Tissue-restricted Expression of Nrf2 and Its Target Genes in Zebrafish with Gene-specific Variations in the Induction Profiles

The Keap1-Nrf2 system serves as a defense mechanism against oxidative stress and electrophilic toxicants by inducing more than one hundred cytoprotective proteins, including antioxidants and phase 2 detoxifying enzymes. Since induction profiles of Nrf2 target genes have been studied exclusively in cultured cells, and not in animal models, their tissue-specificity has not been well characterized. In this paper, we examined and compared the tissue-specific expression of several Nrf2 target genes in zebrafish larvae by whole-mount in situ hybridization (WISH). Seven zebrafish genes (gstp1, mgst3b, prdx1, frrs1c, fthl, gclc and hmox1a) suitable for WISH analysis were selected from candidates for Nrf2 targets identified by microarray analysis. Tissue-restricted induction was observed in the nose, gill, and/or liver for all seven genes in response to Nrf2-activating compounds, diethylmaleate (DEM) and sulforaphane. The Nrf2 gene itself was dominantly expressed in these three tissues, implying that tissue-restricted induction of Nrf2 target genes is defined by tissue-specific expression of Nrf2. Interestingly, the induction of frrs1c and gclc in liver and nose, respectively, was quite low and that of hmox1a was restricted in the liver. These results indicate the existence of gene-specific variations in the tissue specificity, which can be controlled by factors other than Nrf2. Economy, Trade and Industry of Japan (YT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


Introduction
Nrf2 is a transcription factor that binds to the antioxidant response element (ARE) and transactivates cytoprotective genes [1,2]. In basal conditions, Nrf2 is degraded via the Keap1dependent proteasome pathway, while it is stabilized after cells are exposed to electrophilic or oxidative stress, which transactivates its target genes. Many studies have identified the Keap1-Nrf2 system to have multiple sensor sites to a variety of stresses [3,4] and more than one hundred target genes [5,6]. Conservation of the Keap1-Nrf2 system has been demonstrated in vertebrates including zebrafish [7,8], which is a well-established research model.
Through previous studies, we noticed that the expression of gstp1, a major target gene for zebrafish Nrf2, is not ubiquitous as expected, but is rather restricted in the nose, gill, and liver [8][9][10][11][12][13]. It is unclear whether this tissue-restricted induction is specific for gstp1 or a common feature for Nrf2 target genes, since gstp1 was the only gene available in our study that was suitable for WISH analysis. In addition, most studies related to Nrf2 target genes have been performed on cultured cells or a specific murine tissue, and therefore the tissue specificity of Nrf2 target genes has not yet been well characterized, even in other animals. In this paper, we identified seven zebrafish genes appropriate for WISH using microarray analysis and examined their tissue-specific expression in zebrafish larvae. Induced expression of all these genes was restricted to the nose, gill, and/or liver with gene-specific variations in tissue specificity. These results indicate tissuerestricted induction to therefore be a common feature of Nrf2 target genes, which can be a critical issue for both the pharmacological and clinical applications of Nrf2-activating compounds.

RT-PCR and WISH analyses
Zebrafish embryos and larvae were obtained by natural mating. All experiments were carried out using a wild-type AB strain. For induction studies, embryos or larvae were placed in culture dishes containing 100 mM DEM (Wako, Osaka, Japan) or 40 mM sulforaphane (LKT laboratories, St. Paul, MN). RT-PCR analysis was performed as described previously [8] using PCR primers listed in Table S1. WISH analysis was carried out as described previously [14] with slight modifications in the fixation step. Briefly, the larvae were fixed with 4% paraformaldehyde (PFA) in PBS overnight at 4uC, and washed twice in PBS, once in 50% methanol, and twice in 100% methanol, and cooled to -20uC for at least 3 hours. Fixed larvae were then brought back to room temperature (RT), washed twice in PBT (0.1% Tween 20/PBS) for 5 minutes, and immersed for 2 hours in 9% hydrogen peroxide in PBT. After immersion, the larvae were washed twice in PBTw (0.2% bovine serum albumin in PBT), treated for 20 minutes with 50 mg/ml proteinase K (Sigma-Aldrich, St. Louis, MO), and fixed with 4% PFA/PBS for 20 minutes at RT.

Microarray analysis
A microarray analysis was performed using custom-made 16 K MZH chips. The MZH chips contained in a total of 16399 probes of a 65 oligonucleotides in length purchased from Sigma-Aldrich. The collected zebrafish embryos were quickly homogenized with 1 ml of QIAzol reagent (Qiagen, Hilden, Germany), and subsequently stored at -80uC. Total RNA was extracted according to the manufacturer's instructions. Isolated total RNA was then further purified using the RNAeasy mini kit (Qiagen). Amino-allyl-modified amplified RNA was synthesized in one amplification round from 1 mg of purified total RNA using the amino-allyl RNA amplification kit (Sigma-Aldrich). Subsequently, 5 mg of amino-allyl-modified amplified RNA was used for coupling of monoreactive Cy3 and Cy5dyes (GE Healthcare, Little Chalfont, UK) and column purified. Dual-color hybridization of the MZH chips was performed according to the manufacturer's instructions for the AceGene DNA microarray (Hitachi Solutions, Tokyo, Japan). Each experiment was repeated in triplicate. After hybridization, MZH chips were scanned using the Affymetrix 428 array scanner (Affymetrix, Santa Clara, CA). The microarray data were processed from raw data image files with Affymetrix Jaguar (Affymetrix) and normalized. The processed data were subsequently imported into Excel (Microsoft, Redmond, WA) to compare expression profiles of DEM-treated samples to control samples. The cut-offs for significance regarding the ratios of DEM-treated samples vs control samples were set at a 2-fold change.
We have deposited the raw data at GEO under accession numbers (GSM799460, GSM799461, GSM799462, GSM799463), and we can confirm all details are MIAME compliant.

Knockdown and overexpression analyses of Nrf2
Synthetic capped nrf2 RNA was made with an SP6 mMES-SAGE mMACHINE in vitro transcription kit (Ambion, Austin, TX) using pCS2nrf2. The Nrf2-morpholino oligonucleotide (nrf2MO) has been described previously [9]. mRNA or morpholino oligonucleotides were injected into yolk of the zebrafish at the one-cell stage using an IM300 microinjector (Narishige, Tokyo, Japan).

Identification of Nrf2 target genes in zebrafish
In order to investigate tissue-specific expression of Nrf2 target genes, we searched for zebrafish Nrf2 targets other than gstp1 that were able to be used for WISH analysis. Microarray analysis was carried out using cDNA prepared from 4-dpf larvae treated with or without 100 mM DEM for 12 hours. In total, 16,000 zebrafish cDNAs were screened and 42 genes were identified that showed more than a 2-fold induction compared with DEM-treated larvae and untreated larvae (Table S3). The reliability of this screen was demonstrated by the fact that gstp1 produced a top ranking score in this analysis. In microarray analysis, gclm and nqo1 were induced by DEM at levels of only 1.97-and 1.93-fold, respectively. Since they have generally been used as Nrf2 target genes in mammalian cells, we selected them together with gclc and hmox1, for further analysis, in addition to 42 identified genes (Table S4).
We next carried out RT-PCR analysis to confirm the results of microarray analysis. RT-PCR analysis was performed using RNA isolated from 4-dpf larvae treated with DEM and isolated RNA ( Figure 1 and Table S3). Thirty-three genes were analyzed out of 46 selected genes, and 19 genes were identified to be induced by DEM. To clarify whether these inductions were directed by Nrf2, we carried out RT-PCR analysis using Nrf2-knockdown larvae using nrf2-specific morpholino oligonucleotide. As a result, the induction of eleven genes was clearly demonstrated to be Nrf2dependent ( Figure 1).
A WISH analysis was carried out using eleven genes as probes ( Figure 2). Among them, seven genes (gstp1, mgst3b, prdx1, frrs1c, fthl, gclc and hmox1a) showed clear and strong induction in response to DEM, suggesting they will be useful for studying tissue specificity of Nrf2 target genes. The remaining four genes showed either weak induction (sepw2b and gclm) or strong constitutive expression (gstal and bcat1), thus indicating that they were unsuited for gene expression studies. We, therefore, used the seven genes showing a strong induction for further analyses.

Tissue-restricted induction of Nrf2 target genes
Tissue-restricted induction of seven Nrf2 target genes was examined by WISH (Figures 3 and S11), using 5-dpf larvae, since the induction was much clearer compared with that in 4-dpf larvae. As a result, the induction of mgst3b, prdx1, frrs1c, fthl, and gclc was observed in nose, gill, and liver, similar to gstp1, although the expression of frrs1c and gclc in the liver and nose, respectively, was relatively weak. Induction in the intestine was observed in the case of mgst3b, prdx1, frrs1c and gclc, but we did not take them into account since a considerable level of basal expression was detected. Interestingly, hmox1a was only induced in the liver, suggesting the existence of gene-specific differences in tissue specificity. Expression profiles of hmox1a and frrs1c in the liver were confirmed by section analysis in comparison with a liver-specific marker fabp10 ( Figure S11) [26]. These results indicate that tissue-restricted induction is a common characteristic among Nrf2 target genes, not only for gstp1.
The induction time profiles of seven genes were also analyzed in detail ( Figure 4). All genes showed induction beginning three hours after DEM treatment. Interestingly, expression of hmox1a was rapidly reduced beginning six hours after DEM treatment to almost the same level as the non-induced condition, whereas expression levels of the remaining six genes were maintained until twelve hours after treatment. A negative feedback regulation may exist in the case of hmox1a.
To determine whether tissue specificity and induction time profiles of Nrf2 target genes vary according to differences of Nrf2activating compounds, expression profiles of seven genes by sulforaphane were analyzed (Figures S12). Induction profiles of all seven genes were basically identical to those in the case of DEM, suggesting that tissue-restricted induction of Nrf2 target genes is intrinsic properties of each gene.

Gene regulation by Nrf2
Considering that all genes tested showed restricted-expression in the nose, gill, liver, and intestine, gene expression in these four tissues seemed to be a default state for Nrf2 target genes. This may suggest that Nrf2 or its activator dominantly exists in these tissues and directly transactivates the target genes. To test this hypothesis,  we analyzed the expression of nrf2 in 5-dpf larvae by WISH ( Figure 5). As expected, nrf2 is specifically expressed in the nose, gill, liver, and intestine, suggesting that tissue-restricted induction of Nrf2 target genes is based on the tissue-specific expression of nrf2 mRNA.
We previously demonstrated that the expression of gstp1 and gclc (cgcsh) to be strongly induced when Nrf2 is overexpressed in zebrafish embryos [9]. We next investigated whether the other five Nrf2 targets can also be induced by Nrf2 overexpression. Nrf2 was overexpressed by injecting nrf2 mRNA into one-cellstage embryos, and the expression of Nrf2 target genes was analyzed by RT-PCR seven hours after injection ( Figure 6). The results indicate that all seven genes were induced by Nrf2 overexpression. Using the zebrafish genome database (http:// www.ensembl.org/Danio_rerio/Info/Index), we searched the ARE sequences in the 5-kb upstream regions of the deduced transcription initiation sites in these genes (Figure 7). All genes were found to have more than three ARE sites in the 5-kb regions, to which Nrf2 may bind and regulate. All of these results suggest that induction of these seven genes is directly regulated by Nrf2.

Discussion
The zebrafish is a good system to observe the tissue-specific expression of many genes, including gene induction, as described in this study. Indeed, tissue-specific induction in systems other than Keap1-Nrf2, such as the heat-shock genes, has been reported from studies in zebrafish [27,28]. These observations would be difficult to analyze using cultured cell lines, demonstrating a significant advantage for zebrafish system. Zebrafish is also good for drug toxicity screening and testing environmental toxicants [29,30]. Since Nrf2 is activated by a variety of toxic compounds and oxidative stress, it would be worthwhile for these screens and tests to analyze tissue-restricted induction of Nrf2 target genes. The seven target genes of Nrf2 selected in this paper would be useful for such studies.  In our analysis, all tested genes showed tissue-restricted induction. Furthermore, we found gene-specific variation in tissue specificity, e.g., a weak induction of frrs1c and gclc in the liver and nose, respectively, and no hmox1a induction in the nose and gill. The requirements of metabolizing enzymes encoded by each target gene may be different among tissues and the level of enzymes are controlled at a transcriptional level. An important finding in this study is that the expression level of Nrf2 itself is different among tissues, which will be the critical point to exert tissue-restricted induction of its target genes. We hypothesized that the gene expression profiles of the Nrf2 gene defines a default state of tissue specificity of target genes. In cases of frrs1c, gclc and hmox1a, some tissue-and gene-specific transcriptional repressors may be involved to exert their tissue-specific variations. Nrf2-dependent ARE gene regulation has been shown to be negatively regulated by other AREbinding proteins such as Nrf1, Nrf3, Bach1, small Maf proteins, and c-Maf [31][32][33][34][35]. Since we have previously demonstrated that the target genes of Nrf-Maf proteins tend to differ somewhat among family proteins [36,37], it is possible that these ARE-acting factors may bind in a gene-specific manner and interfere with DNA binding of Nrf2. Furthermore, ATF3, c-Myc and p53 have also been reported to repress Nrf2-dependent gene activation [38][39][40]. Some of these transcription factors may bind near the ARE in a gene-and tissue-specific manner and inhibit the transcription activity of Nrf2.
In contrast to Nrf2 regulation at a post translational step, molecular mechanisms of Nrf2 gene regulation have not been extensively studied. The exception is the upregulation of the Nrf2 gene by Nrf2 itself and aryl hydrocarbon receptor in response to their chemical activators [41,42]. We previously reported that the embryonic expression of the Nrf2 gene is quite low both in the mouse and zebrafish and it becomes elevated near birth [12,43]. The downregulation of Nrf2 was found in prostate cancer which may be related to the initiation of cellular transformation [44]. A low Nrf2 expression in the brain has been reported in humans, mice, and chickens [45][46][47], as was also the case in our present study. For the efficient medical applications of the Keap1-Nrf2 system for the treatment of neurodegenerative diseases [48,49], Nrf2 should be considerably expressed in the brain. However, we found the expression of Nrf2 in the brain to be low. It will thus be valuable to find new methods and procedures to elevate the Nrf2 expression in both brain and prostate cancer cells. Figure S1 Phylogenetic tree of Gsta family proteins. Amino acid sequences of full-length proteins were analyzed. The tree was constructed by the neighbor-joining method using the ClustalW program (http://clustalw.ddbj.nig.ac.jp/top-j.html). c, chicken; h, human; m, mouse; r, rat; xt, Xenopus tropicalis; z, zebrafish.