BCL-2 Translation Is Mediated via Internal Ribosome Entry during Cell Stress*

The cellular response to stress involves a rapid inhibition of cap-dependent translation via multiple mechanisms, yet some translation persists. This residual translation may include proteins critical to the cellular stress response. BCL-2 is a key inhibitor of intrinsic apoptotic signaling. Its primary transcript contains a 1.45-kb 5 (cid:1) -untranslated region (UTR) including 10 upstream AUGs that may restrict translation initiation via cap-dependent ribosome scanning. Thus, we hypothe-sized that this 5 (cid:1) -UTR may contain an internal ribosome entry site (IRES) that facilitates BCL-2 translation, par-ticularly during cell stress. Here we show that the BCL-2 5 (cid:1) -UTR demonstrated IRES activity both when translated in vitro and also when m 7 G-capped and polyadenylated mRNA was transiently transfected into 293T cells. The activity of this IRES in unstressed cells was (cid:1) 6% the strength of the hepatitis C virus IRES but was induced 3–6-fold in a dose-dependent manner following short term treatment with either etoposide or sodium arsenite. Thus, the IRES-mediated translation of BCL-2 may enable the cell to replenish levels of this critical protein during cell stress, when cap-dependent translation is repressed, thereby maintaining the balance between pro- and anti-apoptotic BCL-2 family members in the cell and preventing unwarranted induction of apoptosis.

The BCL-2 protein is the prototype member of a conserved superfamily of proteins that regulate the cellular response to various intrinsic stresses, including damage to DNA, the endoplasmic reticulum, or Golgi apparatus (1). BCL-2 prevents apoptosis in response to these stresses through a variety of mechanisms, many of which are now clearly defined. Its predominant isoform (BCL-2␣) contains a C-terminal transmembrane domain and localizes to membranes of the mitochondria, endoplasmic reticulum, and nuclear envelope (2). The anti-apoptotic members of the BCL-2 family, including BCL-2, all contain highly homologous regions (BH domains 1-3), which together form a hydrophobic groove required for anti-apoptotic functionality (3,4). BCL-2 forms heterodimers with a variety of pro-apoptotic proteins, thereby sequestering these proteins to prevent the onset of apoptosis (5). Some of these interactions prevent Bax/Bak-mediated pore formation in the outer mitochondrial membrane and the subsequent release of pro-apoptotic factors, including cytochrome c (6), apoptosis inducing factor (7), Smac/DIABLO (8,9), and Htr2A/Omi (10).
Additional findings suggest that the anti-apoptotic role of BCL-2 extends further than simply protecting mitochondrial integrity. BCL-2 appears to regulate the cell cycle (11) and reportedly prevents a constitutive pro-apoptotic signal from p53 by directly binding this transcription factor (12). In addition, numerous reports (13,14) have shown that BCL-2 controls [Ca 2ϩ ] in the endoplasmic reticulum, thereby controlling apoptotic sensitivity.
With such diverse roles in regulating apoptosis, it is not surprising that the expression of BCL-2 is regulated at multiple levels. First, the transcription of BCL-2 mRNA is transcribed from two different promoters, and its mRNA stability is regulated by an AU-rich element in the 3Ј-UTR 1 (15). In addition, the phosphorylation state of BCL-2 (16,17) may alter its anti-apoptotic efficacy and/or lead to its ubiquitination/degradation (18), whereas caspase-mediated cleavage regulates the half-life of BCL-2 protein during apoptosis (19). Many reports have demonstrated a lack of correlation between the levels of BCL-2 mRNA versus BCL-2 protein, suggesting that BCL-2 expression may be translationally regulated (20 -23). This is supported by a recent report (24) demonstrating that the growth factor FGF-2 up-regulates BCL-2 protein levels via a translational mechanism that cannot be blocked by actinomycin D.
The two BCL-2 promoters, P1 and P2, give rise to transcripts containing 5Ј-UTRs differing in size by ϳ1.4 kb (25,26). The regulation of these two promoters is complex and depends upon both tissue-type and developmental stage (27,28). However, in many cell types the vast majority of transcripts initiate from the upstream P1 promoter, resulting in a 5Ј-UTR of ϳ1.45 kb (25,29). This 5Ј-UTR contains a 219-bp alternatively spliced intron that spans the region from 286 to 505 bases upstream from the translation start site. The splicing frequency of this intron varies among B-cell lines, although both spliced and un-spliced forms are often simultaneously expressed (25). The BCL-2 5Ј-UTR region is highly conserved between human, mouse, rat, and even chicken, suggesting a potential regulatory role for this region (30 -32).
The vast majority of cellular mRNAs initiate translation via m 7 G cap-dependent recruitment of the 40 S ribosomal subunit to the 5Ј end of a mRNA, followed by linear 5Ј-3Ј scanning to the first AUG codon in proper context (33). However, during induced cellular stress this cap-dependent translation is rapidly inhibited by multiple mechanisms (34,35). Even so, a portion of cellular translation persists that is thought to occur via a cap-independent recruitment of ribosomes directly onto certain mRNAs containing internal ribosome entry site (IRES) elements (35,36). IRES-mediated translational initiation was initially described as a mechanism that enables certain viruses to translate effectively viral proteins despite an inhibition of cap-dependent translation in the infected cell (37,38). In eukaryotes, IRES-mediated translation initiation has been most often observed for mRNAs that possess unusually long and thermodynamically stable 5Ј-UTRs with multiple potential uORFs, features that can dramatically inhibit scanning-dependent translation initiation (36). The predominant mRNA transcript coding for BCL-2 possesses both of these characteristics, containing a 1.45-kb 5Ј-UTR with 10 upstream AUGs. Thus, we examined the 5Ј-UTR of BCL-2 for its ability to mediate translational initiation via an unconventional, capindependent mechanism.
Through transfection of m 7 G-capped and polyadenylated reporter mRNAs, we demonstrate that BCL-2 expression is regulated via a stress-inducible IRES located within its 5Ј-UTR. This IRES mediated little reporter gene translation in unstressed cells, yet was induced 3-6-fold during stress induced by treatment with either sodium arsenite or the chemotherapeutic agent etoposide. Thus, IRES-mediated translation of BCL-2 may enable the cell to replenish levels of this critical protein during periods of cell stress, when little cap-dependent translation occurs. Cell Culture-Cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 100 units/ml penicillin, 10 g/ml streptomycin, and 10% fetal bovine serum (Hyclone, Logan UT). Growth was at 37°C in 5.0% CO 2 at 95% relative humidity, and cells were thinned 1:5 every 2-3 days at Ͻ85% confluency.

Materials-Radiolabeled
Construct Assembly-pRL-HL was a gift from S. Lemon (University of Texas Medical Branch, Galveston, TX), and its construction was described previously (39). A unique SacII site was introduced into pRL-HL upstream from FLuc to facilitate insertion of sequences between NotI and SacII sites, and the construct was named pRL-HCV-FL. The BCL-2 5Ј-UTR was amplified via RT-PCR from total RNA of HeLa S3 cells. Gene-specific primers amplified a region from bases Ϫ1146 to ϩ3 of the BCL-2 cDNA (relative to the translational start site) and added NotI (5Ј end) and SacII (3Ј end) sites. This fragment was inserted into a NotI/SacII-digested pRL-HCV-FL backbone to form pRL-BCL2-FL. The original BCL-2 start codon was retained and inserted in-frame with FLuc, creating a FLuc ORF with six extra N-terminal codons. pRL-FL was created by removing the HCV IRES region between NotI and SacII sites. pRL-revH-FL was constructed via PCR of the HCV sequence utilizing primers that swapped the NotI and SacII sites. This fragment was then ligated into the NotI/SacII-cut backbone of pRL-HCV-FL. phpRL-BCL-2-FL included a 143-bp region that forms a stable RNA hairpin (⌬G ϭ Ϫ60 kcal/mol) inserted at a unique NheI site located immediately upstream from RLuc. p(⌬CMV)RL-BCL-2-FL was created using long distance PCR and the 5Ј-phosphorylated primers 5Ј-cagccgcggagaactagtggatcccccggg-3Ј and 5Ј-ctagcggccgcttggtgttacgtttggtttttctttg-3Ј, which allowed amplification of all but the CMV promoter region. The PCR product was then blunt-ligated with T4 ligase. pBCL-2-FL was similarly constructed using the primers 5Ј-gacatccactttgcctttctctcca-3Ј and 5Ј-gtgggttacatcgaactggatctca-3Ј. Construction of phpB-CL-2-FL utilized primers 5Ј-cgttgagcgagttctcaaaagtgaacaataattctagagcgg-3 and 5Ј-ggtggctagcttataaaagcagtgg-3Ј. All clones were verified by both restriction digest and sequencing.
In Vitro Transcription and Translation-Constructs were linearized with either AgeI or XhoI (pRL-revH-FL and pRL-HCV-FL). In vitro transcription was performed using purified T7 polymerase in a typical 100-l reaction. m 7 G(5Ј)ppp(5Ј)G cap analog (m 7 G) (Ambion, Austin, TX) was included at a ratio of 4:1 versus rGTP. Translations were performed in typical fashion using nuclease-treated RRL (Promega) and 10 Ci of Tran 35 S-Label Met/Cys (ICN Biomedicals, Inc.). Reactions were resolved via 12% SDS-PAGE, dried, and then exposed to Kodak Biomax film (Rochester, NY).
RNAi Directed against RLuc-The RNAi vector pBS-RLi was utilized to elicit a small interfering RNA response targeting the RLuc coding region. This vector contains a 50-bp RNA hairpin downstream from the murine U6 promoter, and its construction has been described previously (40). pBS-RLi was transiently co-transfected into cells along with dicistronic reporter constructs at a 1:1 ratio (0.5 g each) using Fu-GENE 6 transfection reagent (Roche Applied Science). The empty vector pBS-ApaI was used as control. 293T cells were plated at 4.0 ϫ 10 5 cells/well on 12-well plates 16 h prior to transfection. Cells were harvested 72 h after transfection, and luciferase activity was measured using the dual luciferase assay kit (Promega) on a Sirius model luminometer (Berthold Detection Systems).
RT-PCR of Transfected BCL-2 DNA Constructs-Total RNA was Trizol-extracted (Sigma) from 293T cells that had been transiently transfected with dicistronic DNA test constructs. RNA was DNase I-treated for 15 min at room temperature, and RT-PCR was then performed by using an upstream primer just downstream from the CMV promoter translation initiation site, 5Ј-cagatcactagaagctttattgcg-3Ј, and a downstream primer just inside the FLuc ORF, 5Ј-tctcttcatagccttatgcagttgc-3Ј. Control PCRs using DNA as template were not DNase I-treated.
Transient Transfection of DNA and mRNA-DNA constructs were transiently transfected into cells using FuGENE 6 reagent (Roche Applied Science). Cells were plated in 12-well plates at 1.25 ϫ 10 5 cells/ well 16 h prior to transfection. The ratio of FuGENE/DNA was 6 l:2 g, and transfections were performed in Dulbecco's modified Eagle's medium ϩ 10% fetal bovine serum. Cells were lysed 8 h post-transfection and assayed for both RLuc and FLuc activities. Transient transfection of in vitro transcribed mRNA was achieved by first cloning a 35-bp poly(A)-containing sequence into all reporter constructs at the unique ApaI site immediately downstream from FLuc. This sequence included a unique AgeI site at its 3Ј end. Each construct was linearized by digesting with AgeI (except for pRL-HCV-FL and pRL-revH-FL, which were linearized with XhoI), followed by phenol/CHCL3/isoamyl extraction and isopropyl alcohol precipitation. Transcription was performed as described above. Cells were plated at 1.25 ϫ 10 5 cells/well in 12-well plates and attached overnight. Transfection proceeded using 1.5 g/well mRNA with 8 l of DMRIE C reagent (Invitrogen). pSV40-Bgal DNA was co-transfected with monocistronic RNAs at 0.1 g/well to control for transfection efficiency. Serum was re-added 3 h post-transfection. Luciferase analysis was as described above. ␤-Galactosidase was quantified using the Beta-Glo TM Assay Reagent (Promega). Treatment with various stress-inducing agents was maintained during transfection. Etopo and Na Ars were both obtained from Sigma.
RNA Analysis-For in vivo assays, additional wells were simultaneously transfected and treated identically for analysis of post-transfection reporter mRNA integrity. Six hours post-transfection, total RNA was harvested with 300 l of Trizol reagent (Sigma). Purified total RNA was denatured using established glyoxal/Me 2 SO methodology and 2.5 g of RNA/well run in a 1% NaHPO 4 -buffered gel. RNA was then transferred to Hybond Nϩ nylon membrane (Amersham Biosciences) and exposed to Kodak X-OMAT TM film.
Northern analysis to assess in vivo expression of pRL-BCL2-FL was performed by transferring RNA to nylon (as described above) and then pre-hybridized with 10 ml of Ultra Hyb TM solution for 1 h (Ambion). A 489-bp FLuc-specific riboprobe was transcribed in vitro using SP6 polymerase, along with 10 Ci of [␣-32 P]CTP. 1 ϫ 10 6 cpm of riboprobe was hybridized for 12 h at 68°C, followed by washing in 2ϫ SCC and autoradiography.
Immunoblotting for BCL2-Cells were lysed in RIPA buffer, and 40 g of total protein was mixed 1:1 with 4ϫ SDS-PAGE sample buffer, boiled 5 min, and then resolved via 12% SDS-PAGE. Proteins were transferred to polyvinylidene difluoride, blocked 1 h in TBST ϩ 3% milk, then probed with 1:1000 anti-BCL-2 monoclonal antibody (clone SC-7382, Santa Cruz Biotechnology) overnight at 4°C, followed by a horseradish peroxidase-conjugated anti-mouse secondary F(abЈ) 2 fragment (Jackson ImmunoResearch). Normalization for total protein loading was accomplished by using a monoclonal antibody specific for ␣tubulin (Sigma).

RESULTS
The predominant BCL-2 promoter (P1) produces a 5Ј-UTR of 1.45 kb that contains numerous structural features that likely BCL-2 Translation Is IRES-mediated regulate BCL-2 translation. For example, an extremely stable secondary structure (⌬G ϭ Ϫ530 kcal/mol) as well as multiple potential uORFs would be expected to largely inhibit conventional ribosomal scanning ( Fig. 1A) (41). Indeed, a uORF that spans Ϫ119 to Ϫ84 bp upstream from the start codon has been reported to inhibit significantly BCL-2 translation, presumably by causing a fraction of scanning ribosomes to translate the uORF, and then disengage from the mRNA (29).
The region of the BCL-2 5Ј-UTR that was examined for this study extended 1146 bp upstream from the start codon and included the alternatively spliced 219-bp intron (Fig. 1B). This sequence was inserted into several reporter constructs in order to assess potential IRES activity, including a dicistronic construct that placed the BCL-2 5Ј-UTR between an upstream RLuc reporter gene and a downstream FLuc gene ( Fig. 2A). The upstream RLuc cistron was immediately followed by two "inframe" stop codons that inhibit ribosomal reinitiation. This dicistronic DNA reporter assay is commonly utilized to assess potential IRES activity because cap-dependent translation of RLuc not only restricts cap-dependent translation of the downstream FLuc but also acts as an internal control to measure relative increases in FLuc expression. A positive control for IRES activity included a 426-bp region containing the HCV IRES between the reporter genes (pRL-HCV-FL), whereas the negative control contained the HCV IRES in an inverted orientation (pRL-revH-FL). For in vitro assays, the negative control (pRL-FL) contained a 27-bp spacer sequence between the cistrons.
The 5Ј-UTR of BCL-2 Functions as an IRES in Vitro-As an initial test for IRES activity, test constructs containing the 5Ј-UTR of BCL-2 were transcribed in vitro in the presence of 7-methyl-G(5Ј)ppp(5Ј)guanosine cap analog (m 7 G). Equimolar amounts of transcripts were then used to program translation in nuclease-treated rabbit reticulocyte lysates (RRL) (Fig. 2B). The empty vector control (Fig. 2B, pRL-FL, lane 1) efficiently translated the upstream RLuc cistron, yet little translation of the second cistron (Fluc) was observed. Similar to previous reports, insertion of the HCV IRES (Fig. 2B, pRL-HCV-FL, lane 2) stimulated expression of downstream FLuc, indicative of IRES-mediated translation. Likewise, expression of FLuc also increased downstream from the BCL-2 5Ј-UTR (pRL-BCL2-FL, lane 5), suggesting that this region supports capindependent translation, and may contain an IRES. Insertion of a sequence that forms a stable RNA hairpin structure (⌬G ϭ Ϫ60 kcal/mol) upstream of RLuc (Fig. 2B, phpRL-BCL2-FL, lane 6) inhibited cap-dependent translation by ϳ98%, yet significant expression of the second cistron (FLuc) persisted. This indicated that increased translation of FLuc mediated by the BCL-2 5Ј-UTR was not due to an increase in ribosomal re-initiation.
Similar results were obtained by using monocistronic RNAs. Translation of the monocistronic pBCL2-FL (Fig. 2B, lane 3) demonstrated that the BCL-2 5Ј-UTR supported ample translation of FLuc, although at a significantly lower level than constructs containing RLuc preceded by only a short leader sequence (Fig. 2B, lanes 1, 2, and 5) or FLuc alone (data not shown). Insertion of the stable RNA hairpin structure in front of the BCL-2 5Ј-UTR (Fig. 2B, phpBCL2-FL, lane 4) inhibited its ability to mediate FLuc translation by ϳ65% but far less than the 98% inhibition of RLuc translation observed when this FIG. 1. Sequence and structural features of the BCL-2 5-UTR. A, predicted RNA secondary structure of the BCL-2 5Ј-UTR as determined using the MFOLD algorithm (version 3.1). The six most thermodynamically conserved domains of predicted secondary structure (labeled I-VI) were determined through multiple folds. Start AUG codons of the 10 uORFs are indicated (filled boxes). The overall folding energy (⌬G) of the most stable predicted structure ϭ Ϫ530 kcal/mol. The star indicates the minor P2 promoter TATA box. B, the full-length BCL-2 5Ј-UTR (␣ isoform) is ϳ1457 bp in length, with 10 upstream AUGs (shown in boldface type). The region encompassing bases Ϫ1138 to Ϫ1 (unshaded) was utilized for this study. The uORF shown previously to inhibit BCL-2 translation is indicated in lowercase letters (bases Ϫ117 to Ϫ78), and the 219-bp alternatively spliced intron is boxed.
hairpin was inserted into phpRL-BCL2-FL. Together, these data strongly indicated that the 5Ј-UTR of BCL-2 could support cap-independent translation, possibly through IRES-mediated recruitment.
In Vivo Assessment of BCL-2 IRES Function in Cells Transfected with DNA Constructs-We sought to confirm our in vitro results by transfecting DNA constructs containing the BCL-2 5Ј-UTR into 293T kidney cells. Following transfection, a similar increase in FLuc:RLuc ratio above background was observed for both pRL-BCL2-FL and pRL-HCV-FL (Fig. 3A), suggesting that the BCL-2 5Ј-UTR contained an IRES of comparable strength to the HCV IRES. HeLa S3 and HepG2 cells were likewise transfected with these constructs, and similar results were obtained (results not shown). To test whether this apparent BCL-2 IRES activity was the result of monocistronic transcripts being generated from a promoter within the BCL-2 5Ј-UTR, a DNA construct lacking the cytomegalovirus (CMV) promoter was created [p(⌬CMV)RL-BCL2-FL]. Upon its transfection into 293T cells, no RLuc or FLuc activity was detected, demonstrating negligible promoter activity (Fig. 3A,  bottom panel).
Next, we utilized several procedures to rigorously evaluate the transcripts produced from pRL-BCL2-FL in vivo. First, we used a new RNAi method to test for spliced transcripts derived from the dicistronic vector (54). An RNAi vector targeted against RLuc was transfected into cells, resulting in silencing of RLuc expression by ϳ80% (results not shown), in accordance with previous observations (40). RLuc-directed RNAi in the cells containing pRL-HCV-FL resulted in equal inhibition of RLuc and FLuc, suggesting the presence of only a single dicistronic transcript (Fig. 3B). However, aberrantly spliced mono-cistronic transcripts coding only for FLuc would not be subject to the same RLuc-specific RNAi silencing. When the RNAi vector was co-transfected with pRL-BCL2-FL, FLuc translation decreased but was 2.3-fold greater than RLuc, suggesting the presence of monocistronic FLuc transcripts (Fig. 3B). Similar results were obtained using a construct containing the XIAP IRES (pRL-XIAP-FL), which was recently demonstrated to contain a splice acceptor (Fig. 3B) (40).
To confirm potential splicing, nonquantitative RT-PCR was performed on total RNA isolated from 293T cells that had been transiently transfected with pRL-BCL2-FL (Fig. 3C). PCR primers were selected to amplify the region between the transcription start site of the CMV promoter to the FLuc ORF, such that control DNA PCR fragments for pRL-HCV-FL (1.9 kb) and pRL-BCL2-FL (2.4 kb) were expected (Fig. 3C, lanes 2 and 5). RT-PCR products were anticipated to be ϳ200 bp smaller due to splicing of the chimeric intron (Fig. 3C, lanes 4 and 6). pRL-HCV-FL-transfected cells produced an RT-PCR product of 1.7 kb, indicating an intact dicistronic transcript. However, cells transfected with pRL-BCL2-FL produced not only a 2.2-kb PCR product representing a dicistronic transcript (visible only upon overexposure, data not shown), but also a smaller product of 350 bp (Fig. 3C). Sequencing showed that the smaller 350-bp product represented a monocistronic FLuc transcript with a 290-bp 5Ј-UTR. Finally, Northern analysis confirmed the presence of both transcripts (Fig. 3D). Although the dicistronic transcript (ϳ4.7 kb) was the predominant species, a significant portion of FLuc expression had resulted from the production of the smaller, monocistronic transcript. Therefore, these data showed that pRL-BCL2-FL generated monocistronic transcripts containing only FLuc in vivo, and thus IRES function within the BCL-2 5Ј-UTR could not be accurately quantified via DNA transfection. Although Northern analysis confirmed the presence of two different transcripts in this instance, the more sensitive techniques described here can detect low level aberrant splicing that is below the sensitivity threshold of Northern blotting (40).
The BCL-2 5Ј-UTR Functions as an IRES in Vivo-We chose to assess potential BCL-2 IRES activity by directly transfecting RNA. All DNA constructs were modified such that in vitro transcription would create transcripts containing a 35-bp poly(A) tail (see "Experimental Procedures"). These constructs (Fig. 4A) were transcribed in vitro in the presence of the m 7 G cap analog, and the resulting transcripts were transfected into 293T cells for 6 h, followed by cell lysis and analysis of luciferase activity. Transfection of dicistronic RNAs containing the BCL-2 5Ј-UTR translated FLuc 8-fold more efficiently versus the negative control pRL-revH-FL, suggesting that this RNA region exhibits modest but significant IRES activity in vivo. (Fig. 4A). This activity was ϳ6% of the activity exhibited by the HCV IRES (pRL-HCV-FL) under the same conditions (Fig. 4A).
Monocistronic constructs containing the BCL2 5Ј-UTR were also examined for IRES activity (Fig. 4B). As expected, the long BCL-2 5Ј-UTR restricted translation and the capped monocistronic pBCL-2-FL RNA produced only 16% as much FLuc as the capped control pFL RNA in untreated cells (results not shown). Most interesting, the lack of a m 7 G cap inhibited the translation of uncapped pFL RNA by 50-fold (Fig. 4C), yet the lack of a cap inhibited the translation of pBCL2-FL RNA by only 10-fold (Fig. 4B). As an alternative blockade to cap-dependent translation, a stable RNA hairpin was placed immediately upstream from the BCL-2 5Ј-UTR (phpBCL2-FL), yet 14% of translation persisted (Fig. 4B). This was significantly higher than the Ͼ97% inhibition of translation typically observed when this hairpin was inserted in front of either RLuc or FLuc (Fig. 2, lanes 4 and 6) (40, 42). Together, these results indicated that about 9 -10% of total translation mediated by the BCL-2 5Ј-UTR had occurred via a cap-independent mechanism, which was consistent with the presence of an IRES in the BCL-2 5Ј-UTR.
To both correct for potential variations in transfection efficiency and monitor integrity of the transfected RNA, [ 32 P]CTPlabeled transcripts were transfected, and total RNA was isolated from cells 6 h later. RNA samples were resolved via denaturing agarose gel electrophoresis, followed by autoradiography. The results indicated that all of the transfected RNAs were stable for the duration of the experiment and that processing or degradation of RNAs had not occurred (refer to Figs. 5D and 6D).

BCL-2 IRES Function Is Induced by Cell Stress-Apoptosis induced by Na
Ars results from its direct interaction with the permeability transition pore complex to induce mitochondrial permeability, whereas Etopo induces apoptosis by inhibiting topoisomerase II to cause DNA breakage (43,44). Most important, apoptosis induced by either of these stressors can be inhibited by BCL-2 (45,46). To test if the stress induced by either Etopo or Na Ars increased BCL-2 IRES function, 293T cells were treated with either Etopo or Na Ars and then transfected with RNA for 6 h prior to harvest, while maintaining drug treatment. Initial studies revealed an IRES activity within the BCL-2 5Ј-UTR that peaked at 8 h for Etopo and 12 h after Na Ars treatment. A dose-dependent stimulation of BCL-2 IRES activity (quantified as increase in FLuc/RLuc) was observed that reached 3.4-fold after 8 h of treatment with 80 M Etopo or 16.8-fold greater than the identically treated negative control (Fig. 5A). Meanwhile, the activity of pRL-HCV-FL dropped by 27% following identical stress treatment (Fig. 5B). As a result, in cells treated with 80 M Etopo, BCL-2 IRES activity reached 26% of the activity of the HCV IRES (Fig. 5B).
In a separate experiment, treatment with Na Ars caused a dose-dependent increase in BCL-2 IRES function, resulting in a 5.6-fold induction in activity after 12 h of treatment with 16 M Na Ars or 12-fold greater than the identically treated negative control (Fig. 5A). This induced BCL-2 IRES activity reached 14.6% of the HCV IRES activity present in untreated cells. However, Na Ars treatment also increased the activity of the HCV IRES (pRL-HCV-FL) by 34% (Fig. 5C). Even so, the background FLuc/RLuc ratio of the negative control pRL-revH-FL remained relatively constant regardless of Na Ars treatment (Fig. 5A). Once again, to ensure transcript integrity [ 32 P]CTP-labeled transcripts were transfected and total RNA was isolated from cells 6 h later. RNA samples were resolved via denaturing agarose gel electrophoresis, followed by autoradiography. The results indicated that all of the transfected RNAs were stable for the duration of the experiment and that processing or degradation of RNAs had not occurred (Fig. 5D).

Cell Stress Induces BCL-2 IRES Function in a Monocistronic
Context-The monocistronic constructs utilized previously to characterized BCL-2 IRES activity in unstressed cells (Fig. 4, B and C) were again utilized to assess the stress inducibility of the BCL-2 IRES in a monocistronic context (Fig. 6A). The translation of these transcripts was measured after treatment with Etopo using the conditions described above. Etopo treatment for 8 h inhibited the translation of capped pFL in a dose-dependent manner by up to 70% (Fig. 6B) However, the translation of pBCL2-FL remained relatively constant, and even increased slightly after treatment with 40 M Etopo (Fig.  6B). Whereas the absence of cap dramatically inhibited the translation of the control pFL transcript under all conditions tested, it did not prevent induction of BCL-2 IRES-mediated translation in response to Etopo treatment. This is clearly shown in Fig. 6C, which depicts the percent change in FLuc translation caused by treatment with 40 M Etopo. This treatment stimulated the translation of un-capped pBCL2-FL by 3.5-fold, whereas the translation of a capped, monocistronic BCL-2 construct containing a stable hairpin structure at its 5Ј end (phpBCL2-FL) was similarly induced by nearly 3-fold.
The stability of the transfected RNA was again assessed 6 h post-transfection (as described previously) to ensure that the differences observed after stress treatment were not due to altered RNA stability (Fig. 6D). The resulting autoradiogram of radiolabeled RNA demonstrated that no change in stability was elicited by either Etopo treatment or the absence of a m 7 G cap on certain mRNAs.

Stress Induction Increases BCL-2 Protein Levels and Can Decrease BCL-2
Half-life-Immunoblot analysis of BCL-2 protein was performed to confirm the induction of BCL-2 protein levels in vivo following the same stress treatments used to induce BCL-2 IRES activity. 293T cells were treated with increasing concentrations of either Etopo or Na Ars in an identical manner as described above (Fig. 7). BCL-2 protein levels were also assessed during simultaneous treatment with the proteosome inhibitor MG-132. An increase in BCL-2 protein levels resulted from treatment with increasing concentrations of Etopo, although at the highest Etopo concentrations BCL-2 levels began to decrease (Fig. 7A, 4th lane). However, when Etopo-treated cells were co-incubated with the proteosome inhibitor MG132, the levels of BCL-2 were induced in a dose-dependent manner, even at the highest concentration of Etopo (Fig. 7A, 8th lane) This correlated well with our earlier data, suggesting that BCL-2 translation was induced at all Etopo concentrations utilized. However, at higher concentrations, an apoptotic threshold appears to have been surpassed that also led to the increased proteosomal degradation of BCL-2.
Treatment with Na Ars also increased BCL-2 protein levels in a dose-dependent manner, reaching a maximum induction of about 2-fold (Fig. 7B). Treatment with the proteosome inhibitor MG132 increased BCL-2 levels by a constant amount, although the dose-dependent induction of BCL-2 levels following Na Ars treatment was unchanged by MG132 treatment (Fig. 7B). This indicates that Na Ars treatment increased BCL-2 levels, yet unlike etoposide treatment, the concentrations of Na Ars we utilized did not decrease BCL-2 half-life by inducing its proteosomal degradation. Collectively, these results correlate well with our data demonstrating inducible BCL-2 IRES activity by showing that BCL-2 levels were increased in a dose-dependent manner in vivo following stress treatments that likewise induce BCL-2 IRES activity. DISCUSSION We demonstrate here that the 5Ј-UTR of BCL-2 mRNA promotes the internal entry of ribosomes both in vitro (using RRL) and also in vivo. Although the activity of this IRES in unstressed cells was low relative to cap-dependent BCL-2 translation (Fig. 4), IRES activity was stimulated 3-6-fold by treatment with either sodium arsenite or etoposide (Figs. 5 and 6). Our data suggest that this translational stimulation is at least partly mediated by an IRES in the BCL-2 5Ј-UTR.
Transfecting DNA Versus RNA-Transfecting DNA constructs is preferable when identifying an IRES, since the re- sulting transcripts are processed in the nucleus, thereby allowing the association of nuclear proteins that might be important for IRES function. It was clear that using DNA transfection to determine potential BCL-2 IRES activity would require stringent RNA analysis because the BCL-2 5Ј-UTR contains both an alternatively spliced intron and a minor promoter. Indeed, analysis revealed an aberrant splicing that was detectable by RNAi directed against RLuc (Fig. 3B) as well as RT-PCR (Fig.  3C). As a result, we were unable to quantify accurately BCL-2 IRES activity via transfection of DNA. It may have been possible to eliminate splicing by deleting/mutating the splice acceptor present within the BCL-2 5Ј-UTR. However, this approach has been attempted previously using another reported cellular IRES (XIAP), with only limited success (40). It is also unclear whether the alterations necessary to prevent splicing may inhibit IRES function.
We instead chose to assess potential BCL-2 IRES function by transfecting capped and polyadenylated RNA directly into cells, a technique that has been utilized previously with success (47). To ensure maximum stability of the transfected RNA, transfected cells were lysed at the earliest time point that produced sufficient luciferase production for accurate quantitation. The subsequent isolation of transfected radiolabeled mRNA (Figs. 5D and 6D) demonstrated that these transcripts were indeed quite stable, even when lacking an m 7 G cap, and were not spliced. This high level of RNA stability may partly have derived from their lack of any native 3Ј-UTR sequence, which often contains specific sequence elements responsible for regulating native mRNA turnover via exosome-mediated mRNA decay (48,49). It is also possible that direct transfection of test transcripts into the cytoplasm prevented recruitment of nuclear proteins that might tag the transcripts for degradation.
Assessing the Strength of BCL-2 IRES-mediated Translation-The activity of the monocistronic BCL-2 IRES in unstressed cells was only about 9 -10% relative to the total translation of the longest BCL-2 5Ј-UTR tested (Fig. 4B). However, this percentage of total BCL-2 translation would likely be higher relative to the translation rate of the full-length BCL-2 5Ј-UTR, which contains an additional ϳ330-bp GC-rich sequence. In addition, a potential lack of nuclear processing re-sulting from direct RNA transfection may have prevented the association of proteins that could enhance IRES function. Alternatively, IRES-mediated translation of BCL-2 in unstressed cells may simply be unnecessary, as it appears the BCL-2 5Ј-UTR mediates adequate cap-dependent translation in a stress-free environment despite its extreme length (Figs. 4 and 6).
The stress-induced activity of the BCL2 IRES may also appear low when compared with the HCV IRES (Figs. 5 and 6). However, it is possible that the artificial context utilized to characterize the BCL-2 IRES may have prevented its full functionality. Indeed, it has been suggested that both the native protein coding sequence and 3Ј-UTR of an mRNA may significantly influence IRES activity present in the 5Ј-UTR (50,51). In addition, the absence of the first ϳ330 bp of the full-length BCL-2 5Ј-UTR may also have had an effect on IRES activity, as this segment was not included in our constructs. Furthermore, an absence of nuclear processing for the transfected transcripts might also have influenced IRES inducibility. The effect of these various factors is currently being studied.
It has been shown previously that the translation of BCL-2 mRNA in unstressed cells is inhibited by a conserved uORF located Ϫ119 to Ϫ84 bp upstream from the BCL-2 translation start codon (29). Presumably, this inhibition results from the failure of most ribosomes to re-initiate following uORF translation. The implication is that a significant percentage of ribosomes must scan through this portion of the 5Ј-UTR in order to initiate BCL-2 translation. Thus, investigating the effect of this uORF on BCL-2 IRES-mediated translation may provide insight into the exact location of ribosome recruitment within the 5Ј-UTR. These studies are currently in progress.
BCL-2 IRES Activity and Cellular Homeostasis-Maintaining the homeostatic balance between BCL-2 (anti-apoptosis) and BH3 (pro-apoptosis) proteins is critical to preventing cellular entry into apoptosis (1,52). The stressed cell appears to compensate in an attempt to maintain this homeostasis. Transcription of BCL-2 often increases after exposure to various stresses, although this induction appears to be predicated upon prior protein synthesis, and can take 4 -6 h (28, 53). Meanwhile, the half-life of the BCL-2 protein is variable and can be Other cells were simultaneously treated with the proteosome inhibitor MG132 (2 M) for 4 h prior to harvest (5th to 8th lanes). Relative protein loading was determined by re-probing with an ␣-tubulin monoclonal antibody (middle panel). Protein band density was quantified, and relative BCL-2 levels (normalized to ␣-tubulin) are expressed in the bar graph (bottom panel). B, BCL-2 levels in untreated cells (U) versus cells treated with 4, 8, or 16 M Na Ars for 12 h. As before, certain cells were co-treated with MG132 for 4 h (5th to 8th lanes). BCL-2 levels were normalized and displayed in the bar graph (bottom panel).
dramatically shortened by induced stress (54,55). In fact, BCL-2 protein t1 ⁄2 is only 45 min in H-510 small cell lung carcinoma cells (24). Thus, constant replenishment of BCL-2 protein in stressed cells may be critical to maintaining the cellular BCL-2/BH3 balance and preventing premature apoptosis. However, once an induced stress surpasses an apoptotic threshold, the cell commits to executing the apoptotic program. Levels of BCL-2 transcript are then rapidly reduced because of a destabilizing AU-rich element in its 3Ј-UTR (15), whereas BCL-2 protein is eliminated via either the proteosome or caspase-mediated cleavage (19,56).
Given this complex regulation of BCL-2 transcript and protein, if the translation of BCL-2 were exclusively cap-dependent, even a transient stress-induced inhibition of cellular translation might lead to a rapid and potentially fatal decrease in BCL-2 levels. However, we found that treatment with either Etopo or Na Ars increased BCL-2 protein levels by 2-fold or more (Fig. 7) in an environment where cap-dependent translation was inhibited by over 85% (data not shown). In addition, inhibiting the proteosome-mediated turnover of BCL-2 by using MG132 revealed that the translation of BCL-2 was increased during stress and that after Etopo treatment this increase in BCL-2 translation helped offset an increase in proteosome-mediated BCL-2 degradation (Fig. 7). Collectively, these results are consistent with an IRES-mediated increase in BCL-2 expression as a part of a rapid protective cellular response that helps maintain BCL-2 protein levels during stress. The relative transcription rate of the BCL-2 gene following these stress treatment was not determined, but given the high level of translational inhibition observed, stress-induced increases in BCL-2 transcript alone would be unlikely to explain the increases in BCL-2 protein we observed.
The discovery that an IRES can mediate stress-induced expression of BCL-2 adds significantly to the repertoire of cellular mechanisms that maintain the balance between pro-and antiapoptotic proteins in the cell, thereby regulating apoptotic sensitivity. Our findings complement other recently published data describing the stress-induced IRES-mediated expression of apoptosis regulatory proteins, including cIAP-1, XIAP, c-Myc, Apaf-1, and DAP-5 (42,(57)(58)(59)(60). However, it is important to note that several of these reports did not test IRES activity by transfecting RNA transcripts into cells. Thus, some reports of stress-induced IRES activity may partly reflect aberrant splicing from expression of DNA constructs, as we have demonstrated recently with the XIAP (40), cIAP-1 (42), and BCL-2 5Ј-UTRs (see Fig. 3). Still, the accumulating evidence suggests that the IRES-mediated translation of apoptosis regulatory proteins may be critical for controlling the fate of cells under stress. However, testing the validity of this hypothesis will require further research into the conditions that regulate IRES-mediated translation, as well as the genes under its control.