Identification of a region within the ubiquitin-activating enzyme required for nuclear targeting and phosphorylation.

The ubiquitin-activating enzyme exists as two isoforms: E1a, localized predominantly in the nucleus, and E1b, localized in the cytoplasm. Previously we generated hemagglutinin (HA) epitope-tagged cDNA constructs, HA1-E1 (epitope tag placed after the first methionine) and HA2-E1 (epitope tag placed after the second methionine) (Handley-Gearhart, P. M., Stephen, A. G., Trausch-Azar, J. S., Ciechanover, A., and Schwartz, A. L. (1994) J. Biol. Chem. 269, 33171-33178), which represent the native isoforms. HA1-E1 is exclusively nuclear, whereas HA2-E1 is found predominantly in the cytoplasm. Using high resolution isoelectric focusing and SDS-polyacrylamide gel electrophoresis, we confirm that these epitope-tagged constructs HA1-E1 and HA2-E1 represent the two isoforms E1a and E1b. HA1-E1/E1a exists as one non-phosphorylated and four phosphorylated forms, and HA2-E1/E1b exists as one predominant non-phosphorylated form and two minor phosphorylated forms. We demonstrate that the first 11 amino acids are essential for phosphorylation and exclusive nuclear localization of HA1-E1. Within this region are four serine residues and a putative nuclear localization sequence (NLS; 5PLSKKRR). Removal of these four serine residues reduced phosphorylation levels by 60% but had no effect on nuclear localization of HA1-E1. Each serine residue was independently mutated to an alanine and analyzed by two-dimensional electrophoresis; only serine 4 was phosphorylated. Disruption of the basic amino acids within the NLS resulted in loss of exclusive nuclear localization and a 90-95% decrease in the phosphorylation of HA1-E1. This putative NLS was able to confer nuclear import on a non-nuclear protein in digitonin-permeabilized cells in a temperature- and ATP-dependent manner. Thus the predominant requirement for efficient phosphorylation of HA1-E1/E1a is a functional NLS, suggesting that E1a may be phosphorylated within the nucleus.

The ubiquitin-activating enzyme (E1) 1 catalyzes the first reaction in the ubiquitin (Ub) conjugation pathway. Activation of Ub occurs by the formation of a high energy thiol-ester bond between E1 and the C-terminal glycine of Ub and the production of AMP and PP i . Activated Ub is then transferred to a ubiquitin-conjugating enzyme (ubiquitin-carrier enzyme, E2). E2 proteins conjugate Ub directly to a target substrate or, alternatively, transfer Ub to a ubiquitin-protein ligase (E3), which then conjugates Ub to a target protein (reviewed in Ref. 1 and 2). Multiple rounds of Ub conjugation result in the rapid degradation of the target protein by the 26 S proteasome (3). Recent published results, however, suggest ubiquitination plays an indirect role in protein degradation as well (reviewed in Ref. 4). Ubiquitination on cell surface receptors such as Ste2 (5), yeast ␣ mating receptor (6), and growth hormone receptor (7) triggers their endocytosis and degradation within the lysosome.
Because Ub requires activation prior to participation in any subsequent reactions, E1 plays a key role in the Ub-conjugating pathway. E1 is localized in both the nucleus and the cytoplasm (8,9) and exists as two isoforms E1a (117 kDa) and E1b (110 kDa) (10,11). To investigate the nature of these isoforms, epitope-tagged cDNA constructs of E1 were generated where the hemagglutinin (HA) epitope tag was placed after the first methionine (amino acid 1; HA1-E1) or after the second methionine (amino acid 40; HA2-E1) (11). HA1-E1 localized exclusively to the nucleus and displayed the same molecular weight as E1a, whereas HA2-E1 localized predominantly in the cytoplasm and displayed the same molecular weight as E1b (11). These observations are consistent with the hypothesis that the E1 isoforms are a result of alternate translational start sites. Of these isoforms, E1a/HA1-E1 is phosphorylated in vivo (11) on a serine residue (12), whereas E1b/HA2-E1 is not phosphorylated. Phosphorylation of E1a occurs in a cell cycle-dependent manner (maximal in G 2 ) and the resultant phosphorylated E1a was concentrated within nuclear extracts (13). On the basis of these observations, we proposed that an increase in phosphorylation of E1a functions to increase the import and/or retention of E1a in the nucleus (13).
The present study uses HA epitope-tagged cDNA constructs of E1 to identify amino acids that are important for nuclear localization and phosphorylation and whether phosphorylation of E1 is a prerequisite for its nuclear localization. We identify a specific serine residue that is phosphorylated in addition to a region of basic amino acids that is required for both nuclear localization and phosphorylation. Our data suggest that phosphorylation is not required for nuclear import, but that E1 may require a functional nuclear localization sequence for efficient phosphorylation.

MATERIALS AND METHODS
Cell Culture and Transfection-HeLa cells were cultured in Dulbecco's modified Eagle's medium and 10% fetal calf serum and maintained at 37°C and 5% CO 2 in a humidified chamber as described previously (9). Transient transfections were performed using a calcium phosphate coprecipitation method as described by Chen and Okayama (14); cells were processed 40 -48 h following transfection.
Metabolic Labeling, Immunoprecipitation, and Immunoblot Analysis of HA-E1 Constructs-HeLa cells were metabolically labeled with [ 32 P]orthophosphoric acid (ICN) as described previously (13). Cells were lysed in 20 mM Tris, pH 7.6, 0.25% Triton X-100, 0.2% DTT containing 0.2 mM phenylmethylsulfonyl fluoride, 2.5 g/ml leupeptin, 1 M pepstatin, 1 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, and 5 mM sodium fluoride. The lysates were incubated on ice for 20 min, then centrifuged at 14,000 rpm for 15 min. Protein concentrations of cleared lysates were determined using the Bio-Rad protein assay with bovine serum albumin as standard. HA-tagged E1 was immunoprecipitated from radiolabeled extracts (200 g of protein) using the 12CA5 monoclonal antibody raised against the HA epitope tag, as described previously (11). The immunoprecipitation buffer contained the following phosphatase inhibitors: 1 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, and 5 mM sodium fluoride. Samples were resolved on a 7.5% reducing gel and then transferred to nitrocellulose. HA-tagged E1 was visualized using the 12CA5 antibody, followed by a peroxidaseconjugated goat anti-mouse IgG (11). Immunoreactive proteins were detected using enhanced chemiluminescence (Amersham) and quantified using a Molecular Dynamics Densitometer. The immunoblot was then exposed in a Molecular Dynamics PhosphorImager cassette, and the radiolabeled proteins were quantified using a Molecular Dynamics Storm Optical Scanner.
Immunofluorescence-Immunofluorescence on transiently transfected HeLa cells was performed as described previously (11). HAtagged E1 was detected using the 12CA5 monoclonal antibody followed by rhodamine-conjugated donkey anti-mouse IgG preabsorbed against bovine, goat, horse, human, rabbit, and sheep serum proteins (Jackson).
Conjugation of Peptides to BSA and Labeling with Cy3-Peptides were synthesized (Washington University) and cross-linked to the nonnuclear protein bovine serum albumin (BSA) (Calbiochem) using sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (sulfo-SMCC, Pierce). This cross-links the N-terminal cysteine residue of the peptide to amino groups on the BSA. Sulfo-SMCC (1 mg/ml) was added at 20-fold molar excess to the BSA and incubated for 30 min at room temperature, excess sulfo-SMCC was removed using a Pharmacia PD10 column equilibrated in 100 mM sodium phosphate, pH 7.0. Peptides with a reduced cysteine residue are added in a 50-fold molar excess to the activated BSA and incubated overnight at 4°C. Free peptide was then removed using a Sephadex G50 column equilibrated in 100 mM sodium phosphate, pH 7.0. The conjugated BSA was dialyzed overnight against 100 mM sodium carbonate, pH 9.2. The dialyzed peptide-BSA conjugate was labeled with Cy3 (indocarbocyanine) using the Fluorolink kit from Amersham. Free Cy3 was removed using a Sephadex G50 column equilibrated in phosphate-buffered saline, pH 7.2. The Cy3-BSA peptide was stored at Ϫ80°C.
Nuclear Import Assays-Cell permeabilization and subsequent nuclear import were based on the methods of Adam et al. (15). HeLa cells were grown until they were subconfluent (18 -36 h) on coverslips (15 mm ϫ 15 mm). Cells were washed once with ice-cold import buffer (20 mM Hepes-KOH, pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 g/ml aprotinin, 2 mM dithiothreitol). Cells were then permeabilized with import buffer containing 50 g/ml digitonin (Sigma) for 5 min at 4°C and washed three times with ice-cold import buffer. Coverslips were inverted on 50 l of import mix (containing 50% (v/v) rabbit reticulocyte lysate (Promega), 2 l of 20 mM ATP, 2 l of 100 mM phosphocreatine, and 2 l of 2 mg/ml creatine phosphokinase, 2 l of 20 mg/ml ovalbumin, and 4 l of Cy3-BSA peptide). Cells were incubated for 30 min at either 30°C or 4°C. For import in the absence of ATP, permeabilized cells were first depleted of ATP with 2 l of 2000 units/ml apyrase for 1 h at 4°C. Nuclear import was then performed as before except the ATP regenerating system in the import mix was replaced with apayrase. After import cells were rinsed twice with import buffer and fixed with 4% paraformaldehyde for 15 min at 4°C and then washed with ice-cold import buffer. Samples were analyzed with an Olympus fluorescence microscope, and photographs were taken using a 50ϫ oil objective and Tmax film (Eastman Kodak Co.).
Two-dimensional Gel Electrophoresis-Isoelectric focusing (IEF) was performed using Immobiline Drystrips, pH 4 -7 (Pharmacia), and electrophoresed using the Multiphor II system (Pharmacia). HeLa cells were metabolically labeled with [ 35 S]methionine/cysteine Tran 35 S-label (ICN) or [ 32 P]orthophosphoric acid as described previously (13). Immunoprecipitated E1 or HA-tagged E1 was eluted from Protein A beads with IEF sample buffer (9 M urea, 0.5% Triton X-100, 2% ampholytes, 1% DTT), loaded onto the Drystrip and focused for 3 h at 300 V and for 13 h at 2200 V. After the IEF run, the Drystrips were equilibrated in SDS buffer (0.125 M Tris, 2% SDS, 10% glycerol, 4.9 mM DTT, pH 6.8) for 20 min. The dry strip was placed on a reducing 7.5% SDS-PAGE gel (18 ϫ 16 cm) and electrophoresed for 5 h at 40 mA as the second dimension. The gel was fixed for 30 min and fluoroenhanced in Amplify (Amersham). The dried gel was then exposed to film for autoradiography at Ϫ80°C. Two-dimensional gels were analyzed using a Molecular Dynamics Storm Optical Scanner.
Dephosphorylation of E1-HeLa cells metabolically labeled with either [ 32 P]orthophosphoric acid or Tran 35 S-label were lysed as described above but without the addition of phosphatase inhibitors. Lysate was then dialyzed overnight into 100 mM Tris/HCl, pH 7.0, 150 mM NaCl, 1 mM dithiothreitol. Dialyzed 32 P-or 35 S-labeled lysate (300 g of protein) was incubated with 1 unit of potato acid phosphatase (Sigma) in 100 mM sodium citrate, pH 5.8, 10 mM MgCl 2 overnight at 4°C (final volume of 80 l). The following day another 1 unit of phosphatase was added and incubated for 2 h at ambient temperature. Parallel incubations were included without the addition of phosphatase. E1 was immunoprecipitated from the lysate as described previously (13). 32 P-Labeled E1 was resolved by SDS-PAGE and transferred to nitrocellulose, and E1 was visualized by an anti-E1 polyclonal antibody. 35 S-Labeled E1 was immunoprecipitated and resolved by two-dimensional gel electrophoresis (as described previously).

Two-dimensional Electrophoresis of HA1-E1, HA2-E1, and
Human E1-E1 exists as two isoforms E1a (117 kDa) and E1b (110 kDa). Previously we prepared E1 cDNA constructs in which an HA-epitope tag was placed after the first methionine (HA1-E1) or the second methionine (HA2-E1) ( Fig. 1A; Ref. 11). HA1-E1 and HA2-E1 were similar in their molecular weight and phosphorylation state to E1a and E1b, respectively (11). Two-dimensional gel electrophoresis analysis of E1 from Chinese hamster (ts20) cells resolved E1 into the two isoforms E1a and E1b; E1a resolved further as three phosphorylated and one non-phosphorylated forms, whereas E1b resolved as one nonphosphorylated form (13). To further determine the relationship between human E1, HA1-E1, and HA2-E1, we analyzed these species by two-dimensional gel electrophoresis. HeLa cells were metabolically labeled with Tran 35 S-label or [ 32 P]orthophosphoric acid, and human E1 was immunoprecipitated with a polyclonal antibody raised against human E1 (11). HeLa cells transiently transfected with HA1-E1 or HA2-E1 were metabolically labeled with Tran 35 S-label or [ 32 P]orthophosphoric acid; HA-tagged proteins were immunoprecipitated with a monoclonal antibody recognizing the epitope tag (12CA5). Immunoprecipitated proteins were then resolved in the first dimension by isoelectric focusing, followed by SDS-PAGE in the second dimension ( Fig. 1). 35 S-Labeled human E1 migrated to its predicted isoelectric point of 5.7, but rather surprisingly E1a resolved as five spots and E1b resolved as three spots (Fig. 1B). In addition to these predominant spots, other less abundant species which migrated toward the anode could be detected (Fig. 1B). When human E1 was labeled with [ 32 P]orthophosphoric acid, E1a resolved as four spots; however, no phosphorylation was observed for E1b (Fig. 1B). This is distinct from our previous observations with hamster E1 (13). We attribute these differences to the better resolving capabilities of our isoelectric focusing used in this study. Previously IEF was performed with 1.5-mm rod gels; however, in this study Pharmacia Immobiline Drystrips were used, in which the Drystrips contain a pre-formed immobilized and non-drifting pH gradient, which can be electrophoresed at much higher voltages and results in superior resolution. These differences were not due to cell type as separation of hamster E1 using this method generated the same pattern of spots (data not shown).
The epitope-tagged forms of E1 migrated to the same isoelectric point as the wild type protein (Fig. 1C). 35 S-Labeled HA1-E1 resolved as five main spots (0, Ϫ1, Ϫ2, Ϫ3, and Ϫ4), consistent with our observations of E1a. 35 S-Labeled HA2-E1 resolved as three predominant spots (Ϫ1, 0, and 1), this pattern was very similar to that of E1b. 32 P-Labeled HA1-E1 resolved as four distinct spots ( Fig. 1C; spots Ϫ1, Ϫ2, Ϫ3, and Ϫ4) in a pattern similar to E1a. 32 P-Labeled HA2-E1 resolved as two spots (Ϫ1 and Ϫ2); however, these were only visible with extended exposures ( Fig. 1C; approximately 8 times longer than for HA1-E1). These species are not very abundant; for instance spot 2 could not be detected when labeled with Tran 35 S-label (Fig. 1C) and spot Ϫ1 only represents 2.5% of the total HA2-E1. These phosphorylated forms were not detected in E1b because they are present at levels too low to detect. In addition to the phosphorylated forms of HA1-E1/E1a and HA2-E1/E1b, there are also some spots that migrate toward the anode (spots 1 and 2); these species probably represent alternative charged forms. The pattern of spots for both HA1-E1 and HA2-E1 is essentially FIG. 1. Two-dimensional gel electrophoresis of HA epitope-tagged and wild type E1. HA epitope tags were placed after the first (HA1-E1) or second (HA2-E1) methionine in the E1 sequence as described under "Materials and Methods." (A). HeLa cells were metabolically labeled with either Tran 35 S-label or [ 32 P]orthophosphoric acid for 5 h. Immunoprecipitated proteins were resolved by IEF in the first dimension, followed by SDS-PAGE in the second dimension (B). HeLa cells transiently transfected with HA1-E1 or HA2-E1 were metabolically labeled with either Tran 35 S-label or [ 32 P]orthophosphoric acid, and immunoprecipitated proteins were resolved by IEF and SDS-PAGE (C).
identical to E1a and E1b and is consistent with our hypothesis that the two E1 isoforms result from alternate translational start sites at the first and second in-frame methionines in the E1 sequence. These results also demonstrate that addition of the HA epitope tag to E1 did not alter its phosphorylation state.
To further confirm our observations, we dephosphorylated human E1 by treatment with potato acid phosphatase. Dephosphorylated 35 S-labeled E1a resolved as a pattern of spots similar to E1b; it was found predominantly as spot 0 with minor species migrating as spot Ϫ1, 1, and 2 (data not shown). There was very little change in the pattern of 35 S-labeled E1b spots after dephosphorylation; only a slight decrease in spot Ϫ1 was observed. Taken together, these data suggest that HA1-E1/E1a exists predominantly as four phosphorylated and a non-phosphorylated form; HA2-E1/E1b exists as two minor phosphorylated forms and as a predominant non-phosphorylated form. Furthermore, in addition to the phosphorylated forms, both HA1-E1/E1a and HA2-E1/E1b can be resolved into other nonphosphorylated charged species. A relatively abundant species (spot 1) is detected in 35 S-labeled HA2-E1/E1b preparations, although the precise nature of these charged variants is currently not known.
Construction of HA Epitope-tagged E1 N-terminal Truncation Mutants-We and others have previously demonstrated that E1a/HA1-E1 are phosphorylated, whereas no detectable phosphorylation has been demonstrated for E1b/HA2-E1 (11)(12)(13). However, in our current study it appears that HA2-E1/E1b is phosphorylated but at levels approximately 100-fold less than HA1-E1/E1a (Fig. 1). Others have recently determined that E1 is only phosphorylated on serine residues (12). As HA1-E1 is 40 amino acids longer than HA2-E1, it is tempting to speculate that this region at the N terminus contains the predominantly phosphorylated serine residues. Within this 40 amino acid region, there 10 serine residues. Thus, to identify which serines are involved in phosphorylation of HA1-E1, a series of N-terminal truncation mutants were generated ( Fig.  2A). Constructs were made where the HA epitope was attached to the N-terminal region of E1 to create proteins that were truncated in segments of about 10 amino acids ( Fig. 2A). These HA-tagged E1 truncation mutants were transiently expressed in HeLa cells and metabolically labeled with [ 32 P]orthophosphoric acid. HA-tagged constructs were then immunoprecipitated using the 12CA5 monoclonal antibody and resolved by SDS-PAGE and transferred to nitrocellulose. The total amount of immunoreactive HA-tagged construct was determined by immunoblot using the 12CA5 antibody (Fig. 2B). The nitrocellulose was subjected to autoradiography to determine the 32 P incorporation (Fig. 2B). Removal of the first 11 amino acids of the E1 sequence (HA1-E1-del-11) resulted in no detectable phosphorylation. Removal of 22 (HA1-E1-del-22), 30 (HA1-E1del-30), and 40 amino acids (HA2-E1) at the N terminus also resulted in no detectable phosphorylation of the HA-tagged constructs. These data thus suggest the first 11 amino acids contain residues that are essential for efficient phosphorylation of HA1-E1/E1a.
Immunolocalization of HA-tagged E1 N-terminal Truncation Mutants-Phosphorylation of certain proteins correlates with their retention in either the nucleus or cytoplasm (reviewed in Ref. 16). For example SWI5 is excluded from the nucleus when phosphorylated (17), whereas STAT-3 is translocated to the nucleus after phosphorylation (18). Indeed this is the case with E1; HA1-E1 is localized exclusively in the nucleus and phosphorylated, whereas HA2-E1 is found almost exclusively in the cytoplasm and is phosphorylated approximately 100-fold less than HA1-E1 (11). Previously we have determined the subcellular localization of HA1-E1-del-11 by immunofluorescence using the 12CA5 monoclonal antibody (11). Removal of the first 11 amino acids resulted in predominant cytoplasmic staining with both positive and negative nuclear staining, as was also observed with HA2-E1 (11). We thus determined the immunofluorescent localization of the HA-tagged E1 truncation mutants, HA1-E1-del-22 and HA1-E1-del-30. These constructs were localized with the same distribution as HA1-E1-del-11 (data not shown). Therefore, in addition to amino acids essen- FIG. 2. Phosphorylation of HAtagged E1 N-terminal truncation mutants. A series of HA-tagged E1 species were generated which had successive 10 or 11 amino acids segments removed form their N terminus (A). HA1-E1 and these constructs were transiently transfected into HeLa cells and metabolically labeled with [ 32 P]orthophosphoric acid for 5 h. Immunoprecipitated proteins were resolved by SDS-PAGE and transferred to nitrocellulose. The HA-encoding constructs were visualized using the 12CA5 antibody and chemiluminescence (B). The nitrocellulose was then exposed to film to detect 32 P-labeled proteins (B). tial for phosphorylation, the first 11 amino acids also contains residues necessary for the exclusive nuclear localization of HA1-E1.
Identification of Amino Acids Which Are Required for Phosphorylation and Nuclear Localization-The first 11 amino acids of the E1 sequence contains, in addition to serines 2, 3, 4, and 7, a putative nuclear localization sequence (NLS), 5 PL-SKKRR 11 . To determine which residues are responsible for phosphorylation or for nuclear localization, two additional HAtagged E1 constructs were prepared: HA1-E1-del-7, where the first 7 amino acids (including serines 2, 3, 4, and 7) were removed, and HA1-E1-del-8 -11, where the basic region of the putative NLS ( 8 KKRR 11 ) was deleted (Fig. 3A). These constructs along with HA1-E1 were transiently transfected into HeLa cells, labeled with [ 32 P]orthophosphoric acid, resolved by SDS-PAGE, and transferred to nitrocellulose. The total amount of HA-tagged construct was determined by immunoblot analysis, and 32 P incorporation was determined by autoradiography (Fig. 3B). To determine the specific phosphorylation of the HA-tagged constructs, the total amount of HA-tagged E1 constructs was quantified by densitometry, and 32 P incorporation was determined by PhosphorImager analysis (Fig. 3B). Removal of the basic amino acids ( 8 KKRR 11 ) in the putative NLS from HA1-E1 drastically reduced the level of specific phospho-rylation to 5% of HA1-E1 (Fig. 3B). Additionally HA1-E1-del-8 -11 was no longer localized exclusively in the nucleus but found predominantly in the cytoplasm (approximately 90% of the cells displayed completely negatively stained nuclei) (Fig.  3C). This distribution is identical to that observed with HA1-E1-del-11. This implies that the basic region ( 8 KKRR 11 ) of the putative nuclear localization sequence at the N terminus of E1 is required for the exclusive nuclear localization of HA1-E1. The dramatic decrease in phosphorylation of HA1-E1-del-8 -11 and the predominant cytoplasmic distribution strongly suggest a correlation between nuclear localization and phosphorylation.
The dramatic reduction in specific phosphorylation and predominant cytoplasmic localization of HA1-E1-del-8 -11 suggests that phosphorylation occurs within the nucleus. An alternative explanation may be that the kinase which phosphorylates HA1-E1 on one or more of the serine 2, 3, 4, and 7 residues binds to the basic region of the NLS ( 8 KKRR 11 ). Thus removal of this region would disrupt both nuclear targeting and phosphorylation. In an attempt to reconcile these two possibilities a further HA-tagged construct was generated where arginine residues 10 and 11 were changed to alanines (HA1-E1-R10A,R11A; Fig. 4A). Removal of these two arginine residues also disrupts nuclear targeting (Fig. 4C) and substantially reduces the specific phosphorylation to levels comparable with HA1-E1-del-8 -11 ( Fig. 4B; 9% of HA1-E1). Although these results are not conclusive, they do suggest that phosphorylation of HA1-E1 may occur within the nucleus.
The E1 NLS Efficiently Imports a Non-nuclear Protein into the Nucleus-The archetypal NLS is that of SV40 large T antigen, which consists of a basic patch of amino acids on an ␣-helix (19,20). The putative E1 NLS ( 5 PLSKKRR 11 ) shows considerable sequence similarity to the SV40 NLS (71% homology over 7 amino acids) (Fig. 5A). Digitonin-permeabilized cells have been used to assay putative NLS motifs for their ability to import non-nuclear proteins to the nucleus (15). This assay system was employed to determine if the putative E1 NLS could import BSA into the nucleus of digitonin-permeabilized HeLa cells. Peptides were synthesized and cross-linked to BSA using sulfoSMCC via a terminal cysteine residue. The BSA peptide was then labeled with the fluorescent compound Cy3. Using reticulocyte lysate as a cytosol source, Cy3-BSA peptides were then assayed for their ability to be imported into the nucleus in digitonin-permeabilized HeLa cells (Fig. 5B). Assays were carried out in the presence of ATP at 30°C, the absence of ATP at 30°C, or the presence of ATP at 4°C. Using this system nuclear import of BSA-NLS SV40 was dependent upon both ATP and temperature (data not shown), consistent with previously published data (15). The E1 NLS imported BSA to the nucleus in the presence of ATP and at 30°C, but not in the absence of ATP nor at 4°C (Fig. 5B). When KKRR were replaced with alanines (BSA-NLS E1mutant ), BSA was no longer imported to the nucleus (data not shown).
Others have demonstrated that phosphorylation at or near a HA-tagged E1 constructs were generated where either arginine 10 and 11 were mutated to alanines or the first 4 amino acids were deleted and serine 7 mutated to an alanine (A). HA1-E1, HA1-E1-del-4,S7A, and HA1-E1-R10A,R11A were transfected into HeLa cells and metabolically labeled with [ 32 P]orthophosphoric acid. The phosphorylation state of each HA-E1 species was determined (B). The specific phosphorylation of each construct was compared with HA1-E1 (B). The subcellular localization of each HA-E1 species was determined by immunofluorescence (C).
NLS may increase the rate of nuclear import or nuclear retention (16,21). Within the putative E1 NLS ( 5 PLSKKRRV 12 ), there is one serine residue. We did not determine whether this serine residue was phosphorylated when this peptide was conjugated to BSA. However, a third BSA peptide was constructed where serine 7 was changed to an alanine (BSA-NLS E1-S7A ; 5 PLAKKRRV 12 ) to determine if this serine played a role in nuclear import. There was no substantial difference in nuclear import between BSA-NLS E1-S7A and BSA-NLS E1 (Fig. 5B). Thus, we conclude that the E1 sequence 5 PLSKKRRV 12 is able to confer nuclear import on a non-nuclear protein, and within this region KKRR is essential, whereas serine 7 is not.
Identification of Individual Serine Residues Which are Phosphorylated-To address which serine residues are phosphorylated within the first 7 amino acids of HA1-E1, a series of serine point mutants were generated where serine residues 2, 3, 4, or 7 were individually changed to an alanine (Fig. 6A). When the specific phosphorylation of each of these constructs was compared with HA1-E1 (Fig. 6B), HA1-E1-S4A was consistently the lowest (approximately 60%). HA1-E1-S2A, HA1-E1-S3A, and HA1-E1-S7A generally had levels of specific phosphorylation similar to the wild type HA1-E1. However, between experiments the level of specific phosphorylation of each construct compared with HA1-E1 varied somewhat. Therefore, to determine definitively which residues are phosphorylated, each construct was analyzed by IEF followed by SDS-PAGE. Each construct was transiently transfected into HeLa cells and metabolically labeled with [ 32 P]orthophosphoric acid; HA-containing proteins were immunoprecipitated and subjected to two-dimensional electrophoresis (Fig. 6C). HA1-E1 migrates as four phosphorylated species when analyzed by two-dimensional electrophoresis. Serine point mutants HA1-E1-S2A, HA1-E1-S3A, and HA1-E1-S7A all migrated as four phosphorylated species (Fig. 6C). However, HA1-E1-S4A migrated as only three phosphorylated species indicating that changing serine 4 to an alanine removed one phosphorylation site from HA1-E1. Thus serine 4 is phosphorylated in HA1-E1/E1a. DISCUSSION Analysis of human E1, HA1-E1, and HA2-E1 by two-dimensional gel electrophoresis results in very similar electrophoretic patterns for both E1a/HA1-E1 and E1b/HA2-E1 (Fig. 1), respectively. This pattern differs from those previously reported for hamster E1 isoforms (13). We attribute these differences to the superior resolution of our isoelectric focusing gels used in this study. However, the data strongly suggest that E1a/ HA1-E1 and E1b/HA2-E1 represent equivalent proteins, with E1a being translated from the first methionine and E1b from the second.
E1b/HA2-E1 exists as two minor phosphorylated forms, a predominant non-phosphorylated form, and several forms which migrated toward the anode (Fig. 1). Densitometric quantification of 35 S-labeled E1b/HA2-E1 revealed that only 2.5-4% is found as the phosphorylated forms. Thus only a very small fraction of E1b is phosphorylated and may explain why previous studies by ourselves (11) and others (12,22) were unable to detect it. E1a/HA1-E1 resolves as four phosphorylated and one non-phosphorylated forms (Fig. 1). Densitometric quantification of the 35 S-labeled E1a/HA1-E1 spots revealed that 92-97% is found in the phosphorylated forms, respectively. Several minor positively charged species were also observed (as described previously for E1b/HA2-E1). Dephosphorylation of E1a changed the electrophoretic migration to one very similar to E1b. This suggests that in the basal non-phosphorylated state E1 exists as a heterogeneous mixture of charged forms. This pattern of charged forms was extremely reproducible and was also observed with hamster E1 (data not shown). This heterogeneous mixture of charged forms may arise from modifications such as deamination; however, further work will be required to resolve this.
In an attempt to identify which serine residues are phosphorylated in HA1-E1, N-terminal truncation mutants were generated ( Fig. 2A). Removal of only the first 11 amino acids completely abolished phosphorylation of the HA-tagged E1 (Fig. 2B). Removal of further segments of the N terminus also FIG. 5. In vitro nuclear import assays using digitonin-permeabilized HeLa cells. Peptides were cross-linked to BSA and the BSA peptides were then conjugated with the fluorescent label Cy3. Residues within the peptide shown in bold correspond to sequence from the wild protein (A). Each Cy3-BSA peptide was assayed for the ability to localize to the nucleus of digitonin-permeabilized cells in either of the following conditions; in the presence of ATP at 30°C, or the absence of ATP at 30°C or the presence of ATP at 4°C (B). resulted in no detectable phosphorylation (Fig. 2B). Thus the first 11 amino acids contain the necessary information for complete phosphorylation of HA1-E1/E1a. HA1-E1 is localized exclusively in the nucleus; however, removal of the first 11 amino acids changes the subcellular localization to predominantly cytoplasmic (11). Consistent with this, HA1-E1-del-22, HA1-E1-del-30, and HA2-E1 are all found predominantly in the cytoplasm (data not shown and Ref. 11). Hence, the first 11 amino acids in the E1 sequence contain the necessary information for exclusive nuclear localization and complete phosphorylation.
For efficient nuclear transport, proteins larger than approximately 50 kDa require a NLS (23). The putative NLS ( 5 PL-SKKRR) at the extreme N terminus of E1 shares 71% homology with the SV40 NLS, which is regarded as the archetype. Deletion of the four basic residues from this putative NLS (HA1-E1-del-8 -11) completely abrogated the exclusive nuclear localization of HA1-E1 (Fig. 3C). In addition, mutation of arginines 10 and 11 to alanines (HA1-E1-R10A,R11A) also abrogated the exclusive nuclear localization (Fig. 4C). Conjugation of this sequence ( 5 PLSKKRRV 12 ) to a non-nuclear protein (BSA) conferred import to the nucleus of digitonin-permeabilized cells (Fig. 5B). This import was both temperature-and ATP-dependent (Fig. 5B). Substitution of 8 KKRR with alanines eliminated nuclear import of the substrate (BSA-NLS E1mutant , data not shown). Therefore by all criteria 5 PLSKKRRV 12 functions as an NLS in E1. However, despite the removal or disruption of this NLS, 5-10% of cells transfected with HA1-E1-del-8 -11, HA1-E1-R10A,R11A, and HA2-E1 still show immunofluorescent staining in the nucleus (Ref. 11 and Fig. 4C). This suggests that there may be other functional NLS motifs within the E1 se-quence. Indeed, within the C-terminal region there is a putative bipartite NLS (KERLDQPMTEIVSRVSKRK). Bipartite NLS motifs are characterized by two interdependent basic domains separated by 10 intervening "spacer" amino acids (24). Future studies will address the role of this region in nuclear localization of our HA-tagged constructs.
Several proteins have been identified whose rate of import to or retention in the nucleus is enhanced by phosphorylation near a NLS (16). For example, phosphorylation by casein kinase II near the SV40 large T antigen NLS increased the rate of nuclear import (21). Within the N-terminal 11 amino acids of the E1a sequence, there are serines 2, 3, 4, and 7. Serine 7 is a predicted protein kinase C site, and serine 4 can be phosphorylated by Cdc2 (22). Both of these residues are close to the NLS. Thus we defined the phosphorylation state of these serine residues and their potential role in nuclear targeting (Fig. 3). When the first 7 amino acids were deleted (HA1-E1-del-7, including the four serine residues), HA1-E1-del-7 was phosphorylated approximately 60% less than HA1-E1 (Fig. 3B) and was localized predominantly in the nucleus (Fig. 3C). However, approximately 3% of the cells also showed some cytoplasmic localization. One explanation for this observation may be that a certain level of phosphorylation is required for exclusive nuclear localization of HA1-E1. Alternatively, deletion of the first 7 residues includes 5 PLS of the putative NLS, and this proline residue may be required for a secondary structural motif necessary for a functional NLS. To address this issue, another construct was generated where the first 4 amino acids were deleted and serine 7 was substituted with an alanine (HA1-E1-del-4,S7A). Phosphorylation of HA1-E1-del-4,S7A was approximately 20% that of HA1-E1 (Fig. 4B) and was FIG. 6. Phosphorylation and two-dimensional gel electrophoresis of HA1-E1 serine point mutations. The first four serine residues at the N terminus of HA1-E1 were individually changed to alanine residues (A). These constructs were transfected into HeLa cells and metabolically labeled with [ 32 P]orthophosphoric acid. The phosphorylation state of each construct was determined, and the specific phosphorylation of each serine point mutant was quantified and compared with wild type HA1-E1. The histogram shows the mean specific phosphorylation of each construct from two separate experiments (B). 32 P-Labeled constructs were resolved by two-dimensional gel electrophoresis as described previously (C). localized exclusively in the nucleus (Fig. 4C). Additionally, in our in vitro nuclear import assays substitution of serine 7 for an alanine (BSA-NLS E1-S7A , Fig. 5) did not alter the efficiency of nuclear import of BSA. Hence, phosphorylation on any of the serine 2, 3, 4, or 7 residues is not essential for nuclear import, whereas it appears that the proline and leucine residues are required for exclusive nuclear targeting of HA1-E1.
The specific phosphorylation level of HA1-E1-del-7 is 60% less than the wild type HA1-E1. This suggests that one or more of the four serine residues within this 7-amino acid region are phosphorylated. Each of these serine residues was substituted individually to an alanine, and their specific phosphorylation compared with the wild type (Fig. 6). HA1-E1-S4A was approximately 50% lower than the wild type. However, the precise level of phosphorylation of each of the constructs varied from experiment. To delineate exactly which serine was phosphorylated, each of the 32 P-labeled constructs were analyzed by twodimensional gel electrophoresis (Fig. 6C). HA1-E1-S2A, HA1-E1-S3A, HA1-E1-S7A, and the wild type all resolved as four phosphorylated spots; however, HA1-E1-S4A resolved as only three spots (Fig. 6C). This suggests that serine 4 is phosphorylated in HA1-E1/E1a and is consistent with observations made by Nagai et al. (22). Our data also suggest that serine 4 is the site that is phosphorylated most predominantly as removal of this site reduces the level of phosphorylation by 50% of the wild type.
Disruption of the NLS by removal of the KKRR (HA1-E1-del-8 -11) or substitution of the arginines for the alanines (HA-E1-R10A,R11A) results in loss of exclusive nuclear localization of HA1-E1 (Figs. 3C and 4C). However, between 5 and 10% of transfected cells still showed some detectable nuclear localization. In addition, HA1-E1-del-8 -11 and HA1-E1-R10A,R11A were phosphorylated to less than 10% of HA1-E1 (Figs. 3B and 4B). This suggests that nuclear localization of HA1-E1 may be required for efficient phosphorylation. Only 3-5% of the total E1b is present in the nucleus (data not shown), and HA2-E1 is phosphorylated approximately 100-fold less than HA1-E1. HA2-E1/E1b also lacks serine 4, which is predominantly phosphorylated in HA1-E1. Thus it appears that the very low levels of phosphorylation of HA2-E1/E1b are due to the absence of serine 4 and its relative abundance outside the nucleus.
Despite these data, which suggest HA1-E1/E1a are phosphorylated in the nucleus, the precise function of E1 phosphorylation is not known. Previous work demonstrated that E1a is phosphorylated in a cell cycle-dependent manner; being maximally phosphorylated in G 2 (13). However, no change in the enzymatic activity of E1a was observed with increased phosphorylation. Phosphorylation is not required for nuclear localization of HA1-E1 (Figs. 3 and 4), but may increase the rate of nuclear targeting and/or nuclear retention, neither of which was determined in our in vitro nuclear import assays (Fig. 5). Alternatively, phosphorylation of E1a may be required or important for stable interaction (or complex formation) with specific nuclear E2 proteins. E1 can be phosphorylated by p34 cdc2 in vivo and in vitro (22). Pines and Hunter (25) have determined the localization of cyclins A and B, both of which associate with p34 cdc2 to form the active kinase. Cyclin B is nuclear from the beginning of mitosis; however, cyclin A is nuclear from S phase onward and may well associate with p34 cdc2 to phosphorylate substrates (including E1) in G 2 .
Many nuclear proteins including transcription factors such as p53, c-Fos, N-Myc, c-Myc (26), and signal transducers such as STAT1 (27) are substrates for the ubiquitin-proteasome pathway. This suggests the requirement for a functional ubiquitin-proteasome system within the nucleus. The proteasome is present in both the cytoplasm and the nucleus (reviewed in Ref. 28), and recent work has identified the functional NLS motifs present within the ␣ subunits (29,30). Interestingly, proteasomes purified from nuclear fractions were found to contain the mammalian homolog to SUG1 (31). In our current study we present evidence demonstrating that nuclear E1 is highly phosphorylated, whereas the cytoplasmic form is not. The enzymes of the ubiquitin-proteasome pathway present in the nucleus may have modifications that distinguish them from their counterparts within the cytoplasm, perhaps necessary for the different substrate specificities of a nuclear ubiquitin-proteasome system.