of the transcription factors acquiring activities recognizing the deviation. Several previous reports suggested a possible heterogeneity in MARE-related sequences; those bound by Maf homodimers and those bound by CNC-Maf heterodimers. The target sequences of large Maf homodimers have been deduced from promoter analysis of the target genes and were found to have well conserved flanking regions rather than core regions (3,14). The target sequences of CNC-Maf heterodimers were reported to be asymmetric, where the flanking region was conserved on one side of the MARE, e.g. the NF-E2 binding site for NF-E2 p45-small Maf heterodimer (TGCTGACTCAT; Ref. 23) and the ARE site for Nrf2-small Maf heterodimer (GCNNNCTCA; Ref. 24). Consistent with these observations, we found that the gene profile activated by Nrf2-small Maf heterodimer substantially differed from that activated by Nrf2(A502Y)-small Maf heterodimer that mimicked Maf homodimer. The present results suggest that the distinct DNA binding specificity of Nrf2-small Maf heterodimer is essential for Nrf2 to function as a key regulator of cytoprotective genes, as most of these genes were ineffectively activated by Nrf2(A502Y). The partially overlapping activated gene profiles of Nrf2-Maf and Nrf2(A502Y)-Maf heterodimers implicated that several genes are co-regulated by Maf homodimer and Nrf2-small Maf heterodimer.

Differential DNA Recognition by Maf-containing Dimers
X-ray crystallography revealed the structure of typical dimeric bZip domains in complex with DNA (25,26). The homodimeric bZip domain of GCN4 and heterodimeric bZip domains of c-Jun and c-Fos were crystallized in the presence of palindromic target DNA molecules. In both cases each bZip domain was found to form a continuous ␣-helix and to recog- Each value was calculated from the raw data of the microarray analysis, and genes induced by Ͼ2-fold after tetracycline treatment are listed. Significant induction (written in bold) was selective by Nrf2 (A), by both Nrf2 and Nrf2(A502Y) (B), and by Nrf2(A502Y) (C). Gene products in bold font have been reported previously as Nrf2-dependent genes.

Differential DNA Recognition by Maf-containing Dimers
nize the DNA sequence corresponding to the core region of the MARE. Compared with these typical bZip factors, the absolute difference in DNA recognition by Maf homodimers is the requirement of the flanking region of the MARE. One previous report showed that glycine and tyrosine residues in the basic region of Maf proteins are critical determinants of their unique binding property (17) and proposed the model that these residues break the ␣-helix and position the EHR closer to the flanking region of the MARE. Mutation of alanine to tyrosine in Nrf2 might have bent the CNC domain (see Fig. 1A) so that it can interact with the flanking region of the MARE. Our previous analysis on the NMR structure of the EHR of MafG revealed that it resembles the CNC domain of Skn-1 (16). The conformational similarity between the EHR of MafG and the CNC domain of Nrf2 might have enabled Nrf2(A502Y) to recognize the flanking region of the MARE with its CNC domain and to recapitulate the binding specificity of Maf. Genes activated by tetracycline-induced Nrf2 in 293 cells included typical Nrf2 target genes involved in cytoprotection (e.g. thioredoxin reductase 1, glutamate-cysteine ligase modifier subunit, and NQO1). Several genes induced by exposure to electrophiles in an Nrf2-dependent manner were not activated in this experiment (e.g. glutathione S-transferases). One reason for this might be the increase in Nrf2 protein in the absence of electrophilic or oxidative stresses in this system. As far as the canonical Nrf2 target genes induced under this condition were concerned, cross-talk between Nrf2-small Maf heterodimer and large Maf homodimer was less probable since Nrf2(A502Y) did not activate these genes. We missed out several genes expected to have been induced in the physiological stress response in this study. Therefore, a rational extension of this study would be to generate a knock-in mouse harboring an A502Y mutation in the nrf2 locus and examine the function of Nrf2(A502Y) in mice with an nrf2-null background.
Small Maf homodimers lacking canonical activation domains are expected to repress transcription when competing with active Nrf2-small Maf heterodimer for the MARE (9). However, most of the cytoprotective genes regulated by Nrf2 were not activated by the Nrf2(A502Y) mutant molecule, suggesting that Nrf2(A502Y)-small Maf heterodimer has a lower binding affinity to the MAREs in the regulatory regions of typical Nrf2 target genes. Because Nrf2(A502Y)-MafG heterodimer displayed a binding profile similar to that of MafG homodimer, this result implies that small Maf homodimer has a lower binding affinity to the MAREs in the regulatory regions of Nrf2 target genes and may not be an efficient repressor of the Nrf2 target genes, assuming that comparable amounts of small Maf homodimer and Nrf2-small Maf heterodimer are present. Therefore, small Maf in excess of Nrf2 would be required for small Maf homodimer to elicit repression of cytoprotective genes.
The genes activated by Nrf2(A502Y) might include some of the large Maf target genes. Expressed under the control of the viral promoter, c-Maf was reported to efficiently transform fibroblasts (27), but the target genes responsible for this transformation have not been pinpointed. Several genes (e.g. ribosomal protein S6 kinase, 90 kDa, polypeptide 4, and serine/threonine/tyrosine kinase 1) reported to be involved in tumorigenesis and cell proliferation (28,29) could be such targets.
Some of the genes induced by both Nrf2 and Nrf2(A502Y), which were interpreted as those co-regulated by Nrf2-small Maf heterodimer and large Maf homodimer, might also be involved in tumorigenesis (e.g. GLIPR1, Ref. 30). A previous study suggested that the cis-regulatory sequence recognized by Jun homodimer is necessary for Maf-induced transformation (31). Although there are several possibilities, we favor the interpretation that the MARE sequence shared by AP-1 (Jun/Fos) and Maf, which should behave as a dual MARE, is involved in Maf-induced transformation. We surmise that Jun/Fos and Maf family proteins may be able to exert their oncogenic activities by regulating common genes through the dual MARE.
An intriguing revelation from our study is that the artificially selected consensus sequence with the highest affinity does not necessarily represent the majority of significant cis-acting elements targeted by a transcription factor in vivo. Both Nrf2 and Nrf2(A502Y) bound to the consensus MARE with high affinity, yet the activated gene profiles in vivo were quite different from each other. The sequences that diverged from the consensus, sometimes with lower affinities, seemed to confer the differential specificities required for diverse gene regulation by related family members of transcription factors.