Cell death inhibiting RNA (CDIR) derived from a 3'-untranslated region binds AUF1 and heat shock protein 27.

Regulators of programmed cell death were previously identified using a technical knockout genetic screen. Among the elements that inhibited interferon-gamma-induced apoptosis of HeLa cells was a 441-nucleotide fragment derived from the 3'-untranslated region (UTR) of KIAA0425, a gene of unknown function. This fragment was termed cell death inhibiting RNA (CDIR). Deletion and mutation analyses of CDIR were employed to identify the features required for its anti-apoptotic activity. Single nucleotide alterations within either copy of the duplicated U-rich motif found in the CDIR sequence abolished the anti-apoptotic activity of CDIR and altered its in vitro association with a protein complex. Further analysis of the CDIR-binding complex indicated that it contained heat shock protein 27 (Hsp27) and the regulator of mRNA turnover AUF1 (heterogeneous nuclear ribonucleoprotein D). In addition, recombinant AUF1 bound directly to CDIR. Furthermore, expression of another AUF1-binding RNA element, derived from the 3'-UTR of c-myc, inhibited apoptosis. We also demonstrate that the level and the stability of p21(waf1/Cip1/sdi1) mRNA, a target of AUF1 with anti-apoptotic activity, were increased in CDIR-transfected cells. The level of mRNA and protein of Bcl-2, another anti-apoptotic gene, containing an AUF1 binding site in its 3'-UTR was also increased in CDIR-transfected cells. Our data suggest that AUF1 regulates apoptosis by altering mRNA turnover. We propose that CDIR inhibits apoptosis by acting as a competitive inhibitor of AUF1, preventing AUF1 from binding to its targets.

Programmed cell death is an evolutionarily conserved process that plays an essential role in development, tissue homeostasis, oncogenesis, and stress response (1,2). One of the approaches used to identify mediators of programmed cell death in mammalian cells is technical knock-out (TKO) 1 (3), which was first successfully applied for the identification of mediators of IFN-␥-induced apoptosis in HeLa cells (4). Several cDNA fragments were rescued through TKO selection and later shown to confer resistance to IFN-␥-induced programmed cell death (5). One of the uncharacterized elements was a 441nucleotide (nt) long U-rich RNA fragment derived from the 3Ј-UTR of the KIAA0425 transcript that we termed cell death inhibiting RNA (CDIR).
KIAA0425 (ZNF262) is a ubiquitously expressed gene that is predicted to encode a 682-amino acid protein containing five novel octacysteine (C8) zinc finger-related motifs known as MYM domains (6). Although other members of the MYM family were demonstrated to be associated with human diseases the role and functions of KIAA0425 are at present unknown (7)(8)(9).
Application of expression cloning previously revealed that expression of fragments of 3Ј-UTRs independently from a mRNA could affect cell behavior (10 -14). The mechanism by which expressing only a portion of a 3Ј-UTR controls cell behavior could be distinct in each case and may mimic the role the 3Ј-UTR plays in the context of an entire mRNA. It was demonstrated that regulatory elements positioned in cis in a 3Ј-UTR affect various aspects of mRNA fate including mRNA transport, localization, stability, and translation (15)(16)(17). Hence, it is possible that effects of a 3Ј-UTR in trans are mediated through the regulation of similar processes.
Multiple cis-acting determinants controlling mRNA fate and the trans-acting proteins that recognize these determinants have been described to date. Among these determinants are Uand AU-rich elements (ARE) that regulate mRNA stability and are positioned in the 3Ј-UTR of many mRNAs including oncogenes and cytokine mRNAs (18,19). Several proteins that bind to U-or AU-rich regions have been identified including mammalian members of the ELAV family such as Hel-N1, HuC, HuD, and HuR as well as the heterogeneous nuclear ribonucleoproteins: A0 A1, C, and D (AUF1) (20). HuR and AUF1 have been shown to directly alter mRNA stability of ARE containing mRNA in vivo (21)(22)(23).
AUF1 is expressed as four related isoforms of 37, 40, 42, and 45 kDa derived by alternative splicing (24 -26). Several lines of evidence demonstrate that binding of AUF1 to AREs often results in acceleration of mRNA decay (27)(28)(29). Association of recombinant AUF1 with AtU-rich RNA results in formation of oligomeric AUF1 complexes on the RNA (30). This suggests that multiple U-rich stretches that are common motifs in AREs could be bound in a cooperative manner by multiple AUF1 molecules. That may explain why multiple U-rich stretches in a 3Ј-UTR are often required for maximum effect on mRNA stability (18).
Another class of proteins associated with mRNA stability is the heat shock proteins. Hsp70 family members were shown to bind to U-rich motifs associated with regulation of mRNA stability (31,32). Hsp70 was also reported to be in the complex with AUF1 (33). Another heat shock protein, Hsp27, regulates the stability of cycloxygenase-2 mRNA through an AU-rich element. Interestingly, this same element forms a complex with AUF1 (29).
In this report we demonstrate that expression of CDIR renders cells resistant to IFN-␥-induced killing. Mutation and deletion analyses revealed that single nucleotide substitutions in U-rich motifs of CDIR abolished its anti-apoptotic activity and these substitutions also altered the specific binding of a protein complex(es) to CDIR as detected by RNA electrophoretic mobility shift assay (REMSA). Analysis of this protein complex(es) by heparin chromatography followed by affinity purification demonstrated that it contained both AUF1 and Hsp27. We further examined whether the binding of AUF1 to CDIR mediates the anti-apoptotic activity of the RNA fragment. We demonstrate that expression of another AUF1-binding fragment derived from c-myc protected cells from apoptosis. Moreover, the stability of mRNA encoding the AUF1 target p21 (34,35), was increased in CDIR-transfected cells. Taken together, our data suggest that the anti-apoptotic role of CDIR is mediated by sequestering AUF1 from cellular targets.

EXPERIMENTAL PROCEDURES
TKO Selection-A fragment of the 3Ј-UTR of the gene KIAA0425 (GenBank TM AB007885, nucleotides 4416 -4856) was identified during TKO selection (3). It was independently rescued three times. One of the fragments showed a C to T change at position 4510.
Transfections and Bioassays-HeLa cells were grown in Dulbecco's modified Eagle medium (Invitrogen) supplemented with 10% fetal calf serum (Hyclone), 4 mM glutamine, 100 units of penicillin/ml, and 0.1 mg of streptomycin/ml. HeLa cells (65 ϫ 10 3 /well in a 6-well plate) were transfected with 5 g of the indicated episomal vector using SuperFect reagent (Qiagen) according to the manufacturer's instructions. Fortyeight hours post-transfection, cells were treated with 200 units/ml hygromycin B (Calbiochem) for 2 weeks. For colony formation assays, 1 ϫ 10 5 cells were plated/10-cm plate in the presence of 200 units/ml IFN-␥ (PeproTech) and 200 g/ml hygromycin B. Cells were treated for 3 weeks and the medium was replaced weekly. Colonies were fixed with methanol and stained with crystal violet. For short term assays, cells were plated and treated similarly, and on the indicated day, the number of viable, trypan blue (Invitrogen) excluding cells were counted.
Fluorocytometric Analysis-To perform cell cycle analysis, HeLa cells were harvested by trypsinization, resuspended in phosphate-buffered saline containing 1% bovine serum albumin (BSA), harvested by low speed centrifugation, and fixed by 70% ethanol for 30 min at 4°C. The cell pellet was incubated with 40 g/ml propidium iodide (Sigma) and 100 units/ml RNase A (Sigma). Cells were sorted by FACS using a BD Biosciences FACSort and the CellQuest program. Analysis of data was carried out by the ModFit program (Verity Software House, Inc.).
Northern Blot Analysis-2.5 ϫ 10 6 cells were plated in a 15-cm plate in the absence or presence of 200 units/ml IFN-␥. Total RNA was isolated using Trizol reagent (Invitrogen) 72 h post-plating. RNA samples (20 g) were separated on 1.2% agarose/formaldehyde gels and transferred onto nylon filters (Amersham Biosciences). The probes were generated by labeling with a Random Priming kit (Amersham Biosciences), and filters were hybridized in ExpressHyb hybridization solution (Clontech). Hybridization signals were visualized by autoradiography using x-ray film (Fuji or Kodak BioMax MS) and quantified by PhosphorImager (Storm 860, Amersham Biosciences). Signals of the mRNA of interest were normalized to the value of glyceraldehyde-3phosphate dehydrogenase (GAPDH) as a loading control. Vector-specific probe was derived from the expression vector pTKO-1 by digestion with BglII and XbaI restriction enzymes. Bcl-2 probe was derived from pBABEPURO-Bcl-2, kindly provided by Dr. Scott Kogan (University of California, San Francisco, CA). For an RNA loading control the filters were hybridized with a GAPDH probe obtained by digestion of pBlue-Script SK-GAPDH with EcoRI (36).
Western Blotting-Western immunoblotting was performed using the ECL chemiluminescence kit (Amersham Biosciences). Proteins were extracted from HeLa cells in Nonidet P-40 lysis buffer (50 mM Tris-HCl, pH 7.4, 75 mM NaCl, 0.1% Nonidet P-40) supplemented with protease inhibitor mixture (Roche Molecular Diagnostics, GmbH, catalog number 1697498). Equal amounts of proteins were separated by SDS-PAGE and transferred onto nitrocellulose membranes. The nitrocellulose membranes were blocked in TBST with 5% nonfat milk, incubated with primary antibody for 1-2 h at room temperature followed by three washes in TBST. Afterward, the membranes were incubated in TBST containing secondary antibody conjugated to horseradish peroxidase for 45 min followed by four washes in TBST. Finally the membranes were developed using ECL (Amersham Biosciences). The membranes were then exposed to x-ray film. Band intensities were quantified by scanning the x-ray films and using the National Institutes of Health Image 1.55 program. The expression level of actin was used as an internal control for loading efficiency. The following primary antibodies were used: anti-Hsp27 (StressGen, SPA-803), anti-AUF1 (24), anti-Waf1 (Calbiochem, OP64), anti-Bcl-2 (DAKO, M0887), and anti-␤actin (Sigma, A5441).
Plasmids-Fragments of CDIR were generated by polymerase chain reaction. The end points of the fragments were as follows: fragment 1-(4416 -4674), fragment 2-(4567-4768), and fragment 3-(4637-4856). To create CDIR mutants, single nucleotide substitutions were introduced by PCR: T to G substitution at position 4493 (5Ј CDIR) and T to C substitution at position 4728 (3Ј CDIR). All fragments were cloned into pTKO-1 and verified by sequencing. The AUF1 binding c-myc fragment (c-myc ARE) was derived from pMycSD3 (27) by digestion with EcoRI and HindIII and recloned into pBlueScript KSII(ϩ). The resulting construct was digested with NotI and XhoI restriction enzymes, and the c-Myc ARE was cloned into a pTKO-1-derived vector between NotI/XhoI.
Analysis of RNA Secondary Structure-RNA secondary structure analysis was performed by using the MFOLD program version 3.1 by Zuker and Turner, available at bioinfo.math.rpi.edu/ϳmfold/rna/ form1.cgi (37). The folding temperature was fixed at 37°C.
REMSA-2.5 ϫ 10 6 HeLa cells were plated in a 15-cm plate in the absence or presence of 200 units/ml IFN-␥. 72 h post-plating, cells were washed in ice-cold phosphate-buffered saline (lacking CaCl 2 and MgCl 2 ), harvested, and resuspended in 30 l of Nonidet P-40 lysis buffer containing standard amounts of protease inhibitors aprotinin, pepstatin, and pefablock. Gel mobility assay was performed as described (38). [ 32 P]CTP-labeled CDIR was obtained by in vitro transcription using the Riboprobe system T7 (Promega, P1440) with [␣-32 P]CTP (Amersham Biosciences). Proteins collected in Nonidet P-40 lysis buffer or as otherwise indicated (15 g) were incubated with ϳ2000 cpm of 32 P-labeled RNA in 30 l of buffer containing 40 mM Hepes (pH 7.2), 20 mM ammonium sulfate, 15 mM potassium acetate, 10% glycerol (v/v), 75 g/ml yeast tRNA, 0.5 mg/ml BSA, and 1 unit/ml of RNase inhibitor (Roche Molecular Biochemicals, catalog number 799017) for 40 min at 4°C. For competition experiments samples were preincubated with unlabeled RNA for 30 min. The reactions were resolved in a gel containing 4% polyacrylamide (29:1, acrylamide/bis-polyacrylamide), 2% glycerol (v/v), 0.5 ϫ Tris borate/EDTA using a loading buffer containing 8% glycerol (v/v) and bromphenol blue. For analysis of specific RNAbinding activity from the heparin column fractions, 2-l aliquots were used per reaction as described above. RNA-binding activity of purified His 6 -AUF1 and His 6 -Hsp27 were assayed as described above. The amount of purified His 6 -AUF1 protein used in the binding reaction is indicated in the figure legends. For the generation of competitor RNAs, CDIR, 5Ј CDIR, or 3Ј CDIR were cloned into pcDNA3 (Invitrogen), linearized by digestion with BamHI, and used as template for in vitro RNA synthesis using the Riboprobe system T7 (Promega, P1440) according to the manufacturer's instructions. The nonspecific competitor corresponding to the EcoRV-AvaI fragment of firefly luciferase was cloned into pBluescript vector downstream of the T7 promoter (39), linearized with EcoRI, and used for in vitro transcription. The following antibodies were used for supershift analysis: anti-mdm2 (K-20) (Santa Cruz, SC-1022), anti-Hsp27 (StressGen, SPA-803), and anti-AUF1 (24).
Heparin Column-Extracts from 4 ϫ 10 9 HeLa cells were prepared as described (40) by using 150 l of high salt extraction buffer per 30 ϫ 10 6 cells (400 mM KCl, 20 mM Tris-HCl, pH 8, 2 mM dithiothreitol, 20% (v/v) glycerol and complete protease inhibitor (Roche Molecular Biochemicals). Cell lysate was adjusted to 100 mM KCl by dilution in ice-cold 20 mM Hepes (pH 7.9), 1 mM EDTA, 10% (v/v) glycerol, 0.2 mM dithiothreitol. Complete protease inhibitor (Roche Molecular Diagnostics) was added to the cell lysate. Cell lysate was loaded onto a 1-ml HiTrap-heparin column (Amersham Biosciences) equilibrated with pu-rification buffer (20 mM Hepes, pH 7.9, 1 mM EDTA, 10% (v/v) glycerol, 0.2 mM dithiothreitol) containing 100 mM KCl. The column was washed with 10 volumes of purification buffer containing 100 mM KCl. Proteins were eluted from the column with purification buffer containing increasing amounts of KCl ranging from 0.25 to 0.75 M in the presence of protease inhibitors. 0.5-ml fractions were collected and analyzed by REMSA or SDS-PAGE.
RNA Affinity Column-Affinity column chromatography was performed using the protocol designed for in vitro binding (38). Wild-type CDIR (200 ng) was incubated with 1 g of the 3Ј biotinylated DNA oligonucleotide, 5Ј-TGATGTGAGGGGCCCATGGCAAAAGCTTCT-3Ј, in the presence of 0.1 ml of annealing buffer (10 mM Tris, 10 mM MgSO 4 , 100 mM NaCl). The mixture was boiled, slowly cooled to room temperature and added to 3 ml of streptavidin-agarose beads (Invitrogen, catalog number 15942-014) prewashed with 10 volumes of annealing buffer. The mixture was rotated at 4°C for 2 h and washed with 10 volumes of equilibration buffer (40 mM Hepes, pH 7.2, 20 mM ammonium sulfate, 15 mM potassium acetate, 10 mM MgSO 4 , 10% glycerol (v/v), 0.5 mg/ml BSA). The beads were transferred into a Talon 2-ml disposable gravity column (Clontech). The fraction eluted from the heparin column with 0.45 M KCl was diluted with equilibration buffer (without BSA) containing 75 g/ml yeast tRNA and a standard amount of protease inhibitors to a final concentration of 15 mM KCl and loaded three times on the affinity column. The column was washed with 15 volumes of equilibration buffer (without BSA). The sample was eluted at 70°C for 2 min with 1 ml of elution buffer (10 mM Tris, 1 mM EDTA). A sample from the eluate was analyzed by SDS-PAGE followed by silver staining.
Purification of His 6 -AUF1 Proteins-Escherichia coli DH5␣ cells were transformed with pBAD/HisB-p37 AUF1 (41). One colony was grown overnight in 3 ml of LB medium containing 10 mM MgCl 2 and 100 g/ml ampicillin. The overnight culture was inoculated into 250 ml of LB containing 10 mM MgCl 2 and 100 g/ml ampicillin. When the culture reached A 600 Ϸ 0.5, arabinose was added to a final concentration of 0.002% followed by an additional 3-h incubation. Cells were collected by centrifugation and the cell pellet was resuspended in 10 ml of sonication buffer (20 mM Tris, pH 8, 100 mM NaCl) containing complete protease inhibitor (Roche Molecular Diagnostics GmbH, catalog number 1697498) and sonicated using a microtip in Sonabox (Artek). The lysate was centrifuged for 20 min at 10,000 ϫ g at 4°C, and the supernatant was loaded on a Talon metal affinity column (Clontech) that had been pre-equilibrated with sonication buffer. The column was washed with 30 column volumes of sonication buffer followed by additional washing with 5 volumes of sonication buffer containing 5 mM imidazole. The protein was eluted in 1 ml of sonication buffer containing 10 mM imidazole.
p21 mRNA Stability Study-2.5 ϫ 10 6 cells were plated in a 15-cm plate and treated with 10 g/ml actinomycin D. At the indicated time points, total RNA was extracted as described earlier and analyzed by Northern blot analysis. p21 probe was kindly provided by Dr. Peter Chumakov, University of Illinois, Chicago. The intensities of the bands were quantified by PhosphorImager (Amersham Biosciences) and normalized to GAPDH. The normalized values were used for the determination of the half-life of p21 mRNA. (3). The 441-nt long CDIR, derived from the 3Ј-UTR of KIAA0425 (zinc finger protein 262, Znf 262, GenBank TM accession AB007885) is presented in Fig. 1A. KIAA0425 mRNA is 6108 nt long with a relatively long 3Ј-UTR of 2261 bases. CDIR is uridylate-rich (36%; 157/441) and contains U-rich clusters reminiscent of AREs.

CDIR Expression Protects Cells from IFN-␥-induced Apoptosis-CDIR was isolated during TKO selection for mediators of IFN-␥-induced apoptosis in HeLa cells
To investigate the role of CDIR in inhibition of IFN-␥-induced killing we generated two populations of HeLa cells stably transfected with either episomal expression vector (pTKO-1) or pTKO-1 containing CDIR. The transfected pools were treated with IFN-␥ for 3 weeks and the long term survival of the cells was determined by colony formation assay. Whereas the control population was efficiently killed by IFN-␥, CDIR-transfected cells survived and proliferate in the presence of IFN-␥, eventually forming visible colonies (Fig. 1B). The difference in response to IFN-␥ between control vector and CDIR-trans-fected cells was detected as early as day 3 post-IFN-␥ treatment. FACS analysis demonstrates that on the third day of IFN-␥ treatment, 52% of control vector-transfected cells were in the sub-G 1 (apoptotic) fraction whereas only 25.5% of CDIRtransfected cells underwent apoptosis (Fig. 1C). Northern blot analysis confirmed the expression of CDIR in CDIR-transfected cells (Fig. 1D). Thus, expression of CDIR inhibits cell killing in long term (2-3 weeks) and in short term assays (3-5 days).
The Expression Level of CDIR Correlates with Cell Resistance to Killing-Because CDIR-transfected cells are a pool of cells expressing various levels of vector-transcribed CDIR, we wished to determine whether there was a correlation between the level of CDIR expression and resistance to killing. For this purpose, we compared the level of CDIR in parallel cultures derived from the same pool of CDIR transfectants. The first culture was propagated in the absence of IFN-␥ treatment, whereas the second culture was treated with IFN-␥ for 2 weeks resulting in selection of cells resistant to killing. Both cultures were further propagated for 2 weeks after release from IFN-␥ to eliminate any direct contribution of IFN-␥ treatment to CDIR expression. Northern blot analysis indicates that CDIR transfectants expressing higher levels of CDIR preferentially survived IFN-␥ treatment. As presented in Fig. 1E, IFN-␥treated cultures of CDIR transfectants (survivors) express 2.5fold higher levels of CDIR than the untreated CDIR-transfected population (original) as quantified by PhosphorImager compared with GAPDH expression. The CDIR survivors were also more resistant to IFN-␥-induced killing than the original population of CDIR-transfected cells (Fig. 1F). Thus, expression of CDIR is sufficient to render HeLa cells resistant to IFN-␥-induced killing, and cells expressing higher levels of CDIR are more resistant to killing.
Structure and Function Analyses of CDIR-To identify the region of CDIR required for its anti-apoptotic activity, we performed structure and function analyses of this RNA fragment. The following features of CDIR were considered to begin a structure-function study of CDIR: the U-rich clusters, a small putative open reading frame (70 amino acids) ( Fig. 2A), and a predicted stable secondary structure (Fig. 3C).
To test whether the U-rich subdomains of CDIR or the open reading frame are sufficient for protection of HeLa cells from IFN-␥ killing, three overlapping fragments of CDIR were generated. As depicted in Fig. 2A each of these fragments harbors a different set of U-rich elements. These fragments of CDIR were generated and cloned into the pTKO-1 vector and used to generate stable pools of HeLa transfectants. Only expression of full-length CDIR conferred resistance to killing, as shown by cell counts taken 3 days post-IFN-␥ treatment (Fig. 2B). This effect was not because of low expression of RNA derived from fragments 1-3. Northern blot analysis utilizing a vector-specific probe that detects all mRNAs that are transcribed from the vector indicated high expression of CDIR and all fragments (Fig. 2C). Thus, neither the hypothetical open reading frame nor the presence of a subset of U-rich stretches is sufficient to render cells resistant to IFN-␥-induced killing.
Because none of the fragments could individually inhibit cell killing, we pursued the possibility that elements present on different fragments are required for anti-apoptotic activity of CDIR. The schematic representation of CDIR with the position of 14 U-rich stretches containing 3 or more Us is shown in Fig. 2A. Regulatory U-rich motifs found in the 3Ј-UTRs of many mRNAs often contain repeated elements that are important for activity. CDIR sequence was examined for a U-rich element that was duplicated in the full-length CDIR but was present only as a single copy in the overlapping fragments. We detected two elements in the CDIR containing motifs that were re- The blot was hybridized to a CDIR-specific probe followed by hybridization to GAPDH. The endogenous transcript containing CDIR is not shown. The experiment was repeated four times, a representative experiment is shown. E, Northern blot analysis peated. The first element, UUUCAUUUUC (nt 4486 -4495), and the second element, AUUUUCAUUUU (nt 4725-4735), correspond to Us2 ϩ Us3 and Us6 ϩ Us7, respectively (as presented in Fig. 2A). Two overlapping motifs were detected inside these sequence elements, AUUUUC (marked in bold) and UUUCAUUUUC (underlined). As depicted in Fig. 2A, the first element (nt 4486 -4495) was present only in fragment 1, whereas the second element (nt 4728 -4735) was present in fragments 2 and 3. Because a single copy of either motif was not sufficient to render cells resistant to apoptosis (Fig. 2B), we tested whether the intact sequence of both elements is required for protection from killing. Two variants of CDIR were generated that had single base substitutions within either copy of these elements such that both motifs (AUUUUC and UUU-CAUUUU) were altered. The first variant, termed 5Ј CDIR, Three days later, cell viability was determined by trypan blue exclusion. Presented is the average of data from two experiments, each including two independent pools of transfectants. C, Northern blot analysis of expression from empty vector or vectors carrying fragments 1, 2, 3, or full-length CDIR in transfected HeLa cells. The blot was hybridized with a vector-specific probe and re-hybridized with GAPDH as a loading control. Note that because of the low resolution of the gel the mobility of CDIR and the fragments appear similar. The experiment was repeated twice with similar results.
of CDIR and CDIR survivors. RNA was extracted from a pool of cells transfected with CDIR and treated with hygromycin B (CDIR original) or after treatment with IFN-␥ and hygromycin B for 2 weeks (CDIR survivor). The experiment was repeated three times, a representative experiment is shown. F, sensitivity of CDIR expressing cells and CDIR survivors to IFN-␥. CDIR survivors, cells transfected with empty vector as well as the original population of CDIR-transfected cells were treated with IFN-␥ and hygromycin B for 5 days. Cell viability was determined by the trypan blue exclusion method. Presented is the average of data from two experiments, each including two independent pools of transfectants.
were cloned into the pTKO-1 vector and transfected into HeLa cells. Transfected pools were treated with IFN-␥ for 3 days. Whereas the cells transfected with wild type CDIR were resistant to killing, the cells transfected with vector only, 5Ј CDIR, or 3Ј CDIR were sensitive to induction of apoptosis (Fig. 3A). The inability of 5Ј CDIR and 3Ј CDIR to protect cells from killing was not because of low level of expression, because CDIR, 5Ј CDIR, and 3Ј CDIR were all expressed at similar levels (Fig.  3B). In addition, a single nucleotide alteration (change of C to T at position 4510) that resides outside of any U-rich motif had no effect on the anti-apoptotic activity of CDIR. This fragment was as efficient as wild type CDIR in protecting cells from IFN-␥-induced killing as determined by a long term assay (data not shown). Thus, the single nucleotide changes in duplicated U-rich motifs abrogated the ability of wild type CDIR to inhibit cell killing.
Computer analysis was employed to predict secondary structures of the wild type and the mutants of CDIR (37) (Fig. 3C). Whereas 5Ј CDIR folds in a structure distinct from the wild type, the 3Ј CDIR is predicted to have the same structure and folding energy as wild type CDIR (Fig. 3C). Because only the expression of wild type CDIR and not 3Ј CDIR prevents killing, it is unlikely that the secondary structure of CDIR alone is sufficient for the death inhibiting activity.
Identification of CDIR-binding Factors-The structure/function study established that the single nucleotides that are altered in 5Ј CDIR and 3Ј CDIR are indispensable for the anti-apoptotic activity of CDIR and that neither the putative open reading frame nor the formation of predicted secondary structures was sufficient for death inhibiting activity. Thus, we propose that cellular factors mediate the anti-apoptotic activity of wild type CDIR by interacting with wild type but not with 5Ј or 3Ј CDIRs.
In vitro RNA-protein binding assay was used to detect protein complexes that specifically bound to CDIR but not to 5Ј or 3Ј CDIR. 32 P-Labeled riboprobe corresponding to CDIR was synthesized in vitro and incubated with cellular protein extracts prepared from HeLa cells. The assay was performed in the presence of excess yeast tRNA (75 g/ml) to ensure adequate specificity and affinity of binding. RNA-protein complex formation was detected by REMSA. The migration of the transcript was retarded (Band shift) when incubated with HeLa cell

FIG. 3. Single nucleotide substitutions in U-rich motifs of CDIR inhibit its anti-apoptotic function.
A, sensitivity of 5Ј CDIR and 3Ј CDIR expressing cells to IFN-␥. HeLa cells transfected with vector, CDIR, 5Ј CDIR, or 3Ј CDIR were treated with IFN-␥ and hygromycin B. Three days later cell viability was determined by trypan blue exclusion. Presented is the average of data from two experiments, each including two independent pools of transfectants. B, Northern blot analysis of expression of 5Ј CDIR, 3Ј CDIR, CDIR, or empty vector in transfected cells. The analysis was repeated twice and a representative result is shown. The analysis was performed as described in Fig. 2C. C, shown is the effect of the mutations on the secondary structure of CDIR as predicted by the MFOLD computer program (37). Arrows indicate the locations of the mutated bases in 5Ј CDIR and 3Ј CDIR. Also shown is stability of the structures in kcal/mol. lysate compared with the control reaction that lacked the protein extract (Free probe) (Fig. 4A). This indicates that CDIRbinding activity was present in the HeLa cell extract. Extract prepared from untreated cells (Ϫ) or cells treated with IFN-␥ (ϩ) displayed similar binding activity (Fig. 4A, compare lanes 2  and 3). To analyze the specificity of the RNA-protein complex, a competition experiment was performed. The RNA-protein binding assay was carried out in the presence of a 100-fold excess of unlabeled CDIR, acting as a specific competitor. CDIR competition decreased the RNA-protein complex formation and restored the rapid migration of the probe. The competition with unlabeled CDIR was observed with extracts prepared from untreated cells (Ϫ) and from cells treated with IFN-␥ (ϩ) (Fig. 4A, lanes 4 and 5). Because there was no difference in CDIR-binding activity between extract from treated and untreated cells, further analysis was performed with extracts from untreated cells. The binding activity could be competed by preincubation with CDIR but not with a nonspecific competitor (luciferase RNA) (Fig. 4B). The majority of the binding activity in cell lysates was competed by a 100-fold excess of either 5Ј or 3Ј CDIR (Fig. 4B). However, some of the binding activity could not be completely competed by either 5Ј CDIR or 3Ј CDIR (marked by asterisk in Fig. 4B).
To identify components of CDIR-binding complex(es) that were not efficiently competed by the biologically inactive 5Ј CDIR or 3Ј CDIR, HeLa cell extract was fractionated on a heparin column by elution with increasing concentrations of KCl and the eluted fractions were analyzed by REMSA. 32 P-Labeled riboprobe corresponding to the CDIR was incubated with the protein fractions eluted at different salt concentrations. The specificity of CDIR binding in these fractions was tested in a competition assay with a 100-fold excess of CDIR (specific competitor), luciferase RNA (nonspecific competitor), 5Ј CDIR, or 3Ј CDIR. The results of analysis of representative fractions are shown in Fig. 5A.
Proteins eluted in the 0.25 M KCl fraction form a clearly recognizable band shift indicative of a strong CDIR-binding activity (Fig. 5A, lane 1). This activity is equally competed by the specific competitor (CDIR) as well as by 5Ј CDIR or 3Ј CDIR (Fig. 5A, lanes 2, 4, and 5). The protein-binding activity was not competed by a nonspecific competitor (luciferase) (Fig. 5A, lane  3). Thus, the binding activity in this fraction is not specific for the biologically active RNA (CDIR) as compared with the biologically inactive RNA (5Ј CDIR or 3Ј CDIR). This fraction was not further analyzed.
Proteins eluted in the 0.7 M KCl fraction do not form distinct complexes and even without competitor RNA there is only a low level of RNA-protein complex formation (Fig. 5A, lane 11). This is despite the fact that the estimated protein concentration in the 0.7 M KCl fraction was higher than in the fractions eluted at 0.25 or 0.5 M KCl. This fraction was not further characterized.
Proteins eluted in the 0.5 M KCl fraction possess a strong CDIR-binding activity that can be competed by specific (CDIR) but not by nonspecific competitor (luciferase) (Fig. 5A, lanes  6 -8). CDIR-binding activity was only partially competed by 5Ј CDIR and 3Ј CDIR (Fig. 5A, lanes 9 and 10). This partial competition likely indicates that this fraction contains two CDIR-binding activities: a set of proteins that bind CDIR as efficiently as 5Ј CDIR and 3Ј CDIR and another set of factors that bind only wild type CDIR. To purify wild type CDIRbinding activity we analyzed additional fractions.
Proteins eluted in the 0.45 M KCl fraction preferentially bind wild type CDIR. CDIR-binding activity observed in this frac- tion was competed by a 100-fold excess of CDIR but not by a 100-fold excess of 5Ј CDIR (Fig. 5B, lanes 1-3). A 100-fold excess of 3Ј CDIR only partially competed, indicating lower affinity binding (Fig. 5B, lane 4). Thus, the 0.45 M KCl fraction contains proteins whose preferential binding to CDIR correlate with its anti-apoptotic activity. Therefore, CDIR-binding complex(es) from the 0.45 M KCl fraction was further purified on a CDIR affinity column. Proteins eluted in the 0.45 M KCl fraction were applied to an RNA affinity column and after several consecutive washes the CDIR-binding proteins were eluted from the column. Proteins eluted from the affinity column, as well as proteins present in the last wash, were analyzed by PAGE followed by silver staining. A prominent band of 25 kDa and a diffuse band(s) of 41-46 kDa were detected only in the eluted fraction (Fig. 5C).
Hsp27 Is a Component of CDIR-binding Complex-A candi-date for the 25-kDa band was the Hsp27 because it migrates as a 25-kDa protein on SDS-PAGE and it has been reported to modulate mRNA fate through the U-rich element in the 3Ј-UTR (29). Therefore, we first tested and confirmed the presence of Hsp27 in the 0.45 M KCl fraction by Western blot analysis (Fig. 6A). Moreover, Hsp27 is part of a specific CDIR complex formed by proteins eluted in the 0.45 M fraction as demonstrated by the formation of the RNA-protein-antibody complex (supershift) upon the addition of different concentrations of anti-Hsp27 antibodies in REMSA (Fig. 6B, lanes 3 and 4). This effect is specific because irrelevant antibodies such as anti-mdm2 antibodies (Fig. 7B, lane 7) and anti-PKR and anti-BAX (data not shown) do not alter the mobility of the CDIR-protein complex. Taken together, our data suggest that Hsp27 is a component of the CDIR-binding complex. However, we failed to detect binding of recombinant Hsp27 to CDIR (data not shown).  KCl (lanes 1-5), 0.5 M KCl (lanes 6 -10), and 0.7 M KCl fractions (lanes [11][12][13][14][15] were incubated with 32 P-labeled CDIR and analyzed by REMSA. The RNA-binding activity specific for U-rich motifs of CDIR was detected by competition experiments, with the following competitors: CDIR (lanes 2,  7, and 12), nonspecific competitor luciferase (lanes 3, 8, and 13), 5Ј CDIR (lanes 4, 9, and 14), or 3Ј CDIR (lanes 5, 10, and  15). The experiment was repeated twice with similar results. B, proteins in the 0.45 M KCl fraction preferentially bind to CDIR. Labeled CDIR was incubated with proteins eluted in the 0.45 M fraction (lanes 1-4). The specific binding was detected by competition with 100-fold excess CDIR (lane 2), 5Ј CDIR (lane 3), or 3Ј CDIR (lane 4) and analyzed as described above. C, analysis of the proteins eluted from the CDIR affinity column. Eluted proteins (eluate) were separated on a SDS-polyacrylamide gel (4 -15%) and visualized by silver stain. The last wash of the column was used as a control sample for nonspecific binding (wash). Molecular weights of polypeptide markers are indicated on the right. Two sets of proteins (25 and 40 -45 kDa) were specifically eluted from the affinity column, whereas nonspecific 60-and 30-kDa proteins were present in both wash and eluted samples. Indicated by asterisk (in the eluate sample) are streptavidin residues eluted from the agarose beads.
It is therefore possible that an additional modification may be required for direct binding of Hsp27 to CDIR and/or Hsp27 may be tethered to the binding complex through protein-protein interactions.
AUF1 Protein Is a Part of CDIR-binding Complex and Binds to CDIR Directly-A candidate for the 41-46-kDa set of proteins was the AϩU-rich binding protein AUF1. AUF1 binds to U-rich sequences often found in 3Ј-UTRs and regulates mRNA turnover.
Western analysis of different fractions indicates that AUF1 is present in fractions 0.45 and 0.5 M KCl that preferentially bind to wild type CDIR (Fig. 7A, lanes 3 and 4). All AUF1 isoforms are enriched in these fractions. The presence of AUF1 in the CDIR-binding complex was demonstrated by a supershift analysis. Addition of anti-AUF1 antibody resulted in a RNA-protein-antibody complex (supershift) detected in REMSA (Fig. 7B, lanes 3 and 4). Alteration in complex mobility was observed in the presence of two different concentrations of anti-AUF1 antibodies. The supershift effect was specific because irrelevant antibodies such as anti-mdm2 antibodies (Fig. 7B, lane 7) do not alter the mobility of the CDIR-protein complex. Moreover, the ability of recombinant His 6 -tagged AUF1 to bind directly to CDIR was detected by REMSA (Fig.  7C). Addition of increasing amounts of recombinant AUF1 results in slower mobility of the complex. This likely reflects the oligomerization of AUF1 that was reported to occur upon binding of AUF1 to RNA targets (30). Thus, AUF1 protein forms a complex with CDIR. The cellular levels of AUF1 protein were not affected by the overexpression of CDIR in untreated or IFN-␥-treated cells (Fig. 7D). Therefore, complex formation between AUF1 and CDIR might result in sequestration of this protein from the cellular pool affecting the accessibility of AUF1 to other targets.
AUF1-binding Fragment of c-myc Protects HeLa Cells from Apoptosis-We propose that the anti-apoptotic activity of CDIR is mediated by AUF1 binding. Because AUF1 binds to other RNA targets, we tested the ability of another AUF1-binding RNA to inhibit IFN-␥-induced cell killing. For this purpose, a fragment of the 3Ј-UTR of c-myc (c-myc ARE), which forms a complex with AUF1 (27), was cloned into pTKO-1 and transfected into HeLa cells. Cells transfected with control vector, CDIR, or c-myc ARE were assayed for their sensitivity to IFN-␥-induced killing. Expression of either CDIR or c-myc ARE inhibits the induction of apoptosis (Fig. 8A). Thus, expression of either AUF1 targets had an anti-apoptotic effect. Because the cellular levels of AUF1 protein were not affected (Fig. 7D), we propose that binding of AUF1 to CDIR results in alteration of AUF1 activity in the cell such as regulation of mRNA turnover.
p21 mRNA Is stabilized in CDIR-transfected Cells-At least 130 AUF1 targets were identified by utilizing AUF1 affinity chromatography (34). Because expression of CDIR protects cells from IFN-␥-induced apoptosis, we decided to determine the abundance and turnover of an AUF1 target that was linked to the IFN-␥ pathway. Cyclin-dependent kinase inhibitor p21 mRNA was examined because it binds AUF1 and protects cells from IFN-␥-induced apoptosis (34,35,42,43). First, the abundance of p21 was examined in control, CDIR-, 5Ј CDIR-, or 3Ј CDIR-transfected cells. Northern blot analysis (Fig. 8B) indicates that p21 mRNA is more abundant in CDIR-transfected cells than in control, 5Ј CDIR, or 3Ј CDIR cells. The level of p21 in 5Ј CDIR-transfected cells is very similar to that in vectortransfected cells, and 3Ј CDIR-transfected cells exhibit an intermediate level (expression levels were normalized to GAPDH and verified by PhosphorImager). This result correlates with the in vitro binding activity of 5Ј CDIR and 3Ј CDIR. As demonstrated (Fig. 5B), 5Ј CDIR has completely lost the ability to bind to (compete) proteins in the 0.45 M KCl fraction, whereas 3Ј CDIR has partially lost binding (competition) activity. This result suggests that the effect on p21 mRNA is correlated with the ability of CDIR to protect cells from killing and with the ability to bind AUF1 and Hsp27. To determine whether the rate of p21 mRNA turnover was altered in CDIR-transfected cells, CDIR and control vector-transfected cells were treated with actinomycin D. Northern blot analysis of the RNA samples collected at different times post-actinomycin D treatment revealed an increase in p21 mRNA stability in CDIR-transfected cells compared with vector-transfected cells (Fig. 8C). calculated for each of three independent experiments and the averages are presented in Fig. 8D. The increase in p21 mRNA half-life also affected the level of p21 protein. As detected by Western blot analysis (Fig. 8E), HeLa cells transfected with CDIR express 25% more p21 protein than vector-transfected cells (calculated using actin levels as a control for loading). Next we tested another anti-apoptotic gene, namely Bcl-2, which has an AUF1 binding sequence and is therefore a potential AUF1 target. The steady state level of Bcl-2 mRNA was measured in vector-transfected cells and in CDIR-transfected cells. The Northern blot in Fig. 8F shows that Bcl-2 mRNA are increased in both of the CDIR-transfected cell populations. PhosphorImager quantification of three experiments using the GAPDH level as a loading control indicated that CDIR transfection increases the level of Bcl-2 mRNA by 36 Ϯ 18%. The actinomycin D experiment revealed that in vector-or CDIRtransfected cells the half-life of Bcl-2 mRNA is longer than 4 h, however, it could not be measured accurately because at longer time points the actinomycin D had deleterious effects on the cells (data not shown). The increased Bcl-2 mRNA levels are accompanied by increased levels of the Bcl-2 protein (Fig. 8G). Thus, we demonstrate that the steady state level of two AUF1targeted mRNAs encoding anti-apoptotic genes are increased in CDIR-transfected cells. The CDIR effect is most likely mediated through diminution of AUF1 from the cellular pool. DISCUSSION This report demonstrates a link between the anti-apoptotic effect of CDIR and the binding of CDIR by a protein complex containing AUF1 and Hsp27. Expression of CDIR, a 441-nt fragment of the 3Ј-UTR of KIAA0425, renders HeLa cells resistant to killing by IFN-␥. A number of studies demonstrate that expression of 3Ј-UTRs may effect cell behavior (10 -14). To our knowledge, the only defined mechanism is the suppression of tumor formation by the 3Ј-UTR of ␣-tropomyosin that is mediated through binding and activation of the doublestranded RNA-activated protein kinase PKR (11,44). Regulation of PKR activity requires double stranded RNA structures rather than specific sequence (45). Structure and function analyses of CDIR revealed that its anti-apoptotic activity is independent of secondary structure but is sequence specific. In addition, we failed to detect PKR in the CDIR-binding complex (data not shown). Thus, CDIR is unlikely to be involved in regulation of PKR.
We here demonstrated that the anti-apoptotic activity of CDIR is highly dependent on the sequence of U-rich elements. Substitution of nucleotides in positions 4493 (5Ј CDIR) or 4728 (3Ј CDIR) that affect both motifs AUUUUC and UUU-CAUUUU abolish the ability of CDIR to protect cells from killing. We demonstrate that alterations at position 4510 that resides outside of any U-rich region had no noticeable effect on CDIR activity. The base substitutions that created 5Ј CDIR and 3Ј CDIR also affect the efficiency of complex formation between CDIR and a subset of proteins. Hence, we propose that the proteins that preferentially form complexes with CDIR but do not bind or bind with low affinity to the biologically inactive CDIR mutants, mediate the anti-apoptotic activity of CDIR.
Fractionation of the CDIR-binding factors by heparin chromatography followed by affinity purification revealed two distinct sets of proteins of 40 -45 and 25 kDa. Further characterization of the CDIR-binding complex by REMSA revealed the presence of both AUF1 and Hsp27. Whether Hsp27 contributes to the anti-apoptotic activity of CDIR is currently unknown. We speculate that Hsp27 might modulate AUF1 activity as it was reported to modulate RNase P (46). Our results implicating Hsp27 as part of the same complex formed by AUF1 and CDIR coincides with the earlier observation that Hsp27 regulates the stability of Cox-2 mRNA through an AUF1-binding element (29). In addition, Hsp27 can form multimeric complexes (47). AUF1 has also been reported to form oligomers (30). It is possible that a complex between CDIR, AUF1, and Hsp27 is formed in a cooperative manner allowing the optimal recruitment of protein molecules into the complex. AUF1 is an RNA as well as a DNA-binding protein (24,48). It is generally acknowledged that AUF1 is involved in the regulation of mRNA turnover through direct binding to the 3Ј-UTRs of the target mRNAs (16,24,29,49). AUF1 binds specifically and with high affinity to c-myc, c-fos, granulocytemacrophage colony-stimulating factor, and other early response mRNAs (24,25,28). To determine whether AUF1 binding to CDIR contributed to CDIR anti-apoptotic activity, we tested another noncoding RNA region that binds AUF1. We demonstrated that the AUF1 binding region of c-myc (c-myc ARE) inhibits IFN-␥-induced killing of HeLa cells as measured in short term assays. It should be noted that the expression of c-myc ARE could not protect HeLa cells from induction of apoptosis in the long term colony formation assay (data not shown). It is possible, that together with AUF1, there are other factors, so far unidentified, that are involved in the sustained protection of HeLa cells by CDIR. Therefore, binding of AUF1 protein to CDIR is one of the events required for the antiapoptotic function of CDIR.
Since no direct effects on AUF1 levels were observed in CDIR-transfected cells, it is possible that expression of AUF1 targets, such as CDIR and c-myc 3Ј-UTR, modulates AUF1 activity through the sequestration of this protein from other FIG. 8. Analysis of AUF1 targets. A, AUF1-binding fragment of c-myc mRNA protects the HeLa cell from apoptosis. Cells transfected with c-myc ARE, CDIR, or vector were treated with IFN-␥ and hygromycin B. Three days later the cell viability was determined by trypan blue exclusion. Presented is the average of two experiments, each including two independent pools of transfectants. B, Northern blot analysis of p21 levels in cells transfected with vector, CDIR, 5Ј CDIR, or 3Ј CDIR. The membrane was hybridized to a p21-specific probe, followed by hybridization to a GAPDH probe as a loading control. C, determination of p21 mRNA half-life in CDIR-transfected cells compared with vector-transfected cells. CDIRtransfected cells as well as vector-transfected cells were treated with actinomycin D (10 g/ml), and total RNA was extracted at the indicated time points. The membrane was hybridized with a p21-specific probe, followed by hybridization to GAPDH for a loading control. D, p21 mRNA is stabilized in CDIR-transfected cells. PhosphorImager derived values of p21 mRNA levels in CDIR-transfected cells (solid diamond) or control vector-transfected cells (open box) were normalized to GAPDH. At each time point the percentage mRNA (relative to time 0) was calculated. An average of three independent actinomycin D experiments is shown. E, Western blot analysis of p21 in CDIR and in control vector-transfected cells. Protein samples were separated on SDS-PAGE. The membrane was probed with an anti-p21 (anti-waf1) antibody followed by an anti-␤-actin antibody. F, Northern blot analysis of Bcl-2 in CDIR or vector-transfected cells. Two different mRNA samples are shown. The membrane was hybridized to a Bcl-2-specific probe, followed by hybridization to GAPDH as a loading control. G, Western blot analysis of Bcl-2 in CDIR and vector-transfected cells. Protein samples were separated on SDS-PAGE. The membrane was probed with an anti-Bcl-2 antibody followed by an anti-␤-actin antibody. cellular targets. We propose that the anti-apoptotic effect of CDIR is mediated through AUF1 targets whose turnovers become critical upon induction of killing. To test this hypothesis, we examined mRNAs that are likely to be regulated by AUF1, cyclin kinase inhibitor p21, and Bcl-2. p21 is reported to inhibit apoptosis in response to various stimuli including IFN-␥ treatment. Anti-apoptotic activity of p21 is believed to be independent from its regulation of cell cycle arrest (reviewed in Ref. 50). Several reports suggest that AUF1 binds and mediates the decay of p21 mRNA (34,35). We observed a stabilization of p21 transcript and an increase in p21 protein levels in CDIR-transfected cells. The abundance of the anti-apoptotic Bcl-2 mRNA and protein were also increased in CDIR-transfected cells.
Deregulation of the turnover of AUF1 mRNA targets is linked to several pathologies possibly associated with a decrease in cell death (17). For example, truncation of an AUF1 destabilization element in the 3Ј-UTR of TNF-␣ mRNA leads to deregulated production of TNF-␣ resulting in chronic inflammatory arthritis (51). Increased stability of Bcl-2 mRNA because of alteration in the U-rich region in the 3Ј-UTR may contribute to the overproduction of the anti-apoptotic Bcl-2 protein that is responsible for transformation of follicular B-cell lymphoma (52,53). Because Bcl-2 mRNA contains an AUF1 binding sequence in the 3Ј-UTR (53) it is likely that expression of CDIR interferes with AUF1 binding, resulting in increased levels of Bcl-2 mRNA.
Here we show that mRNA decay is affected by overexpression of two different AUF1 targets, namely CDIR (a fragment of KIAA0425) and c-myc ARE. Thus, RNA molecules bound by AUF1 can significantly alter the response of cells to the induction of apoptosis. Furthermore, a single base substitution in such RNA can completely abolish this activity by modulating protein complex formation. Thus, we propose that minor alterations in the UTR of a gene might lead to significant consequences in cell behavior that are independent of the coding capacity of the mRNA.