Synthesis and Functional Analyses of Nuclear Clusterin, a Cell Death Protein*

Nuclear clusterin (nCLU) is an ionizing radiation (IR)-inducible protein that binds Ku70, and triggers apoptosis when overexpressed in MCF-7 cells. We demonstrate that endogenous nCLU synthesis is a product of alternative splicing. Reverse transcriptase-PCR analyses revealed that exon II, containing the first AUG and encoding the endoplasmic reticulum-targeting peptide, was omitted. Exons I and III are spliced together placing a downstream AUG in exon III as the first available translation start site. This shorter mRNA produces the 49-kDa precursor nCLU protein. Ku70 binding activity was localized to the C-terminal coiled-coil domain of nCLU. Leucine residues 357, 358, and 361 of nCLU were necessary for Ku70-nCLU interaction. The N- and C-terminal coiled-coil domains of nCLU interacted with each other, suggesting that the protein could dimerize or fold. Mutation analyses indicate that the C-terminal NLS was functional in nCLU with the same contribution from N-terminal NLS. The C-terminal coiled-coil domain of nCLU was the minimal region required for Ku binding and apoptosis. MCF-7 cells show nuclear as well as cytoplasmic expression of GFP-nCLU in apoptotic cells. Cytosolic aggregation of GFP-nCLU was found in viable cells. These results indicate that an inactive precursor of nCLU exists in the cytoplasm of non-irradiated MCF-7 cells, translocates into the nucleus following IR, and induces apoptosis.

The clusterin (CLU) 1 gene was originally isolated by numerous laboratories due to its up-regulation during cell death responses in various tissues and after various toxic stress signals (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). CLU expression is complex, appearing as different forms in different cell compartments. One set of proteins is directed for secretion, and other CLU species are expressed in the cytoplasm and nucleus (see below). The secretory form of the clusterin protein (sCLU) has been extensively studied and is produced by translation from the first AUG codon of the full-length CLU mRNA. sCLU is targeted to the ER by an initial leader peptide. This ϳ60-kDa pre-sCLU protein is further glycosylated and proteolytically cleaved into ␣and ␤-subunits, held together by disulfide bonds (12,13). External sCLU is an 80-kDa protein that appears as an ϳ40-kDa ␣and ␤-protein smear by SDS-PAGE under reducing conditions (14).
The role(s) of CLU in cell death processes is also complex and may be different for the various forms of the protein. Recent data suggest that overexpression of sCLU in human cancer cells caused drug resistance and protection against certain cytotoxic agents that induce apoptosis (9,15,16). Additional confirmatory data suggest that sCLU acts as a molecular chaperone, scavenging denatured proteins outside cells following specific stress-induced injury such as heat shock (17)(18)(19)(20)(21). In fact, sCLU possesses nonspecific binding activity to hydrophobic domains of various proteins in vitro (20), further supporting its role as a molecular chaperone acting to clear cellular debris. sCLU does not associate with Ku70 (14).
Other data, including our own (13), show that overexpression of a specific nuclear form of CLU (nCLU) acts as a pro-death signal, inhibiting cell growth and survival. Our laboratory isolated CLU while searching for Ku70-binding proteins using a yeast two-hybrid screen of a human liver cDNA library (14). The specificity of nCLU-Ku70 interaction was confirmed by co-immunoprecipitation, far Western, and confocal co-localization analyses of the two proteins in transfected or IR-treated MCF-7:WS8 (MCF-7) breast adenocarcinoma cells (13). Confocal microscopy revealed an apparently inactive form of nCLU (i.e. pnCLU) in the cytoplasm of non-irradiated cells (13) that subsequently translocated to the nucleus following clinically relevant low dose IR exposures (Ն1Gy). Co-immunoprecipitation studies demonstrated the presence of the Ku80 protein in the Ku70-nCLU complex, strongly suggesting that nCLU bound to the Ku70/Ku80 dimer (13). Overexpression of nCLU also blocked Ku70/Ku80 DNA end binding activity (14); however, the relationship between nCLU binding to Ku70 and cell death remained undetermined.
Interestingly, an ϳ49-kDa nuclear CLU precursor (pnCLU) was induced and translocated from the cytoplasm to the nucleus after certain cytotoxic events, including IR (13) or transforming growth factor-␤ (22) treatments. This "death" form of the CLU protein was proposed to be synthesized from a second in-frame AUG codon (13,22), although mechanisms of production were not elucidated. The pnCLU protein apparently lacks the ER-targeting leader peptide and does not undergo ␣/␤ cleavage or extensive glycosylation, as observed with sCLU (22). pnCLU functions remain unknown. Following IR treatment of MCF-7 cells, an ϳ55-kDa nCLU protein was expressed at extremely low levels in the nuclei of exposed cells (13). Overexpression of GFP-fused nCLU (initiated from the second AUG start codon in its mRNA) or a C-terminal 120-amino acid fragment (C-120) in MCF-7 cells resulted in significant growth inhibition, caspase 3-independent apoptosis, and lethality (13,23). All attempts to stably express nCLU resulted in lethality (13) or clones expressing G418 resistance without nCLU. 2 In contrast, overexpression of sCLU protein did not affect the survival of MCF-7 cells before or after IR stress (14).
Analyses of the amino acid sequence of full-length rat CLU detected four "myosin tail-like" domains (24) that are synonymous with a "coiled-coil" structure. Coiled-coil domains are the most common motifs used by proteins for oligomerization, and consist of amphipathic ␣-helices usually with a 4-3 pattern of hydrophobic and hydrophilic residues repeated in heptads. To date, no structure/function studies of the CLU protein have been performed, and mechanisms of the production of nuclear forms of this protein has yet to be determined.
The objectives of the following studies were as follows: (a) investigate the mechanism(s) of nCLU production; (b) identify the functional NLS of the protein; (c) identify the death domain of nCLU, and (d) elucidate the Ku70-binding domain of nCLU. These studies may then allow elucidation of the role of nCLU-Ku70 binding in cell death responses. We demonstrate that nCLU has one high affinity coiled-coil Ku70 binding domain within its C terminus. This domain is essential for its cell death function. Furthermore, nCLU contains an additional coiled-coil domain in its N terminus that interacts with its own C-terminal coiled-coil domain. Interestingly, the N-terminal coiled-coil domain of nCLU does not interact with Ku70. Furthermore, an N-terminal nCLU coiled-coil-containing peptide interfered with binding of the nCLU C-terminal coiled-coil domain with Ku70. Based on the apparent physical interactions of N-and C-terminal pnCLU coiled-coil domains, we offer a mechanism by which the inert ϳ49-kDa pnCLU could become activated to an ϳ55-kDa nCLU death protein. Finally, we discuss the cell death function of nCLU in damaged cells, and we propose mechanisms of cell death involving Ku70 binding.

EXPERIMENTAL PROCEDURES
RT-PCR of nCLU-Based on probable splice sites, the following primers were used to investigate whether alternative splice forms of CLU could be found: hCLU5Ј, 5Ј-ACAGGGTGCCGCTGACC-3Ј; hCLU-N-term-rev, 5Ј-TTAGAGCTCCTTCAGCTTTGTCTC TG-3Ј; NCLUforw-3, 5Ј-GCTGACCGAAATGTC-3Ј; hnCLUrev, 5Ј-TCACACCACCC-GGTGCTTTTTG-3Ј. Total RNA from MCF-7 cells was extracted using RNAsol B (Invitrogen). Reverse transcription reactions were performed using the oligo(dT) primer (Promega) and Superscript II enzyme system (Invitrogen). The PCR step was performed using Taq DNA polymerase (Invitrogen). When hCLU5Ј and hCLUN-term-rev primers were used, PCR conditions were as follows: 95°C for 30 s, 62-72°C for 45 s, and 72°C for 45 s. When NCLUforw-3 primer was used in pair with either hCLUN-term-rev or hnCLUrev primers, optimal conditions were as follows: 95°C for 30 s, 64°C for 45 s, and 72°C for 45 s. A Bio-Rad iCycler Thermal Cycler was used for all reactions. RT-PCR products were resolved by 2% agarose gel electrophoresis and purified using the Concept Matrix Gel Extraction System (Invitrogen). DNA sequencing was performed by the Case Western Reserve University DNA Sequencing Service.
Computer Analyses of Protein and Nucleic Acid Primary and Secondary Structures-EMBnet software (www.ch.embnet.org/software/ COILS_form.html) was used to analyze amino acid (aa) sequences of mouse and human CLU proteins for coiled-coil domain structures; coiled-coil domains in proteins are known to be important for proteinprotein interactions (25)(26)(27). Analyses were performed using the algorithm published by Lupas et al. (28). Primary sequences of both human and mouse CLU proteins were retrieved from GenBank TM (accession numbers: S70244 for mouse and M64722 for human CLUs (12,29)).
Construction of PCR primer sets for deletion/mutation analyses of the CLU protein was performed using OligoTech version 1.00 software (Oligos Etc., Inc./Oligos Therapeutics, Wilsonville, OR).
Generation of CLU Mutants and Yeast Two-hybrid Analyses-The Y190 yeast strain (HIS3,lacZ,trp1,leu2 (30,31)) was used in two-hybrid structure/function analyses as described (13,14). Full-length mouse CLU cDNA was a generous gift from Dr. M. Tenniswood (Notre Dame University). Full-length mouse Ku70 cDNA was a generous gift from Dr. M. Abe (National Institute of Radiological Sciences, Japan). Human CLU and Ku70 constructs were used as described previously (14). Mouse CLU deletion mutants, containing aa residues 18 -448, 18 -97, 98 -317, 18 -380, 18 -328, 256 -328, 345-380, and 318 -368, and human CLU deletion fragments 34 -449 and 245-380 were created by PCR using the appropriate primer sets (described above). Deletion mutants were inserted into either pAS2-1 or pACT2 vectors (Clontech Laboratories, Inc., Palo Alto, CA) as prey or bait, respectively, using SmaI sites and the Perfectly Blunt TM Cloning kit (Novagen, Inc., Madison, WI), according to manufacturer's instructions. Mouse CLU deletion mutants containing amino acids 256 -448, 256 -380, and 256 -344 were created using NcoI restriction enzyme digestion in combination with BamHI, ScaI, or PstI, respectively. The resulting fragments were then inserted into the pAS2-1 vector using NcoI, BamHI, SmaI, and PstI restriction sites. Mouse Ku70 cDNA was excised from the original vector using BspLU11I and EcoRI enzymes, and inserted into pACT2 and pAS2-1 vectors, which were pre-digested with NcoI and EcoRI restriction enzymes. Point mutations within mouse CLU cDNAs were introduced using the QuikChange kit (Stratagene, Cedar Creek, TX) following the manufacturer's instructions. The LR mutant (substitution of 9 Leu residues to Arg within the C-terminal coiled-coil domain of mouse nCLU) was generated by direct cloning of a 155-bp DNA fragment containing all required nucleotide substitutions. This was accomplished by first synthesizing a single strand DNA fragment: 5Ј GC CAG GAG  CGG AAC GAC TCG CGC CAG GTG GCC GAG AGG CGGACA GAG  CAG TAC AAG GAG CGG CGG CAG TCC TTC CAG TCG AAGATG  CGC AAC ACC TCA TCC CGG CGG GAG CAG CGG AAC GAC  CAGTTC AAC TGG GTG TCC CAG CGG GCT AAC CGC A 3Ј. This fragment was further amplified by PCR using Pfu DNA polymerase (Stratagene, Cedar Creek, TX) and the following pair of primers: Al-mutPCR1 (30-mer), 5Ј GCT AAC CTG CGC CAG GAG CGG AAC GAC TCG 3Ј, and AlmutPCR2 (38-mer), 5Ј TTA GTA GTA CTT GTC TTC TCC CTG TGT GCG GTT AGC CC 3Ј.
Human CLU cDNA fragments encoding aa residues 84 -449 and 329 -449 in the pACT2 vector, as well as human Ku70 cDNA in the pAS2 vector, were described previously (13,14). All constructs were sequenced to verify in-frame positions with respect to the GAL4 domain. GAL4-Ku70 was co-transfected with various GAL4-CLU fragments into Y190 yeast using the protocol provided in the MATCH-MAKER (Clontech Laboratories, Inc., Palo Alto, CA) manual. Stable transformants were grown for 6 days at 30°C on plates containing His(Ϫ), Trp(Ϫ), Leu(Ϫ) and SD medium (Difco) containing 2% agarose and 15 mM 3-amino-1,2,4-triazole. Colonies were subjected to on-filter ␤-galactosidase assays, with X-gal as a substrate. The reaction was considered positive when colonies turned blue within first 120 min of incubation with X-gal. Negative reactions (i.e. blue colonies were not observed) were indicative of no association, only when protein expression for each CLU or Ku70 mutant could be demonstrated by Western blot analyses (see below). Blue colonies that appeared after 120 min were regarded as false-positive signals. X-gal staining was performed on colony-lift filters as described (14).
Western Blot Analyses-Rabbit polyclonal H-330 antibody (Santa Cruz Biotechnology) was used to detect the 49-kDa species of CLU (pnCLU) in whole cell extracts from MCF-7 cells. A 1:2500 dilution of primary H-330 antibody was used followed by 1:10,000 dilution of horseradish peroxidase-conjugated anti-rabbit secondary antibody (Santa Cruz Biotechnology).
Yeast extracts were prepared using acid-washed glass beads as described in the MATCHMAKER kit manual and were used for Western blot analyses. Protein concentrations were determined using the Bio-Rad protein concentration reagent as per manufacturer's instructions (Bio-Rad). Yeast extracts were mixed with an equal volume of 2ϫ Laemmli buffer, boiled for 5 min, and resolved on either 10 or 15% PAGE gels, depending on the expected molecular weight of the proteins of interest. Proteins were then transferred onto polyvinylidene difluoride-Plus membranes (MSI, Westboro, MA). Membranes were blocked with 5% non-fat dried milk and probed with anti-GAL4BD or anti-GAL4AD monoclonal antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Horseradish peroxidase-conjugated anti-mouse secondary antibody (Amersham Biosciences) was then used, and proteins were visualized by ECL (Amersham Biosciences) and x-ray film. All CLU deletion or point mutants were approximately equally expressed in yeast, and no loss of stability of proteins was noted.
In Vitro GST-Ku70 Binding Assays-Full-length mouse Ku70 cDNA was amplified using PCR and the appropriate primer sets and inserted into the pGEX-2T vector (Amersham Biosciences) at a SmaI restriction site to generate an in-frame glutathione S-transferase (GST)-fused protein. Escherichia coli NovaBlue Single Competent Cells (Novagen, Inc., Madison, WI) were used for plasmid propagation according to the protocol described in the Perfectly Blunt TM Cloning kit manual (Novagen, Inc., Madison, WI). GST or GST-Ku70 fusion proteins were generated and purified using standard methods (32). All genetically manipulated plasmids were confirmed by DNA sequencing prior to use. Glutathione-agarose beads (Amersham Biosciences) were used for purification of GST-fused proteins as described (32) using Binding buffer (20 mM Tris-HCl, pH 7.2, 150 mM NaCl, 1.5 g/liter bovine serum albumin, proteinase inhibitors). Amounts of immobilized GST-Ku70 or GST (alone) proteins were estimated by Coomassie Blue staining of SDS-PAGE gels, with various amounts of bovine serum albumin protein used as standards. Mouse CLU cDNA fragments encoding amino acid residues 18 -448 and 345-380 were amplified by PCR and inserted into the pT7Blue-2 vector (Novagen, Inc., Madison, WI) using the Perfectly Blunt TM Cloning kit (Novagen, Inc., Madison, WI) according to manufacturer's instructions. All cDNA expression vectors were sequenced to confirm cloning before use.
The TNT-coupled wheat germ extract system (Promega, Madison, WI) was used to generate S⅐Tag-fused, 35 S-labeled CLU protein fragments in vitro according to manufacturer's protocols. The amount of 35 S-labeled protein was estimated using the S⅐Tag Rapid Assay kit (Novagen Inc., Madison, WI), and an equal amount of transcription/ translation mixture (40 l) was added to glutathione-containing beads at the same molar amounts (0.2 pmol of CLU to 0.2 pmol of GST or GST-Ku70) and incubated overnight at room temperature. The mixture was then incubated at 4, 23, or 37°C for 1 h, and beads were collected by centrifugation (15,000 ϫ g), washed three times with ice-cold Binding buffer, resuspended in Laemmli buffer, and boiled, and proteins were resolved by 10 -15% SDS-PAGE. 35 S-Labeled nCLU fragments were then visualized by autoradiography as described (33).
Intracellular Trafficking of nCLU and Mutant nCLU Proteins-To study the intracellular targeting of various nCLU fragments, we transfected MCF-7 cells with GFP-nCLU and GFP-nCLU deletion mutants. Effectin reagent (Qiagen Inc., Valencia, CA) was used to transfect 2 ϫ 10 4 log phase MCF-7 cells that were pre-attached for 24 h to coverslips. Forty eight hours (48 h) later, cells were fixed for 1 h in freshly prepared 2% paraformaldehyde at room temperature. Cells were then rinsed twice in phosphate-buffered saline and treated with RNase A in phosphate-buffered saline (1 g/ml) for 20 min at 37°C. Cells were mounted in propidium iodide-containing Vectashield mounting media (Vector Laboratories Inc., Burlingame, CA) and subjected to confocal microscopy using a Bio-Rad MRC 1024 Confocal Laser Scanning microscope (Bio-Rad). All photomicrographs show representative z-sections at ϫ1000 magnification. At least 180 transfected cells (green) were visually screened for distribution of GFP-CLU protein in the cell and its co-localization with propidium iodide-stained nuclei (red). The number of cells with pycnotic propidium iodide versus normal nuclear morphology was manually counted, and percentages of apoptotic cells Ϯ S.E. were calculated (see Table I). Likewise, the number of cells with nuclear only, cytoplasmic and nuclear, or cytoplasmic only localization of the GFP-fused proteins was manually counted. Percentages of cells with nuclear (or cytoplasmic and nuclear) or cytoplasmic only expression Ϯ S.E. were calculated (Table II).

Identification of an nCLU-specific mRNA Splice Variant-
After examining exon/intron sequences of the human CLU gene, we noticed an acceptor splice site two nucleotides upstream from the second in-frame AUG (codon 34) sequence of pnCLU ( Figs. 1 and 2A). Located within exon III, this AUG codon was proposed to be the start site for the intracellular pnCLU protein (13,22). In contrast to previous alternative translation start theories, we hypothesized that nCLU could be translated from low abundance mRNA produced by an alternative splicing event when exons I and III are spliced together. In this case, exon II that contained the first AUG and the ER-targeting signal of sCLU would be omitted. To verify this hypothesis, we used specifically designed primers, hCLU5Ј to exon I and hCLUN-term-rev to exon IV (see "Experimental Procedures"), to amplify the 5Ј ends of various splice variants of CLU cDNA derived from total RNA extracted from MCF-7 cells. The resulting RT-PCR fragments are shown in Fig. 2B. All visible fragments were excised from the agarose gel and sequenced. As expected, the 345-bp major fragment was identical in sequence to the 5Ј end of sCLU. Two minor fragments with sizes of 450 and 550 bp had additional exons in the 5Ј-untranslated region and, therefore, did not contain any alterations in the open reading frame. Sequencing of the 220-bp fragment demonstrated, as we predicted for nCLU mRNA, the lack of exon II and direct junction between exons I and III (see Fig. 2C, sequence).
By using the primer to the 3Ј end of the CLU open reading frame, we amplified full-length nCLU cDNA from MCF-7 cells. Sequence analyses confirmed the expected alternatively spliced full-length nCLU mRNA. No additional changes in the downstream exon composition of the nCLU cDNA were found. 3 Based on the data above, we concluded that human nCLU protein was translated from a separate, minor species of mRNA lacking exon II as a result of an alternative splicing event.
Identification and Analyses of Potential Functional Domains in nCLU-Analyses of the proposed primary aa composition of human and mouse CLU proteins, based upon respective CLU mRNA sequences, revealed two potential coiled-coil domains ( Fig. 1). One coiled-coil domain was located in the N terminus (aa residues 42-98 in human and 42-97 in mouse) of nCLU, and the other coiled-coil domain was located in the C-terminal portion of both human and mouse nCLU proteins (residues 319 -349 in human and 318 -366 in mouse). Both C-and Nterminal domains appeared to be highly conserved and corresponded to "myosin-like" domains I and IV, identified previously by Tsuruta et al. (24). Both N-and C-terminal coiled-coil domains of CLU were rich in leucines and charged aa residues. We compared the characteristic 4-3 repeats of known coiled-coil domain-containing proteins GCN4, c-Jun, c-Fos, and BZLF1 (34)(35)(36)(37)(38) Table I and Fig. 7). and mouse nCLU (Fig. 4A). The C-terminal coiled-coil domain of nCLU contains a 4-3 repeat of predominantly hydrophobic residues, much like the BZLF1 gene product of the Epstein-Barr virus (38). The C-terminal coiled-coil domain of nCLU did not, however, contain a classical leucine zipper heptad. The 4-3 pattern within the predicted N-terminal coiled-coil domain consisted of both hydrophobic and charged residues.
Sequence analyses also indicated that the N-terminal coiledcoil domain of nCLU contained an SV40 large T-antigen-like NLS, LEEAKKKK (residues 74 -81 in human and 73-80 in mouse) (22). In addition, we identified two other potential NLSs. One NLS-like sequence was found only in human nCLU (RRELDESLQVAERLTRK, residues 324 -340) and was similar to the bipartite NLS found in the Xenopus nucleoplasmin protein (39). This potential NLS was not found in rodent nCLU protein. The other potential NLS in nCLU, RKKHR (residues 442-446) in human and RRKSR (residues 442-446) in mouse, was located at the very C terminus of the nCLU protein. Because functional NLSs in nCLU have also not been elucidated, we examined the role of coiled-coil domains and NLS sites for the cell death function of the protein and for association with Ku70.
The C-terminal Coiled-coil Domain of nCLU Is Essential for Binding to Ku70 -A number of deletion mutants of human and mouse nCLU were constructed and tested for Ku70 binding using yeast two-hybrid analyses (Fig. 3A). Only those mutants containing the C-terminal coiled-coil domain, or a portion of the domain, were capable of interacting with the Ku70 protein.  3. A, the C-terminal coiled-coil domain of nCLU mediates Ku70 binding. The relative strength of nCLU-Ku70 interaction for both human and mouse pairs of proteins using yeast two-hybrid analyses are indicated as follows: ϩϩϩ, strong binding, colonies turned blue within 30 min of incubation with X-gal; ϩϩ, moderate binding, yeast colonies turned blue within 60 min; ϩ, weak interaction, colonies turned blue within 120 min; Ϫ, no interaction, no blue colonies within 120 min of incubation with X-gal. nCLU constructs marked with asterisks were tested using yeast two-hybrid assays with GFP moieties on their Nterminal ends. The fusion of GFP to these proteins does not affect their cell death functions (13). Coiled-coil domains are shown as slashed boxes. Numbers represent the respective aa residues. Experiments were performed three or more times. B, expression of GAL4-fused proteins in Y190 yeast. Representative Western blot analyses of the expression of mouse Ku70 protein (left panel) used in two-hybrid analyses shown in A and GAL4AD-and GAL4BD-fused mouse nCLU deletion mutants (right panel) in Y190 yeast. Plus signs indicate binding of the particular nCLU fragment to mouse Ku70 as described in A.
nal fragment (residues 381-448) of the protein did not interact with Ku70. None of these nCLU fragments contained the Cterminal coiled-coil domain. Two nCLU polypeptides encompassing aa residues 256 -344 and 345-380, respectively, contained only a portion of the C-terminal coiled-coil domain. Interestingly, these two nCLU fragments showed strong interaction with Ku70. Steady-state expression of GAL4-fused proteins, both Ku70-binding and non-binding polypeptides, are shown in Figs. 3B and 4B. These findings suggested that the C-terminal coiled-coil domain of nCLU was important for its interaction with Ku70. Furthermore, these data suggested that half, or less, of the coiled-coil domain was sufficient to bind Ku70 in yeast two-hybrid assays.
To study the role of specific residues in nCLU-Ku70 binding, we concentrated on the nCLU polypeptide fragment containing residues 345-380 (nCLU-(345-380)). This fragment displayed strong binding to Ku70 and contained only 22 residues of the predicted C-terminal coiled-coil domain. Because the Ku70binding domain of nCLU contained a putative 4-3 hydrophobic repeat motif, we determined whether specific leucine repeat residues within nCLU-(345-380) were necessary and sufficient to mediate Ku70-nCLU binding. The following point mutations were introduced: (a) L357R,L358Q (a double substitution of leucines 357 and 358 to arginine and glutamine, respectively); (b) E359K; (c) L361R; (d) D363H; (e) L371R; and (f) L374R, L357R,L358Q, E359K, L361R, and D363H were point mutations made inside the nCLU coiled-coil domain, whereas L371R and L374R were substitutions made outside the coiled-coil domain (Fig. 4B). Substitution of leucine residues inside the coiled-coil domain structure of nCLU-(345-380) completely abrogated Ku70 binding. Substitutions of leucine residues outside the coiled-coil domain decreased the association of each nCLU-(345-380) peptide with Ku70 but did not completely abrogate protein-protein interaction. Mutations of two non-leucine residues that were not part of the 4-3 repeat within the same region of the protein (E359K and D363H) did not affect nCLU-(345-380) binding to Ku70. These data strongly suggested that only residues that constituted the 4-3 hydrophobic repeat inside the C-terminal coiled-coil domain of nCLU mediated its interaction with Ku70. Leucines 371 and 374 immediately outside the coiled-coil domain of nCLU may, therefore, stabilize protein-protein interaction but do not appear to be essential for Ku70 binding.
Additional evidence suggesting the importance of the coiledcoil structure of nCLU and its interaction with Ku70 came from  hKu70 and mKu70, respectively). Every 1st, 3rd, and 4th residue is capitalized and boxed. Leucine residues in mCLU Ccoil that were targeted for the LR mutation are underlined. Leucine 343, which was targeted for the L343P mutation, is circled. Rectangles in the bottom indicate positions of ␣-helices 15, 16, and 16Ј as noted (45). The arrow indicates the position of the StuI site, which was used to create a deletion that resulted in a Ku70 mutant protein unable to bind to nCLU (14). Numbers refer to the positions of the first and last residues within the full-length proteins. B, identification of aa residues of nCLU that mediate Ku70 binding. The relative strength of various mutant forms of mouse nCLU associating with mouse Ku70 were tested using yeast two-hybrid analyses as described in Fig. 3 and "Experimental Procedures." Numbers represent the respective aa residues of mouse nCLU protein. Amino acid residues targeted for mutations are marked with asterisks. Residues to the left of the dotted line are inside the predicted C-terminal coiled-coil domain of mouse nCLU. Residues to the right side of the dotted line are outside the functional C-terminal coiled-coil domain. Relative strength of binding was scored as described in Fig. 3A, and experiments were performed three or more times. WT, wild type. C, GAL4-fused mutants of nCLU-(345-380) fragments are expressed in Y190 yeast. Western blot analyses demonstrate expression of mouse nCLU point mutants within aa 345-380 and the corresponding human aa 345-380 nCLU fragment in Y190 yeast. The relative strength of Ku70 binding was described in Fig. 3A, and all experiments were performed three or more times. D, LR mutants (nCLU-LR and NcoLR) of mouse nCLU and mouse nCLU-(256 -448) protein fragment and L343P mutant of mouse nCLU-(256 -448) fragment do not bind to Ku70. Yeast two-hybrid analyses were performed to determine binding of wild-type, L343P, and LR mutants of mouse nCLU and nCLU 256 -448 proteins to Ku70. Mouse nCLU fragment (aa 318 -365) contains C-terminal coiled-coil domain, and wild-type, LR mutant, and L343P (with a P) nCLU proteins are also shown. The relative strength of binding of mouse nCLU LR mutants to mouse Ku70 protein was indicated as described in Fig. 3 and "Experimental Procedures." The experiments were performed three or more times. E, steady-state levels of nCLU wild-type, LR, and L343P mutant proteins are equal when expressed in the Y190 yeasts. The expression and stability of full-length mouse nCLU wild-type and nCLU-LR mutant (top panel), as well as nCLU fragment aa 256 -448, NcoLR, and L343P mutants of the same fragment (lower panel) in Y190 cells were assessed by Western blot analyses using anti-GAL4BD antibody. Expression of proteins in all experiments described in B were confirmed by Western blotting. deletion analyses using human nCLU cDNA/protein and human Ku70 as bait in yeast two-hybrid analyses (Fig. 3A). Our limited analyses suggest that aa residues 326 -365 are minimally essential for the association of nCLU with Ku70. The human homolog portion of the mouse nCLU-(345-380) polypeptide fragment was unable to bind human Ku70 (Fig. 3A). This fragment of human nCLU contained leucines in positions 357, 358, and 361 that were essential for mouse nCLU binding to Ku70. However, the predicted C-terminal coiled-coil domain of human nCLU was shorter than in mouse nCLU. This shortened length of the C-terminal coiled-coil domain of human nCLU was apparently due to the presence of a tryptophan residue at position 350 in human nCLU that presumably interferes with ␣-helical secondary structure. Thus, human nCLU appears to have a more tightly compact Ku70 interaction domain than the comparable domain present in the mouse protein.
To investigate the role of the ␣-helical structure of the Cterminal coiled-coil domain in Ku70 binding, we introduced a single proline residue at position 343 (L343P mutant) in the context of the nCLU aa 256 -448 C-terminal half of the protein.
When tested by yeast two-hybrid analyses, this mutant did not show detectable binding to Ku70. Equal expression of wild-type and L343P mutant proteins in yeast was confirmed by Western blot analyses (Fig. 4E).
GST-Ku70 Pull-down Analyses Confirm Yeast Two-hybrid Results-To confirm the yeast two-hybrid data, we utilized GST-Ku70 in vitro binding (pull-down) assays. GST-Ku70 and GST proteins were expressed in bacteria and immobilized on glutathione-Sepharose beads as described under "Experimental Procedures" (Fig. 5B). 35 S-Labeled mouse nCLU polypeptide was used to determine optimal Ku70-binding conditions (Fig. 5A). Minimal nCLU binding to GST was detected in a temperature-insensitive fashion as described by Humphreys et al. (20). Significant interaction of nCLU with GST-Ku70 was detected when incubation temperatures were between 23 and 37°C, whereas nCLU binding to an equal amount of GST alone did not change with temperature (Fig. 5A). At 4°C, nCLU binding to GST-Ku70 was not significantly higher than background. These data indicated a temperature-dependent association of nCLU binding to Ku70, and that a temperature-sensitive conformational change was necessary for protein-protein association.
Point mutations in mouse nCLU fragment 345-380 at L357R,L358Q, E359K, L361R, D363H, and L371R were then examined using GST-Ku70 in vitro binding assays (Fig. 5C). The results were identical to those obtained by yeast twohybrid analyses, except that mutant E359K showed unexpectedly high binding to GST-Ku70 compared with wild-type nCLU, suggesting that some non-4-3 repeat residues inside the coiled-coil domain might also modulate Ku70 binding.
The Coiled-coil Domains of nCLU Bind Each Other-The longest Ku70-binding nCLU polypeptide (residues 84 -449), isolated from the human liver cDNA library by yeast twohybrid screening, excluded most of the N-terminal coiled-coil domain (14). However, all nCLU polypeptides that interacted with Ku70 contained the C-terminal coiled-coil domain. Interestingly, full-length nCLU cDNA containing C-and N-terminal coiled-coil domains was not isolated in the initial yeast twohybrid screen (14). The reasons for the absence of full-length nCLU were unclear. One possible explanation for the lack of isolation of full-length nCLU cDNA was that the N-terminal coiled-coil domain might bind to the C-terminal coiled-coil domain of nCLU and interfere with Ku70 binding (see below). Yeast two-hybrid analyses of a number of nCLU deletion mutants showed that whenever two coiled-coil domains of nCLU were present in the same molecule (e.g. polypeptides containing aa residues 18 -380 of mouse nCLU and human nCLU fragment containing aa residues 34 -449), binding of these fragments to Ku70 was notably lower compared with nCLU polypeptides containing a single C-terminal coiled-coil domain (i.e. see mouse nCLU mutants containing residues 256 -448, and the human nCLU polypeptide fragment containing residues 84 -449) (Fig. 3A). These data suggested to us that the N-terminal coiled-coil domain of nCLU might compete with Ku70 for binding to the C-terminal coiled-coil domain of nCLU.
Because the N-terminal coiled-coil domain of nCLU did not interact with Ku70 (Fig. 3A), we hypothesized that this Nterminal domain of nCLU might act as a suppressor of nCLU-Ku70 protein-protein interaction. Therefore, we examined the ability of coiled-coil domains of nCLU to bind each other. The C-terminal coiled-coil domain of human nCLU (i.e. a fragment containing aa residues 329 -449) was used as "bait." The Cterminal coiled-coil domain of nCLU was also able to interact with the N-terminal domain of the protein (aa 18 -97) and with itself (aa 256 -380) (Fig. 6A). However, neither the central portion of nCLU (aa 98 -317) nor a C-terminal fragment of nCLU missing the coiled-coil domain (aa 381-448) interacted with either the C-or N-terminal coiled-coil domains of nCLU. In addition, when tested as bait, the N-terminal coiled-coil domain interacted only with the C-terminal coiled-coil domain (Fig. 6B). The N-terminal coiled-coil domain did not interact with itself nor with central or C-terminal fragments of nCLU that lacked coiled-coil structures. These data strongly support the theory that the coiled-coil domains of nCLU may indeed be involved in either intra-or intermolecular interactions.
Intracellular Targeting of nCLU Domains and Induction of Apoptosis by Specific nCLU Proteins-Previous analyses of the nCLU aa sequence identified one NLS within its N-terminal coiled-coil domain (40) and another within the C-terminal fragment, but outside its coiled-coil domain (13). To investigate the role of these NLSs in the subcellular localization of the nCLU protein and in the cell death functions of the protein, we transfected MCF-7 breast cancer cells with GFP-fused constructs of wild-type and NLS mutants of the nCLU protein (used above). MCF-7 cells were chosen because we characterized previously (13) nCLU translocation in these cells following IR. Confocal microscopic analyses of the subcellular localization of various GFP-CLU fragments in MCF-7 cells 48 h post-transfection were then noted (Fig. 7), and the percentage of apoptosis Ϯ S.E. was quantified (Table I). Surprisingly, exogenously overexpressed full-length GFP-nCLU fusion protein was mostly cytosolic with minor expression noted in the nuclei of some transfected (green) cells (Fig. 7, B and C). In fact, the GFP-nCLUpositive population could be divided into two types. One population contained cells with normal nuclear morphology and very bright GFP-nCLU aggregates in the cytosol and perinuclear space (Fig. 7C), similar to untreated MCF-7 cells stained for endogenous nCLU protein (13). These aggregates did not co-localize with the ER, the Golgi apparatus, or mitochondria when co-stained with the following organelle-specific fluorescent markers: Texas Red-conjugated concanavalin A, Texas Red X-conjugated wheat germ agglutinin), and Mito-Tracker Red (Molecular Probes, Eugene, OR). Virtually no GFP-nCLU was detected in nuclei of cells within this subpopulation of transfectants, and virtually none of these cells showed classical apoptotic morphology. Another subpopulation of GFP-nCLU transfectants (ϳ15% of GFP-nCLU-positive cells as determined by visual screening) had diffuse distribution of GFP-nCLU with some apparent GFP-nCLU expression in nuclei (Fig. 7B). Most importantly, all cells that had visible nuclear GFP-nCLU expression had condensed chromatin, characteristic of apoptotic cells (41). Expression of GFP-Nterm protein in MCF-7 cells was diffusely distributed and located in both the cytosol and nucleus of transfected cells (Fig. 7D). In contrast, expression of the GFP-Center fusion protein within MCF-7 cells resulted in the formation of cytosolic aggregates (Fig. 7E), similar to the pattern of protein expression observed in the non-apoptotic subpopulation of the GFP-nCLU MCF-7 transfectants, as well as to the endogenous nCLU expression pattern in non-irradiated MCF-7 cells (13). Few, if any, GFP-Centertransfected MCF-7 cells exhibited apoptosis or any other abnormal morphology. These data strongly suggested that the apoptotic effects of nCLU are specific and that cytosolic localization and aggregate formation are attributed to protein regions separate from the domain responsible for nCLU-induced cell death. In contrast, the distribution of GFP-Ccoil (Fig. 7F) protein in both cytosol and nucleus resembled the subpopulation of apoptotic cells transfected with GFP-nCLU (Fig. 7B). As with GFP-nCLU, the presence of apoptotic nuclei following GFP-Ccoil transfection coincided with GFP-positive nuclear staining, and few apoptotic, non-transfected cells were noted within the transient cell population. No GFP-positive cells with large aggregates were detected in the GFP-Ccoil transfectants, as observed in the non-apoptotic GFP-positive subpopulation present within 48 h following GFP-nCLU transfection (Fig.  7C). Transfection of MCF-7 cells with GFP-End (that lacks a coiled-coil domain but has a C-terminal NLS) resulted in very few cells with large aggregates in their cytoplasm, and the majority of protein was expressed in the nuclei of GFP-positive MCF-7 cells (Fig. 7G). Unlike transfection with GFP-nCLU ( Fig. 7B and Table I), few cells with apoptotic morphology were detected among the GFP-End transfectants. A single L343P mutation within the C-terminal coiled-coil domain of GFP- FIG. 6. N-and C-terminal coiled-coil domains bind to each other in yeast two-hybrid system. The human nCLU deletion mutant containing aa 329 -449 (A) or the mouse nCLU deletion mutant containing aa 18 -97 (B) coiled-coil domains were used as bait to analyze their abilities to bind to various deletion mutants of mouse nCLU. Numbers represent corresponding amino acid residues. The potential coiled-coil regions are illustrated as shaded boxes. Relative strength of protein-protein binding (plus signs) was indicated as described in Fig.  3A, and all experiments were performed three or more times. nCLU (Fig. 7H, Table I) resulted in little or no detectable cell death and punctated cytoplasmic localization of the GFP-nCLU L343P protein. Transfection of MCF-7 cells with vector expressing GFP alone resulted in cells with diffusely distributed protein expression, mostly in the cytoplasm of transfected cells (Fig. 7A). Few apoptotic or other abnormal morphologies were noted in GFP-positive expressing MCF-7 cells after transfection with GFP alone, as found with GFP-Nterm, GFP-Center, and GFP-End (Fig. 7, D, E, and G and Table I).
nCLU Protein Has Two Functional NLS Motifs-It was proposed previously (13,22) that the nCLU protein contains two conserved potential NLSs among species. We define the Nterminal NLS as NLS1 and C-terminal as NLS2. To identify the functional NLS of nCLU, we introduced point mutations in the respective NLS sequences. We mutated NLS1, 77 KKKK 80 to 77 KAVK 80 in the context of mouse GFP-Nterm protein (NLS1mut), and NLS2, 442 RRKSR 446 to 442 VVKSR 446 in the context of mouse GFP-End protein (NLS2mut). The nCLU pro-tein fragments were chosen because they elicited no apoptotic responses, which could influence the trafficking analyses. We transfected MCF-7 cells with NLS1mut, NLS2mut, GFP-Nterm wild-type, GFP-End wild-type or GFP alone, and we determined the intracellular localization of the GFP-fused protein 48 h post-transfection (Fig. 8). The percentage of cells with cytoplasmic versus nuclear localization of GFP-fused proteins are summarized in Table II. Although GFP-alone control cells expressed GFP predominantly in the cytoplasm (Fig. 8A), we noticed uniform distribution of the wild-type GFP-Nterm in 57 Ϯ 5% cells with expression in both the cytoplasm and nucleus (Fig. 8B). Mutation in the N-terminal NLS led to an ϳ2.5-3-fold decrease in percentage of cells with detectable nuclear expression of GFP-Nterm NLS1mut (Fig. 8C, Table II). As in Fig. 7G, localization of GFP-End was predominantly nuclear (98 Ϯ 0.2% cells with nuclear expression, Fig. 8D). In contrast, GFP-End NLS2mut was detected only in the cytoplasm in 78 Ϯ 5% of GFP-positive cells (Fig. 8E). Based on these data, we conclude that both NLS1 and NLS2 of nCLU are functional; however, NLS2 is a relatively stronger site, resulting in greater differences in nuclear localization of GFP-End versus GFP-End NLS2mut. DISCUSSION The exact role(s) of alterations in CLU gene expression before and after cytotoxic stress remain(s) a mystery. Our data indicate that human cells synthesize CLU protein forms that lack ER-signaling peptides and are destined for functions in the cytoplasm and nucleus. This is in contrast to the sCLUs that are currently thought to exert cytoprotective effects on exposed cells as well as effects on non-damaged bystander cells as discussed previously (42,43). Collectively, our data suggest that nCLU contributes to cell death. Previous data (13,22) indicated the existence of a mature ϳ55-kDa nCLU protein.
The nCLU protein did not appear to be either glycosylated or cleaved at its ␣/␤-site, a site cleaved during maturation of the 60-kDa sCLU precursor protein. nCLU was induced by rela-  Table I for quantitation of apoptotic responses.

TABLE I Induction of apoptosis by exogenously expressed GFP-fused nCLU
protein fragments Log phase MCF-7 cells were transfected with either GFP-or GFP-nCLU-fused cDNAs and analyzed 48 h later by confocal microscopy. z-sections were then analyzed for protein localization (Fig. 7), and percentages of apoptotic cells among GFP-positive cells were quantified. Apoptotic cells were defined as cells with condensed and fragmented pycnotic nuclei (Fig. 7). Values are mean Ϯ S.E., representing three independent transient transfections. At least 200 GFP-positive cells were counted for each GFP construct per experiment.
tively high levels of cytotoxic stress (e.g. IR) in direct proportion to lethality, compared with sCLU protein increases that were very sensitive to low, nontoxic, and growth stimulatory doses of IR (13). Utilization of a second downstream AUG start site was proposed to result in the synthesis of an ϳ49-kDa precursor nCLU protein that resided in the cytoplasm of undamaged cells by confocal microscopy (13). nCLU greatly accumulated and translocated to the nuclei of cells 48 -72 h after IR exposure.
Overexpression of nCLU, even without IR treatment, caused cell death (i.e. apoptosis). The mechanism of synthesis of this nuclear protein from the second AUG translation start site, and how translation could bypass the much stronger first translation CCAUG start site present in full-length CLU mRNA, was not elucidated. Previous investigations did not address how the precursor ϳ49-kDa nCLU protein (pnCLU) might reside in the cytoplasm of undamaged cells in an apparent nonfunctional form (13), and then dramatically accumulate in the nuclei of damaged cells as an ϳ55-kDa mature nCLU death protein 48 -72 h after exposure (13). Finally, no previous study attempted to elucidate the functional domains of the nCLU protein and, in particular, to identify the portions of the protein required for cell death, Ku70 binding, and/or nuclear localization. In fact, the need for nuclear localization in the cell death function of the protein was not established. The data presented in this paper begin to address these processes and are consistent with the proposed hypothetical model of nCLU synthesis and cell death functions shown in Fig.  9, and discussed below. Our data show that a unique, shortened form of the nCLU mRNA is present at low levels (detected only by RT-PCR) in undamaged cells, and presumably increases after IR exposure in direct proportion to accumulation of fulllength CLU mRNA. The low abundance, truncated form of the nCLU mRNA is the result of alternative splicing that eliminates exon II, including the first strong AUG start site. Translation from this truncated nCLU mRNA results in the production of an ϳ49-kDa nCLU precursor protein (i.e. pnCLU) that is initiated from a second downstream AUG translational start site (located in exon III), and the protein produced (pnCLU) lacks the ER signal peptide. The ϳ49-kDa pnCLU appears to contain at least two functional NLSs that targeted the protein to the nucleus. Only overexpression of nCLU proteins that have C-terminal coiled-coil domains and a functional NLS caused cell death. The nCLU protein has two coiled-coil domains, one at its N terminus that is unable to bind Ku70, and a C-terminal coiled-coil domain that is uniquely able to associate with Ku70 and is minimally required for cell death. The N-terminal coiledcoil domain of nCLU can associate with the C-terminal coiled-  Table II for quantitation of the percentage of the cells with nuclear versus cytoplasmic localizations of GFP-fused proteins.

TABLE II
Mutations in NLS1 or NLS2 result in cytoplasmic localization of GFP-nCLU constructs Log phase MCF-7 cells were transfected with either GFP-or GFP-nCLU-fused cDNAs with or without mutations in NLS1 and NLS2, as described in the text, and analyzed 48 h later by confocal microscopy. z-sections were then analyzed for protein localization (Fig. 8), and the percentages of cells with cytoplasmic or cytoplasmic and nuclear localization of GFP-fused proteins were quantified. The numbers with standard errors are the result of three independent transient transfections. At least 200 GFP-positive cells were counted for each GFP construct per experiment.  9. A model for nCLU synthesis in human epithelial cancer cells. CLU promoter activity is stress-inducible (T. L. Criswell, D. A. Boothman, K. S. Leskov, and D. Y. Klokov, unpublished data). The precursor transcript of the CLU mRNA is spliced in various ways, producing the highly abundant sCLU mRNA (with exon II) and a much lower abundance pnCLU mRNA (without exon II). sCLU protein is targeted to the ER and secreted outside the cell. In contrast, the ϳ49-kDa pnCLU protein stays in the cytoplasm and can be modified following stress to produce mature 55-kDa nCLU protein when the radiation dose is equal or exceeds ϳ1 Gy (42). nCLU is imported into the nucleus where it binds Ku70 via its C-terminal coiled-coil domain. Overexpression of nCLU results in caspase-3-independent apoptosis (13). The C-terminal, Ku70-binding coiled-coil domain is required for the apoptotic function of nCLU. coil domain of the nCLU protein. Specific mutations in, or deletions of, the C-terminal coiled-coil domain of nCLU simultaneously abrogated Ku70 binding and prevented cell death, even though these mutant forms of the nCLU protein accumulated in the nuclei of transfected MCF-7 cells (Fig. 7). Our data further support a role of Ku70 binding in the cell death function of mature nCLU. We show that only nCLU proteins containing both the C-terminal coiled-coil domain and a functional NLS can kill MCF-7 cells after transfection. Alterations in, or deletions of, the C-terminal NLS or C-terminal coiled-coil domain abrogated cell death. However, attempts to date to show conclusively that Ku70 is necessary for cell death caused by nCLU have yielded inconclusive results due largely to the enhanced apoptotic indices of Ku70 Ϫ/Ϫ cells, observed with SV40-transformed murine embryonic fibroblasts (44). The generation of mutant nCLU proteins with alterations in the C-terminal coiled-coil domain or NLS should allow us to elucidate the role of nCLU-Ku70 binding in the cell death functions of nCLU in the future.
Synthesis of nCLU-Our data show that a low abundance nCLU mRNA is constitutively made in MCF-7 cells, in direct proportion (but in low abundance) to total CLU mRNA. Translation from this nCLU mRNA produces an ϳ49-kDa nCLU precursor protein (Fig. 9, pnCLU). Detection of this truncated nCLU mRNA in human cells argues against the more unlikely production of the ϳ49-kDa pnCLU through internal ribosome entry site-mediated translational initiation from a second inframe AUG sequence in full-length CLU mRNA as suggested (13,22). There is precedence for alterations in the subcellular localization of protein species via alternative splicing, as proposed for nCLU (Fig. 9). Expression of the rat renin protein undergoes alternative splicing, producing either secreted or intracellular forms of the protein from two unique mRNAs (42). We are investigating the mechanisms of alternative splicing of the CLU mRNA. In particular, we are examining the possible inducibility of nCLU-specific splicing events in response to IR or other cellular stresses. Previous data showed that mature ϳ55-kDa nCLU protein could be modestly induced (ϳ2-fold) following 1 Gy of IR, compared with more dramatic Ͼ50-fold increases in sCLU protein forms. We propose that this reflects the relative abundance of the two mRNA species. Because the truncated, alternatively spliced form of nCLU mRNA was apparent in undamaged cells, our data suggest that the ϳ49-kDa precursor nCLU protein is constitutively synthesized in cells and stored in a reservoir. The stored protein can then be activated by an as yet unidentified post-translational modification after certain threshold doses of IR (Ͼ1 Gy), resulting in an ϳ55-kDa mature nCLU protein (Fig. 9). Such a mechanism would explain expression of the protein without triggering cell death responses. Localization of pnCLU in the cytoplasm of non-treated cells by confocal microscopy, with no detectable levels in the nuclei of untreated MCF-7 cells (13), appears to support a possible mechanism of sequestration of the pnCLU protein to prevent cytotoxicity.
Functional Domains of the nCLU Protein-Our data strongly suggest that nCLU-Ku70 binding was facilitated by strong leucine zipper-like hydrophobic interactions, mediated by the C-terminal coiled-coil domain in nCLU and a comparable coiled-coil domain in Ku70. Deletion of more than one hydrophobic residue in the full-length C-terminal coiled-coil domain of nCLU was not sufficient to abrogate nCLU-Ku70 binding (Fig. 3A, deletion mutants 256 -344 and 345-380), apparently due to redundancy of hydrophobic residues required for association with this protein. Introduction of a single proline residue that was inconsistent with the ␣-helical coiled-coil domain (L343P mutation) completely abrogated Ku70 binding (Fig. 4D) and cell death (Fig. 7H). Thus, the coiled-coil domain is absolutely required for interaction with Ku70 and initiation of cell death. It is important to note that the overall length of the C-terminal coiled-coil domain of nCLU is almost twice as large as the dimerization domains of GCN4, c-Fos, c-Jun, and BZLF1. This could explain our finding that deletion of a significant portion of the C-terminal coiled-coil domain of nCLU did not affect protein-protein interaction.
Our previous studies showed that in-frame deletion of aa residues 492-583 of Ku70 abrogated nCLU-Ku70 binding. Recent structural data on the Ku complex (43) indicated that this deletion affected ␣-helices 15, 16, and 16Ј that also contained 4-3 patterns of hydrophobic amino acids (Fig. 4A). The ultimate role of the 4-3 repeated hydrophobic residues in mediating Ku70 binding was elucidated using point mutation analyses of the 345-380 minimal Ku70-interacting fragment of nCLU. Thus, the interaction between Ku70 and nCLU appears to occur through the C-terminal coiled-coil domain of nCLU, a region of the protein also required for cell death (only in conjunction with the C-terminal NLS of nCLU) in transfected cells.
At the present time, little is known about the regulation of nCLU induction and translocation. Translocation of pnCLU from the cytoplasm to the nucleus was accompanied by an electrophoretic mobility shift of the nCLU protein (13). In its inactive form, pnCLU was an ϳ49-kDa protein present in the cytoplasm of control, non-irradiated MCF-7 cells; little, if any, protein was found in the nuclei of control MCF-7 cells (13). After IR exposure, ϳ55-kDa nCLU protein levels increased and translocated from the cytoplasm into the nuclei of damaged cells (13). The IR-induced, post-translational modification processes that activate nCLU are not known, but we hypothesize that these modifications are involved in exposing either of the functional NLS signals, allowing subsequent translocation of pnCLU to the nucleus, as the mature nCLU death protein (Fig.  9). It is also possible that other proteins may be involved as specific inhibitors of pnCLU maturation in much the same way as IB blocks NF-B activation (44). The nature of the apparent post-translational modification of pnCLU is not known, but mass spectrometry studies are ongoing to elucidate these modification processes.
Comparison of the subcellular distribution of GFP-nCLU and GFP-Center suggests that cytosolic pnCLU may be actively prevented from entering the nucleus or actively excluded. The non-coiled central region of the pnCLU protein (amino acids 98 -317) appears to be responsible for the inhibition of nuclear import and the formation of the characteristic aggregates of nCLU in the cytoplasm; this region of the protein may bind to a certain cytosolic factor(s), may be required for oligomerization, or needed for folding of the pnCLU protein in the cytosol. The distinct nuclear localization of the End fragment of nCLU (aa 368 -448) and significantly lower nuclear expression of the GFP-End NLS2mut (Fig. 7G and Fig. 8, D and E; Table II) indicate that the C-terminal NLS was the functional nuclear import sequence for nCLU. N-terminal NLS1 may contribute to the nuclear translocation of nCLU, but our data on Fig. 8, B and C, and Table II demonstrate that NLS1 function is not as prominent as that of NLS2.
Because the C-terminal coiled-coil domain of mouse nCLU (aa 318 -367) did not have an apparent NLS, very little of the protein was detected in the nucleus of cells with apoptotic morphology. These data indicate that enough of this protein may have entered the nucleus due to forced, cytomegalovirusdirected overexpression to cause cell death. An alternative hypothesis is that the cell death function of nCLU may not require nuclear translocation. Our data suggest that dissocia-tion of pnCLU coiled-coil domains may allow modification, translocation, and cell death responses. We are currently exploring the mechanism of nCLU transport into the nucleus, as well as the role of its nuclear internalization in nCLU-mediated cell death responses, including the apparent triggering of apoptotic responses.
Function of nCLU as a Death Gene-The precise intracellular functions of nCLU, as well as the significance of its binding to Ku70 protein, remain to be determined. As demonstrated in our previous (13) and current studies, overexpression of GFP-nCLU induces apoptosis in an apparent C-terminal coiled-coiland NLS-dependent fashion. For example, overexpression of GFP-Ccoil, the Ku70-binding fragment of nCLU, resulted in short term apoptotic cytotoxicity ( Fig. 7 and Table I), as well as long term clonogenic cytotoxicity (13). Although overall apoptosis measurements did not exceed 25-30% at any time, the clonogenic survival of the GFP-nCLU-positive MCF-7 cells was ϳ10-fold lower than GFP-alone transfected MCF-7 cells. Our data suggest, but do not prove, that interaction and interference with Ku70-dependent end joining, even in the absence of overt IR damage, can cause significant cell death. Superficially, this appears to mimic the enhanced apoptosis and lowered plating efficiency of Ku70 Ϫ/Ϫ mouse embryonic fibroblasts (44). Elucidation of mechanisms by which CLU gene expression and nCLU production occur after DNA damage in tumor cells may augment various cancer therapies.