A Novel Nuclear Localization Signal in the Auxiliary Domain of Apobec-1 Complementation Factor Regulates Nucleocytoplasmic Import and Shuttling*

C to U editing of the nuclear apolipoprotein B (apoB) transcript is mediated by a core enzyme containing a catalytic deaminase, apobec-1, and an RNA binding subunit, apobec-1 complementation factor (ACF). ACF expression is predominantly nuclear, including mutant proteins with deletions of a putative nuclear localization signal. We have now identified a novel 41-residue motif (ANS) in the auxiliary domain of ACF that functions as an authentic nuclear localization signal. ANS-green fluorescence protein and ANS-β-galactosidase chimeras were both expressed exclusively in the nucleus, whereas wild-type chimeras or an ACF deletion mutant lacking the ANS were cytoplasmic. Nuclear accumulation of ACF is transcription-dependent, temperature-sensitive, and reversible, features reminiscent of a shuttling protein. ACF relocates to the cytoplasm after actinomycin D treatment, an effect blocked by the CRM1 inhibitor leptomycin B. Heterokaryon assays confirmed directly that ACF shuttles in vivo. ACF binds to the protein carrier, transportin 2 in vivo, and colocalizes to the nucleus as determined by confocal microscopy. Co-immunoprecipitation experiments revealed that transportin 2 binds directly to the ANS motif. These data suggest that directed nuclear localization and compartmentalization of the core complex of the apoB RNA editing enzyme is regulated through a dominant targeting sequence (ANS) contained within ACF.

C to U editing of the nuclear apolipoprotein B (apoB) transcript is mediated by a core enzyme containing a catalytic deaminase, apobec-1, and an RNA binding subunit, apobec-1 complementation factor (ACF). ACF expression is predominantly nuclear, including mutant proteins with deletions of a putative nuclear localization signal. We have now identified a novel 41-residue motif (ANS) in the auxiliary domain of ACF that functions as an authentic nuclear localization signal. ANSgreen fluorescence protein and ANS-␤-galactosidase chimeras were both expressed exclusively in the nucleus, whereas wild-type chimeras or an ACF deletion mutant lacking the ANS were cytoplasmic. Nuclear accumulation of ACF is transcription-dependent, temperature-sensitive, and reversible, features reminiscent of a shuttling protein. ACF relocates to the cytoplasm after actinomycin D treatment, an effect blocked by the CRM1 inhibitor leptomycin B. Heterokaryon assays confirmed directly that ACF shuttles in vivo. ACF binds to the protein carrier, transportin 2 in vivo, and colocalizes to the nucleus as determined by confocal microscopy. Coimmunoprecipitation experiments revealed that transportin 2 binds directly to the ANS motif. These data suggest that directed nuclear localization and compartmentalization of the core complex of the apoB RNA editing enzyme is regulated through a dominant targeting sequence (ANS) contained within ACF.
Apolipoprotein B (apoB) 1 mRNA is the target of a site-specific deamination reaction that introduces a C to U change in the edited transcript (1). Two proteins are required for C to U deamination of apoB mRNA. These include apobec-1, an RNAspecific cytidine deaminase that represents the catalytic subunit of the core enzyme complex and which binds to a competence factor, ACF, that likely functions as the RNA binding subunit (2,3). Physical interaction of these two components is required in order for this minimal complex to mediate C to U RNA editing of a synthetic apoB transcript in vitro (4,5). Physiological activity of these two core components and in vivo apoB mRNA editing efficiency may be modified through higher order interactions with other proteins that represent elements of a larger, holoenzyme complex (6 -8).
The tissue-specific expression pattern and subcellular distribution of both ACF and apobec-1 has been explored in some detail (2,3,9,10). Independent studies using epitope-tagged protein expression in vitro have established that ACF localizes almost exclusively to the nucleus (4,7). These findings were corroborated in recent studies, demonstrating a marked increase in endogenous, nuclear ACF immunostaining and protein content accompanying metabolic induction of apoB mRNA editing activity after insulin or thyroid hormone administration in vivo in rodent hepatocytes (11,12). On the other hand, apobec-1 is found in a predominantly cytoplasmic or perinuclear distribution in transfected cells (4,7,8,13). In examining the cytoplasmic distribution of apobec-1, recent studies concluded that neither the putative bipartite nuclear localization sequence (NLS) nor the predicted nuclear export signal in the carboxyl terminus of apobec-1 was functional in directing import or export of a reporter (14). Thus, the mechanism by which apobec-1 is transported to the nucleus and the signals by which this process is directed remain unknown. Because C to U editing of the nuclear apoB transcript requires a physical interaction between ACF and apobec-1 (4), these two proteins must presumably exist in the same subcellular compartment (i.e. the nucleus) to effect deamination of the targeted base in apoB mRNA (13). This assumption is supported by the demonstration that co-expression of apobec-1 with ACF results in the redistribution of apobec-1 to the nucleus (4, 7). By contrast, co-expression of an ACF mutant that lacks the apobec-1 interaction domain along with wild-type apobec-1 reveals that apobec-1 remains cytoplasmic, whereas the ACF mutant localizes to the nucleus (4). These observations raise the possibility that, in addition to its requisite function in the enzymatic catalysis of C to U deamination, ACF may also regulate apoB mRNA editing by controlling nuclear accumulation of apobec-1 through proteinprotein interactions and directed intracellular trafficking.
A putative, SV-40 type NLS motif (PKTKKRE) is present in the amino-terminal portion of ACF, overlapping the second RNA recognition motif (RRM2). Point mutations or deletion of this motif, however, failed to alter nuclear distribution of the mutant ACF, suggesting that another active NLS domain directs nuclear import (4). In this study we have characterized a functional NLS located in the auxiliary domain of ACF. The function of this motif and the regulation of nuclear-cytoplasmic transport of ACF and the ACF-apobec-1 complex is the focus of this report.

Plasmid Constructs and Protein Expression-
The construction of the ACF deletion mutants has been previously described (4). The ⌬ANS-ACF deletion mutant was cloned into the eukaryotic expression vector pCMV2B using the restriction sites BamHI and XhoI. All the green fluorescence protein (GFP) fusion proteins were constructed following a two-step PCR method (15). The full-length PCR products were sequenced and subcloned into the pEGFP-N1 vector (Clontech) using KpnI-BamHI sites. This cloning generates an NH 2 -terminal GFP fusion protein. The ANS-␤-galactosidase fusion was constructed following a two-step PCR strategy using pTYB1-ACF and pSV-␤-galactosidase vector (Promega) as template. The full-length PCR product was inserted into pCMV2B vector (Stratagene) using the BamHI-XhoI restriction sites. Importin-␣, transportin 1, and transportin 2 (Trn2) were cloned following PCR amplification and inserted into pCMV2B or -3B vectors (Stratagene). This cloning resulted, respectively, in the expression of an NH 2 -terminal FLAG or Myc-tagged fusion protein. Trn2 recombinant protein was expressed as a GST fusion (a generous gift of I. Gallouzi, McGill University) and purified to homogeneity using glutathione-Sepharose 4B affinity chromatography as recommended by the manufacturer (Amersham Biosciences). After extensive washes, the recombinant protein was eluted in the presence of 50 mM Tris-HCl, pH 8.0, 10 mM reduced glutathione, 120 mM NaCl, 0.1% Triton X-100, and protease inhibitors and dialyzed against Dignam D buffer (7).
Protein-Protein Interaction Studies-COS-7 cells were transiently transfected with 2 g of plasmid encoding FLAG-tagged wild-type ACF, ⌬ANS-ACF, or HuR (a generous gift of I. Gallouzi, McGill University). 48 h post-transfection, cell lysates were prepared in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, and protease inhibitor mix. Each extract was treated with RNase A for 30 min at room temperature and loaded on GST or GST-Trn2-coupled glutathione-Sepharose beads for 12 h at 4°C. The unbound material was washed with 20 mM HEPES, pH 8.0, 100 mM KCl. Bound proteins were eluted with 2ϫ Laemmli loading buffer, fractionated through SDS-10% PAGE, and analyzed by Western blotting using rabbit anti-FLAG antibody (Affinity BioReagents, Inc.). In vitro translation of epitopetagged constructs encoding importin-␣, transportin 1, transportin 2, FLAG-tagged ACF, ⌬ANS-ACF, or FLAG-ANS was conducted as described previously using TNT coupled reticulocyte lysate (Promega). 35 S-radiolabeled Trn2 protein was mixed with unlabeled FLAG-tagged ACF, ⌬ANS-ACF, or ANS, and the complex was recovered by immobilization on immune-coupled protein A beads (4,7).
Transfection, Immunofluorescence, and Confocal Microscopy-COS-7 cells, HeLa cells, and NIH3T3 (ATCC) were maintained in Dulbecco's modified Eagle's medium as described previously (7). Where indicated, cells were incubated with actinomycin D (AMD) (Sigma) at 10 g/ml for 3 h or as indicated in the relevant figure legend. In some experiments, cells were treated with leptomycin B (LMB) at 10 ng/ml (a generous gift from Dr. Minoru Yoshida, RIKEN Institute, Japan) (16, 17) for 6 h before fixation and staining. Cycloheximide (Sigma) was used where indicated for 3 h at 20 g/ml to inhibit protein synthesis. COS-7 cells were grown to 50 -60% confluency on glass coverslips. The cells were transiently transfected with 2 g of appropriate eukaryotic expression vector using 6 l of FuGENE 6 (Roche Molecular Biochemicals). 48 h after transfection, cells were fixed with 10% formalin solution (Sigma), permeabilized with 0.5% Triton X-100, and probed with rabbit anti-FLAG IgG and/or mouse monoclonal anti-Myc IgG followed by Cy3 secondary antibody or fluorescein isothiocyanate-conjugated secondary IgG (Jackson ImmunoResearch). For standard immunofluorescence microscopy, nuclei were identified using 4,6-diamidino-2-phenylindole (DAPI, Vector). Slides were examined using a Zeiss Axiashop 2 MOT microscope equipped with a 40ϫ plan neofluar objective and a 3-CCR camera (DAGE-MTI, Inc.). A Zeiss Attoarc variable intensity lamp was used with filter sets designed for Cy3 and 4,6-diamidino-2-phenylindole. For confocal microscopy, nuclei were detected using TO-PRO3 iodide (Molecular Probes, Oregon). Preparations were visualized using a 63ϫ Zeiss plane apochromatic objective and a Bio-Rad MRC 1024 confocal adaptor. A krypton-argon laser was used with epifluorescence filter sets designed for Texas Red (Cy3), fluorescein isothiocyanate, and cyanine (Cy5). The confocal aperture was set at 1.8. Usually, 4 -8 images at planes separated by 0.5 m were collected. Images were processed using Adobe Photoshop 4.0 software.
Heterokaryon Assay-The interspecies human-mouse heterokaryons were generated as described previously (18) with minor modifications. Briefly, HeLa cells grown on 60-mm dishes were transfected with 3 g of plasmid expressing amino-terminal FLAG-tagged ACF using 25-l Polyfect (Qiagen). 24 h after transfection, HeLa cells were seeded onto coverslips in 35-mm dishes at 3 ϫ 10 5 cells/coverslip. After overnight incubation, NIH3T3 cells preincubated for 30 min in the presence of 100 g/ml cycloheximide were seeded onto the same coverslips at 3 ϫ 10 5 cells/coverslip. The co-cultures were grown for 3 h with 100 g/ml cycloheximide and then were fused in the presence of 50% polyethylene glycol 3350 (Sigma) for 3 min. Cells were washed with phosphatebuffered saline, cultured in the same medium containing 100 g/ml cycloheximide for 3 h, and fixed for immunostaining as described above. To visualize the HeLa-NIH3T3 heterokaryon, Hoechst dye 33258 (Sigma) was used at 1 g/ml during the incubation with the secondary antibody to permit discrimination between human and murine nuclei (18).

RESULTS
ACF Has a Non-conserved Nuclear Import Sequence in the COOH-terminal Auxiliary Domain-Previous studies demonstrated that deletion of the putative SV-40 type NLS in the amino-terminal domain of ACF (2, 3) (Fig. 1A) failed to alter its nuclear distribution (4). Thus, to identify the active NLS within ACF we constructed a series of deletion mutants spanning the entire protein (Fig. 1B). Each mutant was transiently expressed in COS-7 cells, and its intracellular distribution was examined by immunofluorescence microscopy. Most of these deletion mutants revealed a nuclear staining pattern indistinguishable from wild-type ACF (Fig. 1B). However, there were two exceptions, mutants ⌬-(331-385) and ⌬-(380 -402), which appeared cytoplasmic (Fig. 1B). These results suggested that the region spanning residues 331-402 may contain an NLS. Further deletion mapping (data not shown) indicated that the region spanning residues 360 -401 contains the minimal nuclear localization sequence. A mutant ACF construct with an internal deletion spanning residues 360 -401 showed a cytoplasmic, perinuclear localization pattern when expressed in COS-7 cells (data not shown). Sequence alignment of the region spanning residues 360 -401 revealed no clusters of basic residues suggestive of a canonical SV-40 type or bipartite NLS. Furthermore, this region revealed no homology with either the NLS consensus of heterogeneous nuclear ribonucleoproteins (hnRNP) A1 or hnRNP K (M9 and KNS, respectively) or with the HNS (HuR nuclear localization signal) in HuR (Fig. 1C). We will heretofore refer to this apparently novel motif (residues 360 -401 of ACF) as ANS for ACF nuclear localization sequence.
The Nuclear Localization Signal of ACF Is Functional-To confirm that the ANS domain is functional, we examined its capacity to restrict nuclear expression of GFP. Expression of wild-type GFP revealed diffuse fluorescence throughout the cell, including the cytoplasm and nucleus (Fig. 2a). When expressed as a (331-402)-GFP fusion protein, fluorescence was restricted to the nucleus (Fig. 2b). This observation suggests that the ANS domain is biologically active in nuclear targeting. We further investigated if the minimal ANS (i.e. residues 360 -401) could function to target a large, heterologous protein such as ␤-galactosidase to the nucleus. Accordingly, an epitopetagged ANS-␤-galactosidase chimera was expressed, and its subcellular localization was determined by examining both protein activity and immunohistochemical distribution. The wildtype construct demonstrated the expected perinuclear distribution, whereas the ANS-␤-galactosidase chimera was efficiently targeted to the nucleus (Fig. 3, A and B). These findings collectively suggest that the ANS functionally directs protein import into the nucleus, both in cis as well as in trans.
ACF Can Redistribute between the Nucleus and Cytoplasm-Sequence alignment and phylogenetic analysis suggest that ACF is distantly related to hnRNPs, some of which continuously shuttle between the nucleus and cytoplasm (19). These studies used the RNA polymerase II inhibitor AMD to demonstrate transcription dependence of hnRNP shuttling (20). To examine the transcription dependence of ACF distribution, COS-7 cells were treated with AMD for 3 h, which revealed a shift in distribution of ACF from the nucleus to the cytoplasm (Fig. 4, panel A). There was no effect on the nuclear localization of ACF when cells were treated with cycloheximide (20 g/ml) (Fig. 4B). These findings suggest that the cytoplasmic accumulation of ACF after AMD treatment was likely the result of transcription inhibition rather than defects in protein synthe-  (a and b). A pCMV empty vector was used as control of background ␤-galactosidase activity (e). B, determination of nuclear localization of ANS-␤-galactosidase by immunofluorescence staining. COS-7 cells were transfected with plasmids expressing either amino-terminal FLAG-tagged ␤-galactosidase (a and b) or FLAGtagged ANS-␤-galactosidase (c and d). Cells were fixed and immunostained using anti-FLAG IgG. Nuclei were identified by DAPI staining. This is representative of three independent assays. sis. To further determine whether the cytoplasmic accumulation of ACF observed after AMD treatment reflects increased nuclear export, we treated cells with LMB, a specific inhibitor of CRM1 (21,22), both alone and in the presence of AMD. These experiments revealed that LMB abolished the cytoplasmic redistribution of ACF previously observed with AMD (Fig. 4A). These findings, thus, suggest that ACF translocates across the nuclear envelope via a CRM1-dependent pathway. In view of the recent findings by Smith and co-workers (14) that apobec-1 does not accumulate in the nucleus of LMB-treated cells, the current results strongly imply that the nuclear export of ACF is at least partially independent of apobec-1. We will return to this point below.
Cytoplasmic redistribution of ACF was abrogated when AMD treatment was performed at 4°C (Fig. 4C), suggesting that nuclear export of ACF is an active, energy-requiring process. Furthermore, the cytoplasmic accumulation of ACF was reversible upon restoring transcription, as revealed in studies where AMD-treated cells were incubated in an AMD-free medium at 37°C for 3 h (Fig. 4C). In summary, these data indicate that nuclear accumulation of ACF is a transcription and energydependent process.
ACF Shuttles between the Nucleus and Cytoplasm-To formally demonstrate nuclear-cytoplasmic shuttling of ACF, we analyzed its migration between nuclei in a human-mouse heterokaryon assay (18,20). HeLa cells transfected with FLAG-ACF were fused to mouse NIH3T3 cells to produce heterokaryons. Three hours after fusion, ACF was localized by immunofluorescence microscopy within murine nuclei of a representative heterokaryon (Fig. 5, a-c), establishing shuttling as a mechanism to account for ACF migration from the human to the mouse nucleus. Human nuclei were distinguished from the mouse nuclei using Hoechst dye (18). These data demonstrate that ACF, like HuR (Fig. 5, d-f), shuttles between the nuclear and cytoplasmic compartments.
Nuclear Transport of ACF Involves Interaction with a Member of the Transportin Family-Distinct shuttling/NLS/NES motifs are recognized by specific protein carriers. Proteins containing a classical NLS motif are bound by the importin ␣/␤ complex and directed to the nuclear pore complex (23). Motifs like M9 or KNS in hnRNPs proteins or the HNS motif in HuR interact with transportin family members transportin 1 and transportin 2, respectively (24 -26). Our findings to this point suggest that ACF uses a novel NLS for nuclear import. Accordingly, we asked whether ACF could interact with the known carrier proteins, importin-␣, transportin 1, and transportin 2. We first performed immunoprecipitation assays in which recombinant ACF was mixed with a panel of [ 35 S]Met-labeled transport proteins. The results of co-immunoprecipitation assays revealed an interaction only with Trn2 (data not shown). To demonstrate a biologically relevant ACF-Trn2 interaction in whole-cell extracts we applied lysates from ACF-transfected COS-7 cells to GST-Trn2 glutathione beads and examined the retained material by Western blotting. This approach demonstrated ACF in the eluate from the GST-Trn2 beads (Fig. 6A,  lane 6). No ACF was detected in the eluted fraction from the control GST beads (Fig. 6A, lane 5). Importantly, an ACF mutant lacking the ANS motif did not bind Trn2 (Fig. 6A, lane

FIG. 4. Characterization of ACF nuclear localization activity.
A, nuclear localization activity of ACF depends on RNA polymerase II transcription. FLAG-tagged ACF was transiently expressed into COS-7 cells. Forty-eight hours post-transfection cells were incubated with either AMD at 10 g/ml (c-d), LMB at 10 ng/ml (e-f), or both actinomycin D and leptomycin B (g-h). The cellular distribution of ACF was then determined by immunofluorescence staining using anti-FLAG IgG. Nuclei were identified by DAPI staining. A control performed without AMD and LMB is shown (a and b). B, cycloheximide (CHX) alone has no effect on the nuclear localization of ACF. As described above, COS-7 cells transiently expressing FLAG-ACF were incubated in the absence (a) or presence (b) of cycloheximide at 20 g/ml for 3 h. Cells were fixed and immunostained with anti-FLAG IgG. DAPI staining identifies nuclei (c and d). C, cytoplasmic accumulation of ACF upon AMD treatment is temperature-dependent and reversible. b, COS-7 cells transiently expressing FLAG-ACF were incubated in the presence of AMD at 10 g/ml for 3 h at 37°C and then fixed and analyzed with anti-FLAG IgG. c, cells were treated following the conditions described for b except that incubation with AMD was performed at 4°C. d, cells were incubated with AMD for 3 h at 37°C. After removal of the growth medium, cells were washed and placed in fresh medium without AMD for another 2 h before fixation and immunostaining. Nuclei were identified by DAPI staining (e-h). A control performed at 37°C in the absence of AMD is shown (a). This is representative of three independent assays.
FIG. 5. ACF migrates between nuclei in an interspecies heterokaryon assay. HeLa cells were transiently transfected with FLAG-ACF, fused with NIH3T3 cells to form heterokaryons in the presence of 50% polyethylene glycol, and incubated for an additional 3 h in the presence of cycloheximide. The co-culture was then fixed and immunostained with anti-FLAG IgG. Mouse nuclei (indicated with arrows) were distinguished from human nuclei by Hoechst dye, which reveals mouse nuclei through their punctate pale blue staining compared with the homogenous staining revealed in human nuclei. FLAG-HuR shuttling activity was examined as a positive control.

7)
and as expected, accumulated in the cytoplasm of transfected cells (see Fig. 3B). By way of a positive control, the binding of HuR to Trn2 was confirmed in this assay (Fig. 6A, lane 8).
To further confirm the interaction between Trn2 and ACF, we transiently expressed epitope-tagged proteins in COS-7 cells and confirmed their nuclear colocalization (Fig. 6B). The merged images show an overlapping signal at the border of the nuclear envelope, suggesting the presence of ACF associated with Trn2, possibly at the nuclear pore complex. Clearly, however, further studies will be required to examine this possibility in detail. Nevertheless, these data allowed us to examine the hypothesis that transport of ACF requires a physical interaction between the ANS motif and the Trn2 carrier. Accordingly, FLAG-tagged fusion proteins were prepared using wild-type and ⌬-ACF mutants and a FLAG-ANS chimera (Fig. 7, A and B). Transportin 2 was expressed in vitro as a 35 S-radiolabeled protein and mixed with the indicated unlabeled protein in an immunoprecipitation pull-down experiment to examine the physical interaction between the ANS and Trn2 (Fig. 7C). The data reveal a strong interaction with both wild-type ACF and with the FLAG-ANS recombinants but not with the ⌬-ANS mutant of ACF (Fig. 7C). We interpret these findings as evidence of a physical interaction between Trn2 and the ANS domain of ACF.

DISCUSSION
In this study we have characterized a novel 41-residue motif in ACF that functions as an authentic NLS in directing ACF as well as heterologous proteins to the nucleus. In addition, we demonstrate that the transport of ACF into and out of the nucleus is an energy-dependent process that requires active transcription. We demonstrate that ACF shuttles between the nucleus and cytoplasm and that this shuttling likely involves at least two candidate transport proteins, transportin 2 and CRM1.
ApoB mRNA editing is almost certainly a nuclear event. Evidence from earlier in vivo studies using rat liver (27) as well as more recent approaches using reconstituted cell-based assays with splicing-competent or -defective reporter constructs have demonstrated that C to U editing occurs on a spliced but intranuclear transcript (28). These features are consistent with the nuclear staining reported for epitope-tagged ACF (4,7) and the finding that induction of apoB mRNA editing activity in vivo is associated with increased nuclear abundance of ACF (12). The distribution of endogenous apobec-1, on the other hand, remains unknown since its expression is below the detection limits of currently available reagents (13). This said, an unexplained paradox is that the distribution of apobec-1, as inferred from expression studies using epitope-tagged protein, has been reported to be predominantly cytoplasmic (4,7,13). It remains possible that the cytoplasmic distribution of apobec-1 after transient transfection may reflect saturation of the nuclear import machinery in COS cells, specifically ACF, and we will return to this possibility below.
The current report identifies a novel motif (ANS) that directs nuclear localization of ACF itself as well as heterologous proteins, including GFP and ␤-galactosidase. This motif has no homology to other known NLSs, including the hnRNP-type elements found in known shuttling proteins and the NLS found in the third double-stranded RNA binding domain of the RNAediting enzyme ADAR1 (29). Among the potentially informative features examined within this motif, the ANS contains a cluster of glycine residues, reminiscent of those in the M9 motif of hnRNP A1 (Fig. 1C). The importance of these residues emerged from studies in which mutation of a specific glycine (Gly-274) in the M9 motif disrupted the nuclear localization of hnRNP A1 (19). Accordingly, we mutated each glycine to an alanine residue and tested the localization of the ACF mutants in transiently transfected COS-7 cells. However, none of these mutations affected the nuclear distribution of ACF, 2 suggesting that these residues alone do not confer targeting specificity to the ANS. Comparison of the putative NLS motifs of ACF orthologs indicates that the ANS is not conserved between Drosophila (accession number AAL13706), Caenorhabditis elegans (accession number F58D5.1), and human ACF (accession number AF 209192), suggesting that the ANS may have been acquired after the divergence of C. elegans and mammalian lineages. By contrast, the amino-terminal, SV-40-type putative NLS is highly conserved throughout evolution (30,31). These findings suggest the ANS motif may have evolved in connection with other functions, including site-specific C to U deamination of apoB, which emerged relatively recently.
Several lines of evidence suggest that the ANS directs nuclear localization and shuttling of ACF independent of binding to apoB RNA. First, the current studies were undertaken in COS and HeLa cells, neither of which express endogenous apoB (7). In this regard, it is important to emphasize that a nuclear distribution of ACF has been previously reported in both human HepG2 cells and in rat McA7777 cells, suggesting that nuclear distribution is not a function of either endogenous apoB RNA expression nor of mRNA editing (7). Secondly, the location of the ANS is within the auxiliary domain of ACF, distant from the major functional domains required for its complementation activity in vitro. Conversely, mutants of ACF that disrupt RNA binding, such as ⌬RRM2 and ⌬-(257-331), appear nuclear even though they lack the apoB mRNA binding domain (4).
The current findings demonstrate that in addition to its nuclear localization, ACF continuously shuttles across the nuclear envelope. Upon AMD treatment, ACF redistributes to the cytoplasm in a time-and temperature-dependent manner. Cotreatment with AMD and LMB restores the nuclear localization of ACF, suggesting that nuclear export occurs via CRM1, the canonical LMB-sensitive pathway (32,33). The transport of proteins across the nuclear envelope occurs through nuclear pores and involves several soluble carriers, particularly transportins and other factors recently identified as key regulators of mRNA export (23). CRM1 is a member of the importin-␤ family and a major RNA export receptor in eukaryotes (34). CRM1 is responsible for the export of shuttling proteins containing a leucine-rich NES, with the most fully characterized being within human immunodeficiency virus 1, Rev (35). No such leucine-rich region or NES-like motif could be identified in the primary sequence of ACF, and thus, the domain responsible for CRM1 interaction remains to be identified. We considered the possibility that apobec-1 contains the interacting leucinerich NES, although previous work by Yang and Smith (14) demonstrates that apobec-1 failed to accumulate in the nucleus after LMB treatment. Although this possibility will require formal evaluation, our preliminary observations indicate that the nuclear accumulation of ACF observed upon AMD and LMB co-treatment is diminished when apobec-1 is co-expressed; both proteins demonstrated a predominantly cytoplasmic colocalization pattern. 2 We interpret these preliminary observations to suggest that the interaction of apobec-1 with ACF renders the CRM1-interacting domain inaccessible. Under these conditions the cytoplasmic distribution of apobec-1 and ACF implies the existence of an alternative, CRM1-independent pathway for nuclear export, a possibility that will require more complete evaluation in future studies.
A further question raised by our findings concerns the possibility that ACF and apobec-1 are coordinately directed for nuclear import. Bidirectional transport functions, for example, have been ascribed to shuttling motifs within hnRNP A1 and hnRNP K (19,25). The current findings suggest that the ANS motif is responsible for the nuclear targeting of ACF but leave open the possibility that this motif directs import of the ACFapobec-1 core enzyme complex. Clearly, the relationship between the shuttling of free ACF versus ACF functionally contained within an apobec-1 core enzyme complex will require further study. However, our findings permit us to conclude that the shuttling of free ACF (Fig. 5) occurs through an LMBsensitive, CRM1-dependent pathway (Fig. 4).
The current findings also suggest that the ANS domain is responsible for an interaction with transportin 2 and further demonstrate that ACF colocalizes with Trn2 in the nucleus (Fig. 6). These findings raise the possibility that ACF may also exit the nucleus via a CRM1-independent mechanism. In this regard it bears emphasis that the interaction between the ANS of ACF and Trn2 in no way precludes the possibility that Trn2 is involved in the nuclear export of ACF. Precedent exists for diverse pathways of nuclear export of RNA-binding proteins, exemplified by findings with HuR (36). HuR interacts with two nuclear phosphoproteins, pp32 and APRIL, that contain leucine-rich NES motifs and which shuttle via a CRM1-dependent pathway. In addition, HuR is exported to the cytoplasm via a motif referred to as HNS, that interacts with Trn2 and which may be involved in HuR nuclear import as well as its export (26). Further work will be required to resolve the importance of Trn2 in the nucleocytoplasmic distribution of ACF and its relationship to the compartmentalization of the ACF-apobec-1 complex.
Finally, the results of these studies bear directly on the physiological significance of directed nuclear import of apobec-1. Previous studies have established that forced transgenic overexpression of apobec-1 is associated with promiscuous C to U RNA editing, suggesting that mechanisms likely exist in vivo through which the expression or localization of this protein is constrained. These issues take on greater relevance in the context of findings that both apobec-1 and ACF are AU-rich RNA binding proteins with targets (at least for apobec-1) that extend beyond apoB (37). These and other questions concerning the nucleocytoplasmic distribution of both ACF and the ACFapobec-1 complex will be the focus of future reports.