Nuclear Heme Oxygenase-1 (HO-1) Modulates Subcellular Distribution and Activation of Nrf2, Impacting Metabolic and Anti-oxidant Defenses*

Background: A 28-kDa HO-1 isoform is induced by oxidative stress and cancer and accumulates in the nucleus. Results: Nuclear HO-1 interacts with Nrf2 and alters expression of its target genes. Conclusion: HO-1 modulates Nrf2 function. Significance: Exploiting the synergistic benefits of the HO-1·Nrf2 protein complex is important for developing therapeutic strategies against oxidative stress or cancer. With oxidative injury as well as in some solid tumors and myeloid leukemia cells, heme oxygenase-1 (HO-1), the anti-oxidant, anti-inflammatory, and anti-apoptotic microsomal stress protein, migrates to the nucleus in a truncated and enzymatically inactive form. However, the function of HO-1 in the nucleus is not completely clear. Nuclear factor erythroid 2-related factor 2 (Nrf2), a transcription factor and master regulator of numerous antioxidants and anti-apoptotic proteins, including HO-1, also accumulates in the nucleus with oxidative injury and in various types of cancer. Here we demonstrate that in oxidative stress, nuclear HO-1 interacts with Nrf2 and stabilizes it from glycogen synthase kinase 3β (GSK3β)-mediated phosphorylation coupled with ubiquitin-proteasomal degradation, thereby prolonging its accumulation in the nucleus. This regulation of Nrf2 post-induction by nuclear HO-1 is important for the preferential transcription of phase II detoxification enzymes such as NQO1 as well as glucose-6-phosphate dehydrogenase (G6PDH), a regulator of the pentose phosphate pathway. Using Nrf2 knock-out cells, we further demonstrate that nuclear HO-1-associated cytoprotection against oxidative stress depends on an HO-1/Nrf2 interaction. Although it is well known that Nrf2 induces HO-1 leading to mitigation of oxidant stress, we propose a novel mechanism by which HO-1, by modulating the activation of Nrf2, sets an adaptive reprogramming that enhances antioxidant defenses.

With oxidative injury as well as in some solid tumors and myeloid leukemia cells, heme oxygenase-1 (HO-1), the anti-oxidant, anti-inflammatory, and anti-apoptotic microsomal stress protein, migrates to the nucleus in a truncated and enzymatically inactive form. However, the function of HO-1 in the nucleus is not completely clear. Nuclear factor erythroid 2-related factor 2 (Nrf2), a transcription factor and master regulator of numerous antioxidants and anti-apoptotic proteins, including HO-1, also accumulates in the nucleus with oxidative injury and in various types of cancer. Here we demonstrate that in oxidative stress, nuclear HO-1 interacts with Nrf2 and stabilizes it from glycogen synthase kinase 3␤ (GSK3␤)-mediated phosphorylation coupled with ubiquitin-proteasomal degradation, thereby prolonging its accumulation in the nucleus. This regulation of Nrf2 post-induction by nuclear HO-1 is important for the preferential transcription of phase II detoxification enzymes such as NQO1 as well as glucose-6-phosphate dehydrogenase (G6PDH), a regulator of the pentose phosphate pathway. Using Nrf2 knock-out cells, we further demonstrate that nuclear HO-1-associated cytoprotection against oxidative stress depends on an HO-1/Nrf2 interaction. Although it is well known that Nrf2 induces HO-1 leading to mitigation of oxidant stress, we propose a novel mechanism by which HO-1, by modulating the activation of Nrf2, sets an adaptive reprogramming that enhances antioxidant defenses.
The nuclear factor erythroid 2-related factor 2 or Nrf2, 2 a member of the cap'n'collar family of basic leucine zipper tran-scription factors, plays a crucial role in antioxidant defenses (1,2). Nrf2 is constitutively expressed in the cytoplasm, and its accumulation and activation in the nucleus are favored in oxidative injury and also in cancer cells (3)(4)(5)(6). In the nucleus, Nrf2 in a heterodimer with small Maf proteins binds to the antioxidant response elements (ARE) in the promoter regions of Nrf2 target genes. This signaling facilitates transcription of numerous antioxidant enzymes and growth factors, thereby promoting tolerance to oxidative stress (2,6,7). In addition, Bach1, which competes with Nrf2 for binding to the ARE, suppresses ARE-mediated gene transcription (8). One important step in Nrf2 signaling is stabilization of the constitutive Nrf2 from Kelch-like ECH-associated protein 1 (KEAP1), which acts as an adaptor protein between Nrf2 via the Neh2 domain at the N terminus and Cul3 E3 ligase targeting Nrf2 for ubiquitin-proteasomal degradation. Many studies show that, with oxidative or electrophilic stress, both Nrf2 and Keap1 undergo posttranslational modification, favoring release of Nrf2 by the degradation of Keap1 thereby facilitating stabilization and nuclear transport of Nrf2 (9 -13). However, the regulation of Nrf2 is complex and involves a network of intricate mechanisms beyond Keap1 (14 -19). More recently it was reported that GSK3␤-mediated phosphorylation of Ser 344 and Ser 347 within the phosphodegron motif of the Neh6 domain targets Nrf2 protein for proteasomal degradation by another E3 ligase, ␤-TrCP, in a Keap1-independent manner (20 -22). This process is thought to occur in the nucleus post-induction.
It is well documented that heme oxygenase (HO)-1, a 32-kDa protein, is positively regulated by Nrf2 (1,23). As a protective and adaptive response, most tissues exhibit robust activation of the highly inducible HO-1 (24 -28). The beneficial effects of HO-1 in cytoprotection have long been attributed to its enzy-* This work was supported, in whole or in part, by National Institutes of Health matic action in heme degradation and also to the catalytic byproducts including carbon monoxide and biliverdin (29). More recently subcellular localization of HO-1 has been studied in the context of oxidative stress and cancer (30). Although HO-1 is an integral protein of the smooth endoplasmic reticulum (31,32), it can localize to other compartments including caveolae (33), mitochondria (34), and the nucleus (35,36), where it can mediate signaling functions (30).
We have previously described a more rapidly migrating isoform of HO-1 in the nuclear extracts of hypoxia-exposed cells transfected with N-terminally FLAG-tagged HO-1 (30). A negative reactivity for an antibody directed to the C terminus identified this isoform as a 28-kDa fragment of HO-1 missing 52 amino acids from the C terminus (36). This nuclear HO-1 was also found to be enzymatically inactive; however, it was associated with enhanced activation of transcription factor AP-1 (37). Earlier Hori et al. (38) showed that a mutated, enzymatically inactive form of HO-1 was still able to protect against oxidative injury. In addition, a HO-1 mutant with a deletion of the C-terminal amino acids was shown to bind to heme but could not degrade it to biliverdin (39). We documented that the enzymatically inactive HO-1 can alter its own transcription through activation of AP-1 (36,37). Because an AP-1 consensus sequence is found within the ARE, and since both ARE and API sequences are found on the human NAD(P)H:quinone oxidoreductase 1 (NQO1), HMOX1, and sulfiredoxin genes responding to both AP-1 and Nrf2 activation (40), we wondered whether the nuclear isoform of HO-1 could also regulate Nrf2 activation and what would be the consequences of this activation.
To evaluate this, we used hyperoxia exposure, a clinically relevant oxidative stress known to activate Nrf2 signaling (41)(42)(43) and induce HO-1, resulting in its migration to the nucleus (42). In the present study, mouse embryonic fibroblasts (MEFs) were used, as was a prostate cancer cell line (LnCap) because this cell line and human prostate cancer tissues have enhanced nuclear localization of HO-1 when compared with the normal prostate tissue (35). The present study demonstrates that nuclear HO-1 specifically interacts with Nrf2 in the nucleus to facilitate its sustained stabilization from GSK3␤-mediated proteolytic degradation. This leads to preferential activation of cytoprotective pathways for sustained tolerance against oxidative injury, as well as cell survival.

EXPERIMENTAL PROCEDURES
Cell Lines-MEFs, HO-1-knock-out (HMOX1 Ϫ/Ϫ , KO) and wild type (HMOX1 ϩ/ϩ , WT), were generated from embryonic day 13.5 embryos of the same littermates. These cells were immortalized by transfection of plasmids pSV3-neo (ATCC 37150) expressing the simian virus 40(SV40) large T-antigen and were maintained in 5% CO 2 , at 37°C, in DMEM with 10% (v/v) FCS and 1% P/S (100 units/ml penicillin as well as 100 g/ml streptomycin). To express full-length HO-1 (FL), murine HO-1 cDNA was tagged with 3ϫFLAG epitope at the C terminus and then cloned in a retroviral vector, pMX-puro, under the control of SV40 promoter. HO-1-KO cells were reconstituted with retrovirus containing empty vectors (control) or FL HO-1. This plasmid was further used as a template to generate and express truncated HO-1 (TR). In brief, the C-ter-minal 53 amino acid residues of the FL HO-1 were deleted, and three copies of a nuclear localization signal derived from SV40 large T-antigen were inserted to express truncated HO-1 (TR) specifically in the nucleus. High titer retroviruses were packaged with the helper plasmid containing the packaging sequence (psi) in 293T cells. Immortalized HO-1-KO and Nrf2-KO MEFs were infected with the retroviruses of empty vector (V) or vector containing FL or TR HO-1 cDNA and further selected with 2 g/ml puromycin. All retrovirus-complemented cells were maintained within 10 passages. In some experiments, we also used prostate cancer cell lines (LnCap), gifts from Dr. William J Fredericks, University of Pennsylvania. The Nrf2 null mutant MEF cell lines were also transfected with a cDNA expressing Nrf2 fused with an HA tag at the N terminus (44), gifts from Drs. Melpo Christofidou-Solomidou and Alan Diehl, respectively, of the University of Pennsylvania.
Stable Knockdown of HO-1 with shRNA-To knock down endogenous HO-1 expression, lentiviruses carrying shRNA targeting HO-1 mRNA were produced as per the manufacturer's instructions (RMM4534-NM-010442, Thermo Fisher Scientific Open Biosystems products). Based on the highest knockdown efficiency, clone 6 (71758) was chosen for the knockdown experiments, and clone 11, which did not reduce HO-1 expression, was used as a control (nonspecific) shRNA.
Preparation of Cell Lysates-Cells were harvested and homogenized in the presence of protease and phosphatase inhibitors either to extract whole cell homogenate using the M-PER mammalian protein extraction reagent (78501) or to extract cytoplasmic and nuclear fractions using NE-PER nuclear and cytoplasmic extraction reagents (78835). Both extraction reagents were obtained from Thermo Fisher Scientific.
Protein Content Estimation-The protein content in the whole cell lysate or nuclear and cytosolic fractions was measured by a 96-well plate based Bradford assay.
Immunohistochemistry-Cells (1 ϫ 10 4 /well in 2 ml) were seeded on coverslips placed in 6-well plates and grown overnight. The cells on coverslips after fixation with 3% paraformaldehyde were detected for compartmental localization of proteins as described previously (41,42) and visualized by fluorescent microscopy.
Immunoprecipitation (IP)-IP of FLAG fusion HO-1-FL and HO-1-TR from O 2 or air-exposed stably infected MEF cells was performed using the anti-FLAG M2 affinity gel (Sigma, A2220) following the manufacturer's protocol. In brief 100 -200 g/200 -300 l of total protein either from whole cell lysate or from cytoplasmic and nuclear fractions was incubated overnight with 20 l of packed volume of M2 beads in a cold room under slow rotation. After washing repeatedly we used 3ϫFLAG peptide (Sigma, F4799) at a final concentration of 150 ng/l in 50 l of TBS to elute the proteins. IP of endogenous Nrf2 was performed by overnight incubation with anti-Nrf2 (Santa Cruz Biotechnology, sc-722) and anti-pSer 40 -Nrf2 (Epitomics, 2073-1) antibodies at a concentration of 2-3 g/100 g of total protein. This was followed by further incubation with protein A-agarose beads (Invitrogen, 15918-014) at room temperature for an hour. After washing the beads repeatedly, the bound proteins from the beads were eluted with 0.1 M glycine HCl, pH 3.5, and immediately restored in neutral pH by adding 0.5 M Tris-HCl, pH 7.4, containing 1.5 M NaCl. Similarly, IP of recombinant HA-tagged Nrf2 in the nuclear fractions was performed by incubation with anti-HA antibody (Santa Cruz Biotechnology, sc-805) followed by protein-A-agarose beads, and the elution was performed by low pH glycine as described above.
Expression and Purification of Recombinant GST Fusion HO-1-FL and HO-1-TR-GST fusion constructs of HO-1-FL and HO-1-TR were expressed in competent Escherichia coli strain BL21 (Invitrogen) as described previously (36). Bacteria were grown to an optical density of 0.6 -0.8 at 600 nm. The fusion proteins thereafter were induced in the presence of 100 FIGURE 1. Oxidative stress induces nuclear enrichment of a 28-kDa HO-1 and a simultaneous enrichment of Nrf2 in the nucleus. A, the WT MEF cells were evaluated for HO-1 and Nrf2 signal after exposure to air or O 2 . B and C, the cytoplasmic and nuclear fractions of the above cells were further evaluated for HO-1 and Nrf2 signals (B), and normalization of these signals to calnexin and lamin B, respectively, is represented in the bar diagrams (C). D, the compartmental localization of HO-1 (green) and DAPI staining (red) with exposure to air (left column), O 2 (middle column), or incubation with N-acetyl-L-cysteine (NAC) (60 nM) in O 2 (right column) is shown in WT MEF cells; the yellow staining demonstrates co-localization in the overlay (bottom panel). E, after exposure to air or O 2 , the immunosignal of HO-1, Nrf2, and calnexin is shown in whole cell lysates of WT MEF cells silenced for HO-1 and in control cells incubated with nonspecific shRNA. F, the signal of Nrf2 (from E) is quantitated after normalization to calnexin and represented in bar diagrams. G, after exposure to O 2 for 2, 6, and 18 h, the subcellular proportion of HO-1 and Nrf2 and loading controls in the cytoplasm (lowermost three panels) and in the nucleus (upper three panels) of WT MEF cells silenced for HO-1 or control cells with nonspecific shRNA is shown in a representative Western blot. ૺ, p ϭ 0.05; values are the mean Ϯ S.E. of three separate determinations.
M isopropyl 1-thio-␤-D-galactopyranoside at 30°C for 4 h. The fusion proteins in the bacterial lysate were purified using a GST purification module (GE Healthcare, catalog number 18-1128-13AC). The eluted proteins were evaluated for protein content and stored in aliquots at Ϫ80°C until used.
In Vitro Translation of Full-length Nrf2 and C-terminally Truncated Nrf2 Fragments (⌬C1 to ⌬C4)-Mice Nrf2 ORF cloned in Pmx-Puro vector were used as a template to generate deletion mutants of Nrf2 (⌬C1 to ⌬C4, see Fig. 4B) serially truncated from the C terminus. The specific forward primer containing T7 promoter and the reverse primers are shown in a table (see Fig. 4D). DNA constructs were amplified by PCR. The PCR products of Nrf2 and the Nrf2 fragments were verified on agarose gel and purified using the QIAquick PCR purification kit (Qiagen, catalog number 28106) following the manufacturer's instructions. The pure PCR-generated constructs were transcribed in vitro by using the T7 high yield RNA synthesis kit (New England Biolabs, catalog number E2050) in the presence of T7 promoter enhancer according to the manufacturer's instructions, and the transcripts were detected on agarose gel. For in vitro translation, the transcripts were added to an aliquot of the TNT T7 coupled reticulocyte lysate system, (Promega, catalog number L4610) and incubated in a 50-l reaction volume containing [ 35 S]methionine instead of methionine for 60 -90 min at 30°C. After being resolved on SDS-PAGE, the radiolabeled Nrf2 and Nrf2 fragments were detected on the dried gel by autoradiogram. An in vitro translation reaction in the absence of DNA template was also carried out as a specificity control.
In Vitro Binding of Nrf2 to HO-1 Captured on GST SpinTrap Columns-Equal amounts of the purified GST fusion HO-1-FL and HO-1-TR were captured on several GST SpinTrap columns (GE Healthcare, catalog number 18-1128-13AC). The in FL and TR cells is shown. D, HO-1 signal in FLAG IP of total cell lysate from V, FL, and TR exposed to air or O 2 for 2, 6, and 18 h (upper panel) and Nrf2 co-precipitation (lower panel) are shown. E, co-staining of Nrf2 in FLAG-IP from cytoplasmic and nuclear fractions of V, FL, and TR cells exposed to O 2 is shown in the upper panel. A reverse IP of endogenous Nrf2 using anti-Nrf2 antibody (sc722) from the cytoplasmic and nuclear fractions was counterstained for HO-1 (middle panel), and input of Nrf2 and HO-1 is shown in 10% of the nuclear and cytoplasmic fractions (bottom two panels). F, the transient expression of HA-tagged Nrf2 (HA-N) in Nrf2-KO MEF whole cell lysate is shown. G, the HA-specific IP of HA-Nrf2 (HA-N) from these cells after O 2 exposure was first stained for Nrf2 (upper panel) followed by a counterstain for HO-1 (lower panel). The input of Nrf2 and HO-1 is also shown in 5% of the lysate. columns were washed repeatedly with phosphate buffer to remove any unbound protein. Each of these columns was then incubated for 10 min individually with 20 l of 35 S-labeled Nrf2 and fragments (⌬C1-⌬C4). The columns were then washed with copious amounts of phosphate buffer. The Nrf2 and Nrf2 fragments bound to trapped GST HO-1-FL or HO-1-TR were then eluted using 30 l of 10 mM reduced glutathione (supplied with the SpinTrap). Once the eluates were resolved on SDS-PAGE, the Nrf2 signal was evaluated by autoradiogram of dried gel. The nonspecific binding was verified by incubating radiolabeled Nrf2 on a GST SpinTrap without any HO-1 bound to it.
Use of Inhibitors-The cells were exposed to O 2 for 18 h followed by an additional 30-min incubation with an inhibitor of AKT (MK2206, ChemieTek, catalog number CT-MK2206) alone at 1.0 M or after co-incubation with an inhibitor of GSK3␤ (SB 216763, Sigma-Aldrich, catalog number S3442) at 10 M. The cell pellets were evaluated for Nrf2 and HO-1 signals by Western blot.
Evaluation of Oxidative Stress-After air or O 2 exposure, cells were incubated with 10 M dihydroethidium (DHE; Molecular Probes) for 30 min. The amount of oxyethidium generated was measured by fluorescent microscopic analysis. The number of fluorescence-positive cells was counted in an average of 10 high power fields and expressed as a ratio of the total number of cells.
Evaluation of Protein Carbonyls-The cell lysates were also analyzed for the protein carbonyls using OxiSelect protein carbonyl immunoblot kit (Cell Biolabs Inc., STA 309) as per the manufacturer's instructions. In brief, the carbonyl groups were derivatized by dinitrophenylhydrazine followed by immunoblotting with an anti-dinitrophenyl antibody.
G6PDH Enzyme Activity-Constitutive activity of glucose-6phosphate dehydrogenase (G6PDH) in whole cell lysate was detected in a 96-well plate using a kit (Sigma, MAK015). In brief, 10 l of lysate from 1 ϫ 10 6 cell homogenate in 100 l of PBS was transferred in triplicates in 40 l/well of master reaction mix. This assay detects oxidation of glucose 6-phosphate generating NADH, which is measured every 5 min colorimetrically at 450 nm using a standard curve of NADH. After subtracting the values from the background, the nmol amounts of NADH obtained between T initial and T final were finally expressed in milliunits/ml by employing the following formula where B ϭ amount (nmol) of NADH generated between T initial and T final , Reaction time ϭ T final Ϫ T initial (minutes), and V ϭ sample volume (ml) added to the well. Measurement of Cell Viability by XTT Assay-The XTT assay (Biotium, 30007) was performed following the manufacturer's instructions to measure the proliferation rate/metabolic activity of cells over a period of 6 h. Briefly the V, FL, and TR cells (20,000/well in a 96-microtiter plate) were grown in DMEM with high glucose (25 mM) or without any energy substrate in the presence of XTT reagent. XTT is a tetrazolium analog with a negatively charged inner salt that more readily enters cells. In metabolically active cells XTT is rapidly reduced by NAD(P)H-dependent oxidoreductases and dehydrogenases and produces solubilized formazan that is colorimetrically measured at 405 nm (45,46).
Statistical Analysis-The values in figures were expressed as means Ϯ S.E. Unpaired t tests were used to compare two groups. Values of p Ͻ 0.05 and p Ͻ 0.005 were considered as statistically significant.

RESULTS
Hyperoxia Induces HO-1, Leading to Nuclear Localization of a 28-kDa Isoform and Simultaneous Accumulation of Nrf2 in the Nucleus-The WT MEF cells exposed to air or O 2 were analyzed for HO-1 and Nrf2 proteins in Western blot (Fig. 1). In air, HO-1 in the whole cell lysate appeared as a single band at 32 kDa, whereas in O 2 , an additional faster migrating band also appeared at 28 kDa, which was accompanied with a simultaneous increase of Nrf2 (Fig. 1A). The 32-kDa isoform constitu- tively was predominant in the cytoplasm, whereas in O 2 , the 28-kDa HO-1 was predominant and primarily localized to the nucleus. The latter was accompanied with a concomitant increase of nuclear Nrf2 (Fig. 1, B and C). The immunohistochemistry further verified an O 2 -driven nuclear enrichment of HO-1 (Fig. 1D). N-Acetyl-L-cysteine, an antioxidant, reversed the O 2 -mediated nuclear shift of HO-1, suggesting that the migration of HO-1 to the nucleus is driven by oxidative stress.
Because we observed a coordinated nuclear enrichment of Nrf2 along with the enrichment of the 28-kDa HO-1, the latter was silenced (Fig. 1E). This in turn significantly reduced both basal and O 2 -induced Nrf2 (Fig. 1, E and F). Furthermore, nuclear HO-1 increased in a time-dependent manner with exposure to O 2 being maximal at 18 h (Fig. 1G). This subsequently was accompanied with a nuclear accumulation of Nrf2, which was reversed when HO-1 was silenced (Fig. 1G). In the cytoplasm of the HO-1-silenced cells, Nrf2 was decreased as well (Fig. 1G). These data indicate that HO-1 modulates stabilization and nuclear accumulation of Nrf2.
Nuclear HO-1 Physically Interacts with Nrf2-Nrf2 accumulation is predominantly regulated by post-translational mechanisms (17,44,47). To understand whether the nuclear accumulation of Nrf2 impacted by nuclear HO-1 was dictated by a physical interaction between the two proteins, HO-1-KO-MEF cells reconstituted with HO-1-FL or HO-1-TR and the empty vector (V) were used ( Fig. 2A). The HO-1-TR was expressed in the nucleus due to the presence of three nuclear localization sequences, whereas the HO-1-FL expression was restricted to the cytoplasm (Fig. 2B). The lack of migration of FL protein to the nucleus could have been due to a resistance to proteolytic cleavage of the protein from the C terminus and release from the membrane. Alternatively, this may have resulted from a structural change due to the presence of the N-terminal Myc and C-terminal FLAG tags. The cytoplasmic and nuclear compartment-specific expression of HO-1-FL and HO-1-TR, respectively, was further documented by immunostaining using three different antibodies directed at HO-1, N-terminal Myc, and C-terminal FLAG tags (Fig. 2C).
The Nuclear HO-1⅐Nrf2 Complex Appears Late in Oxidative Stress-After exposure of the FL, TR, and the control (V) cells to air or O 2 for 2, 6, and 18 h, the FLAG fusion proteins were immunoprecipitated from whole cell lysate. In comparison with the air-exposed cells, the IP of the O 2 -exposed cells showed increased abundance of both HO-1-FL and HO-1-TR at all time points (Fig. 2D). Counterstaining indicated co-precipitation of Nrf2, which in O 2 appeared most abundantly with HO-1-TR at 18 h, whereas a comparatively weak signal was observed with HO-1-FL. Regardless of exposure, Nrf2 signal was absent in the IP from control cells lacking HO-1 (Fig. 2D). These data suggested that a preferential interaction exists between nuclear HO-1 and Nrf2. The absence of the Nrf2 signal at early time points (2 and 6 h) further indicated that this interaction occurs late during oxidative stress.
The HO-1⅐Nrf2 Complex Localizes to the Nucleus-To further assess compartment-specific localization of the nuclear HO-1⅐Nrf2 complex, FLAG IP from cytoplasmic and nuclear fractions of cells exposed to O 2 for 18 h was performed. The co-precipitation of Nrf2 was strongest in the nucleus with HO-1-TR, indicating localization of HO-1-TR⅐Nrf2 complex predominantly in the nucleus (Fig. 2E). The IP of HO-1-FL, abundant in the cytoplasm, weakly counterstained with Nrf2, and the control (V) cells were completely devoid of Nrf2 signal. This further supported the preferential interaction between 28-kDa nuclear HO-1 with Nrf2 and predominant localization of the complex in the nucleus. A reverse IP of the endogenous Nrf2 using anti-Nrf2 antibody (sc772) further confirmed this preferential interaction between Nrf2 and nuclear HO-1 (Fig.  2E). We did observe a considerable amount of HO-1-TR in the cytoplasm, as shown by its co-precipitation with Nrf2 and in the input lane as well. Although this may have resulted from contamination of cytoplasm with the nuclear portion during fractionation, this could also suggest that Nfr2 interacts with HO-1 in the cytoplasm prior to migration to the nucleus as a complex. An IP of HA-tagged Nrf2 (HA-N) ectopically expressed in Nrf2-KO MEF cells (Fig. 2F) further confirmed the interaction of endogenous 28-kDa HO-1 with Nrf2 (Fig.  2G). As expected, this interaction was absent in Nrf2-KO-MEF cells.

The LnCap Prostate Cancer Cells Constitutively Are Enriched with the 28-kDa HO-1 That Resides in the Nucleus and Forms a
Complex with Nrf2-To further evaluate whether the HO-1/ Nrf2 interaction is physiological, we used LnCap prostate adenocarcinoma cell lines. The whole cell lysate exhibited predominantly a 28-kDa HO-1 isoform (Fig. 3, A and B), which upon fractionation of cytoplasm and nucleus exclusively localized to the nucleus (Fig. 3, C and D). The Nrf2 signal as well was significantly increased in the nucleus when compared with cytoplasm (Fig. 3, C and D). The Nrf2 IP from LnCap cells showed co-precipitation of HO-1 particularly of the 28 kDa (Fig. 3E). Because phosphorylation of Nrf2 at Ser 40 is known to modulate Nrf2 activity (17), an antibody specific to pSer 40 (Epitomics, 2073-1) was used to immunoprecipitate Nrf2. Although the IP showed abundant Nrf2, HO-1 co-precipitation was absent (Fig.  3E). These data indicated that Nrf2, if phosphorylated at Ser 40 , does not interact with nuclear HO-1 and hence does not participate in Ser 40 -mediated Nrf2 signaling.
Nuclear HO-1 Interacts with Nrf2 via the Neh4 Transactivation Domain-The purified GST fusion HO-1-FL and HO-1-TR proteins expressed in bacteria are presented in Fig. 4A. Nrf2 and Nrf2 fragments (⌬C1 to ⌬C4) generated with an in vitro translation system are shown stepwise in Fig. 4, B-F. The binding of Nrf2 and Nrf2 fragments with HO-1 was verified with a binding assay using purified GST fusion HO-1-FL and HO-1-TR proteins trapped on GST columns (Fig. 4G). The binding of the [ 35 S]methionine-radiolabeled Nrf2 fragments to GST-HO-1 revealed that Nrf2 could bind to HO-1-TR and that this binding did not occur with HO-1-FL. Furthermore, the presence of binding signal for ⌬C1 to ⌬C3 and the absence of this signal with ⌬C4 indicated that the transactivation domain Neh4 is crucial for the interaction between Nuclear HO-1 and Nrf2.
Nuclear HO-1 Stabilizes Nrf2 from GSK3␤-mediated Degradation-Active GSK3␤ phosphorylates serine moieties (343 and 347) in the Neh6 domain of Nrf2 (Fig. 5A) and targets Nrf2 for ubiquitination and degradation (20). To understand whether HO-1-TR protects Nrf2 from GSK3␤-mediated proteolytic degradation, AKT, the negative regulator of GSK3␤, was inhibited by MK2206. This in turn results in the activation of GSK3␤ by preventing phosphorylation at Ser-9 (48,49). The phosphorylation signals of AKT and GSK3␤ were verified with specific antibodies (pSer 473 -AKT, Cell Signaling, 9272; and pSer 9 -GSK3␤, Cell Signaling, 9832) in FL and TR cells after exposure to O 2 followed by an additional 30-min incubation with MK2206 (Fig. 5B). In the presence of MK2206, when GSK3␤ activity is increased, Nrf2 signal was reduced in FL, whereas it remained unaltered in TR. This observation was further verified in the next experiment where FL and TR cells exposed to O 2 were treated similarly with MK2206 alone or in co-incubation with SB16763, an inhibitor of GSK3␤. When compared with the Nrf2 level in FL cells incubated only with DMSO (Fig. 5, C and D), the FL cells incubated with MK2206 showed significant loss of Nrf2, indicating that activation of GSK3␤ results in degradation of Nrf2. Restoration of this loss by co-incubation with SB16763 verifies that this effect is dependent on GSK3␤. On the contrary, in TR cells, Nrf2 level was relatively higher and retained regardless of inhibitor treatment (Fig. 5, C and D). FIGURE 6. Cells enriched with nuclear HO1 exhibit a significant increase in transcription and activation of Nrf2 target genes. A, HO-1-silenced and control MEF cells with scrambled shRNA were exposed to air or O 2 and analyzed for steady state mRNA levels of HO-1, Nrf2, NQO1, and G6PDH by quantitative real time PCR. B, similarly, Nrf2, NQO1, and G6PDH mRNA levels were evaluated in V and TR cells. C and D, constitutive G6PDH enzyme activity in control (scrambled shRNA) and HO-1-silenced-MEF cells (C) as well as in V, FL, and TR cells (D) is shown. E and F, cell viability or rate of metabolic activity of V, FL, and TR as evaluated using XTT reagent for indicated times in medium devoid of energy substrates (E) or in glucose-enriched medium (25 mM) is shown (F). ૺ, p ϭ 0.05; ૺૺ, p ϭ 0.005. Values are the mean Ϯ S.E. of three separate measurements.
This validates the role of HO-1-TR in rescuing Nrf2 from GSK3␤-mediated degradation.
Because ubiquitination is a general mechanism to target proteins for proteasomal degradation, this was evaluated in FL, TR, and control cells (Fig. 5E). Interestingly, maximum Nrf2 coprecipitation signal with FLAG-HO-1-TR IP was associated with minimal polyubiquitination signal, and the FLAG IP from FL and control cells showed less Nrf2 co-precipitation yet maximum polyubiquitination signal. The input samples of total cell lysates from TR also exhibited less polyubiquitin signal when compared with FL and control cells (Fig. 5E). These data could reflect sustenance of protein homeostasis and limited ubiquitination in TR. Collectively, these data support the hypothesis that HO-1-TR prevents Nrf2 ubiquitination by physically interacting with the protein and thus preventing it from GSK3␤mediated degradation, thereby facilitating Nrf2 accumulation in the nucleus.
Nuclear HO-1-mediated Stabilization of Nrf2 Regulates Transcription of Specific Downstream Antioxidants and Metabolic Genes-To understand the physiologic importance of HO-1-mediated nuclear accumulation of Nrf2, we evaluated steady state mRNA levels of HO-1, Nrf2, and a group of Nrf2 downstream target genes including NQO1 and G6PDH in WT-MEF and HO-1-silenced MEF cells (Fig. 6A). As expected, after exposure to air or O 2 , HO-1 mRNA was suppressed in the HO-1-silenced cells. In O 2 -exposed WT cells, levels of NQO1 mRNA and G6PDH mRNA increased concomitantly with the increase in HO-1. Disruption of this signal in HO-1-silenced cells indicated that regulation of the Nrf2 downstream genes is dependent on HO-1. Of note, Nrf2 mRNA induction was also increased in O 2 exposure, indicating that nuclear HO-1 may influence Nrf2 transcription as well (Fig. 6A). In addition, in HO-1-TR cells, there was significant induction of NQO1 and G6PDH mRNA levels both at baseline and after O 2 exposure when compared with control cells (Fig. 6B).
Cells Enriched with Nuclear HO-1 Exhibit Increased G6PDH Activity-Because HO-1-TR was associated with enhanced transcription of G6PDH in TR cells and disruption of HO-1 in WT cells diminished it, constitutive G6PDH activity was assessed in these cells. In fact, the basal G6PDH activity observed was indeed decreased significantly with the disruption of HO-1 by shRNA (Fig. 6C). Additionally, G6PDH activity was significantly higher in TR when compared with FL and control cells (Fig. 6D). These data further indicate that modulation of G6PDH activity by Nrf2 is particularly influenced by nuclear HO-1.
Cells Enriched with Nuclear HO-1 Are Metabolically More Viable in Glucose-enriched Medium-Increased activity of G6PDH should improve growth and proliferation of cells in glucose medium. This was tested in V, FL, and TR cells. Cells grown in medium without any energy substrate as expected did not grow (Fig. 6E). In glucose-enriched (25 mM) medium, TR cells showed proliferation significantly at a higher rate when compared with FL and control cells (Fig. 6F).
Nrf2 Requires Participation of Nuclear HO-1 to Induce Antioxidant Defenses-Because NQO1 by utilizing NAD(P)H and NADH plays a major role in scavenging toxic oxidized substrates, we measured the oxidation burden in both air-exposed and O 2 -exposed V, FL, and TR cells by measuring the intensity of 2-hydroxyethidium fluorescence in cells after incubation with DHE and also assessed protein carbonyls as indices of protein oxidation. All cell groups showed low DHE fluorescence in air, whereas in O 2 , only TR showed low levels of DHE fluorescence when compared with V and FL cells (Fig. 7, A and B).
To understand whether the enhanced anti-oxidant protection in the TR cells was by virtue of nuclear HO-1 interaction with Nrf2, we used Nrf2-WT and KO cells as such and also infected with cDNA constructs of V, FL, and TR. Fig. 7, C and D, indicate the levels of HO-1-TR and HO-1-FL expression in both Nrf2 WT and KO cells. We then evaluated levels of carbonylated proteins in these cells after exposure to air or O 2 (Fig.  7E). The protein carbonyls did not differ much between the types of cells in air. However, in O 2 , Nrf2-WT cells with HO-1-TR showed reduced levels of carbonylated proteins when compared with those having HO-1-FL and empty vector, whereas the Nrf2-KO cells exhibited enhanced protein carbonyls, which were not reversed even by HO-1-TR or HO-1-FL.
Furthermore, Nrf2-WT cells containing HO-1-TR showed significantly increased transcription of NQO1 (Fig. 8A) and G6PDH (Fig. 8C) when compared with those with HO-1-FL and empty vector. The fact that NQO1 and G6PDH transcription was lower in Nrf2-KO background even in the presence of HO-1-TR (Fig. 8, B and D) further supports our hypothesis that nuclear HO-1 partners with Nrf2 or vice versa to enhance antioxidant defenses. Nevertheless, steady state levels of SOD2 were not modulated by the interaction of nuclear HO-1 and Nrf2 (Fig. 8, E and F).

DISCUSSION
We demonstrate in various ways that oxidative stress mediates induction of HO-1 along with nuclear localization of a 28-kDa isoform of HO-1, which physically interacts with Nrf2. This interaction results in the stabilization of Nrf2 in the nucleus. We further established that under physiologic conditions, Nrf2 interacts with the 28-kDa endogenous nuclear HO-1. Many have described how Nrf2 results in the induction of HO-1 via binding at the ARE (25,50), but this is the first report of the converse, that is, the modulation of Nrf2 function by HO-1.
Participation of various kinase signaling pathways has been documented in ARE-mediated transcription (52). In particular, phosphorylation of Ser 40 by protein kinase C has been implied for the modulation of Nrf2 in a Keap1-dependent manner (17,18). In the present study, neither isoform of HO-1 interacted with Nrf2-pSer 40 . Nevertheless, activation of PI3K/Akt is well documented to provide protection from oxidative stress-medi- ated injury via Nrf2 activation (1). The PI3K-Akt-GSK3␤ axis may regulate Nrf2 activation in several ways. Active GSK3␤ can directly phosphorylate the phosphodegron motif in the Neh6 domain of Nrf2 protein, leading to Keap1-independent and ␤-TrCP-dependent proteasomal degradation of Nrf2 protein (20 -22). Alternatively, GSK3␤ can phosphorylate and activate Fyn kinase, leading to phosphorylation and nuclear exclusion of Nrf2 protein (53). Therefore, phosphorylation of Nrf2 might dictate its accumulation and nucleocytoplasmic trafficking (52).
A cellular model where the truncated form of HO-1 is constitutively overexpressed in the nucleus of cells devoid of endogenous HO-1 allowed for documentation of a distinct function of nuclear HO-1, specifically that it preferentially binds to Nrf2 and influences its nuclear abundance. Accumulation of nuclear HO-1 and consequent interaction with Nrf2 may take up to 18 h, indicating a specific role for nuclear HO-1 in the regulation of Nrf2 post-induction. Nuclear enrichment of Nrf2 was observed in TR cells even with GSK3␤ activation, whereas in FL cells lacking nuclear HO-1, this resulted in loss of Nrf2. These data document specifically that by interacting with Nrf2, nuclear HO-1 stabilizes it from GSK3␤-mediated proteolytic degradation. The observed decreased polyubiquitin signal in the lysates of O 2 -exposed TR cells could reflect sustained protein homeostasis by limited ubiquitination, although this would need to be systematically evaluated.
HO-1 has long been identified as proto-oncogenic due to its anti-apoptotic and pro-angiogenic properties (54,55), and there is a positive relationship between high nuclear HO-1 content and severity of prostate cancer (35), Nonetheless, the precise role of HO-1 in cancer biology is far from being completely understood. Cancer cells display altered metabolic circuitry and nutrient uptake to increase proliferation preferentially, leading to glucose-dependent ATP production (Warburg theory). Increased expression and activity of G6PDH are frequently observed in various cancers (56 -58). It is the first key rate-limiting enzyme in the pentose phosphate pathway, which provides a readily available supply of pentose sugars for RNA and DNA synthesis (59,60). Also, by regulating the pentose phosphate pathway, G6PDH generates NAD(P)H, which provides reducing equivalents for the maintenance of a pool of reduced glutathione to balance redox state (61). Additionally, by utilizing NAD(P)H, NQO1 scavenges a wide variety of oxidized and toxic substrates such as quinones (62,63). This should provide protection against oxidative stress. In accordance, the cells enriched with nuclear HO-1 show increased NQO1 transcripts and decreased oxidative damage of proteins and an increased G6PDH activity as well, and this indicates that these effects are attained by the interaction of nuclear HO-1 with Nrf2.
Interestingly, nuclear HO-1-mediated activation of Nrf2 led to the transcription of specific downstream genes including Nrf2 itself. This is explained by the fact that the pattern of resultant downstream genes depends on the alignment of the ARE and AP1 primary core sequences that Nrf2 binds to (51,64). How the binding of Nrf2 to HO-1 mediates preferential transcription of NQO1 and G6PDH is still not clear. In addition, the fact that nuclear HO-1 may also regulate the transcription of Nrf2 needs further verification.
In summary, we demonstrate that with oxidative stress, there is increased expression of nuclear HO-1. This regulates subcellular distribution and activation of Nrf2 post-induction, resulting in transcriptional regulation of late phase II antioxidant enzymes and enabling a feed forward adaptive reprogramming for recovery and a survival advantage in oxidative stress.