Acetylation of the Adenovirus-transforming Protein E1A Determines Nuclear Localization by Disrupting Association with Importin- (cid:1) *

Posttranslational modifications may alter the biochemical functions of a protein by modifying associations with other macromolecules, allosterically altering intrinsic catalytic activities, or determining subcellular localization. The adenovirus-transforming protein E1A is acetylated by its cellular targets, the co-activators CREB-binding protein, p300, and p300/CREB-binding protein-associated factor in vitro and also in vivo at a single lysine residue (Lys 239 ) within a multifunctional carboxyl-terminal domain necessary for both nuclear localization and interaction with the transcriptional co-repressor carboxyl-terminal binding protein (CtBP). In contrast to a previous report, we demonstrate that acetylation of Lys 239 does not disrupt CtBP binding and that 12 S E1A-mediated repression of CREB-binding pro-tein-dependent transcription does not require recruitment of CtBP. Instead we find that the cytoplasmic fraction of E1-transformed 293 cells is enriched for acetylated E1A with relative exclusion from the nuclear compartment. Whereas wild type 12 S E1A binds importin- (cid:1) 3, binding affinity was markedly reduced both by single amino acid substitution mutations and acetylation

Posttranslational modifications may alter the biochemical functions of a protein by modifying associations with other macromolecules, allosterically altering intrinsic catalytic activities, or determining subcellular localization. The adenovirus-transforming protein E1A is acetylated by its cellular targets, the co-activators CREB-binding protein, p300, and p300/CREB-binding protein-associated factor in vitro and also in vivo at a single lysine residue (Lys 239 ) within a multifunctional carboxyl-terminal domain necessary for both nuclear localization and interaction with the transcriptional corepressor carboxyl-terminal binding protein (CtBP). In contrast to a previous report, we demonstrate that acetylation of Lys 239 does not disrupt CtBP binding and that 12 S E1A-mediated repression of CREB-binding protein-dependent transcription does not require recruitment of CtBP. Instead we find that the cytoplasmic fraction of E1-transformed 293 cells is enriched for acetylated E1A with relative exclusion from the nuclear compartment. Whereas wild type 12 S E1A binds importin-␣3, binding affinity was markedly reduced both by single amino acid substitution mutations and acetylation at Lys 239 . This is the first demonstration that acetylation may alter nuclear partitioning by direct interference with nuclear import receptor recognition. The finding that the cytoplasmic fraction of E1A is acetylated indicates that E1A may exert its pleiotropic effects on cellular transformation in part by affecting cytoplasmic processes.
Viral oncoproteins often target important steps cellular metabolism to exert effects on cellular proliferation and differentiation. The E1 early region of human adenoviruses is the first portion of the adenovirus genome expressed on infection of a host cell and encodes the prototypical oncoproteins E1A and E1B. Whereas both E1 region products are necessary for full viral oncogenic transformation, the E1A products alone can induce S phase, immortalize cells, and cooperate with other cellular and viral oncogenes to transform primary rodent cells (1)(2)(3)(4).
The cellular effects of E1A are thought to be mediated primarily by effects on gene transcription, modulating both cellular and viral gene expression through physical associations with cellular transcriptional regulatory proteins (5). The E1A region of adenovirus encodes two major mRNAs (12 and 13 S), generated through alternative splicing of two exons from a single primary transcript. For the C-type adenovirus (serotypes 2 and 5), these mRNAs encode proteins of 243 and 289 amino acids, respectively (Fig. 1A). Amino acid sequences of E1A isoforms from each of the adenovirus serotypes differ substantially, except within CRs 1 that act as interaction motifs for E1A target proteins. Both forms interact with cellular proteins including pRB and the transcriptional co-activators CBP and p300 through CR2, and CR1 and N-terminal regions, respectively. The E1A 13 S form (E1A 289R) differs from the 12 S form (E1A 243R) in the presence of an additional first exonderived sequence (CR3), which serves as an interaction domain for ATF-2 and TBP/TFIID (5). Whereas both of the two major forms of E1A are necessary for productive infection and viral transformation, the CR3 region is dispensable for immortalization and cellular transformation in conjunction with cellular oncogenes (1,2,6).
The 12 S E1A protein inhibits CBP-and p300-dependent transcription by interacting with and repressing the activities of these co-activators, although the mechanism of repression remains controversial. E1A repression of the co-activator function of CBP depends on a direct physical interaction between the N terminus of E1A and at least one domain of the CBP and p300 co-activators (7)(8)(9). This region of CBP/p300 serves as an interaction site for a large number of transcription factors, including c-Fos, E2F, MyoD, and the basal factor TFIIB (10). Models proposed for inhibition of CBP/p300 function have included sequestration of CBP/p300 from active transcription complexes (11)(12)(13) and displacement of positive transcription factors by competitive binding (14,15). Whereas some reports suggest that E1A may repress the HAT activity of CBP/p300 (16 -18), others suggest that E1A may stimulate CBP/p300 HAT activity in vivo (19) and may enhance in vivo acetylation of pRB (20).
The second exon-encoded functions of E1A are likewise functionally complex. Deletion mutations within the C terminus of E1A relieve negative effects on transformation and correlate in part with a loss of binding of E1A to CtBP, a 48-kDa phosphoprotein (21,22). CtBP1 and a second isoform, CtBP2 (23,24), bind E1A through a conserved -PLDLS-motif near the common C terminus of both E1A isoforms (Fig. 1A). This motif and variant sequences (-PXDLS-) are also found in a number of cellular transcriptional repressor proteins important in cellular proliferation, growth, and differentiation (25). In some of these CtBPinteracting proteins, mutation of a core -PXDLS-motif both decreases binding of CtBP and correlates with either abolished or reduced repressor activity (24, 26 -30), whereas in others (31), mutation of this motif has little or no effect on repression.
The biochemical consequences of CtBP binding to E1A are less well defined. One model implicates CtBP in repression of CR1-dependent transactivation through E1A (32). Alternatively, E1A may act to displace or sequester CtBP from cellular targets that contain the -PXDLS-motif (33), a model more consistent with proposed mechanisms for E1A regulation of p300/CBP and pRB activities.
A recent report suggests that CtBP may directly participate in the activity of E1A and mediate repression of CBP-dependent co-activation by E1A. Zhang et al. (34) report that p300-or P/CAF-dependent acetylation of 12 S E1A at Lys 239 within the -PLDLSCK 239 -motif attenuates E1A repression of CBP activity by impairing the interaction of CtBP with E1A. In this model, recruitment of the CtBP co-repressor complex is essential for the repression of CBP-dependent co-activation by E1A. Whereas several reports have indicated that sequences outside of the -PLDLS-might influence the interaction of CtBP with its targets, in E1A this lysine has been shown to be dispensable in peptide competition studies (35), and examination of -PXDLSand surrounding sequences of the known probable CtBP cellular targets indicates that this lysine is not invariably conserved (36). Furthermore, previous studies demonstrate that a sequence carboxyl-terminal to the -PLDLS-motif, including Lys 239 , is essential for nuclear partitioning of E1A (37)(38)(39)(40).
In this report, we demonstrate that like P/CAF and p300, CBP acetylates lysine 239 of 12S E1A in vitro and that a small fraction of E1A is acetylated at this lysine in vivo in E1transformed HEK293 cells and E1A-transfected cells. In our experiments, single amino acid substitution mutations and bona fide acetylation of full-length E1A at Lys 239 do not disrupt binding to CtBP in vitro. Whereas substitution mutations at this position abrogate 12 S E1A repression of phospho-CREBdependent, CBP-mediated co-activation of transcription due to impaired nuclear localization, mutations within the -PLDLSmotif that completely disrupt CtBP binding have no effect on E1A-mediated repression of CBP function, suggesting that the association of CtBP with E1A is not necessary for repression of CBP activity. Instead, consistent with a previous study (39), we find that the lysine at position 239 is invariably essential for the nuclear localization of 12 S E1A, and further that E1A acetylated at this lysine resides predominantly in the cytoplasm of E1Aexpressing cells. 12 S E1A binds strongly to importin-␣3 in vitro, while Lys 239 substitution mutations and acetylation of E1A by CBP abrogate this interaction. These results indicate that acetylation of E1A determines subcellular localization by interfering with nuclear import. Acetylation of E1A may act to either attenuate the nuclear functions of E1A or redirect a portion of 12 E1A to cytoplasmic targets. These results raise the possibility that E1A may exert its pleiotropic effects on cellular transformation in part by affecting cytoplasmic processes.

EXPERIMENTAL PROCEDURES
Plasmids-Mammalian expression pRc/RSV-derived expression plasmids for wild type adenovirus type 2 12 S E1A protein (243R), CBP, CREB 341 , and the catalytic subunit of protein kinase A have been previously described (8). Individual cDNAs encoding wild-type adenovirus type 5 13 S E1A protein (289R) and deletion mutants mCR1 (deletion of amino acids 38 -65), mCR2 (deletion of amino acids 125-133), and mCR3 (deletion of amino acids 140 -185), obtained from Dr. Ganes Sen (Cleveland Clinic), were inserted into pRc/RSV (Invitrogen) by PCR (41). Site-directed mutagenesis at lysine 239 in 12 S E1A or the corresponding lysine 285 in 13 S E1A was performed by using the QuikChange TM system (Stratagene). Eukaryotic Rc/RSV expression vectors for wild type 12 S E1A and K239A 12 S E1A with the carboxylterminal fusion of the SV40 T-antigen nuclear localization signal (-PKKKRKV) (42) were produced by PCR. A plasmid for expression of histidine-tagged E1A in Escherichia coli was constructed by cloning the 12 S E1A cDNA in-frame into pET23d (Novagen) by PCR. A bacterial expression plasmid for expression of a GST fusion protein with fulllength CtBP1 (pGST-CtBP1) was constructed by cloning the full-length cDNA for human CtBP1 (derived from pT7-CtBP1; generously provided by Dr. G. Chinnadurai, St. Louis University) in frame into pGEX-GK (43) by PCR. The full-length cDNA for human importin-␣3 (44), provided by Dr. Mattias Kohler, was inserted into pGEX-GK by PCR. DNA constructs and mutations were confirmed by sequencing.
The full-length murine CBP cDNA (46), with two copies of the FLAG epitope at the carboxyl terminus (CBP-2ϫFLAG), was cloned into pFastBac1, and recombinant baculovirus producing epitope-tagged fulllength CBP was produced in Sf9 cells according to the manufacturer's instructions (Bac-to-Bac; Invitrogen). CBP-2ϫFLAG was purified by ␣-FLAG M2 immunoaffinity chromatography, followed by elution with FLAG peptide (Sigma) as previously described (47). The purity of eluted CBP was greater than 95% as determined by Coomassie staining.
Acetylation Assays-E1A was acetylated in vitro in a reaction containing 5 M purified 12 S E1A, 300 nM CBP-2ϫFLAG, 10 M Production of E1A Acetyl-lysine 239-specific Monoclonal Antibody (␣-AcK-E1A)-A synthetic peptide corresponding to the C terminus of Ad2 E1A, designed with an additional N-terminal cysteine (-CEDLLNESGNPLDLSCK 239 RPRP 243; acetylated at the ⑀-amino position of lysine 239) was obtained from Sigma, purified by reverse phase high pressure liquid chromatography. Peptide was coupled to maleimide-activated keyhole limpet hemocyanin (Imject; Pierce), mixed with Freund's adjuvant, and injected into mice. Sera from injected mice were screened for immunoreactivity toward recombinant E1A acetylated in vitro by CBP. Mice with high titer immunoreactivity for acetylated E1A were sacrificed to obtain splenocytes for hybridoma fusion by standard procedures (49). Clonal hybridoma supernatants were screened by Western blot for selective immunoreactivity toward acetylated E1A.
Cell Culture, Transfection, and Reporter Assays-Monolayer cultures of U2OS cells, COS-7 cells, and HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. COS-7 and U2OS cells were transfected with E1A expression vectors using the TransIT reagent (PanVera, Madison, WI). Assay of E1A-mediated repression of CBP-dependent co-activation of the somatostatin promoter was performed as previously described (8). In brief, F9 teratocarcinoma cells were transfected by the calcium phosphate method (Calcium Phosphate Transfection Kit; Invitrogen) using 4 g of p(Ϫ71) SRIF-CAT, 4 g of pRc/RSV-FLAG-CREB 341 , 4 g of pRSV-PKA, and 2 g of pRSV-Luciferase, with or without 15 g of pRc/RSV-CBP per 10-cm plate, with 1, 5, or 10 g of plasmids encoding pRc/RSV-12 S E1A and 12 S E1A mutants as indicated. Results are expressed as the relative CAT activity (mean Ϯ S.E., n ϭ 3) compared with CAT activity in the presence of CREB and protein kinase A alone (CREB alone column). CAT activity values were normalized to luciferase activity as previously described (50).
Immunocytochemistry and Western Blotting-Transfected COS-7 cells and 293 cells were washed with phosphate-buffered saline, fixed with 4% paraformaldehyde for 4 min at room temperature, incubated with ice-cold methanol/acetone (1:1) for 2 min, permeabilized with 0.1% Triton X-100 in phosphate-buffered saline, and blocked with 3% bovine serum albumin in phosphate-buffered saline prior to incubation with E1A-specific antibodies. E1A was detected with either M73 alone (a carboxyl-terminal, E1A-specific monoclonal antibody), a mixture of E1A-specific antibodies (M1, M29, M37, and M73) that recognize multiple E1A epitopes (48), or the ␣-AcK-E1A monoclonal antibody and then detected with an Oregon Green 488-goat anti-mouse IgG conjugate (Molecular Probes, Inc., Eugene, OR). Nuclei were counterstained with TOPRO-3 (Molecular Probes). Coverslips were washed in phosphatebuffered saline and viewed by either conventional fluorescence microscopy or on a Nikon TE 300-inverted microscope with a Bio-Rad MRC 1024 confocal imaging system.
Whole cellular extracts were prepared as previously described (8). For preparation of nuclear and cytoplasmic extracts, cells were harvested and washed in phosphate-buffered saline and then resuspended in 10 mM Tris-HCl, 3 mM MgCl 2 (TM) with 250 mM sucrose and lysed with a Dounce homogenizer (B pestle). Nuclei were collected by centrifugation at 250 ϫ g, and the supernatant (cytosol) was centrifuged at 100,000 ϫ g for 30 min (Beckman TLA100.2) to pellet cellular membranes. Nuclei were washed twice and then brought to 0.6 M sucrose and layered over a solution of 1.7 M sucrose in TM and centrifuged at 28,000 rpm in a SW-28 rotor (Beckman) for 2 h. The purified nuclear pellet was washed twice in TM ϩ 250 mM sucrose. Nuclear extracts were prepared as previously described (51). Cytoplasmic and nuclear extracts were probed with a monoclonal ␣-tubulin antibody to confirm adequate fractionation. Total protein concentrations were determined by Bradford Reagent (Bio-Rad). Cellular proteins were fractionated by SDS-PAGE, transferred to polyvinylidene difluoride membrane (Immobilon-P; Millipore Corp.), and detected by incubation with antibodies as described for each experiment, by using a goat anti-mouse IgG-horseradish peroxidase conjugate (Bio-Rad) and enhanced luminol reagent (Renaissance; PerkinElmer Life Sciences).

E1A Is Acetylated at a Single Lysine Residue Both in Vitro and in
Vivo-Previous studies have demonstrated that the nuclear histone acetyltransferases acetylate not only free and nucleosomal histones but also nonhistone transcriptional regulatory proteins (reviewed in Refs. 52 and 53). E1A is also a substrate for the acetylase activity of CBP, p300, and P/CAF in vitro (17,18,34). CBP-dependent acetylation of 12 S E1A shows similar substrate concentration saturation kinetics as purified H3 and H4 histones (not shown), consistent with the ability of E1A to competitively inhibit acetylation of purified histones in vitro (16 -18).
To determine the consequences of acetylation on the function of 12 S E1A, we mapped the site of acetylation by mutagenesis of three potential target lysine residues (Lys 162 , Lys 207 , Lys 239 ). Consistent with previous results obtained with p300 and P/CAF (34), we find that E1A is predominantly acetylated at a single lysine residue (Lys 239 ) in vitro at the C terminus (Fig.  1B). Recombinant wild type 12 E1A or E1A proteins with single amino acid substitutions at each of three lysine residues were incubated with baculovirus-expressed recombinant CBP in the presence of [ 14 C]acetyl-coenzyme A, and radiolabeled E1A was detected by autoradiography. Wild type E1A, E1A K162A, and E1A K207A are acetylated by CBP in vitro, whereas E1A K239A is not significantly acetylated by CBP (Fig. 1B).
Since mutation of Lys 239 could impair E1A acetylation by CBP indirectly by disturbing acetylation at another site, we developed a mouse monoclonal antibody specific for E1A acetylated at Lys 239 ([⑀-acetyl-K239]-E1A) to confirm that Lys 239 is a bona fide acetylation site both in vitro and in vivo. Mice were immunized with a peptide corresponding to the carboxyl-terminal 16 amino acids of Ad2 E1A synthesized with ⑀-acetyllysine at a position corresponding to Lys 239 (Fig. 1C). Clonal hybridoma lines were identified that were specifically reactive to in vitro acetylated full-length 12 S E1A. Fig. 1D shows the specificity of one of these antibodies (monoclonal antibody 327. antibodies (54) or the ␣-AcK-E1A, demonstrating that E1A is acetylated in vivo (Fig. 1E). This antibody fails to recognize other cellular proteins in untransfected cellular extracts (Fig.  1E) and also does not recognize purified in vitro acetylated proteins, including acetylated histones or acetylated CBP (not shown). Because of phosphorylation, 12 S E1A from cellular extracts migrates as a doublet by conventional SDS-PAGE, with the hypophosphorylated protein running at a higher mobility (55)(56)(57). Whereas CBP acetylates E1A in vitro, binding of E1A to CBP is not necessary for acetylation, since a deletion mutant that lacks N-terminal CBP/p300 binding determinants (⌬2-36) is also acetylated in vivo (Fig. 1F). Mutation of the -PLDLS-CtBP interaction motif alters recognition by both the ␣-AcK-E1A antibody and the M73 monoclonal antibody but not antibody M37, M1, or M29 ( Fig. 1F and data not shown). Furthermore, both wild type 13 S E1A and 13 S forms containing deletion mutations within each of the conserved regions (CR1-3) are acetylated in vivo in transfected cells at a level essentially equivalently to 12 S E1A (Fig. 1F). Whereas these results do not exclude the possibility of acetylation of Lys 162 and/or Lys 207 in vivo, these experiments support the identification of Lys 239 as the major site of E1A acetylation by CBP as has been previously demonstrated for P/CAF and p300 (34).
Acetylation at Lys 239 Does Not Disrupt the Association of E1A with CtBP-CtBP interacts with E1A and a variety of cellular transcriptional repressors through variations of a motif related to -PXDLS-. Whereas the acetylation site is adjacent to the consensus binding motif for CtBP (in boldface type; -GNPLDLSCK 239 RPRP), Lys 239 is within a previously identified pentapeptide nuclear localization signal (37) (underlined) conserved in all E1A isoforms (38,40). Although the proximity of Lys 239 to the CtBP binding motif and conservation of this sequence (Fig. 1B) suggests a probable conserved function, previous binding studies (35,58) do not support a significant role for this lysine in directly mediating the interaction between E1A and CtBP.
These contradictory reports prompted us to investigate whether mutations at Lys 239 in the context of full-length 12 S E1A altered the interaction with full-length CtBP in a direct binding assay (Fig. 2). Wild type 12 S E1A interacts with GST-CtBP with high affinity. Consistent with previous studies (35), we find that substitution mutations of K239 (K239A, K239R, and K239Q) in the context of full-length 12 S E1A do not substantially impair binding, whereas a double substitution mutation within the -PLDLS-motif (D235A/L236S) completely abolishes CtBP binding (Fig. 2). We next asked whether bona fide acetylation of E1A disturbed its direct interaction with CtBP. First, we acetylated E1A with recombinant CBP and [ 14 C]acetyl coenzyme A and asked whether GST-CtBP could bind radiolabeled E1A. Under these conditions, [acetyl-14 C]E1A binds to GST-CtBP, as detected by autoradiography of the captured E1A (Fig. 3A). Western analysis of in vitro acetylated E1A bound to GST-CtBP using the ␣-AcK239-E1A monoclonal antibody confirms that E1A acetylated at Lys 239 binds CtBP to an extent similar to unmodified E1A (Fig. 3B). In these experiments, ϳ20 -30% of E1A was acetylated by CBP in vitro, as determined by two-dimensional gel electrophoresis and Western blotting for total E1A (data not shown). While these results do not preclude the possibility that acetylation of E1A results in a modest reduction in the affinity for CtBP, acetylation does not substantially impair the interaction of intact E1A with CtBP.
C-terminal Mutants of E1A and Repression of CBP-dependent Transactivation-We and others have demonstrated that 12 S E1A represses CBP-and p300-dependent co-activation of expression of a cAMP-dependent protein kinase activated transcription (8,9). Since the CtBP interaction motifs found in a number of cellular repressors have been shown in part to be necessary for the activities of these proteins as transcriptional repressors (36), Zhang et al. (34) suggested that CtBP recruitment mediates some of the repressive activity of E1A on CBP/p300-dependent coactivation. In their model, acetylation or mutation of Lys 239 terminates this interaction to attenuate transcriptional repression. This model also predicts that mutation of residues within the -PLDLS-motif known to be necessary for the recruitment of CtBP will reduce the activity of E1A as a repressor of CBP/p300dependent co-activation. Previous studies, however, have indicated that deletion of the carboxyl-terminal portion of E1A (37) or substitution mutations within the terminal 5 residues (39, 40) disrupts nuclear partitioning of E1A.
To test the function of carboxyl-terminal E1A mutants in the repression of CBP-dependent co-activation, we employed a transfection assay devised to measure the co-activator activity of CBP in mediating the activation of the cAMP-dependent protein kinase activation of a CREB-dependent promoter (50). As previously demonstrated (8), expression of wild type 12 S E1A represses CBP-dependent co-activation, whereas an amino-terminal deletion mutant of E1A that fails to interact with CBP/p300 (⌬2-36) is less effective as a repressor (Fig. 4A). All E1A mutants with amino acid substitutions at Lys 239 were ineffective as repressors of CBP co-activation. Compared with wild type 12 S E1A, E1A K239R had an intermediate activity as a repressor (Fig. 4A), compared with the relative inactivity of K239A and K239Q substitutions. The reason for this partial activity as a repressor may be due to partial restoration of nuclear localization by basic amino acid substitution; however, this conclusion is not supported by immunocytochemistry experiments (see below). Finally, E1A D235A/L236S, a double amino acid substitution mutant within the -PLDLS-motif that completely disrupts CtBP binding both in vivo and in vitro (21) (see also Fig. 2), is a potent repressor of CBP-dependent coactivation (Fig. 4A). Mutant proteins were expressed at equivalent levels in whole cellular extracts of transfected cells by immunoblot analysis (data not shown).
By immunofluorescence, we find wild type E1A shows predominantly nuclear localization as has been previously described (Fig. 4B). Lys 239 mutants lack wild type nuclear localization, however, consistent with a previous report indicating that Lys 239 is essential for function of the NLS (39). Furthermore, the well characterized CtBP binding mutant, E1A D235A/L236S, also a potent repressor of CBP-dependent transcriptional activation (Fig. 4A), resides principally in the nucleus (Fig. 4B), indicating that neither CtBP binding nor CBP/ p300 binding (⌬2-36; Fig. 4B) is necessary for nuclear retention. We obtained identical results in transfected U2OS, HeLa, and NIH 3T3 cells (not shown).
These results do not exclude an independent role for Lys 239 in the nuclear functions of E1A. Thus, we asked whether directing E1A to the nucleus by fusing the SV40 nuclear localization signal (-PKKKRKV) in frame to the C terminus of 12 S E1A K239A would rescue the defect in repression of CBPdependent co-activation. Redirection of K239A 12 S E1A to the nucleus restores its ability to repress CBP-dependent co-activation of a cAMP-response element reporter (Fig. 5, A and C). Interestingly, both the wild type E1A-NLS fusion and the K239A mutant seem to have even more potent repressor activity when corrected for the relatively lower levels of E1A protein expression in this experiment (Fig. 5B). The wild type 12 S  Fig. 4 with expression vectors for CREB, protein kinase A, and CBP. E1A-dependent repression of CBPaugmented transcriptional activity was measured following cotransfection of 1, 5, or 10 g of wild type 12 S E1A, K239A 12 S E1A, or SV40 T-antigen NLS fusions of wild type E1A and K239A E1A, as indicated. Transfections were performed in triplicate and analyzed for expression of CAT relative to the expression of RSV-luciferase as a control for transfection efficiency. B, Western blot analysis of 12 S E1A expression in transfected F9 cells. Total protein concentration loaded for each lane was normalized for luciferase activity. E1A was detected with pooled monoclonal antibodies against E1A. C, immunolocalization of wild-type 12 S E1A, K239A 12 S E1A, or SV40 T-antigen NLS fusions of wild type E1A and K239A E1A in transfected COS-7 cells. E1A (green) was localized as described for Fig. 4. D, acetylation of wild type 12 S E1A, K239A 12 S E1A, or SV40 T-antigen NLS fusions of wild type E1A and K239A E1A expressed in COS-7 cells, using ␣-AcK-E1A.
We conclude that the ability of E1A mutants to repress the co-activator function of CBP correlates more closely with the nuclear localization of these proteins rather than with the ability to recruit CtBP. These results indicate that CtBP binding to E1A is neither necessary nor sufficient for repression of CBP-dependent co-activator activity, but Lys 239 mutations interrupt the carboxyl terminus dependent nuclear localization function of E1A necessary for transcriptional repression.
Acetylated E1A Has Altered Nuclear Localization-Based on the stringency of the requirement for a lysine at position 239, we considered the possibility that acetylation might influence the function of the E1A carboxyl-terminal NLS and determine subcellular localization of E1A. Previous studies demonstrated that whereas 13 S E1A co-purifies with nuclear matrix fractions and thus is relatively tightly associated with the nucleus, 12 S E1A is relatively soluble and is easily extracted from the nucleus under conventional fractionation methods (59). Despite this, immunocytochemical experiments demonstrate that both forms of E1A under steady-state conditions reside primarily in the nucleus.
By Western blotting, ␣-AcK-E1A recognizes E1A isoforms in extracts of the adenovirus type 5 E1-transformed HEK293 cell line (60) (Fig. 6A) and individually expressed 12 and 13 S isoforms in U2OS cells (Fig. 1F), indicating that both major forms are acetylated in vivo. In contrast to E1A detected by antibodies that recognize all forms, acetylated E1A is relatively depleted from nuclear extracts of 293 cells (Fig. 6A).
Nevertheless, by immunocytochemistry, the E1A monoclonal antibody (M73) recognizes primarily nuclear E1A, as previously demonstrated (Fig. 7). In contrast, the ␣-AcK-E1A monoclonal antibody demonstrates diffuse cytoplasmic and nuclear staining. HeLa cells, COS-7 cells, and U2OS cells failed to show significant staining as a negative control (data not shown).
Previous studies have demonstrated that 13 S E1A contains a second independent signal for nuclear targeting or retention, probably in the CR3 region, which is not present in the 12 S form (61), although other studies indicate that this signal may not function in all mammalian cells (62). To determine whether the nuclear staining observed for 293 cells using the ␣-AcK-E1A antibody might arise from nuclear localization of acetylated 13 S E1A, we transfected COS-7 cells with wild type 13 S E1A and 13 S with a deletion of conserved region 3 (mCR3) and a mutation of the carboxyl-terminal acetylation site (K285Q) (Fig. 8). Both wild type and deletion mutants of 13 S E1A are acetylated at Lys 285 in vivo essentially equivalently to the 12 S isoform (Fig. 1F). Under these conditions, 13 S E1A is primarily nuclear by immunolocalization with ␣-E1A (Fig. 8). Cells transfected with pRc/RSV-K285Q 13 S E1A show both nuclear and cytoplasmic localization, indicating that CR3 can function in a dominant and independent manner in directing nuclear partitioning of 13 S E1A. The deletion mutant of CR3 (13 S E1A mCR3) shows nuclear partitioning that depends on the integrity of Lys 285 . Mutation of both CR3 and Lys 285 (mCR3 K285Q) abolished nuclear localization (Fig. 8). These results suggest that acetylation of Lys 239 within the carboxyl-terminal nuclear localization signal differentially localizes the 12 S E1A isoform to the cytoplasm. The sequence of CR3 lacks obvious homology to a NLS and appears to only weakly allow partitioning and possibly in a cell type-specific manner (61,62).
Acetylation of Lys 239 Disrupts Recognition of 12 S E1A by Importin-␣-Acetylation could alter the localization of E1A by promoting nuclear export or inhibiting nuclear import. E1A does not apparently undergo active CRM1-dependent nuclear export, since treatment of 293 cells with leptomycin B does not increase nuclear concentration of acetylated E1A (data not shown). NLS-containing protein recognition in the cytoplasm is mediated by the heterodimeric importin-␣/␤ receptor, in which the importin-␣ subunit functions as the energy-independent, saturable component of the receptor, with the ␤ subunit targeting the NLS-containing cargo to the nuclear pore complex (63,64). Recent experiments using E1A carboxyl-terminal peptides fused to bovine serum albumin suggest that E1A entry to the nucleus might be mediated by importin-␣ isoforms (␣1, ␣3, ␣5, ␣7), showing some preference for importin-␣3-dependent transport based on competition by nucleoplasmin (65). Thus, we asked whether E1A directly bound importin-␣3, a ubiquitous member of the importin-␣ family, and whether acetylation of E1A disrupted this interaction in vitro. As shown in Fig. 9A, 12 S E1A interacts strongly with GST-importin-␣3, saturating immobilized importin-␣3 at an E1A concentration between 10 and 50 nM, consistent with the affinity (K d of ϳ10 Ϫ8 M) of bona fide NLS sequences for importin-␣ or -␣/␤ heterodimers (66), whereas substitution mutations at Lys 239 substantially diminish this interaction. Furthermore, 12 S E1A acetylated by re-combinant CBP in vitro binds poorly to GST-importin-␣3 (Fig.  9B), remaining below saturation binding at 100 nM. These results demonstrate that Lys 239 is essential for recognition of E1A by the nuclear import receptor importin-␣, providing a molecular basis for the stringent requirement of lysine at this position for the function of this nuclear localization signal. DISCUSSION The effects of E1A on cellular proliferation, transformation, and apoptosis have been largely attributed to interactions with cellular nuclear factors including CBP/p300, the retinoblastoma tumor suppressor, pRB, and more recently the transcriptional co-repressor CtBP. In this report, we demonstrate that 12 S E1A is acetylated within the functionally dense carboxylterminal domain that includes both a nuclear localization signal and an interaction domain for CtBP. Rather than interfering with CtBP recruitment, we find that acetylation determines the subcellular location of E1A by interfering with the recognition of the nuclear localization signal by the importin-␣ family of nuclear import receptors raising the possibility that acetylation may direct E1A to cytoplasmic targets.
What Is the E1A Acetylase?-Although the experiments presented in this report indicate that Lys 239 /Lys 285 is a major site of acetylation of the 12 and 13 S E1A isoforms in vivo and that CBP can acetylate E1A at this lysine in vitro, these studies do not exclude the possibility that another acetyltransferase modifies E1A in vivo. By aligning acetylation sites of known nonhistone substrates of p300, Thompson et al. (67) speculate that charged residues at Ϫ3 and ϩ4 relative to the substrate lysine may identify physiologically relevant protein substrates; however, the E1A site (-PLDLSCKRPRP 243 ) lacks this characteristic. Furthermore, the E1A sequence only marginally resembles the substrate recognition sequence (-GKXP-) suggested by the structure of the tetrahymena GCN5-peptide substrate complex (68). The suggestion that one of the nuclear acetylases modifies E1A is plausible. E1A associates with CBP, p300, and P/CAF, and effectively competes for histone acetylation in vitro. Nevertheless, the ⌬2-36 mutation that prevents association of E1A with the C/H3 region of CBP does not alter E1A acetylation ( Fig. 1F) or nuclear localization (Fig. 4B). Thus, binding to CBP is a prerequisite for neither acetylation nor nuclear retention of E1A. Thus, whereas CBP, p300, and P/CAF may acetylate E1A in vitro, these may not be the preferred enzymes in vivo. Definitive identification of the E1A acetylase awaits quantitative kinetic comparison of various HAT enzymes using E1A as a substrate, including the possibility that E1A may be the target of one of the cytoplasmic HATs.
Implications for the Mechanism of Repression of CBP by E1A-How does 12 S E1A repress CBP-or p300-dependent co-activation, and does CtBP participate in E1A-dependent transcriptional repression? In these experiments, we demonstrate that the interaction of CtBP with E1A is dispensable for the repression of CBP-mediated co-activation of the CRE-dependent somatostatin promoter (Fig. 4A). Thus, whereas CBP, E1A, and CtBP may form a ternary complex in vitro (not shown), this complex has no apparent role in regulation of CBP-dependent co-activation under these conditions. This conclusion conflicts with a previous report indicating that CtBP is necessary for a component of E1A repression of CBP-dependent co-activation (34). These investigators based this conclusion on differences between the extent of repression seen at different amounts of transfected expression vector for wild type and mutant E1A, implying that K239A was less able to recruit CtBP at low concentrations, whereas at higher levels of expression other mechanisms of E1A-dependent repression were operative. This model of repression is at odds with the biological and biochemical behavior of amino-terminal E1A mutants. E1A may have general effects on transcription by interacting with components of TFIID (69,70), which may explain the observed nonspecific effects of the high levels of expression achieved in transient expression experiments. We employed three experimental conditions that allowed us to ensure the specificity of this assay for CBP-dependent repression and demonstrate that recruitment of CtBP is not necessary for repression by E1A. First, in our experiments we have used the amino-terminal ⌬2-36 E1A mutant that fails to interact with CBP and p300 (14) and the well characterized -PLDLS-mutation D235A/L236S that completely abrogates CtBP binding (22,33). Finally, whereas levels of expression of E1A mutants are comparable in whole cell extracts of transfected cells in the functional assay shown in Fig. 4A (data not shown), they differ in subcellular localization, a possibility not considered in this previous study (34). It seems probable that the very high levels of expression that may be achieved by transient plasmid-based expression may nonspecifically repress transcription even with nuclear localization-defective substitution mutants at Lys 239 .
Some studies have demonstrated that E1A represses CBP-or p300-dependent acetylase activity toward histones in vitro (16 -18), whereas others demonstrate stimulation of HAT activity (19). Still others have demonstrated that cellular transcription factor interactions through the C/H3 domain may allosterically regulate CBP and p300 HAT activity (71). The in vitro inhibition of CBP HAT activity depends on the carboxylterminal portion of E1A and micromolar concentrations of protein. Since the K m of full-length p300 for histone H4 is on the order of 1-3 M (67), this observation is most consistent with E1A competing for the active site of the enzyme as a substrate. Finally, repression of reconstituted p300-dependent transcription in vitro by E1A depends only on amino-terminal, exon 1-encoded determinants (72), consistent with the observed role of the amino terminus in repression of CBP-dependent functions under physiological conditions.
What Is the Role of Acetylation in the Function of E1A?-A previous report indicated that acetylation of E1A might diminish binding to the CtBP co-repressor (34) in contrast to this study. The different E1A proteins used for in vitro binding assays probably explain the discrepancy between this previous report and the results reported here. Zhang et al. (34) utilized an E1A model peptide encompassing the carboxyl-terminal 14 amino acids of E1A in competition binding experiments, whereas we utilized full-length protein in direct binding experiments. Our experiments and those of others (35) have indicated that mutation of Lys 239 has little effect on this interaction. Furthermore, the affinity of full-length E1A for CtBP is more than 2 orders of magnitude greater than that of an isolated peptide containing this sequence (35,73). These results suggest that structural determinants within the intact E1A protein may contribute to either proper folding or stability of a biologically relevant CtBP interaction domain. Nevertheless, our results do not exclude the possibility that in some contexts, acetylation of a lysine with this proximity to a -PXDLS-motif may quantitatively and functionally alter recruitment of CtBP. Comparison of the affinities of CtBP for peptides derived from a number of CtBP target proteins indicate that sequences outside the core -PXDLS-motif may influence binding (58). In fact, a recent report indicates that acetylation of the nuclear receptor co-repressor RIP140 may terminate its interaction with the CtBP co-repressor (74). In this case, acetylation of RIP140 at a site adjacent to an interior PXDLS motif (-440 PIDLSCK 446 -) seems to exclude CtBP from the RIP140 complex in vivo. In this central position, acetylation might have more significant effects on the structure of the CtBP interaction domain than that observed on the binding motif located near the carboxyl terminus of E1A. In the simplest model, E1A acetylation might terminate the interactions of E1A with nuclear targets like CBP and p300 and thus the transcriptional effects by changing its subcellular localization. Treatment of 293 cells with leptomycin B does not increase nuclear concentration of acetylated E1A (data not shown), suggesting that E1A exits the nucleus via a CRM-1 independent mechanism. However, because of its small size, acetylated E1A could theoretically passively diffuse from the nucleus (75) but clearly requires the NLS for nuclear entry. Alternatively, acetylation of E1A cotranslationally or immediately posttranslationally in the cytoplasm might inhibit recognition by importin-␣. Acetylation as a regulator of transcription factor nuclear localization has precedence (76), although previous studies have not demonstrated that the acetylation of a nuclear localization signal may directly modify its importin interaction. The sequence of the E1A NLS resembles a monopartite NLS, similar to the consensus motif -K(K/R)X(K/R)-(residues designated P 2 to P 5 ) (77), where X is any amino acid (78). Recognition of the NLS depends on complex interactions between the NLS and importin-␣, with affinity mediated by hydrophobic and hydrogen bonding between the side chains and backbone of the NLS and importin-␣, whereas binding specificity is dictated by complementary electrostatic interactions between importin-␣ and the basic side chains of the NLS (64,77,79). The binding pocket for the charged residues (P 2 , P 3 , and P 5 ) retains flexibility to accommodate both lysine and arginine residues in the NLS. Nevertheless, the E1A NLS is somewhat atypical in that it contains only a single lysine residue critical for its function (Lys 239 ), corresponding to the conserved P 2 position. Thus, the interaction of 12 S E1A with importin-␣ may be particularly sensitive to inhibition by acetylation. Because of the importance of a conserved lysine at P 2 in a variety of NLSs, this may represent a general mode of regulation of nuclear import.
The steady-state concentration of acetylated E1A relative to total cellular E1A in E1-transformed 293 cells is low. Previous estimates based on immunocytochemistry suggest that as much as 5-10% of E1A is cytoplasmic in both adenovirusinfected cells and E1-transformed 293 cells (40, 59), a level consistent with our estimates of the fraction of acetylated E1A. Interestingly, inhibition of protein deacetylases by treating 293 cells with sodium butyrate increased neither the steady-state levels of acetylated E1A nor the extent of cytoplasmic partitioning (not shown), suggesting that E1A is either a poor substrate for deacetylases or perhaps that acetylated E1A undergoes rapid degradation following acetylation. In favor of the latter possibility, previous studies indicate that E1A undergoes relatively rapid turnover in both adenovirus-infected and -transformed cells (80) and that nuclear E1A may undergo translocation to the cytoplasm prior to degradation (40). Recent reports indicate that E1A interacts with components of the 26 S proteasome both in the nucleus and cytoplasm and itself undergoes proteasome-mediated degradation (81,82). Thus, acetylation may be one of several signals that target E1A for degradation.
A more intriguing possibility is that acetylated E1A has cytoplasmic functions distinct from its capacity as a transcriptional regulatory protein. As discussed above, 12 S E1A alone stimulates DNA synthesis and immortalization, but complete transformation requires co-expression of another oncogene such as ras or E1B (2). Interestingly, these functions of E1A seem to differentially require nuclear localization. Immortalization and induction of DNA synthesis requires an intact nuclear localization signal. The second exon of E1A, including the nuclear localization signal, while necessary for the immortalization functions and cotransformation with myc or E1B, is dispensable for ras-dependent cotransformation of primary baby rat kidney cells (38,39,(83)(84)(85). A variety of cytosolic targets for E1A have been identified, including Rack1 (86), a membraneassociated scaffolding protein implicated in cytoskeletal dynamics and growth factor signaling (87,88), Yak-related kinases (89), and the RII␣ subunit of protein kinase A (90). Among these potential cytoplasmic targets is the CtBP isoform, BARS/CtBP3, implicated in Golgi function (91). The observation that the carboxyl-terminal nuclear localization signal differentially partitions the 12 and 13 S isoforms suggests that any cytoplasmic function may be specific to the 12 S isoform.