A Nuclear Localization Signal of Human Aryl Hydrocarbon Receptor Nuclear Translocator/Hypoxia-inducible Factor 1β Is a Novel Bipartite Type Recognized by the Two Components of Nuclear Pore-targeting Complex*

Aryl hydrocarbon receptor nuclear translocator (ARNT) is a component of the transcription factors, aryl hydrocarbon receptor (AhR) and hypoxia-inducible factor 1, which transactivate their target genes, such as CYP1A1 and erythropoietin, in response to xenobiotic aromatic hydrocarbons and to low O2concentration, respectively. Since ARNT was isolated as a factor required for the nuclear translocation of AhR from the cytoplasm in response to xenobiotics, the subcellular localization of ARNT has been of great interest. In this investigation, we analyzed the subcellular distribution of ARNT using transient expression of a fusion gene with β-galactosidase and microinjection of recombinant proteins containing various fragments of ARNT in the linker region of glutathioneS-transferase/green fluorescent protein. We found a clear nuclear localization of ARNT in the absence of exogenous ligands to AhR, and identified the nuclear localization signal (NLS) of amino acid residues 39–61. The characterized NLS consists of 23 amino acids, and can be classified as a novel variant of the bipartite type on the basis of having two separate regions responsible for efficient nuclear translocation activity, but considerable deviation of the sequence from the consensus of the classical bipartite type NLSs. Like the well characterized NLS of the SV40 T-antigen, this variant bipartite type of ARNT NLS was also mediated by the two components of nuclear pore targeting complex, PTAC58 and PTAC97, to target to the nuclear rim in an in vitro nuclear transport assay.

the expression of various genes via binding to its cognate binding sequence named XRE (xenobiotics responsive element) (1)(2)(3). CYP1A1, one of the target genes of AhR, plays an important role in the metabolism of procarcinogens, such as benzo-(a)pyrene in cigarette smoke, resulting in formation of activated DNA binding derivatives (4). Among various classes of ligands for AhR, the most potent is a well known environmental pollutant, tetrachlorodibenzo-p-dioxin (TCDD). Using an animal model system, the toxicity of TCDD has been found to cause wasting syndrome, immunodeficiency, tumor promotion, and teratogenesis (5,6).
The aryl hydrocarbon receptor nuclear translocator (ARNT) was identified as a factor that rescues the aryl hydrocarbon hydroxylase activity in response to xenobiotics in Hepa-1 c4 mutant cells (7). Since the ligand binding subunit of AhR is present in the cytoplasm of Hepa-1 c4 mutant cells, this led to the notion that ARNT is required for ligand-dependent nuclear translocation of AhR. Sequence analysis showed that ARNT is a 90-kDa protein possessing the basic-helix-loop-helix (bHLH) domain as well as the PAS domain, showing similarity with two Drosophila proteins called Per (for period) important for circadian rhythms and Sim (single-minded) required for the formation of the central nervous system (7). Subsequent cloning of the ligand-binding subunit of AhR also showed structural similarities with ARNT in the bHLH and PAS domains (8,9). Various mutational analyses revealed that the formation of heteromeric complex of ARNT and AhR mediated by bHLH and PAS domains is required for it to have DNA binding activity (10 -14). Recently, ARNT has also been identified as a component of another transcription factor called HIF-1 (hypoxiainducible factor 1). HIF-1 consists of two subunits called HIF-1␣ and HIF-1␤ (15); the former is a new member of the bHLH/PAS protein family, while the latter was found to be identical to ARNT (16). Physiologically, HIF-1 responds to a low O 2 concentration and transactivates many genes, including erythropoietin (17) and vascular endothelial growth factor (18).
Since ARNT was first cloned as a factor required for the nuclear translocation of AhR from the cytoplasm to the nucleus, the subcellular localization of ARNT was believed to be cytoplasmic. In fact, most of ARNT were recovered in the cytosolic fraction by cell fractionation. However, recent immunohistochemical analysis has shown that ARNT is localized predominantly in the nucleus, regardless of the presence or absence of ligands (19,20). It has also been reported that overexpressed ARNT is present in both the cytoplasm and the nucleus in the insect cell system (21). These observations led to the notion that ARNT is localized mainly in the nucleus, but some fractions might also be localized in the cytoplasm under certain circumstances. The increasing importance of the physiological role of ARNT in the regulation of gene transcription in response to various signals, such as oxygen (22), prompted us to investigate its subcellular localization in detail.
Active transport of protein from the cytoplasm to the nucleus requires the presence of a short amino acid moiety named nuclear localization signal (NLS) in any part of the protein (23). NLSs of various proteins identified so far can be classified mainly into two classes: 1) a single cluster of basic amino acids represented by the SV40 large T antigen NLS, and 2) a bipartite type in which two sets of adjacent basic amino acids are separated by a stretch of approximately 10 amino acids (24). The NLS-dependent nuclear translocation process depends on the cytosolic fractions and can be separated mainly into two steps: energy-independent targeting to the nuclear pore and the energy-dependent entrance to the nucleus. Recently, four soluble factors have been purified and implicated in nuclear protein import (24,25): importin-␣ (26 -30), importin-␤ (31)(32)(33)(34)(35), the GTPase Ran/TC4 (36,37), and pp15 (38). Importin has also been called karyopherin or nuclear pore-targeting complex (PTAC). The first step in the selection of protein targeting the nuclear pore is thought to be recognition of the NLS by the 58-kDa cellular protein importin-␣, which associates with the 97-kDa cellular factor importin ␤ (39,40). This NLS recognition complex docks to the nuclear pore complex via importin-␤ (41) and subsequently is translocated through the pore by an energydependent, Ran-dependent mechanism (37). Although the association of importin-␣ (PTAC58) and importin-␤ (PTAC97) with SV40 T NLS has been investigated extensively, not much is known about other NLSs of nucleoproteins, including transcription factors (42).
In the present study, we investigated the subcellular localization of ARNT using a transient expression of chimeric constructs of ARNT and ␤-galactosidase (␤-Gal), clarifying nuclear localization of the ARNT protein. Subsequent analysis of various portions of ARNT using ␤-Gal fusions as well as fusion protein with GST-GFP gave the minimum NLS consisting of amino acids 39 -61 of ARNT. The identified region differed from the classical type of NLS reported so far, which prompted us to analyze interaction with PTAC58 and PTAC97.

EXPERIMENTAL PROCEDURES
Cell Cultures-Cell lines used for the study were mouse hepatoma Hepa-1 clone Hepa1c1c7, Hepa-1 c4 mutant which lacks ARNT expression (generously provided by Dr. O. Hankinson, UCLA), and HeLa cells. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum at 37°C with a 5% CO 2 atmosphere.
Plasmid Construction-The human ARNT cDNA was prepared by PCR amplification of reverse-transcribed products of total RNA from HepG2 cells using specific primers and Pfu DNA polymerase (CLON-TECH), and was inserted in the pGEM-7Zf(ϩ) vector (Promega). The sequence of the construct was confirmed by sequencing using fluorescein-labeled SP6 and T7 primers, AutoRead Sequencing kits, and A.L.F. II DNA sequencer (Pharmacia Biotech Inc.). For subsequent cloning into the ␤-Gal expression vector, the NcoI site at the initiation codon of ARNT was modified to the BglII site using pBglII linker (Takara), and another BglII site was created in front of the stop codon by PCRmediated mutation. To construct the expression vectors of ␤-Gal fusion proteins with various portions of ARNT, an artificial BglII site was created in front of the stop codon of ␤-galactosidase gene of pSV␤-Gal vector (Promega) by PCR-mediated mutation. The BglII-BglII fragment of the ARNT cDNA was ligated to the BglII site of the modified ␤-Gal control vector to generate ␤-Gal/ARNT-(1-789) vector.
Various portions of ARNT cDNA were amplified by PCR using ␤-Gal/ ARNT-(1-789) vector as a template and Pfu DNA polymerase with specific sets of primers to generate artificial BglII sites at both ends. To construct GST-ARNT-GFP fusion genes, the GST-GFP cassette vector was prepared as follows. After PCR amplification of the GFP cDNA (a gift of Dr. Roger Y. Tsien, University of California, San Diego) to generate SmaI and EcoRI sites at its 5Ј and 3Ј flanks, the resulting fragments were subcloned into the pGEM-7Zf(ϩ) vector. The vector was cleaved with SmaI and XhoI and subcloned into pGEX-5X-2 vector (Pharmacia). For construction of in-frame fusion proteins, the resultant vector was cleaved with XmaI, treated with Klenow fragment (Takara) in the presence of 0.1 mM dNTPs to blunt the ends, followed by religation and transformation to generate the GST-GFP vector. Various portions of ARNT cDNA described above were inserted at the BamHI site of the GST-GFP2 vector. The direction of inserts was determined by sequencing. To construct GST-NLSc-GFP vector, the core sequence of NLS of SV40 large T antigen generously provided by Dr. Tsuneoka (Kurume University, Fukuoka, Japan), the coding sequence of which was 5Ј-AAG CTT GCC ATG GGG TGG CCC ACT CCT CCA AAA AAG AGA AAG GTA GAA GAC CCC GGG-3Ј, was ligated with GFP cDNA at the SmaI site and subcloned into the pGEM-7Zf(ϩ) vector. The BamHI-EcoRI fragment of the resultant was ligated to the pGEX 2T (Pharmacia) vector to give the GST-NLSc-GFP vector.
DNA Transfections into HeLa, Hepa-1, and Hepa-1 c4 Mutant Cells-Electroporation was carried out using 15 g each of ␤-Gal fusion protein expression vectors and cells (3.5 ϫ 10 6 ) in 400 l of K-PBS buffer at 960 microfarads/450 V with Gene Pulser (Bio-Rad). The electroporated cells were seeded onto a 10-cm plastic dish and incubated at 37°C under an atmosphere with a 5% CO 2 content for 48 h. In situ staining of ␤-Gal was carried out as follows; the cells were washed twice with PBS, followed by fixation with 0.2% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.0), 1 mM MgCl 2 for 15 min; they were stained with 0.2% 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside in 10 mM sodium phosphate buffer (pH 7.0) containing 1 mM MgCl 2 , 150 mM NaCl, and 3.3 mM each of potassium ferrocyanide and ferricyanide at 37°C overnight.
Preparation of GST-ARNT-GFP Fusion Proteins-The GST-ARNT-GFP vectors described above were introduced into the Escherichia coli strain BL21. A single colony was picked and cultured in LB broth/Amp until A 600 reached 1.2, then isopropyl-1-thio-␤-D-galactopyranoside was added to 1 mM and incubated at 20°C for 14 h with vigorous shaking. The cells were collected by centrifugation at 3,500 rpm for 15 min, washed with saline, and resuspended in a lysis buffer, 50 mM Tris-HCl buffer (pH 8.3) containing 500 mM NaCl, 1 mM EDTA, 2 mM dithiothreitol, and 0.2 mM phenylmethylsulfonyl fluoride. These cells were lysed by two rounds of freeze/thaw treatment, then subjected to sonication. The resultant samples were centrifuged at 12,000 rpm at 4°C for 30 min, and the soluble fractions were collected and subjected to batch purification procedures using glutathione-Sepharose 4B resin (Pharmacia). The purified protein was dialyzed against buffer containing 20 mM HEPES (pH 7.3), 100 mM potassium acetate, and 2 mM dithiothreitol.
Microinjection-Microinjection experiments were performed essentially as described previously (43). After microinjection of samples into the cytoplasm of HeLa cells, the cells were incubated at 37°C for 30 min before fixation with 3.7% formaldehyde. Localization of injected GST-ARNT-GFP fusion proteins was examined by fluorescent microscopy.
Recombinant Expression and Purification of PTAC58 and PTAC97-Recombinant PTAC58 and PTAC97 were expressed in BL21 as GST fusion protein as described previously (28). The fusion proteins were purified using glutathione-Sepharose affinity chromatography. Finally, recombinant proteins of PTAC58 and PTAC97 were obtained by cleavage with thrombin to release the GST portion.
Cell-free Import Assay-Preparation of total cytosol of Ehrlich ascites tumor cells was conducted as described previously (40). Digitonin-permeabilized MDBK cells were prepared based on the method of Adam et al. (44) as described previously (45). The testing solution (10 l) consisted of GST-ARNT-(39 -61)-GFP and transport buffer (20 mM HEPES (pH 7.3), 110 mM potassium acetate, 2 mM magnesium acetate, 5 mM sodium acetate, 0.5 mM EGTA, 2 mM dithiothreitol, and 1 g/ml each of aprotinin, leupeptin, and pepstatin). Transport assay was performed in the presence or absence of cytosol with 1 mM ATP, 5 mM creatine phosphate, and 20 units/ml creatine phosphokinase at 37°C for 30 min. For the nuclear-binding assay, recombinant PTAC58 and PTAC97 proteins were added and incubated at 4°C for 30 min. After incubation, the cells were fixed and the location of GST-ARNT-GFP fusion proteins was examined.

Transient Expression of Chimeric Constructs of ␤-Gal and
Full-length ARNT in HeLa, Hepa-1, and Hepa-1 c4 Cells-To determine the subcellular localization of human ARNT, we constructed the fusion genes ␤-Gal and ARNT under the control of the SV40 enhancer/promoter. Since the molecular mass of the ␤-Gal is large enough (120 kDa) to prevent passage through the nuclear pore by diffusion, bacterial ␤-Gal has been widely used as a reporter gene for the determination of subcellular localization of expressed protein. The expression vector of ␤-Gal/ARNT-(1-789) was transfected to three cell lines, including HeLa, Hepa-1, and ARNT-deficient Hepa-1 c4 mutant cells, by means of electroporation. Representative profiles of expressed fusion proteins visualized by in situ staining of ␤-Gal with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside are shown in Fig. 1. As expected, no staining of the nucleus was observed for the expressed ␤-Gal alone (Fig. 1A), while the fusion of ␤-Gal with the NLS of SV40 large T antigen (␤-Gal/ SV40 NLS) gave strong nuclear localization in transfected cells (Fig. 1B). When the chimeric gene of ␤-Gal/ARNT-(1-789) was expressed, the fused product was clearly localized in the nucleus of all three cell lines tested (Fig. 1C), which agreed well with the results obtained by immunohistochemical analysis (19,20).
Requirement of the N-terminal Portion of ARNT for Nuclear Localization-To identify the region of ARNT required for nuclear localization, various portions of cDNA for ARNT were synthesized using PCR and ligated to the modified ␤-Gal vector as described above. The chimeric constructs were introduced into HeLa cells, and their localization was examined (Fig. 2). Deletion of the transactivation domain located in the C-terminal portion (Fig. 2B) and/or PAS domain (Fig. 2C) did not affect the nuclear localization of ␤-Gal/ARNT fusion proteins. In contrast, fusion proteins of either ␤-Gal/ARNT-(166 -485) (Fig.  2D) or ␤-Gal/ARNT-(486 -789) (Fig. 2E) showed cytoplasmic localization, confirming the absence of NLS in these regions. On the other hand, the fusion protein containing the bHLH region of the ARNT-(1-165) showed strong nuclear staining, suggesting the presence of NLS in this region (Fig. 2C). Since the NLS identified so far contained a cluster(s) of basic amino acids, two candidates of NLS of ARNT can be estimated, one of which is located between 39 and 45 (Arg-Ala-Ile-Lys-Arg-Arg-Pro), while the other is between 99 and 104 (Arg-Arg-Arg-Arg-Asn-Lys) in the bHLH domain. To identify which segments are involved in the NLS activity of ARNT, we further divided ARNT-(1-165) into two fragments, ARNT-(1-88) (Fig. 2F) and ARNT-(89 -165) (Fig. 2G). Fusion protein containing the former fragment gave intense nuclear staining, while those containing the latter fragment did not. Finally, ARNT NLS was identified as a region between the 39-and 61-amino acid residues (Figs. 2I and 3A).

Region of ARNT Necessary and Sufficient for Nuclear Import Comprises a 23-Amino Acid Domain in the N-terminal Por-
tions-To confirm the capacity for nuclear localization, we next examined the fate of recombinant proteins microinjected into the cytoplasm of HeLa cells. The cDNA of GFP(S65T), possessing amino acid substitution from Ser 65 to Thr to give a stronger fluorescence intensity (46), was inserted into the region downstream of GST gene to give a fusion gene of GST-GFP. The fusion protein was obtained by expression in BL21 in the presence of isopropyl-1-thio-␤-D-galactopyranoside. When collected by centrifugation, the bacteria showed a slightly greenish color indicating the expression of GFP fusion proteins. The cells were disrupted, and the cell lysates were subjected to affinity purification using a glutathione-Sepharose resin. The bound protein was eluted by addition of buffer containing glutathione. The fusion protein was analyzed using 7.5% acrylamide gel and showed a major protein of molecular mass 55 kDa (data not shown), which would be large enough to prevent its diffusion into the nucleus. Actually, when microinjected into the cytoplasm of HeLa cells, the GST-GFP protein was localized to the cytoplasm even after incubation for 2 h (Fig. 3B, a). Next, we constructed a plasmid by insertion of the SV40 NLSc fragment into the junction of the fusion gene (GST-NLSc-GFP), and the gene product was prepared as above. Microinjected GST-NLSc-GFP protein revealed the efficient nuclear import within 30 min of incubation at 37°C (Fig. 3B, b). Using this system, various portions of the N-terminal region of ARNT were ana-lyzed (Fig. 3B, c-f). As was seen for the transient expression of ␤-Gal fusions (Fig. 2), only the GST-GFP fusion protein that contains ARNT-(39 -61) showed efficient nuclear localization (Fig. 3B, c), confirming that the fragment serves as an NLS. As was expected from its similarity to the consensus sequence of SV40 large T-like NLS, deletion of the amino acid residues between 39 and 45 diminished its ability as an NLS, suggesting the absolute requirement of these basic amino acids for NLS activity (Fig. 3B, d). On the other hand, the single cluster of basic amino acids was not sufficient for the full NLS activity, since GST-ARNT-(39 -55)-GFP (Fig. 3B, e) and GST-ARNT-(39 -45)-GFP (Fig. 3B, f) did not translocate to the nucleus, implying the importance of the rest of the amino acids between 46 and 61 for full NLS activity.
NLS of ARNT Is a Novel Bipartite Type-To identify which amino acid residue(s) are important for the accessory activity to give a full NLS, mutational analysis in the region of 46 -61 was performed. Among these, we developed great interest in the Ser residue at 57, since various transcription factors possess a Ser residue in their NLS, which is often phosphorylated in the activated form (47). Substitution of the Ser 57 to Ala (S57A; Fig.  4A) or Thr (S57T; Fig. 4B) was performed, and the mutated fragments were analyzed by microinjection of GST-ARNT-GFP fusion proteins. As shown in Figs. 3 and 4, nuclear localization of these mutants and the wild type proteins were indistinguishable, indicating that Ser 57 does not participate in the NLS activity.
We next asked possible roles of basic amino acids Lys 58 and Arg 61 by replacing them to Ala (K58A and R61A, respectively). Notably, K58A mutant lost efficient nuclear translocation activity in microinjection assay (Fig. 4C). R61A mutant also drastically reduced the NLS activity; the fluorescence intensity was almost equal between the cytoplasm and nucleus (Fig. 4D). The double mutant K58A/R61A completely lost its translocation activity in the microinjection assay (Fig. 4E). These findings suggest that these two basic amino acid residues are necessary for efficient nuclear translocation activity.
We are also interested in amino acid residues 48 -54 of ARNT, since this region is apparently rich in acidic amino acids (5 out of 7). To evaluate the possible requirement of the negative charge for efficient NLS activity, we have also introduced amino acid substitutions of Asp 48 for Asn, Asp 52 for Asn, and Glu 54 for Gln (Fig. 5A). No apparent effect of these mutations was observed, suggesting that the negative charge of the cluster of acidic amino acids in the center of ARNT-(39 -61) is not important for efficient NLS activity. We also changed five out of five negative residues to neutral residues and found the same observation (data not shown). However, deletion of this region abolished the NLS activity (Fig. 5B), indicating that these amino acids were also required as spacers to give enough length between the two separated clusters of basic amino acids.
Amino Acid Residues 39 -61 of ARNT Interact with PTAC58/PTAC97 to Target to the Nuclear Pore-We next investigated whether the NLS of ARNT is recognized by the components of the nuclear pore targeting complex, PTAC58/ PTAC97. To evaluate the involvement of these factors in the nuclear targeting process, we analyzed various recombinant fusion proteins of GST-ARNT-GFP using an in vitro nuclear transport assay (Fig. 6). We used GST-NLSc-GFP fusion protein as a control substrate and observed clear nuclear accumulation or targeting to the nuclear rim incubated with cytosol in the presence of ATP (Fig. 6B) or with purified PTAC58/PTAC97 (Fig. 6C), respectively. When incubated with cell extracts of Ehrlich tumor in the presence of ATP at 37°C, GST-ARNT-(39 -61)-GFP localized to the nucleus, confirming that the prepared recombinant protein can be a good substrate for this assay as well (Fig. 6E). In addition, incubation of GST-ARNT-(39 -61)-GFP with PTAC58/PTAC97 enabled targeting of the recombinant protein to the nuclear rim (Fig. 6F), suggesting that the inserted fragment ARNT-(39 -61) is recognized by these factors as was the classical SV40-like NLS.
Since the two basic amino acids Lys 58 and Arg 61 are needed for efficient nuclear translocation as judged by microinjection  (Fig. 4), the effect of double mutation of these two amino acids to Ala on the nuclear transport activity in vitro was investigated. When incubated with cell extract in the presence of ATP, mutated GST-ARNT-GFP did not accumulate in the nucleus at all (Fig. 6G). Furthermore, ARNT with double mutations K58A/R61A drastically reduced the nuclear rim targeting activity (Fig. 6H). DISCUSSION We have shown that human ARNT is localized to the nucleus when analyzed by transient expression of the chimeric constructs of ARNT cDNA and bacterial ␤-galactosidase gene. Our observation confirmed the previous reports using immunohistochemical techniques as shown by Pollenz et al. and Hord and Perdew (19,20). Since the heterodimeric partner AhR presents in the cytoplasm in the absence of ligands and translocates to the nucleus upon binding of ligands in the ARNT deficient cell line, Hepa-1 c4 (20), these two subunits may translocate independently to the nucleus, where they may form a heterodimer to bind to the cognate DNA sequence. On the other hand, Reisz-Porszasz et al. (50) reported that the mouse ARNT deletion mutant named bHLHAB, which contains amino acids 70 -474, can transactivate the CAT reporter construct under the control of the XRE-containing promoter. Apparently, the bHL-HAB lacks the corresponding region of the NLS identified for human ARNT in this study, thus the bHLHAB would not translocate to the nucleus independently. Since they have shown that the bHLHAB forms a heterodimer with AhR in the presence of TCDD in vitro (50), it is likely that the formation of heterodimer of bHLHAB and AhR occurs in the cytoplasm and then the bHLHAB-AhR complex may utilize the NLS of AhR to translocate from the cytoplasm to the nuclei. Wood et al. (22) have recently shown that the bHLHAB is also capable of transactivating HIF-1 activity in the Hepa-1 c4 mutant cell lines, suggesting that the heterodimerization of bHLHAB with another partner of HIF-1␣ could also occur in the cytoplasm. Taken together, ARNT/HIF-1␤ might heterodimerize with either AhR or HIF-1␣ in both the cytoplasm and the nucleus, depending on the subcellular distribution of ARNT/HIF-1␤ at the preactivated stage.
Although it has been pointed out previously that the basic amino acid sequence (Arg-Ala-Ile-Lys-Arg-Arg-Pro; amino acids 39 -45) of human ARNT, which is similar to the NLS of SV40 large T antigen, might be a candidate for an NLS (20), our analysis presented here clearly showed that the basic amino acids were necessary, but not sufficient alone for the full activity needed for nuclear localization of ARNT. Most importantly, we have shown the absolute requirement of the additional amino acids spanning from 46 to 61 (Gly-Leu-Asp-Phe-Asp-Asp-Asp-Gly-Glu-Gly-Asn-Ser-Lys-Phe-Leu-Arg) for efficient nuclear localization. Both deletion and point mutation analyses of the C-terminal part of this region revealed that two basic amino acids, Lys at 58 and Arg at 61 were important for NLS activity. We also analyzed the possible involvement of the acidic amino acids present in the central region (positions 48 -54) of the NLS of ARNT. Apparently, deletion of this region diminished its NLS activity; thus, these amino acids were also found out to be necessary for the nuclear localization activity. On the other hand, when their negative charges were altered to neutral, the mutant did not show any changes in the nuclear localization activity. These results strongly suggest that the central region (48 -54) of the minimal NLS of ARNT is required as a spacer to give enough length between the two separated portions of the clusters of basic amino acids as was present in the nucleoplasmin NLS. Although the role of negative charges in the central part of the ARNT NLS is still unknown, they may enhance the exposure of NLS to the surface of the whole protein. In this sense, the NLS of the human ARNT resembles the consensus of bipartite NLS (48), but a clear difference from the consensus did exist in the C-terminal portions, where two basic amino acids required for the full NLS activity were separated by a two amino acid insertion. Here, we propose that the ARNT NLS (39 -61) should be categorized to be a novel variant of the bipartite type. To date, two protein sequences closely related to the human ARNT have been reported; one is the mouse homologue of ARNT (49,50), while the other is a novel member of this family named ARNT2, which is expressed in limited organs in both adult and developing embryo (51). As is shown in Fig. 7, all the amino acids required for the NLS activity of ARNT were conserved among these three proteins, confirming the importance of these amino acids.
The novel variant bipartite type of ARNT NLS was also recognized by the two components of nuclear targeting complex, PTAC58 and PTAC97. Concomitant with the deficiency of nuclear localization activity of the K57A/R60A mutant of ARNT NLS, the mutant did not target the nuclear rim in the presence of PTAC58 and PTAC97, confirming the requirement of these basic amino acids for the first step of nuclear localization: the recognition of NLS by its receptor to bring it to the nuclear pore. These observations also suggested that PTAC58 (importin-␣) can recognize various types of NLSs, including a variant bipartite NLS of ARNT, to target the nuclear pore. Of course, we can not exclude the possibility that NLS of ARNT may be recognized by an unknown specific receptor resulting in efficient targeting of the nuclear pore. Further investigation will be required to clarify the contribution of PTAC58 to the nuclear pore targeting of ARNT in the cells.
In summary, we have identified the NLS of human ARNT/ HIF-1␤ to consist of 23 amino acid residues spanning amino acids 39 -61, which was shown to be a novel bipartite type of NLS. Our observations combined with others led to the notion that either AhR or HIF-1␣ might form a heterodimer with ARNT/HIF-1␤ in both the nucleus and cytoplasm, depending on the subcellular localization of ARNT/HIF-1␤ in the preactivated state.