The Aryl Hydrocarbon Receptor Nuclear Transporter Is Modulated by the SUMO-1 Conjugation System*

The aryl hydrocarbon receptor nuclear transporter (ARNT) is a member of the basic helix-loop-helix/PAS (Per-ARNT-Sim) family of transcription factors, which are important for cell regulation in response to environmental conditions. ARNT is an indispensable partner of the aryl hydrocarbon receptor (AHR) or hypoxia-induc-ible factor-1 (cid:1) . This protein is also able to form ho-modimers such as ARNT/ARNT. However, the molecular mechanism that regulates the transcriptional activity of ARNT remains to be elucidated. Here, we report that ARNT is modified by SUMO-1 chiefly at Lys 245 within the PAS domain of this protein, both in vivo and in vitro . Substitution of the target lysine with alanine

The aryl hydrocarbon receptor nuclear transporter (ARNT) 1 belongs to the basic helix-loop-helix (bHLH)/PAS (Per-ARNT-Sim) family of proteins. These transcription factors are required for cell regulation to respond to various environmental conditions (1,2). The bHLH/PAS proteins include the aryl hydrocarbon receptor (AHR) and hypoxia-inducible factor-1␣ (HIF1␣). ARNT is an indispensable partner of these proteins for the formation of heterodimers such as AHR/ARNT and HIF1␣/ARNT. Both transcription pathways are not only biologically significant, but represent remarkable regulatory mechanisms in vivo. First, polycyclic aromatic hydrocarbons such as 2,3,7,8-tetrachlorodibenzo-p-dioxin and 3-methylcholanthrene are exogenous ligands for AHR and induce the formation of the AHR⅐ARNT complex (3). In the absence of ligands, AHR is generally found in the cytoplasm in association with hsp90 and other molecules (3,4). Upon the binding of a ligand, AHR is converted to a functional DNA-binding species in a multistep process involving nuclear translocation, dissociation from the hsp90-containing complex, and dimerization with ARNT. The resulting AHR⅐ARNT complex binds a specific cis-acting regulatory DNA sequence, termed the xenobioticresponsive element (XRE), upstream of its target genes encoding drug-metabolizing enzymes such as cytochrome P450 (CYP1A1 and others), quinone reductase, and the glutathione S-transferase (GST) Ya subunit. Xenobiotic-activated AHR is then degraded by the ubiquitin/proteasome system after being exported from the nucleus to the cytoplasm (5). Mice deficient in the AHR gene demonstrate dioxin-induced cytotoxicities, including developmental defects and tumorigenesis, depending on the AHR/ARNT pathways (6,7). Second, ARNT, also known as HIF1␤, forms a HIF1␣/ARNT heterodimer in response to oxygen tension in the cells. Under hypoxic conditions, it activates the transcription of a number of target genes whose promoters contain the binding motif termed the hypoxia-responsive element (HRE) (8). These include genes encoding erythropoietin, vascular endothelial growth factor, glycolytic enzymes, tyrosine hydroxylase, inducible nitric-oxide synthase, and heme oxygenase-1, all of which allow the cells to cope with lower oxygen levels. In addition, HIF1␣/ARNT controls the gene expression involved in iron metabolism, pH regulation, cell proliferation and apoptosis, and tumorigenesis. The HIF1␣ protein is rapidly degraded by the ubiquitin/proteasome pathway under normoxic conditions (9,10). In a hypoxic state, it becomes stable and translocates into the nucleus. It then dimerizes with ARNT to activate transcription. Thus, bHLH/ PAS family members have significant features of their regulatory mechanisms, and ARNT is implicated in many signaling pathways mediated by AHR or HIF1␣ as its partner.
In addition to the heterodimerization with AHR or HIF1␣, ARNT is likely to form a homodimer with itself to bind the E-box core sequence CACGTG with high specificity and affinity, suggesting a physiological role of the ARNT⅐ARNT complex (3,39). Furthermore, the t(1,12)(q21;p13) translocation of human acute myeloblastic leukemia results in a fusion protein containing the amino-terminal part of TEL (translocated ETS leukemia, also known as ETV6) and almost all of ARNT (40). The activity of ARNT may directly contribute to leukemogenesis. It is therefore of considerable interest to study the function of ARNT and the mechanism that regulates the transcriptional role of this protein. During the investigation of ARNT in cultured mammalian cells, we discovered that ARNT is modified by SUMO-1 chiefly at Lys 245 within the PAS domain, which is required for forming the complexes with bHLH/PAS and other molecules (3,41,42). ARNT associated with PML or PML-NBs, and the sumoylation of ARNT inhibited both the ability of ARNT to interact with PML and the positive effect of PML on the transactivation by ARNT. These data suggest the importance of the SUMO-1 conjugation system in modulating ARNT-mediated gene expression. Cell Lines and Cultures-MCF-7, HeLa, and Hepa-1 cells were grown in Dulbecco's modified Eagle's medium and Ham's F-12 medium (Invitrogen) supplemented with 10% (v/v) heat-inactivated fetal calf serum.
Western Blot Analysis-Cell lysates were prepared with SDS sample buffer (60 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 0.005% brom-phenol blue, and 100 mM dithiothreitol). Samples were separated on an 8% SDS-polyacrylamide gel and transferred to nitrocellulose filters with a constant current of 140 mA for 2 h. The filters were blocked with phosphate-buffered saline (PBS) containing 10% skim milk and then incubated with the appropriate antibodies in PBS containing 0.03% Tween 20 for 2 h and washed three times for 7 min with PBS containing 0.3% Tween 20. The filters were incubated with horseradish peroxidase-conjugated secondary antibodies for 40 min, and then specific proteins were detected using the ECL system (Amersham Biosciences).
Nickel Affinity Pull-down Assay-MCF-7 cells (30 -50% confluent in six-well plates) were transfected with 1 g of each of the expression vectors for His 6 -FLAG-ARNT (wild-type (WT) or K245A) and HA-SUMO-1 per well using FuGENE 6 (Roche Molecular Biochemicals). At 48 h after transfection, the cells were washed with ice-cold PBS, harvested in 200 l of Gua8 buffer (6 M guanidine HCl, 100 mM NaCl, 10 mM Tris, and 50 mM NaH 2 PO 4 (pH 8.0)) (22) per well, briefly sonicated, and then centrifuged. The cell lysates were incubated for 2 h with 20 l of ProBond nickel-chelating resin (Invitrogen). Bound proteins were washed twice with Gua8 buffer; three times with buffer containing 8 M urea, 100 mM NaCl, and 50 mM NaH 2 PO 4 (pH 6.5)) (22) with 10 mM imidazole; and once with cold PBS before being eluted by boiling in SDS sample buffer. Transfected cells in one well were lysed directly with SDS sample buffer as an input sample. All samples were electrophoresed on an 8% SDS-polyacrylamide gel, followed by Western blot analysis.
Luciferase Assay-MCF-7 cells were plated at a density of 6ϳ8 ϫ 10 5 cells/ml in a six-well plate and cultured for 24 h prior to transfection. The cells were introduced with 0.5 g of reporter plasmid and 0.1 g of pRL-CMV (Promega), which was used for monitoring the transfection efficiency, together with the indicated expression plasmids by Trans-Fast (Promega). The transfected cells were treated with 10 M 3-methylcholanthrene (Aldrich) and 200 M cobalt chloride (Wako Bioproducts) or solvent alone for 12 h during the incubation. At 48 h after transfection, the cells were lysed in the lysis buffer provided by the manufacturer (Promega). The insertless pcDNA3 was used as a mock vector. The luciferase activities were determined with the dual-luciferase reporter assay system. Values are the means Ϯ S.D. of the results from at least three independent experiments.
Confocal Laser Scanning Microscopic Analysis-For immunofluorescence analysis, MCF-7 and Hepa-1 cells were transfected with the indicated vectors using FuGENE 6. At 48 h after transfection, the cells were fixed with 4% paraformaldehyde in PBS for 10 min and permeabilized with 0.2% Triton X-100 in PBS for 5 min. After washing with PBS, the cells were incubated with specific primary antibodies at room temperature for 60 min. The samples were incubated with fluorescein isothiocyanate (BIOSOURCE)-or Cy3 (Amersham Biosciences)-conjugated secondary antibodies for 60 min and visualized with an Olympus confocal laser scanning microscope. To avoid bleed-through effects in the double-staining experiments, each dye was independently excited, and images were electronically merged.
GST Pull-down Assay-The pcDNA3-FLAG-ARNT or mock vector was transfected into MCF-7 cells with or without pCGN-HA-SUMO-1 and pCGN-HA-UBC9 using FuGENE 6. At 48 h after transfection, the cells were lysed with binding buffer (10 mM HEPES (pH 7.5), 50 mM KCl, 2.5 nM MgCl 2 , 50 M ZnCl 2 , 2.5 mM dithiothreitol, 0.025% Nonidet P-40, and 5% glycerol), briefly sonicated, and centrifuged. The supernatants were incubated with GST-PML or GST at 4°C for 2 h. 25 l of glutathione-agarose beads as a 50% slurry in binding buffer was added to the mixture. After incubation at 4°C for 1 h, the beads (complexed with proteins) were washed five times with binding buffer and then boiled in SDS sample buffer.

Sumoylation of ARNT at Lysine 245 in Vivo and in Vitro-
The fact that several transcription factors are modified by SUMO-1 prompted us to test whether ARNT possesses the consensus sequence for SUMO-1 conjugation. Because amino acids 244 -247 and 545-548 of murine ARNT coexist with the minimal consensus sequence, Lys 245 and Lys 546 seemed to be putative SUMO-1 acceptor sites in ARNT (Fig. 1A). Both lysine residues of ARNT are perfectly conserved among species, including human, murine, and rat. To test whether ARNT can be conjugated with SUMO-1 and to determine the target site(s) of the SUMO-1 modification in ARNT, we generated three plasmids to express ARNT mutants in which Lys 245 and/or Lys 546 was substituted with alanine: ARNT K245A, ARNT K546A, and ARNT K245A/K546A. Wild-type ARNT and its mutants fused to FLAG were each expressed together with HA-tagged SUMO-1 in MCF-7 cells. Both MCF-7 and HeLa cells, used in this study, have been proven to maintain ARNT-related signaling pathways to target genes ( Fig. 1B) (46). HA-SUMO-1 was efficiently utilized to modify substrate proteins in the cells (data not shown). Western blot analysis of whole cell extracts using anti-FLAG monoclonal antibodies revealed that a high molecular mass band of a covalently modified form (ϳ150 kDa) was detected in both wild-type ARNT and ARNT K546A upon coexpression of HA-SUMO-1. In contrast, the ARNT K245A and ARNT K245A/K546A modifications were rarely found. Wild-type ARNT and its mutants were equally expressed as shown by the presence of their unconjugated form (ϳ90 kDa). These data suggest that a fraction of ARNT is modified, probably by SUMO-1, chiefly at Lys 245 , which exists in the PAS domain (Fig. 1A).
To confirm that the covalently modified band in Fig. 1B is a sumoylated form of ARNT, we expressed His-FLAG-tagged wild-type and mutant ARNT in the presence or absence of HA-SUMO-1 to examine the precipitates on nickel beads using anti-FLAG and anti-HA antibodies (Fig. 1C). The 150-kDa modified band of wild-type ARNT (but not of the K245A mutant) was seen in the cells expressing HA-SUMO-1 (left panel). Western blot analysis of the nickel bead precipitates using anti-HA antibodies indicated that the slowly migrating band originated in sumoylated ARNT (right panel). Therefore, these data suggest that Lys 245 of ARNT is the major site of SUMO-1 modification in the cells. In addition, the size of the apparent SUMO-modified ARNT did not match that expected upon addition of a single SUMO protein. As described below, sumoylated ARNT in Fig. 1 (B and C) corresponded to band b in Fig. 2.
To further clarify the sumoylation of ARNT, we constituted

SUMO-1 Conjugation of ARNT
in vitro the SUMO-1 transfer using recombinant proteins. Wild-type ARNT and ARNT K245A (amino acids 145-345) were prepared as GST fusion proteins. The reaction was performed in the presence of ATP using Sua1/UBA2, UBC9, GST-ARNT (WT or K245A), and His-FLAG-SUMO-1. GST pulldown assay was performed using glutathione-agarose beads (Fig. 1D). In addition to the unconjugated form, the sumoylated form of wild-type ARNT was detected by anti-GST antibodies (left panel). The use of anti-FLAG antibodies showed that wildtype ARNT formed an adduct with FLAG-SUMO-1 (right panel). The ARNT K245A mutant showed little acceptance for the SUMO-1 molecule, although unconjugated GST-ARNT WT and GST-ARNT K245A were comparably present. Collectively, these data provide evidence for SUMO-1 modification of ARNT at Lys 245 . ARNT Accepts at Least Three SUMO-1 Molecules in Vivo-As mentioned above, UBC9 and PIAS1 are believed to be key factors in the SUMO-1 conjugation system. To investigate the effect of these enzymes on the modification of ARNT, we expressed HA-tagged UBC9 and/or FLAG-fused PIAS1 in combination with FLAG-ARNT and HA-SUMO-1 in MCF-7 cells ( Fig. 2A). The unmodified form of ARNT appeared as an ϳ90-kDa band (lanes 2-8). As was the case in Fig. 1, ARNT was covalently conjugated upon expression of SUMO-1 (band b (150 kDa) in lane 3). The fact that overexpressed ARNT alone did not induce the sumoylated form might be explained by the limited amount of free endogenous SUMO-1 (as described under "Discussion"). In agreement with this, the coexpression of UBC9 or PIAS1 did not induce the modified form of ARNT (lanes 4 and 5). Furthermore, combination of SUMO-1 with UBC9 or PIAS1 enhanced the SUMO modification of ARNT (lanes 6 and 7), although the coexistence of UBC9 and PIAS1 appeared not to produce a synergistic effect (lane 8). In addition to the modified band (band b), the presence of two minor species were observed in lanes 6 -8, as shown by bands a (120 kDa) and c (180 kDa).
To test whether these bands (bands a-c) are associated with the sumoylation of ARNT, we performed an immunoprecipitation analysis in HeLa cells expressing FLAG-ARNT in combination with HA-SUMO-1 and HA-UBC9 (Fig. 2B). All species of FLAG-ARNT were collected by FLAG immunoprecipitation for subsequent Western blot analysis with anti-FLAG and anti-SUMO-1 antibodies. FLAG-ARNT was equally expressed in lanes 2-4. The expression of FLAG-ARNT and HA-SUMO-1 produced the faint FLAG-reactive band (band a) between the unconjugated and sumoylated forms (band b) of ARNT (lane 3). Moreover, the most slowly migrating band (band c) was seen in the cells cotransfected with FLAG-ARNT, HA-SUMO-1, and HA-UBC9 (lane 4). In addition, these three forms of ARNT (bands a-c) were clearly recognized by anti-SUMO-1 monoclonal antibodies (lanes 8 and 9). Although the calculated molecular mass of the SUMO-1 protein is ϳ11.5 kDa, SUMO-1conjugated proteins run on SDS-polyacrylamide gel as if they have an additional 20 -30 kDa for each covalently attached SUMO-1 molecule (20,47). Therefore, bands a-c are most likely to be mono-, di-, and trisumoylated forms of ARNT, respectively. Unlike ubiquitin, SUMO-1 is thought to be conjugated to a single lysine in a monomeric form (11,48). Collectively, our results suggest that 1) ARNT potentially accepts at least three SUMO-1 molecules; 2) at least three lysines in ARNT may be modified by covalent attachment of SUMO-1;

SUMO-1 Conjugation of ARNT
and 3) Lys 245 is primarily important as a SUMO-1 acceptor site. However, we did not exclude the other possibilities because these data were obtained from overexpression of ARNT, SUMO-1, and SUMO-conjugating enzymes.
We further examined the sumoylation status of endogenous ARNT in MCF-7 cells without any overexpression (Fig. 2C). The above-mentioned experiments suggested that a fraction of ARNT is modified by SUMO-1 in vivo. Western blot analysis with anti-ARNT monoclonal antibodies showed the presence of the unmodified form of ARNT (input). We then immunoprecipitated endogenous ARNT by anti-ARNT antibodies from the cell lysate and tested it by Western blot analysis with anti-ARNT and anti-SUMO-1 antibodies (IP). Two high molecular mass bands were immunoreactive to both antibodies, and they migrated as did bands a and b in Fig. 2 (A and B). Because the intensity of these sumoylated forms was much weaker than that of unmodified ARNT even after a long time exposure, the steady-state population of SUMO-modified ARNT seemed to be limited in the cells. In addition, the band corresponding to the trisumoylated form of ARNT was not found under this assay condition. Taken together, our data indicate that endogenous ARNT protein is modified by SUMO-1 in vivo.
Transcriptional Role of Sumoylation of ARNT-To investigate the significance of the SUMO-1 conjugation of ARNT, we examined whether sumoylation affects the transcriptional activity of ARNT. Because Lys 245 is the major sumoylated site, we investigated the expression of wild-type ARNT and ARNT K245A. ARNT sumoylation was induced by a combination of SUMO-1 and UBC9, and appropriate luciferase reporter constructs were used to check the transactivation by ARNT. To test the effect of sumoylation of ARNT on AHR/ARNT-dependent transcription in MCF-7 cells, we employed the pGL3-promoter-XRE reporter, in which the luciferase gene is driven by six XREs and a minimal promoter, in the coexpression of AHR, ARNT (WT or K245A), SUMO-1, and UBC9 (Fig. 3A). Transfected cells were treated either with solvent or with 10 M 3-methylcholanthrene for 12 h before cell lysis at 48 h after transfection. Upon 3-methylcholanthrene treatment (black bars), the expression of SUMO-1 augmented AHR/ARNT-mediated transcription, regardless of whether wild-type or mutant ARNT was used. In addition, UBC9 synergistically enhanced the effect of SUMO-1 on transcription by AHR/ARNT. To eliminate the influence of endogenous ARNT, similar experiments were carried out in Hepa-1-c4 cells, which are defective in ARNT; and the results showed that transcription by AHR/ ARNT was stimulated by SUMO-1 and UBC9, despite the ARNT status (data not shown). We further utilized the pGL3basic-Cyp1A1 reporter (containing ϳ1.5 kb of the promoter region of the mouse Cyp1A1 gene), resulting in very similar data (data not shown).
Next, we examined the effect of SUMO-1 and UBC9 on HIF1␣/ARNT-mediated transcription. The pGL3-promoter-HRE reporter (49), containing six HREs from the erythropoietin gene and a minimal promoter, was employed in combination with HIF1␣, ARNT (WT or K245A), SUMO-1, and UBC9. Transfected cells were treated with 200 M cobalt chloride, which is known to mimic hypoxia and to induce HIF1␣/ARNTmediated transcription (50), for 12 h before cell lysis (Fig. 3B). As was the case in Fig. 3A, SUMO-1 and UBC9 synergistically augmented hypoxia-induced transcription, and there was little difference between wild-type and mutant ARNT. Accordingly, these results suggest that the SUMO-1 conjugation system positively regulates both AHR/ARNT-and HIF1␣/ARNT-mediated transcription, independent of the sumoylation of ARNT at Lys 245 . This finding might be explained by the predominant roles of AHR and HIF1␣ in these transcription pathways, com-

SUMO-1 Conjugation of ARNT
pared with the involvement of ARNT (51,52). The sumoylated forms of AHR and HIF1␣ were not detected under our experimental conditions (data not shown).
To demonstrate whether SUMO-1 directly affects the transcriptional activity of ARNT, we used the luciferase reporter plasmid pG5luc, containing Gal4-binding sites and a minimal promoter, together with the expression of ARNT fused to the Gal4 DBD. This enabled us to elucidate the transcriptional activity of ARNT or probably the ARNT/ARNT homodimer without any influence of other proteins. The ability of Gal4-ARNT (WT or K245A) to activate transcription was first tested (Fig. 3C). Gal4-ARNT WT increased the luciferase activities by ϳ7-fold compared with the Gal4 DBD alone. In contrast, Gal4-ARNT K245A stimulated transcription over 3-fold in comparison with wild-type ARNT, suggesting that sumoylation inhibits the transcriptional capacity of ARNT itself. To further investi-gate the transcriptional role of the sumoylation of ARNT, Gal4fused wild-type and mutant ARNT were expressed in MCF-7 cells with increasing amounts of SUMO-1 (Fig. 3D). The expression of SUMO-1 augmented transcription by wild-type ARNT and ARNT K245A in a dose-dependent manner. However, the transactivating activities of ARNT K245A were significantly higher than those of wild-type ARNT. Taken together, these data suggest that ARNT sumoylation per se represses its transcriptional ability and that certain cofactor(s) subjected to sumoylation may be involved in ARNT-related transactivation.
Localization of Part of ARNT in PML Nuclear Bodies-It has been shown recently that the sumoylation of some target proteins results in alteration of their intracellular distribution (11)(12)(13). We tested whether sumoylation influences the localization of ARNT in the nucleus. To address this question, we FIG. 3. Transcriptional role of SUMO-1 conjugation of ARNT. A and B, transcription by AHR/ARNT or HIF1␣/ARNT is enhanced by SUMO-1 and UBC9, independent of the sumoylation of ARNT. MCF-7 cells were transfected with pGL3-promoter-XRE or pGL3-promoter-HRE (0.5 g), in which the luciferase gene is driven by six XREs or six HREs and a minimal promoter, respectively, together with plasmids expressing AHR or HIF1␣ (0.5 g), ARNT (WT or K245A; 0.5 g each), SUMO-1 (1 g), and UBC9 (1 g). The transfected cells were treated with solvent alone (0.1% Me 2 SO; white bars) or with 10 M 3-methylcholanthrene (3MC; black bars) for 12 h (A) and with solvent alone (white bars) or with 200 M cobalt chloride (black bars) for 12 h (B). At 48 h after transfection, luciferase activities were determined, and the relative luciferase activities with mock vectors in the absence of treatment were normalized to 10. C, the enhanced transcriptional capacity of the sumoylation-resistant ARNT mutant. The pG5luc reporter (0.5 g), which contains Gal4-binding sites and a minimal promoter, was introduced into MCF-7 cells together with a plasmid expressing Gal4-ARNT (WT or K245A) or the Gal4 DBD (DB; 1 g each). At 48 h after transfection, luciferase activities were determined. The relative luciferase activities of the Gal4 DBD were normalized to 10. D, the effect of SUMO-1 on the transcriptional activities of wild-type and mutant ARNT. Gal4-ARNT (WT or K245A) or the Gal4 DBD was expressed in MCF-7 cells together with increasing amounts of SUMO-1 (0 -2 g of the vector). The total amount of transfected DNAs was balanced by the addition of mock vector. Values represent the means Ϯ S.D. from at least three independent experiments.

SUMO-1 Conjugation of ARNT
generated two plasmids that express wild-type and mutant ARNT as a fusion with green fluorescent protein, termed GFP-ARNT WT and GFP-ARNT K245A, respectively, and transfected them into MCF-7 cells for confocal laser scanning microscopic observation (Fig. 4A). Wild-type ARNT showed a few patterns of distribution in the nucleus: a diffuse localization (panels a) and discrete nuclear dots (10 -50 in each single nucleus) (panels b and c). However, most of the transfected cells showed the diffuse pattern (panels a), whereas the focus formations (panels b and c) were seen in ϳ10 -30% of the cells. The K245A mutant similarly localized in the nucleus compared with wild-type ARNT (panels d) (data not shown). These observations seem to be consistent with a recent report (53).
Because several SUMO-1-modified proteins associate with nuclear domain structures known as PML-NBs (29 -34), we investigated whether GFP-ARNT colocalizes with endogenous PML-NBs using immunostaining with anti-PML antibodies (Fig. 4A). The foci produced by GFP-fused wild-and mutant ARNT partly colocalized with PML-NBs in MCF-7 cells (panels b-d), suggesting the possible relationship of PML with ARNT. Although ARNT was diffusely present in the nucleus (panels a), PML is known to localize not only in NBs, but also throughout the nucleus. Next, to investigate whether PML mediates the translocation of ARNT into NBs, PML was expressed in MCF-7 and Hepa-1 cells (Fig. 4B). Increased PML induced relatively large sized NBs in transfected cells, leading to the translocation of both GFP-fused and endogenous ARNT into the NBs (panels a-c). In the absence of PML overexpression, endogenous ARNT showed a diffuse localization in the nuclei, similar to the frequent case of GFP-ARNT (Fig. 4A, panels a) (data not shown). Thus, a portion of ARNT colocalized with PML in NBs, but the sumoylation of ARNT appeared not to affect the intracellular distribution in the nucleus. The coexistence of ARNT with PML in NBs suggests the possibility that PML modulates the transcriptional activities of ARNT.
Sumoylation of ARNT Alters Its Ability to Interact with PML-To examine the effect of PML on the transcriptional ability of ARNT, we expressed Gal4-ARNT WT or Gal4-ARNT K245A in combination with PML and SUMO-1 in MCF-7 cells. Luciferase analysis was performed in a similar way as shown in Fig. 3 (Fig. 5A). PML increased the transcriptional activities of both wild-type and sumoylation-resistant ARNT. Interestingly, the coexistence of SUMO-1 significantly repressed the effect of PML on transcription by wild-type ARNT, but much less by ARNT K245A. Thus, PML positively regulates ARNTdependent transcription, and the sumoylation of ARNT at Lys 245 inhibits the transcriptional role of PML.
The SUMO-1 modification of substrates can affect proteinprotein interactions (20,22). It is of interest that the SUMO-1 acceptor site of ARNT at Lys 245 resides in the PAS domain, which is required for dimerization with bHLH/PAS partner proteins (3,42). We therefore examined whether the sumoylation of ARNT affects its ability to dimerize with AHR in vivo (Fig. 5B). For immunoprecipitation, Myc-tagged AHR was expressed in MCF-7 cells in combination with FLAG-ARNT (WT or K245A), HA-SUMO-1, and HA-UBC9, and the cells were lysed at 48 h after transfection. The Myc precipitates were subjected to Western blot analysis with anti-FLAG antibodies (right panel). Unconjugated forms of both wild-type ARNT and ARNT K245A were similarly immunoprecipitated with AHR, and sumoylated ARNT also complexed with AHR in proportion to the original amount in the input lysate. Nonspecific bands reacting to anti-FLAG antibodies were detected between the unconjugated and conjugated forms of ARNT, indicating the presence of equal amounts of cellular proteins in each lane.
To further investigate whether the sumoylation of ARNT affects the interaction of ARNT with PML, we employed a GST pull-down assay (Fig. 5C). Bacterially expressed GST-PML (amino acids 5-452) was incubated with whole cell lysates from MCF-7 cells expressing FLAG-ARNT, HA-SUMO-1, and HA-UBC9. The precipitates on glutathione-agarose beads were analyzed by immunoblotting with anti-FLAG antibodies. Interestingly, unconjugated ARNT (WT or K245A) bound PML, but sumoylated ARNT did not associate with PML. Collectively, these data suggest that PML enhances transcription by ARNT and that the sumoylation of ARNT abolishes the interaction with PML to suppress the enhancement of ARNT-mediated transcription by PML. DISCUSSION In this study, we have reported that ARNT is a substrate for SUMO-1 modification at the major target Lys 245 within the PAS domain. UBC9 and PIAS1 stimulated the sumoylation of ARNT when they coexpressed with ARNT and SUMO-1. Because free SUMO-1 is known to be sparsely present in mammalian cells, the increase in SUMO-1 and its conjugating activities facilitated our initial detection of the sumoylated form of ARNT. PIAS1 is a member of the E3-like ligase family, involved in the sumoylation of some target proteins (27,35). Our data imply that PIAS1 functions as a SUMO-1 ligase of ARNT. Both the ubiquitin and SUMO-1 conjugation systems show very similar enzymatic cascades, but these pathways seem to be different in some aspects discussed below. First, in contrast to ubiquitin ligases, PIAS proteins are not always required for the sumoylation of substrates, and they appear to function as factors that stimulate the SUMO-1 conjugation reaction via UBC9 (54,55). Overexpressed UBC9 was sufficient for inducing the sumoylation of ARNT in the absence of PIAS1. Second, there is a difference in substrate specificity between ubiquitin and SUMO-1 ligases. Ubiquitin ligases strictly recognize their substrates for ubiquitination, whereas PIAS proteins generously determine the targets. PIAS1 has been reported to mediate the sumoylation of p53 and c-Jun (27,35) and to interact with the androgen receptor, which is one of the SUMO-1 substrates (56). From our results, there may be two other possible target sites for sumoylation in addition to Lys 245 , which are unlikely to be located within a sequence matching the consensus motif. SUMO-1 conjugation at these sites may depend on the sumoylation at Lys 245 . Although many SUMOmodified proteins possess more than one sumoylated site, such coordinated attachment of SUMO-1 molecules to the different target sites has been demonstrated in only a few substrates such as c-Myb. As shown in this study, a relatively small proportion of ARNT was detected in the conjugated form, even when sumoylation was stimulated by UBC9 and PIAS1. This may be explained by the idea that the sumoylation of ARNT is transient and that there is a dynamic equilibrium between SUMO-1-conjugated and -unconjugated forms. Another possibility is that the presence of certain limiting factors affects this sumoylation. We detected a small amount of a SUMO-1-modified form of endogenous ARNT in MCF-7 cells by immunoprecipitation. Sumoylation of some substrates has been shown to be enhanced in the presence of ligand and upon cell treatment (for example, the androgen receptor and heat shock transcription factor-1). The sumoylation of endogenous ARNT might be precisely controlled by certain regulatory mechanisms in vivo.
The sumoylation of transcription factors and other related proteins produces diverse influences on their transcriptional activities (11,23,24,26,28,29,57,58). In our study, the use of the luciferase reporter constructs together with the expression of ARNT (WT or K245A) indicated that the lack of the major sumoylation of ARNT had little effect on AHR/ARNT-and HIF1␣/ARNT-mediated transcription. On the other hand, transactivation by Gal4-ARNT was evidently stimulated by the sumoylation-resistant K245A mutation compared with wildtype ARNT. One possible explanation is that the difference in transcriptional activities between wild-type ARNT and ARNT K245A was masked by the involvement of the partner, AHR or HIF1␣. In fact, it has been reported that AHR plays a more prominent role than ARNT in AHR/ARNT-mediated transcrip- tion (51,52). ARNT is required for transcription by the AHR/ ARNT heterodimer, but basically plays a supportive role in this pathway.
Alternatively, SUMO-1 conjugation to ARNT at target site(s) except for Lys 245 , even if it is infrequent and unstable, may have a sufficient effect on AHR/ARNT-and HIF1␣/ARNT-mediated transcription. Furthermore, SUMO-1 and UBC9 synergistically enhanced both AHR/ARNT-and HIF1␣/ARNT-mediated transcription, independent of the Lys 245 sumoylation of ARNT. SUMO-1 also enhanced the transcriptional activities of both Gal4-ARNT WT and Gal4-ARNT K245A in a dose-dependent manner. In other words, the activation of the SUMO-1 conjugation system induces the sumoylation of many substrates as well as ARNT. In agreement with our data, molecules in the SUMO conjugation system, including SUMO-1, UBC9, and PIAS, have been shown to modulate the transcriptional activities of p53, p73␣, the androgen receptor, and lymphoid enhancer factor-1, even when they were mutants lacking their major SUMO-1 acceptor site (27,31,57,58). The sumoylation of certain key factor(s) such as PML might be crucial for the general mechanism of transcriptional control. Moreover, it was reported that UBC9 modulates transcription by ETS-1 and TEL independent of its enzymatic ability to conjugate them with SUMO-1 (59,60) and that UBC9 mediates the nuclear localization of Vsx-1 without its SUMO-1-conjugating activity (61). Thus, UBC9 may possess some other functions in addition to the role of a SUMO-1-conjugating enzyme.
The Lys 245 target of ARNT resides in the PAS domain, which has been implicated in dimerization with bHLH/PAS members and with interacting molecules such as the transcription factor Sp1 (3,41,42). As shown in Fig. 5, sumoylation did not affect the ability of ARNT to dimerize with AHR. This is consistent with the result that the sumoylation of ARNT had little effect on transcription by AHR/ARNT. Recent reports have shown that the SUMO-1 modification of PML, heat shock factor-1 and -2, and lymphoid enhancer factor-1 localizes to nuclear domain structures known as PML-NBs (29 -34). In addition, sumoylation leads to the localization of homeodomain-interacting protein kinase-2 and TEL to another type of nuclear subregion (62,63). These lines of evidence suggest that sumoylation functionally regulates substrate proteins through changing their intracellular distribution. Our observations show that part of ARNT colocalizes with PML or PML-NBs. Furthermore, sumoylation significantly reduced the interaction with PML, resulting in a weakening of the ability of PML to enhance transcription by ARNT. Although the overall distribution of the ARNT K245A mutant was similar to that of wild-type ARNT, we did not exclude the possibility that sumoylated ARNT dislocates from PML-NBs because a small part of ARNT interacting with PML is modified by SUMO-1 in the cells. Thus, the SUMO-1 conjugation of ARNT changes the affinity for PML, but not for AHR. There is a possibility that the association of ARNT with some other proteins may be affected by this modification. In conclusion, our study has shed light on the significant role of the sumoylation of ARNT in modulating its transcriptional roles through affecting the interaction with cooperative proteins.