Sequence Elements in Both Subunits of the DNA Fragmentation Factor Are Essential for Its Nuclear Transport*

DNA cleavage is a biochemical hallmark of apoptosis. In humans, apoptotic DNA cleavage is executed by DNA fragmentation factor (DFF) 40. In proliferating cells DFF40 is expressed in the presence of its chaperone and inhibitor DFF45, which results in the formation of the DFF complex. Here, we present a systematic analysis of the nuclear import of the DFF complex. Our in vitro experiments demonstrate that the importin α/β-heterodimer mediates the translocation of the DFF complex from the cytoplasm to the nucleus. Both DFF subunits interact directly with the importin α/β-heterodimer. However, importin α/β binds more tightly to the DFF complex compared with the individual subunits. Additionally, the isolated C-terminal regions of both DFF subunits together bind importin α/β more strongly than the individual C termini. Our results from in vivo studies reveal that the C-terminal regions of both DFF subunits harbor nuclear localization signals. Furthermore, nuclear import of the DFF complex requires the C-terminal regions of both subunits. In more detail, one basic cluster in the C-terminal region of each subunit, DFF40 (RLKRK) and DFF45 (KRAR), is essential for nuclear accumulation of the DFF complex. Based on these findings two alternative models for the interaction of importin α/β with the DFF complex are presented.

DNA cleavage is a biochemical hallmark of apoptosis. In humans, apoptotic DNA cleavage is executed by DNA fragmentation factor (DFF) 40. In proliferating cells DFF40 is expressed in the presence of its chaperone and inhibitor DFF45, which results in the formation of the DFF complex. Here, we present a systematic analysis of the nuclear import of the DFF complex. Our in vitro experiments demonstrate that the importin ␣/␤heterodimer mediates the translocation of the DFF complex from the cytoplasm to the nucleus. Both DFF subunits interact directly with the importin ␣/␤-heterodimer. However, importin ␣/␤ binds more tightly to the DFF complex compared with the individual subunits. Additionally, the isolated C-terminal regions of both DFF subunits together bind importin ␣/␤ more strongly than the individual C termini. Our results from in vivo studies reveal that the C-terminal regions of both DFF subunits harbor nuclear localization signals. Furthermore, nuclear import of the DFF complex requires the C-terminal regions of both subunits. In more detail, one basic cluster in the C-terminal region of each subunit, DFF40 (RLKRK) and DFF45 (KRAR), is essential for nuclear accumulation of the DFF complex. Based on these findings two alternative models for the interaction of importin ␣/␤ with the DFF complex are presented.
Apoptosis is an evolutionarily conserved process that plays an important role in development and tissue homeostasis (1). Apoptotic signaling is ultimately transduced by a family of cysteine proteases called caspases, which are the central executioners of apoptosis (2). One of the biochemical hallmarks of apoptosis is nucleosomal DNA fragmentation (3) mainly caused by the DNA fragmentation factor 40 (DFF40, 2 also known as caspase-activated DNase, CAD) (reviewed in Ref. 4). In proliferating cells DFF40 is expressed in the presence of DFF45 (inhibitor of CAD), which has a dual role as chaperone and as inhibitor of DFF40 (5,6). During apoptosis active DFF40 is released from the DFF complex due to cleavage of DFF45 executed by caspase-3 and caspase-7 (7,8). A structural feature of both DFF40 and DFF45 is the N-terminal cell death-inducing DFF45-like effector domain that mediates the interaction between DFF40 and DFF45 (9 -11). Recently, it was shown that the DFF complex is composed of two DFF40 and two DFF45 subunits (12). The tetrameric DFF complex is located in the nucleus (13,14), and sequence analysis revealed two putative nuclear localization signals (NLSs). According to this prediction, the C-terminal portion of murine DFF40 contains a monopartite NLS, and the C-terminal sequences of murine and human DFF45 exhibit a bipartite NLS (5,14). Although previous studies showed that both C-terminal regions are involved in nuclear translocation of the DFF complex (13,14), the nuclear transport mechanism has not been characterized so far.
The nuclear transport of proteins harboring a NLS in their amino acid sequence is mediated by soluble nuclear import receptors also known as importins or karyopherins. NLSs can be generally categorized in classic (cNLS) and non-classic (ncNLS) signals. Although ncNLSs are directly recognized by members of the importin ␤-family receptors, binding of importin ␤ to cNLS requires an importin ␣ adapter protein (15,16). In humans there are six importin ␣ variants, which can be assigned to three subclasses: NPI-1, Rch1, and Qip1 (17,18). All importin ␣ subtypes share a large NLS binding domain and a flexible N-terminal importin ␤-binding (IBB) domain (19,20). The NLS binding domain comprises ten tandem armadillo repeats (21) that form an array of binding pockets (22)(23)(24). Armadillo repeats 2-4 create a larger N-terminal major binding site, whereas armadillo repeats 7 and 8 form a smaller C-terminal minor binding site. cNLSs are generally characterized as short stretches enriched in positively charged amino acids (25) and can be classified as monopartite and bipartite cNLSs. Monopartite cNLSs contain one cluster of basic amino acids, as in the case of the SV40 large T antigen (26), whereas bipartite cNLSs like that of nucleoplasmin are composed of two basic regions connected by a 10-to 12-amino acid spacer (27,28). Monopartite cNLSs and the larger downstream basic cluster of bipartite cNLSs bind to the major binding site of importin ␣, whereas the smaller upstream basic cluster of bipartite cNLSs binds to the minor binding site (reviewed in Refs. 29 and 30). Importin ␤ is a superhelical molecule composed of 19 HEAT repeats (31). HEAT repeats 7-19 (cargo binding site I) associate with the IBB domain of importin ␣ (32), whereas HEAT repeats 2-11 comprise a second cargo binding site (33).
Association and dissociation of cargo-import receptor complexes relies on the asymmetric distribution of RanGTP with a high concentration in the nucleus and a very low concentration in the cytoplasm (34,35). This RanGTP gradient across the nuclear membrane is maintained by the chromatin-bound exchange factor RanGEF (RCC1) and the GTPase-activating protein (RanGAP), which is localized on the cytoplasmic side of the nuclear pore. Importins bind their cargoes in the absence of RanGTP in the cytoplasm, whereas dissociation of the cargoreceptor complex occurs upon RanGTP binding in the nucleus (36,37).
In this study, we have analyzed the mechanism underlying the nuclear transport of the DFF complex. Our results show that the DFF complex is imported into the nucleus via the classic importin ␣/␤-pathway. We determined that at least one basic stretch in the C-terminal region of each subunit, DFF40 (RLKRK) and DFF45 (KRAR), is required for nuclear accumulation of the DFF complex. Moreover, the basic patch KRAR in the C-terminal region of DFF45 was also identified as key element for nuclear transport of monomeric DFF45. Additional experiments strongly suggest that DFF45 exhibits a classic monopartite cNLS recognized by importin ␣/␤.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-HeLa cells were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ no. ACC57) and were cultured in modified Eagle's medium (Biochrom). HeLa P4 cells (38) were cultured in Dulbecco's modified Eagle's medium (Amersham Biosciences). Both media were supplemented with 10% (v/v) fetal bovine serum (Biochrom) and antibiotics. Cells were maintained in a humidified incubator with 5% CO 2 atmosphere at 37°C.
Transfection into HeLa P4 cells was performed with the Effectene TM transfection reagent (Qiagen) according to the manufacturer's instructions using 0.3 g of plasmid DNA for each construct. The cells were fixed 24 h after transfection with 3% paraformaldehyde in phosphate-buffered saline for 15 min. The nuclei were visualized with Hoechst33258 (Molecular Probes).
Expression Constructs-The coding regions of the respective genes were amplified from plasmid DNA using specific primer pairs with appropriate restriction sites (for details see supplemental "Experimental Procedures"). All constructs were verified by DNA sequencing.
Recombinant Protein Expression and Purification-Epitopetagged DFF complexes were generated as follows: DFF40 and DFF45 were coexpressed in Escherichia coli BL21(DE3). The cultures were grown at 37°C to an optical density of 0.9 at 600 nm. After shifting the temperature to 18°C bacterial protein expression was induced with 0.2 mM isopropyl-␤-D-thiogalactopyranoside, and the cultures were grown for 18 h. The collected bacteria were resuspended in buffer A (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 2 mM ␤-mercaptoethanol) and lysed by sonication, and the recombinant DFF40⅐DFF45 complexes were purified on nickel nitrilotriacetic acid-agarose (Qiagen). In the case of the GST-DFF40⅐His-DFF45 complexes a second purification step on glutathione-Sepharose 4B (GE Healthcare) followed.
c-Jun and c-Fos complexes were generated by coexpression of the respective subunits in E. coli for 3 h at 25°C with 1 mM isopropyl-␤-D-thiogalactopyranoside. The collected bacteria were resuspended in buffer B (50 mM Tris, pH 7.5, 300 mM NaCl, 2 mM ␤-mercaptoethanol) and lysed by sonication, and the recombinant complexes were tandem-purified on nickel nitrilotriacetic acid-agarose and glutathione-Sepharose 4B.
His-tagged caspase-3 was expressed as described in Goebel et al. (39), and affinity purification was performed on nickel nitrilotriacetic acid-agarose according to the manufacturer's instructions.
DNA Cleavage Assay-Increasing amounts of recombinant, purified epitope-tagged DFF40⅐DFF45 complexes were incubated with recombinant purified His-caspase-3, 1 g of circular plasmid DNA, 5 M MgCl 2 , and 4.4 M dithiothreitol for 2 h at 37°C in buffer C (20 mM HEPES-KOH, pH 7.5, 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol) in a total volume of 50 l. The reactions were stopped on ice, and NaCl was added to a final concentration of 1 M. After phenol-chloroform extraction, the DNA was analyzed by agarose gel electrophoresis.
GST Pulldown Assays-GST fusion proteins immobilized on glutathione-Sepharose were used as affinity matrix for binding experiments. Appropriate amounts of affinity matrix were incubated for 3 h at 4°C with either bacterial lysates containing expressed import receptors or the corresponding purified recombinant import factor in buffer D (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 5 mM MgCl, 5% (v/v) glycerol, 2 mM ␤-mercaptoethanol). The binding experiments were performed in the absence or presence of 2 M RanGTP. After washing three times with ice-cold buffer D, the affinity matrix was boiled in SDS-PAGE sample buffer, and the matrix-bound proteins were analyzed by SDS-PAGE. Import receptor-binding intensities were quantified using the program 1Dscan EX (Scanalytics). The amount of bound import receptor was normalized over the amount of immobilized protein. The amount of import receptor bound to DFF complexes was normalized to the amount of either DFF40 or DFF45 depending on the individual DFF subunit to which it was compared.
In Vitro Nuclear Import Assays-Import reactions were performed as described previously (42) based on the method established by Adam et al. (46). Briefly, digitonin-permeabilized HeLa cells were incubated at 37°C for 25 min with 20 l of a transport reaction mix consisting of 0.35 M substrate, either 10 l of reticulocyte lysate (Promega, Madison, WI) or 0.5 M recombinant import receptor and an energy-regenerating system (0.5 mM ATP, 0.5 mM GTP, 10 mM creatine phosphate, 50 g/ml creatine kinase) in buffer E (20 mM HEPES-KOH, pH 7.4, 110 mM potassium acetate, 5 mM magnesium acetate, 0.5 mM EGTA, 2 mM dithiothreitol, 250 mM sucrose). For performing reconstitution experiments with recombinant transport factors, a Ran mix (3 M Ran(GDP), 0.5 M NTF2) was added to the transport reaction mix. For negative controls the assay was done in the absence of any transport factors, i.e. without reticulocyte lysate and without recombinant transport factors. Import reactions under inhibitory conditions were carried out either after pretreating the cells with 50 g/ml wheat germ agglutinin in buffer E or by replacing the energy-regenerating system with apyrase (100 units/ml). HeLa cells were fixed with 3% paraformaldehyde for 15 min and were mounted by using Vectashield with 4Ј,6diamidino-2-phenylindole (Vector Laboratories). Import reactions were visualized by fluorescence microscopy with a Zeiss microscope (Axioskop 20).

RESULTS
The Nuclear Import of the DFF Complex Is Energy-dependent and Requires Access to Nuclear Pore Proteins-To characterize the nuclear import of the DFF complex in vitro, we generated a recombinant fluorescently labeled DFF complex by fusing EGFP to the N terminus of DFF40. For purification of the DFF complex DFF45 was N-terminally Histagged. Because DFF45 is necessary for correct folding and activity of DFF40, both subunits were coexpressed in E. coli and affinity-purified. A plasmid DNA cleavage assay was performed to examine whether the N-terminal tags of DFF40 and DFF45 interfere with proper folding and complex formation. The plasmid DNA was efficiently cleaved upon incubation with EGFP-DFF40⅐His-DFF45 and purified caspase-3 ( Fig. 1A, lanes 3-7). This demonstrates that the EGFP-DFF40⅐His-DFF45 complex can be activated by caspase-3 leading to the release of nucleolytically active EGFP-DFF40. In contrast, the plasmid DNA was not cleaved in the presence of caspase-3 or fluorescently labeled DFF complex alone (Fig. 1A, lanes 2 and 8). For technical reasons, we had to clone His-tagged DFF45 lacking the amino acid residues 1-7. The DNA cleavage assay showed that the deletion of amino acids 1-7 in DFF45 did not interfere with its function as chaperone and inhibitor of DFF40.
The nuclear uptake of the EGFP-DFF40⅐His-DFF45 complex was examined using in vitro nuclear import assays (46). For that purpose, HeLa cells were grown on glass coverslips, and the plasma membranes were selectively permeabilized with digitonin. EGFP-DFF40⅐His-DFF45 was only imported into nuclei of permeabilized HeLa cells in the presence of rabbit reticulocyte lysate as a source of import receptors and an energy-regenerating system (Fig. 1B). The lack of reticulocyte lysate (buffer) and the addition of wheat germ agglutinin or apyrase to the transport system strongly inhibited nuclear accumulation of the DFF complex. Wheat germ agglutinin binds to glycosylated residues of nucleoporins, thereby inhibiting receptor-mediated transport processes through the nuclear pore complexes while pas- After phenol-chloroform extraction, the DNA was analyzed on a 1% agarose gel. The bands in lane 1 (plasmid control) represent the three topological forms of circular plasmid DNA. Purified EGFP-DFF40⅐His-DFF45 complex used for the DNA cleavage assay was analyzed by SDS-PAGE and Coomassie stained (shown on the right of the agarose gel). B, nuclear transport of the EGFP-DFF40⅐His-DFF45 complex is energy-dependent and requires access to nucleoporins. Digitonin-permeabilized HeLa cells were incubated with 0.35 M substrate, rabbit reticulocyte lysate (retic), and an energy-regenerating system for 25 min at 37°C (for details see "Experimental Procedures"). For a negative control, retic was replaced by transport buffer (buffer). Import reactions under inhibitory conditions were carried out either after pretreating the cells with wheat germ agglutinin (WGA) or by replacing the energy-regenerating system with apyrase as indicated. Cells were fixed and studied by direct fluorescence (EGFP). The DNA was counterstained with 4Ј,6-diamidino-2-phenylindole (DAPI). EGFP-DFF40⅐His-DFF45 tended to slightly diffuse into nuclei in the absence of import receptors (buffer). Scale bar, 10 m.
sive diffusion remains unaffected (47). Hence, nuclear import of the DFF complex requires access to nucleoporins. In addition, the loss of nuclear uptake under conditions where the energyregenerating system was replaced by apyrase underlines the energy dependence of the nuclear transport and excludes passive diffusion. These results demonstrate that the DFF complex, as the majority of nuclear proteins, traverses the nuclear membrane in a receptor-mediated fashion.
The Importin ␣/␤-Heterodimer Is Responsible for Nuclear Transport of the DFF Complex-To identify potential nuclear transport receptors that mediate nuclear import of the DFF complex, in vitro binding studies were performed. GST-tagged DFF40 and His-tagged DFF45 were coexpressed in E. coli and tandem affinity-purified. The proper folding of the subunits and complex formation was again analyzed in a plasmid DNA cleavage assay. Upon activation of the purified GST-DFF40⅐His-DFF45 complex by caspase-3, GST-DFF40 was able to cleave plasmid DNA (supplemental Fig. S1A). This demonstrated that the GST-DFF40⅐His-DFF45 complex is functional. GST-DFF40⅐His-DFF45 was immobilized on glutathione-Sepharose and incubated with importin ␣, importin ␣/␤-heterodimer, importin ␤, transportin, importin 5, importin 7, and importin 13, all from bacterial lysates ( Fig. 2A). After washing, the bound proteins were analyzed by SDS-PAGE followed by Coomassie staining. Only the importin ␣/␤-heterodimer was efficiently bound to the DFF complex ( Fig. 2A). The binding of importin ␤ and transportin was very weak, and none of the other import receptors was bound. In the nucleus direct binding of RanGTP to ␤-family import receptors disintegrates the receptor-substrate interaction. The binding of importin ␣/␤ to the DFF complex was abolished in the presence of RanGTP, demonstrating its specificity (Fig. 2B). To examine whether the importin ␣/␤-heterodimer is a functional import receptor for the DFF complex, we reconstituted nuclear import in vitro by using purified recombinant importin ␣ and importin ␤ (Fig.  2C). The EGFP-DFF40⅐His-DFF45 complex was carried into the nuclei of permeabilized HeLa cells in the presence of importin ␣/␤, although the nuclear uptake was not as efficient as with reticulocyte lysate. Because the DFF complex can be recognized by different importin ␣ variants (supplemental Fig. S2), the better nuclear uptake in the presence of reticulocyte lysate may be explained by the broader range of suitable importin ␣ molecules. In addition, weak nuclear import of EGFP-DFF40⅐His-DFF45 was also observed when just importin ␣ was added (Fig.  2C). That is probably due to formation of an ␣/␤-heterodimer with not completely washed out endogenous importin ␤. According to the results of the in vitro binding studies, import receptors that did not bind to the immobilized DFF complex were also not able to translocate the fluorescently labeled DFF complex into the nuclei of permeabilized HeLa cells (supplemental Fig. S1B).
Basic Amino Acids in Both Subunits Are Essential for Nuclear Import of the DFF Complex-For the initial characterization of sequence elements necessary for nuclear accumulation of the DFF complex the subcellular localization of coexpressed, fluorescently tagged DFF subunits was examined in HeLa P4 cells (Fig. 3). Because EGFP-or RFP-tagged DFF40 and DFF45 tend to diffuse into the nucleus, one subunit was expressed as an EGFP-GST fusion to further increase the size of the protein. In doing so, passive diffusion of the DFF complex was excluded. Fusion of tags to the DFF subunits did not interfere with DFF complex formation as analyzed by coprecipitation of the subunits from transfected HeLa cells (data not shown). Firstly, C-terminal deletions of both DFF subunits were coexpressed with the corresponding wild-type subunit to test whether the two C termini contribute to nuclear targeting of the DFF complex ( Fig. 3B) as previously shown by Lechardeur et al. (14). In contrast to the nuclear accumulation of cotransfected wildtype EGFP-GST-DFF40⅐RFP-DFF45 (Fig. 3B, top panel), deletion of either amino acids 324 -338 of DFF40 or amino acids 306 -331 of DFF45 led to a dominant cytoplasmic colocalization of both subunits (Fig. 3B, middle and bottom panels). However, red fluorescent staining of uncomplexed RFP-tagged DFF45 was also visible in the nucleus. This is probably caused by nuclear uptake of full-length RFP-DFF45 (see below, Fig. 4) and passive diffusion of C-terminally truncated RFP-DFF45. C-terminal truncations of the two DFF subunits had similar effects on the nuclear transport of the DFF complex when the fluorescence tags were exchanged among the DFF subunits (supplemental Fig. S3). This clearly confirmed the importance of both C termini for nuclear accumulation of the DFF complex. To further delimit the amino acid sequences responsible for importin ␣/␤-mediated nuclear import of the DFF complex, we generated mutants substituting basic amino acids in the C-terminal regions of DFF40 and DFF45 (Fig. 3A). The EGFP-GST- tagged mutant subunit was coexpressed with the corresponding RFP-tagged wild-type subunit. The results of these in vivo transfection studies are summarized in Fig. 3 (C and D). Nuclear accumulation of the DFF40⅐DFF45 complex was blocked when either the basic cluster RLKRK in DFF40 (DFF40mutA, Fig. 3C, top panel) or the basic cluster KRAR in DFF45 (DFF45mutB, Fig. 3D, bottom panel) was mutated. In contrast, mutation of the basic cluster RKR of DFF40 (DFF40mutB, Fig. 3C, bottom panel) or substitution of the two positively charged amino acids at positions 307 and 313 of DFF45 (DFF45mutA, Fig. 3D, top panel) had no effect on the nuclear localization. Similar results were obtained with RFP fused mutants with the exception of DFF45mutA. In that case, mutation of the basic amino acids Arg-307 and Lys-313 in RFP-DFF45 abolished nuclear localization of EGFP-GST-DFF40/RFP-DFF45 (supplemental Fig. S3C). To resolve this discrepancy EGFP-GST-DFF40mutB, which did not affect nuclear accumulation (see Fig. 3C, bottom panel), was coexpressed with RFP-DFF45mutA. In this transfection experiment, the dominant nuclear localization pattern of the DFF40⅐DFF45 complex did not change (supplemental Fig.  S3D). Hence, amino acids Arg-307 and Lys-313 of DFF45 are presumably not essential for nuclear translocation of the DFF complex. In summary, one basic cluster in the C-terminal region of each subunit, DFF40 (RLKRK) and DFF45 (KRAR), is required for nuclear accumulation of the DFF complex.
Additionally, GST pulldown assays with DFF complexes C-terminally truncated in either DFF40 or DFF45 were performed. For that, DFF40⌬324 -338 or DFF45⌬306 -331 were coexpressed with the corresponding wild-type subunit in E. coli, purified, and immobilized on glutathione-Sepharose as bait for import receptor binding. As shown in Fig. 3E, deletion of the C-terminal region of either DFF40 or DFF45 strongly decreased the binding of importin ␣/␤ to the DFF complex. Thus, this observation supports the results of the in vivo transfection experiments underlining the importance of the C-terminal regions of both subunits for nuclear import of the DFF complex.
Both DFF Subunits Contain an NLS in Their C-terminal Region-The identification of positively charged amino acids essential for nuclear transport of the DFF complex in each sub- unit raised the question whether both basic clusters represent independent cNLSs. To answer this question, the nuclear transport of the individual DFF subunits was analyzed particularly focusing on the capability of the C-terminal regions to mediate nuclear uptake. First, the DFF45 mutants described above (Fig.  3A), along with wild-type DFF45 were expressed as EGFP-GST fusion proteins in HeLa P4 cells. Although wild-type DFF45 showed a clear nuclear localization, deletion of the amino acids 306 -331 (⌬306 -331) or substitution of the basic cluster KRAR (mutB) blocked nuclear uptake (Fig. 4A). Mutation of Arg-307 and Lys-313 only had a minor effect on nuclear accumulation, leading to a homogenous distribution of monomeric DFF45 (Fig. 4A, mutA). In addition, the exclusively cytoplasmic localization of an EGFP-EGFP-GST (EEG) fusion protein changed upon fusion to amino acids 306 -331 of DFF45 (EEG-DFF45-(306 -331)), now being largely nuclear (Fig. 4C). These results demonstrate that the C-terminal region of DFF45 exhibits a NLS that is necessary and sufficient for nuclear uptake of monomeric DFF45. In more detail, besides playing an essential role for the importin ␣/␤-mediated nuclear import of the DFF complex, the basic cluster KRAR was also identified as key element of the NLS in DFF45 itself. Second, in vivo transfection experiments with EGFP-GST-tagged DFF40 were performed. As shown in Fig. 4B, EGFP-GST-DFF40 was not transported into the nucleus of HeLa cells. However, the cytoplasmic EGFP-EGFP-GST fusion protein accumulated in the nucleus of transfected cells upon fusion to the C-terminal amino acids 301-338 of DFF40 (EEG-DFF40-(301-338) (Fig. 4C)). This suggests that the C terminus of DFF40 also harbors an NLS. The cytoplasmic localization of monomeric EGFP-GST-DFF40 may be due to misfolding of DFF40 overexpressed in the absence of exogenous DFF45 chaperone.
The Individual DFF Subunits Are Recognized by Importin ␣/␤-Based on the in vivo data, it was reasonable to assume that importin ␣/␤ also binds to the individual DFF subunits. To verify this assumption, interaction studies between recombinant importins and the single DFF subunits were performed. For the DFF40 subunit, immobilized GST-DFF40⅐His-DFF45 was treated with purified caspase-3 first to cleave off DFF45. The activated GST-DFF40 and GST-DFF45 were incubated with import receptors. Among them, importin ␣/␤ was bound RanGTP-sensitive to DFF40 and DFF45; however, importin ␣/␤-binding was less efficient than to the DFF complex (Fig. 5,  A and B). For DFF45 much more protein had to be immobilized to detect importin ␣/␤-binding (Fig. 5B). Because of this weak binding we analyzed whether other import receptors are able to interact with GST-DFF45. However, none of the other importins (importin ␣, importin ␤, transportin, importin 5, importin 7, importin 9, and importin 13) bound to immobilized GST-DFF45 (data not shown). In vitro binding studies with the C terminus of DFF45 fused to GST (GST-DFF45-(296 -331)) confirmed the specific interaction of importin ␣/␤ with this region (Fig. 5C). In conclusion, DFF45 harbors a classic, presumably monopartite cNLS in its C-terminal region.
In contrast to the DFF complex and monomeric DFF45, activated GST-DFF40 was additionally bound by importin ␤ alone (Fig. 5A). GST pulldown assays with amino acids 314 -338 of DFF40 revealed that its C terminus is directly recognized by importin ␤ rather than by importin ␣/␤ (Fig. 5D). Therefore, we analyzed whether importin ␣ contributes to the binding of importin ␤ to DFF40 at all. For that purpose, GST pulldown assays were performed with importin ␣ lacking amino acids 1-55 that contain the IBB domain (importin ␣⌬IBB). This importin ␣ mutant is able to interact with cargo proteins in the absence of importin ␤ (24,48). The results of these pulldown assays are summarized in Fig. 5E. Binding of importin ␣⌬IBB to activated DFF40 demonstrated that DFF40 directly interacts with importin ␣. This interaction was approximately half as efficient as importin ␣⌬IBB-binding to the DFF complex, as already observed for importin ␣/␤-binding (Fig. 5A). Similarly, binding of importin ␣⌬IBB to DFF45 was much weaker (Fig.  5E). Furthermore, deletion of the C-terminal region of either DFF40 or DFF45 in the DFF complex clearly reduced importin ␣⌬IBB-binding compared with the wild-type DFF complex.
In conclusion, in vivo transfection and in vitro binding experiments show that the C-terminal regions of both subunits, DFF40 and DFF45, exhibit NLSs. However, importin ␣/␤ and importin ␣⌬IBB bind the DFF complex about two times stronger than DFF40 alone, whereas monomeric DFF45 hardly binds to importin ␣/␤ using similar amounts of protein. To explain these different binding efficiencies, the interaction between the DFF complex and importin ␣/␤ was analyzed in more detail. Binding Models for the Interaction between the DFF Complex and Importin ␣/␤-The binding of importin ␣/␤ to the C terminus of DFF45 (Fig. 5C) and the binding of importin ␤ to the C terminus of DFF40 (Fig. 5D) could point to the following binding model: recognition of the DFF complex by importin ␣/␤ involves direct interactions with both importin ␣ and importin ␤. Although the fact that importin ␣⌬IBB bound more tightly to the DFF complex compared with its subunits in the absence of importin ␤ (Fig. 5E) does not favor this hypothesis, we exam-ined it further. Because in this scenario direct interactions between importin ␤ and the DFF complex can only occur via the second cargo binding site (HEAT repeats 2-11) GST pulldown assays with N-terminally truncated importin ␤ were performed (supplemental Fig. S4). N-terminally truncated importin ␤ only slightly reduced binding of importin ␣/␤ to the DFF complex. A similar reduction was observed for the control substrate nucleoplasmin that interacts with importin ␣/␤ via a bipartite cNLS. Therefore, a direct interaction of the DFF complex with both importin ␣ and importin ␤ was excluded.
Because at least one basic cluster in each DFF subunit is required for nuclear import of the DFF complex, another binding model for increased DFF-import receptor interaction could be the following: both essential basic clusters in DFF40 and DFF45 interact simultaneously with importin ␣, mimicking a bipartite cNLS. To test this assumption, we generated an importin ␣ mutant with a nonfunctional minor binding site. It was shown that substitution of glutamate at position 402 with arginine in Saccharomyces cerevisiae importin ␣ (SRP1p) strongly decreased importin ␣/␤-binding to bipartite cNLSs (49). Hence, we substituted arginine for the corresponding glutamate at position 388 of Xenopus laevis importin ␣1 (␣E388R) and performed GST pulldown assays with this importin ␣ mutant (Fig. 6). Amino acids 94 -135 of SV40 large T antigen (SV40NLS) containing its monopartite cNLS and nucleoplasmin exhibiting a bipartite cNLS were used as control substrates. Binding of importin ␣/␤ containing importin ␣E388R to GST-SV40NLS was reduced to ϳ75% compared with wild-type importin ␣, whereas GST-nucleoplasmin was hardly recognized (ϳ2%) by this importin ␣ minor binding site mutant (Fig. 6B). This demonstrated that the importin ␣E388R mutant was still capable to bind monopartite cNLS but had almost entirely lost its ability to bind bipartite cNLSs. Like the GST-SV40NLS, GST-DFF45 was recognized by importin ␣E388R/␤ half as efficiently (56.6%) as wild-type importin ␣/␤ (Fig. 6C). These data are in line with the results described above and confirmed that DFF45 harbors a FIGURE 5. The C-terminal regions of DFF40 and DFF45 both contribute to the interaction between the DFF complex and importin ␣/␤. GST pulldown assays were performed in the absence or presence of 2 M RanGTP as described in the legend to Fig. 2. Bound fractions were analyzed by SDS-PAGE and Coomassiestained or transferred onto nitrocellulose membrane, Ponceau-stained, and detected using mouse anti-His antibody. The amount of bound importin ␣, importin ␣⌬IBB, or importin ␤ was quantified using the program 1Dscan EX and normalized to the amount of immobilized protein (for details see "Experimental Procedures"). A value of 100% was assigned to the binding of importin ␣ (⌬IBB) to the DFF40⅐DFF45 complex or the binding of importin ␤ (in complex with importin ␣) to monomeric DFF40, as indicated. A and B, importin ␣/␤ specifically bound also to the individual DFF subunits, DFF40 and DFF45. Immobilized GST-DFF40⅐His-DFF45 complex was incubated with caspase-3 for 30 min at 30°C to cleave off DFF45. GST-DFF40⅐His-DFF45, activated GST-DFF40, and GST-DFF45 were incubated with importin ␣/␤ or importin ␤ (A). Additionally, 12.5 times the amount of GST-DFF45 used in A was analyzed by SDS-PAGE and immunoblotted (B). C and D, importin ␣/␤ binding to amino acids 296 -331 of DFF45 is RanGTP-sensitive, and amino acids 314 -338 of DFF40 are specifically recognized by importin ␤. Immobilized GST-DFF45-(296 -331) (C) and GST-DFF40-(314 -338) (D) were incubated with importin ␣/␤ or importin ␤. E, immobilized GST-DFF40⅐His-DFF45, GST-DFF40 (activated as described in A), GST-DFF40⅐His-DFF45⌬306 -331, GST-DFF45, and GST-DFF40⌬324 -338⅐His-DFF45 were incubated with importin ␣⌬IBB (⌬IBB). Mw, molecular weight; aa, amino acids; imp (and i), importin. monopartite cNLS. In contrast, binding of importin ␣/␤ to GST-DFF40⅐His-DFF45 was strongly reduced (6.9%) using the importin ␣ minor binding site mutant (Fig. 6D) suggesting that the DFF complex indeed exhibits a bipartite cNLS. However, a similar effect was observed for the interaction between activated DFF40 and importin ␣E388R/␤ (Fig. 6D). Hence, besides the non-classic NLS in the C terminus, DFF40 probably contains a second NLS. This signal resembles the classic bipartite type and explains the binding of importin ␣/␤ to activated GST-DFF40 observed in pulldown assays (Fig. 5, A and D). The reduction of importin ␣/␤-binding to the DFF complex using the importin ␣ minor binding site mutant is not distinguishable from the reduced binding to activated DFF40. Therefore, we used a second approach to analyze the formation of an intermolecular cNLS: the C-terminal regions of DFF40 (amino acids 301-338) and DFF45 (amino acids 296 -331) were fused to c-Jun-(257-338) and c-Fos-(139 -200), respectively. These regions of the activator protein-1 transcription factor subunits contain basic leucine-zippers and therefore dimerize (50). The respective Fos and Jun constructs were coexpressed in E. coli, purified, and immobilized on glutathione-Sepharose as bait for importin ␣⌬IBB binding (Fig. 7). The Jun/Fos complex containing the C-terminal regions of both DFF subunits (His-Jun-DFF40-(301-338)⅐GST-Fos-DFF45-(296 -331)) bound importin ␣⌬IBB roughly 5-fold stronger than the complex containing only the C terminus of DFF45 (His-Jun⅐GST-Fos-DFF45-(296 -331)) and 8-fold stronger than the complex containing the C terminus of DFF40 (His-Jun-DFF40-(301-338)⅐GST-Fos). These results demonstrate a great amount of importin ␣⌬IBB binding cooperativity between the isolated, but complexed, C-terminal regions of the DFF subunits. In addition, the reconstitution of efficient importin ␣/␤-binding to the complexed C-terminal regions compared with the single C termini was also observed when the histone fold motifs of negative cofactor (NC) 2␣ and NC2␤ were used as dimerization domains (supplemental Fig.  S5). In conclusion, these data support the binding model in which the basic amino acids in the C-terminal regions of DFF40 and DFF45 form an intermolecular cNLS.

DISCUSSION
Classic NLSs were initially described as short sequences enriched in positively charged amino acids (25,28). Based on several structural studies the originally defined consensus sequences of monopartite and bipartite cNLSs have been modified. The current view of a monopartite cNLS follows the consensus K(K/R)X(K/R) (P2-P5) (51), and amino acids that are compatible with sharp turns upstream of the consensus motif and a small hydrophobic amino acid (P6) followed by an acidic residue (P7) downstream of the consensus are preferred (30). In contrast to monopartite NLSs, sequence requirements for bipartite NLSs are less stringent (25,52). Bipartite cNLSs consist of two basic clusters connected by a 10-to 12-amino acid spacer following the consensus 2(K/R)-X 10 -12 -3(K/R) (22,25,28). In certain cases the spacer might exceed a length of up to 32 amino acids (53). Simultaneous interaction of the two basic stretches with the importin ␣ binding sites characterizes a bipartite cNLS (27).
In this study, the nuclear import pathway of the human DFF complex was analyzed. We could demonstrate that importin ␣/␤ mediates nuclear import of the DFF complex in permeabilized cells and that the C-terminal regions of both subunits are necessary for efficient binding of importin ␣/␤ to the DFF complex. Moreover, the results of our in vivo studies reveal that one basic cluster in the C terminus of each subunit, DFF40 (RLKRK) and DFF45 (KRAR), is essential for nuclear accumulation of the DFF complex.
In addition, the basic cluster (KRAR) in the C-terminal region of DFF45 was also identified as key element for nuclear translocation of monomeric DFF45. This amino acid stretch fulfills the consensus motif (P2-P5) for a monopartite cNLS, but additional sequence requirements are not fully satisfied. For instance, position P6 is occupied by a polar glutamine instead of a small hydrophobic amino acid. Differences from preferred amino acids for monopartite cNLSs may well explain why monomeric DFF45 only weakly interacted with importin ␣/␤. Furthermore, our results of in vivo and in vitro experiments argue against a bipartite cNLS in the C terminus of DFF45 as previously proposed by Lechardeur et al. (14). First, substitution of additional basic residues in the C-terminal region of DFF45 (Arg-307 and Lys-313) only slightly decreased nuclear localization of DFF45 overexpressed in HeLa cells. Second, the interaction between monomeric DFF45 and importin ␣/␤ did not depend on the importin ␣ minor binding site, which is necessary to bind to bipartite cNLSs. Among different species, the C-terminal regions of murine and bovine DFF45 harbor a stretch of basic amino acids very similar to the monopartite cNLS of human DFF45. Thus, also in these species nuclear entry of monomeric DFF45 may be mediated by a monopartite cNLS. The presence of monomeric DFF45 in the nucleus could result in binding of accidentally activated DFF40 and therefore prevent DNA degradation in healthy cells.
GST pulldown experiments revealed that importin ␣/␤ preferably binds to full-length DFF40, and this interaction depends on the importin ␣ minor binding site. In contrast, the C terminus of DFF40 interacts with importin ␤ only. Additionally, the C-terminal region of DFF40 proved to be sufficient for nuclear accumulation of a cytoplasmic reporter protein. Together, these results indicate that the DFF40 subunit contains two different types of NLSs: a bipartite classic NLS distributed among the DFF40 subunit and a non-classic NLS within the C-terminal region of DFF40 recognized by importin ␤. Surprisingly, fulllength DFF40 overexpressed in HeLa cells was not translocated into the nucleus. However, misfolding of DFF40 overexpressed in the absence of its chaperone DFF45 could bury the NLSs and therefore explain this observation.
We have demonstrated that only importin ␣/␤ mediates efficient nuclear import of the DFF complex in permeabilized cells and that each DFF subunit alone can be recognized by the importin ␣/␤-heterodimer. These results suggest that nuclear transport of the DFF complex is mediated by multiple importin ␣/␤ molecules. The solubility of import receptor⅐cargo complexes in the permeability barrier of the nuclear pore is determined by the number of import receptors bound to the cargo protein (54). Accordingly, the presence of more than one NLS in a large protein would lead to a more rapid translocation through the nuclear pore. Independent interactions of importin ␣/␤ with the cNLSs in DFF40 and DFF45 would therefore confer efficient nuclear import of the DFF complex. Lechardeur et al. (14) previously reported that putative cNLSs in the C-terminal regions of DFF40 and DFF45 have an additive effect on the nuclear targeting efficiency of the DFF complex. In contrast, our in vivo studies demonstrate that the cNLS of one subunit alone is not sufficient for nuclear accumulation of the DFF complex (Fig. 3B). Also, the results of our binding studies do not support the proposed additive effect. First, importin ␣/␤ binds the DFF complex about two times stronger than DFF40 alone, whereas importin ␣/␤ binding to DFF45 was hardly detected using comparable amounts of protein (Fig. 5A). Second, importin ␣/␤ binding to the DFF complex is approximately four times reduced when the C-terminal region of either DFF40 or DFF45 is missing (Fig. 3E). These observations raise the possibility of another binding model in which the C-terminal regions of DFF40 and DFF45 interact with only one importin ␣/␤ molecule. Direct interactions with both importin ␣ via the cNLS in DFF45 and importin ␤ via the ncNLS in DFF40 are excluded, because (i) N-terminal truncations in the second cargo binding site of importin ␤ did not significantly affect importin ␣/␤binding to the DFF complex and (ii) the DFF complex interacted more tightly with importin ␣ lacking its IBB domain compared with the individual subunits. The latter results combined with the fact that nuclear import of the DFF complex requires the C-terminal regions of both subunits rather point to following binding mechanism: the essential basic amino acids in the C-terminal regions of DFF40 and DFF45 are both simultaneously recognized by importin ␣. Support for the formation of this intermolecular cNLS derived from binding experiments showing that the isolated, but complexed, C-terminal regions of Nuclear Import of DNA Fragmentation Factor DECEMBER 7, 2007 • VOLUME 282 • NUMBER 49 the DFF subunits bind importin ␣⌬IBB (importin ␣/␤, respectively) much more efficiently than the individual C-terminal regions.
However, further experiments are necessary to distinguish whether the C-terminal regions of DFF40 and DFF45 together interact with one importin ␣/␤-heterodimer or are independently recognized by importin ␣/␤ molecules. Regarding the nuclear import of a complex due to sequence elements distributed on its subunits, Liku et al. (55) recently demonstrated that nuclear transport of the Mcm2-7 complex is mediated by two potential cNLSs on the Mcm2 and Mcm3 subunits. Each of the two cNLSs is required but not sufficient for nuclear accumulation of the Mcm2-7 complex. Taken together, DFF45 is not only chaperone and inhibitor of DFF40 but also plays an essential role in nuclear targeting of the nuclease.