A Potent Cell Death Activity Associated with Transient High Level Expression of BCL-2*

The BCL-2 proto-oncogene contains unusually long untranslated 5′ and 3′ sequences. Deletion of the sequences flanking the BCL-2 open reading frame dramatically increases the level of protein expression. Transient high level BCL-2 protein expression mediated by plasmid transfection or by infection with recombinant adenovirus results in potent apoptosis of several cell lines. Detailed mutational (deletion and add-back) analysis reveals that both 5′- and 3′-flanking sequences contribute to the negative modulation of protein expression from the BCL-2 open reading frame. It appears that these sequences exert the negative regulatory effect in an orientation-dependent manner. Analysis of BCL-2RNA levels indicate that elevated levels of mRNA may be the primary cause of elevated levels of protein expression. Apoptosis induced by adenovirus vectors expressing elevated levels of BCL-2 can be readily inhibited by the caspase inhibitor z-VAD-fmk, suggesting that high levels of BCL-2 expression induce apoptosis via the caspase cascade. Mutational analysis of BCL-2 indicates that its pro-apoptotic activity is separable from its anti-apoptosis activity. Our results raise the possibility that oncogenic conversion of BCL-2 may require somatic mutations in the pro-apoptotic activity, in addition to other activating mutations that result in enhanced expression. Consistent with this hypothesis, a somatic mutation of BCL-2 observed in multiple human tumors results in reduced apoptosis activity.

It is well established that the BCL-2 proto-oncogene suppresses apoptosis induced by a multitude of stimuli in a variety of cell types (1,2). In a seminal study by Vaux et al. (3), BCL-2 was shown to extend the survival of growth factor-dependent hematopoietic progenitor cells after factor withdrawal. In subsequent studies by Hockenbery et al. (4), it was shown that BCL-2 promotes survival of such factor-dependent cells by suppressing apoptosis. Most studies on the anti-apoptosis activity of BCL-2 have been based on stable cell lines (selected after transfection of BCL-2 expression vectors) that ectopically express elevated levels of BCL-2. In addition to the anti-apoptosis activity, recent studies have revealed that BCL-2 exhibits other novel activities. For example, BCL-2 has been shown to inhibit cell proliferation (5)(6)(7)(8)(9), which is a function separable from the anti-apoptosis activity (6,9). BCL-2 has also been shown to promote cell proliferation in the central nervous system (10).
The human BCL-2 gene is characterized by unusually long 5Ј-and 3Ј-untranslated regions (11)(12)(13), including the entire first exon (14). The functions of the unusually long untranslated regions are not known. Interestingly, the 5Ј-untranslated region contains a sequence element that negatively regulates translation of BCL-2 (15). Here, we show that deletion of the 5Јand 3Ј-flanking sequences results in a large increase in BCL-2 protein expression, which results in cell death via the CED-3/ ICE family caspase cascade.
Adenovirus Expression Vectors-The BCL-2 expression cassettes, from plasmids pcDNA3-Bcl2 and pcDNA3-Bcl2⌬UTR, and the lacZ expression cassette, from plasmid pCMV-␤-galactosidase, were cloned into the adenovirus transfer vector pAdLTR (17) between unique XhoI and BglII sites. To construct the recombinant adenoviruses, 5 g of each of the transfer plasmid was cotransfected with 5 g of the adenovirus genomic plasmid pBHGE3 (18) into 293 cells by the calcium phosphate method. Seven days after transfection, plaques were picked and screened for the presence of the transgene by restriction analysis. Positives plaques were further purified through a second plaque assay, amplified and titrated in 293 cells. In these recombinants, the transgenes are expressed under the control of CMV-IE promoter and interleukin-2 poly(A) site.
Cell Death Assays-Transient cell death assays were carried out using the human 293 cell line. Cells were transfected at 60% confluency with 1.6 g of various expression plasmids and 0.4 g of the reporter plasmid pCMV-␤-galactosidase by the calcium phosphate method in 12-well plates. Eight hours after transfection cells were fixed and stained with X-gal (19). The number of blue cells was counted, and the number of round (apoptotic) cells compared with the total blue cell count was determined.
Two different cell death assays were used to determine the effect of adenovirus vectors. In the first assay, cell viability was determined by vital stain exclusion (trypan blue). In the second assay, cell viability was determined by measuring mitochondrial viability (MTT assay). In the vital dye exclusion assay, 5 ϫ 10 5 cells/35-mm 2 dish were infected with 100 plaque-forming units of various viruses in 2 ml of DMEM containing 2% fetal calf serum. Twenty-four hours after infection, adherent cells were collected by treatment with 0.5 ml of trypsin-EDTA solution and combined with the floating cells collected from the culture media. The suspension was mixed with an equal volume of trypan blue (0.4% solution in phosphate-buffered saline). Cells that excluded the dye were counted. For MTT assays (20), cells were plated in 96-well tissue culture plates in DMEM containing 10% fetal calf serum and incubated overnight. The following day, triplicate wells of cells at approximately 50% confluency were infected with various multiplicities of infection resulting in an infection rate of Ն90% (as determined by X-gal staining of companion cultures infected with Ad-lacZ). Viruses were added in 50 l of DMEM containing 2.5% fetal bovine serum for 4 h at 37°C. An additional 150 l of DMEM containing 2.5% fetal bovine serum were then added, and the cells were cultured overnight. After an overnight infection, the media were replaced with 200 l of DMEM containing 10% fetal calf serum. At 45-48 h post-initiation of infection, 50 l of MTT (Sigma M-5655, 5 mg/ml stock in phosphate-buffered saline, diluted to 1 mg/ml of culture medium) were added and incubated for additional 4 h. Media were aspirated, cell monolayers were solubilized in 150 l of Me 2 SO, and the absorbance at 545 nm was determined. The data are plotted as percentages of the average of 12 wells of noninfected cells used as the control.
Analysis of DNA Fragmentation-293 cells (1 ϫ 10 6 cells/60-mm dish) were infected at 100 plaque-forming units/cell. Twenty-four hours later, adherent and floating cells were collected and lysed, and small molecular weight DNA was prepared by Hirt extraction (21), treated with RNase, and analyzed on a 1% agarose gel.
Western Blot Analysis-3.4 ϫ 10 4 MCF-7 cells were plated in 24-well plates in DMEM containing 10% fetal calf serum. Twenty-four hours after plating, cells were infected at 300 plaque-forming units/cell in 250 l of serum-free DMEM for 4 h at 37°C. Virus-containing media were aspirated and replaced with fresh DMEM containing 2.5% fetal calf serum. At various times post-infection, adherent cells from triplicate wells were harvested and pooled, and lysed in SDS-polyacrylamide gel electrophoresis sample buffer, and the equivalent of 7.5 ϫ 10 4 adherent cells were loaded per lane of a 16% minigel (Novex). Following SDSpolyacrylamide gel electrophoresis, the proteins were electrotransferred onto a 0.2-m nitrocellulose membrane and probed with a mouse anti-human BCL-2 primary antibody (Dako, MO887) at 1:300 dilution, followed by a sheep anti-mouse-horseradish peroxidase secondary antibody (Amersham NA931 used at 1:5000 dilution) and chemiluminescent detection (Amersham Corp.). For the analysis of transiently transfected cells, 293 cells were transfected at 60% confluency in 35-mm dishes with 4 g of total DNA. Nine and a half hours later, cells were lysed in 100 l of electrophoresis sample buffer, and 10-l samples were loaded onto a 15% gel SDS-polyacrylamide gel electrophoresis, followed by transfer to a nitrocellulose membrane. A rabbit anti-glutathione S-transferase-BCL-2 polyclonal serum (1:1000) and donkey anti-rabbithorseradish peroxidase (Amersham NA934 at 1:3000) were used to detect BCL-2.
Northern Blot Analysis-Total cytoplasmic RNA was isolated using RNAzol B solution (TEL-TEST Inc.) and purified by CsCl density gradient centrifugation. 12 g of RNA were loaded in each lane, subjected to electrophoresis in a formaldehyde-agarose gel, transferred to a Duralon nylon membrane (Stratagene), and hybridized with the BCL-2 probe. The probe was prepared by random primed extension (DuPont) and the HindIII/XbaI fragment from pcDNA3-Bcl2⌬UTR as the template.

Transient Expression of BCL-2-
The human BCL-2 cDNA contains 1.4 kb of 5Ј-and 3.9 kb of 3Ј-untranslated sequences (Fig. 1A). A widely used cDNA clone (clone no. 58) (14) contains 58 bp of 5Ј-and 900 bp of 3Ј-untranslated sequences. We constructed a derivative of this cDNA clone by PCR that lacks the untranslated sequences (pcDNA3-Bcl2⌬UTR). We observed that, after transfection of pcDNA3-Bcl2⌬UTR, the number of G418-resistant colonies induced was substantially lower than that induced either by the pcDNA3 empty vector or the vector expressing the parental cDNA (pcDNA3-Bcl2). This finding prompted us to examine whether the clone lacking the untranslated regions (pcDNA3-Bcl2⌬UTR) caused any cell death in transfected cells. For this purpose, 293 cells were transfected with a reporter plasmid expressing the Escherichia coli lacZ gene along with the pDNA3 vector or pcDNA3-Bcl2 or pcDNA3-Bcl2⌬UTR. Cells expressing ␤-galactosidase were detected by X-gal staining, and the percent of apoptotic cells was calculated from the number of round and fragmenting blue cells out of the total number of blue cells (19). Transfection of the pcDNA3 vector or the vector expressing the parental cDNA (pcDNA3-Bcl2) did not induce significant cell death (Fig. 1B).
In contrast, about 80% of the cells transfected with pcDNA3-Bcl2⌬UTR were round. The observed cell death activity of BCL-2 was confirmed with two different independently derived clones that lack the flanking sequences (not shown). None of the pcDNA3-Bcl2⌬UTR clones contained any mutations compared with parental cDNA. Similar results were also obtained with a different human BCL-2 cDNA (11) and a mouse BCL-2 cDNA (22).
Adenovirus Vector-mediated Expression of BCL-2-To further substantiate the above results, we constructed recombinant adenoviruses that express the two versions of the BCL-2 cDNA. The effects of the recombinant adenoviruses on cell viability of three different human cell lines were determined by the MTT assay. Infection of cell lines MRC-5, MCF-7, and DU145 with an adenovirus vector expressing the E. coli lacZ gene (Ad⌬E1-lacZ) did not result in significant cell death compared with uninfected control cells (Fig. 2). Similarly, infection with the vector that expresses the parental cDNA clone (Ad⌬E1-Bcl2) also did not result in any significant cell death. In contrast, infection with the adenovirus vector expressing the BCL-2 cDNA lacking the flanking sequences (Ad⌬E1-Bcl2⌬UTR) resulted in severe cell death in MRC-5 ( Fig. 2A). Similarly, infection of MCF-7 and DU145 cells with Ad⌬E1-Bcl2⌬UTR also resulted in significant cell death (Fig. 2, B and  C), although the effect on DU145 was less pronounced (Fig.  2C). These results indicate that the BCL-2 cDNA lacking the 5Ј-and 3Ј-flanking sequences induce cell death during transient expression. It should be noted that the DU145 cell line is resistant to apoptosis induced by Fas and tumor necrosis factor-␣ (23).
BCL-2 mRNA and Protein Expression-To determine the possible mechanism by which the flanking sequences dictate the opposing activities of BCL-2, we determined the levels of expression of BCL-2 mRNA and protein. These levels were determined from MCF-7 cells infected with Ad⌬E1-Bcl2⌬UTR or Ad⌬E1-Bcl2. Total cytoplasmic RNA was extracted from infected cells (adherent and floating) and analyzed by Northern blot analysis. In cells infected with Ad⌬E1-Bcl2, an mRNA band of 2.6 kb (that corresponds to the BCL-2 ORF and flanking vector sequences) was observed (Fig. 3A). In cells infected with Ad⌬E1-Bcl2⌬UTR, significantly higher levels of a 1.2-kb mRNA were observed. These results suggest that deletion of the flanking sequences of the BCL-2 cDNA significantly increases the level of BCL-2 mRNA expression.
The levels of BCL-2 protein expression in cells infected with the adenovirus vectors expressing the two versions of the BCL-2 cDNA were also analyzed (Fig. 3B). Adherent and floating cells were collected at various times after infection and analyzed by Western blot analysis. At various times after infection, the level of BCL-2 protein expression was consistently higher in cells (both adherent and floating) infected with Ad⌬E1-Bcl2⌬UTR. Similar results were also obtained with two other cell lines infected with adenovirus vectors and in cells transiently transfected with plasmid expression vectors (results not shown). These results clearly indicate that deletion of the flanking sequences of the BCL-2 cDNA dramatically increases the level of mRNA and protein expression.
Mutational Analysis of the Flanking Sequences-To determine the role of the 5Ј-and 3Ј-flanking sequences separately, we constructed a series of add-back mutants by adding either the 5Ј-or the 3Ј-flanking sequences (by PCR) to pcDNA3-Bcl2⌬UTR (Fig. 4A). Similarly, a series of 5Ј-or 3Ј-deletion mutants of pcDNA3-Bcl2 were also constucted by exploiting conveniently located restriction sites (Fig. 4B). The cell death activities of the various mutants were then determined by transient transfection (Fig. 4D) as described in the legend to  Fig. 4D) reduced the cell death activity of pcDNA3-Bcl2⌬UTR. Similarly, addition of the 3Ј-flanking sequences also reduced cell death activity (ϩ3Ј in Fig. 4D). Addition of both 5Ј-and 3Ј-flanking sequences significantly reduced the cell death activity, close to the level of pcDNA3-Bcl2. Comparable results were obtained by analysis of deletion mutants (Fig. 4B). Deletion of the 5Ј-untranslated region (Ϫ7; EcoNI) increased the apoptosis activity of pcDNA3-Bcl2. Similarly, deletion of the 3Ј-flanking sequences (ϩ752; PstI) also increased the apoptotic activity. Deletion of both 5Ј-and 3Јflanking sequences increased the apoptotic activity in an additive manner. These results provide evidence that both 5Ј-and 3Ј-flanking sequences have a combinatorial effect in modulating the cell death activity of BCL-2.
To determine if the flanking sequences have any orientationindependent effect, we examined the effect of a construct in which the orientation of the 3Ј-flanking sequence was inverted (ϩ3Ј-UTR-I in Fig. 4C). Similarly, we also examined the effect of a construct in which we substituted the 3Ј-flanking sequence with a heterologous DNA (ϩ3Ј-UTR-H in Fig. 4C). The apoptotic activities of both these constructs were similar to that of the construct that lacks the 3Ј-UTR (⌬3Ј-UTR), suggesting that the flanking sequences may function only in transcriptional orientation (Fig. 4E). Effect of BCL-2 Mutations-To determine if the apoptotic activity of BCL-2 is linked to its anti-apoptotic activity, we examined the effect of two previously characterized mutants, G145E (BH1) and W188A (BH2) which were found to be severely defective in the anti-apoptotic activity (24). The BH2 mutant (W188A) did not significantly affect the apoptotic activity (Fig. 5). In contrast, the BH1 mutant (G145E) exhibited significantly reduced apoptotic activity (Fig. 5). These results suggest that the pro-and anti-apoptotic activities of BCL-2 are separable and that a protein region located around the BH1 domain may be important for the apoptotic activity.
Effect of Caspase Inhibitor-Since CED-3/ICE family caspases are activated during apoptosis induced by several different stimuli (reviewed in Refs. 25 and 26), we determined if the cell death induced by higher levels of BCL-2 expression is also dependent on caspases (Fig. 6, A and B). 293 cells infected with Ad⌬E1-BCL-2⌬UTR induced extensive internucleosomal DNA fragmentation (Fig. 6A, left panel). In contrast, there was no significant DNA fragmentation in mock-infected cells or in cells infected either with Ad⌬E1-lacZ or Ad⌬E1-BCL-2. We then examined if the caspase inhibitor z-VAD-fmk can inhibit DNA fragmentation induced by Ad⌬E1-BCL-2⌬UTR. Addition of 25 or 50 M z-VAD-fmk significantly inhibited the extent of DNA fragmentation induced by Ad⌬E1-BCL-2⌬UTR. Similarly, addition of z-VAD-fmk also inhibited the extent of cell death (determined by vital dye exclusion) induced by Ad⌬E1-BCL-2⌬UTR (Fig. 6B). These results suggest that a distal step in the cell death pathway activated by infection with Ad⌬E1-Bcl2⌬UTR may involve caspases.
Effect of Oncogenic Mutant-Expression of BCL-2 is activated in a number of human tumors. Our results raise the possibility that oncogenic conversion of BCL-2 may require somatic mutations in the pro-apoptotic activity, in addition to other activating mutations that result in enhanced expression. To test this hypothesis, we examined the apoptotic activity of a BCL-2 mutant, P59S, located within the nonconserved region of BCL-2. This mutant was chosen since mutations involving the Pro 59 residue (Pro 59 3 Ser or Pro 59 3 Leu) have been detected in three different human non-Hodgkin's lymphomas (27,28). The P59S mutant has previously been shown to efficiently protect against apoptosis (27). 2 Transient transfection of the plasmid containing this mutant (pcDNA3-Bcl2⌬UTR/ P59S; lacking the flanking sequences) or pcDNA3 Bcl2⌬UTR expressed elevated levels of BCL-2 compared with cells transfected with pcDNA3-Bcl2 (wild type) (Fig. 7A), as expected. Transfection of pcDNA3-Bcl2⌬UTR/P59S consistently exhibited less pronounced cell death compared with pcDNA3-Bcl2⌬UTR (Fig. 7B). These results suggest that at least one of the somatic mutations observed in human tumors exhibit a reduced apoptosis activity.

DISCUSSION
In this communication, we have identified a surprising cell death activity of BCL-2 manifested during high levels of transient protein expression. We believe that a similar activity has not thus far been identified since most assays used to determine the activity of BCL-2 have employed cell lines that have been selected to grow in tissue culture. It should be noted that cell lines established by transfection of pcDNA3-Bcl2⌬UTR also protect against cell death similar to cell lines established by transfection of pcDNA3-Bcl2 (not shown), suggesting that 2 E. J. Uhlmann and G. Chinnadurai, unpublished results. cells transfected with pcDNA3-Bcl2⌬UTR get selected for survival during the selection procedure. We believe that our results may, at least partially, explain the effect of BCL-2 reported by Pietenpol et al. (5). These investigators have observed that transfection of a BCL-2 expression vector in certain tumor cell lines strongly inhibited G418-resistant colony formation.
Employing one of the tumor cell lines, SW480, used by these investigators, we have observed that transfection of pcDNA3-Bcl2⌬UTR strongly reduced the G418-resistant colony number compared with cells transfected with pcDNA3 or pcDNA3-Bcl2 (not shown).
Our study does not address the physiological relevance of the cell death activity of BCL-2. However, the observation that CED-9, the Caenorhabditis elegans anti-cell death protein also contains a minor cell death activity (29) raises the possibility that the cell death activity of BCL-2 may also be physiologically relevant under certain yet unidentified settings. It could be speculated that the unusually long flanking sequences of the BCL-2 gene might have evolved to regulate the opposing activities of BCL-2 protein. Under certain physiological or pathological conditions, shorter transcripts may be generated, and such transcripts may be preferentially translated leading to cell death rather than survival. There is some basis for such a speculation, because the BCL-2 gene has also been shown to be transcribed from a minor promoter (30), resulting in a transcript that would have a 5Ј end similar to the 5Ј ends of the mutants described here. The presence of a separate cis-acting translational inhibitory sequence in the 5Ј-untranslated sequences (15) also lends support for negative regulation of BCL-2 protein expression by flanking sequences. The modulation of BCL-2 expression by the flanking sequences would be similar to the role of the 3Ј-flanking sequences in regulation of p53 protein expression in response to DNA damage (31).
In the present study, we have observed that deletion of the UTRs enhanced BCL-2 mRNA accumulation and protein expression. It is possible that transcriptional and post-transcriptional regulation such as mRNA stability may contribute to increased accumulation of BCL-2 mRNA. Although we have observed both elevated levels of mRNA and protein expression, the role of flanking sequences in negative modulation of translational efficiency cannot be ruled out. Interestingly, we have observed that both 5Ј-and 3Ј-flanking sequences of the BCL-2 mRNA could potentially form extensive secondary structures (not shown). It is possible that such secondary structures may also suppress translational efficiency of the BCL-2 mRNA.
The molecular basis by which elevated levels of BCL-2 induce cell death is not known. It is possible that, during high level expression of BCL-2, N-terminal-deleted versions are expressed from internal translational initiation sites. Under certain conditions, we have observed a 24-kDa polypeptide (see Fig. 3B), in addition to the 26-kDa polypeptide. We believe that the 24-kDa polypeptide is translated from an internal initiation codon corresponding to Met 16 since a BCL-2 mutant ⌬UTR/ M16L does not produce the 24-kDa band (not shown). The mutant M16L (Met 16 3 Leu) still induces cell death, suggesting that the cell death activity observed in our studies may be primarily contributed by the 26-kDa BCL-2. However, we cannot fully rule out the role of other truncated versions generated during high levels of expression being pro-apoptotic. A hypothesis would be that higher levels of BCL-2 might result in formation of dimers and oligomers that may be pro-apoptotic. Under conditions of relatively lower levels of protein expression, BCL-2 may function as a monomer with anti-apoptosis activity. In this context, it has been shown that BCL-x L folds as a monomer in solution (32). Recent studies indicate that both BCL-x L (32) and BCL-2 form membrane pores (33,34), and BCL-2 has been shown to block leakage of cytochrome c into the cytosol (35,36). BCL-2 has been shown to form channels with larger conductances, suggesting the formation of BCL-2 dimers and oligomers in the membrane (33). It is possible that, above a threshold level, BCL-2 may cause constitutive loss of cytochrome c or other apoptotic factors from the mitochondria by forming megapores. BCL-2 retains the BH3 domain essential for the death activity of the pro-apoptotic proteins of the BCL-2 family (37-39). Our mutational analysis suggests that pro- FIG. 5. Effect of BCL-2 mutants on cell death. ⌬UTR versions of mutants G145E and W188A were constructed by PCR using appropriate mutagenic, and the cell death activities were determined in 293 cells.
FIG. 6. Effect of caspase inhibitor. A, DNA fragmentation assay. 293 cells were infected with recombinant adenoviruses, and the low molecular weight DNA was analyzed by agarose gel electrophoresis. B, cell viability assay. Cell viability was determined by trypan blue exclusion assay. apoptotic and anti-apoptotic activities of BCL-2 are separable. It remains to be determined if the cell death activity of BCL-2 is mechanistically similar to the known pro-apoptotic proteins. Recent studies from several laboratories suggest that CED-9 promotes cell survival by simultaneously sequestering the cell death proteins CED-4 and CED-3 (reviewed in Refs. 40 -42). If BCL-2 also functions by a similar mechanism, it is possible that homodimerization of BCL-2 (at higher levels of expression) may result in unsequestered BCL-2 interacting pro-apoptotic proteins resulting in cell death.
The two opposing activities of the BCL-2 gene appear to be similar to the activities of other oncogenes, such c-myc and E1a, which promote cell proliferation as well as cell death. Interestingly, two pro-apoptotic proteins of the BCL-2 family, BAK (43) and BAX (44) have been reported to suppress apoptosis under certain conditions (45,46). As suggested by Evan (47), coupling the two opposing effects induced by a single oncogene gene may be a defense against cancer. It would be expected that the loss of the cell death activity of BCL-2 as a result of somatic mutations may also play a role in oncogenesis in addition to other activating mutations. In a sense, mutations that inactivate the apoptosis activity of BCL-2 could be considered as gain-offunction mutants that contribute to enhanced oncogenesis. Consistent with this prediction, at least one of the somatic mutations expressed at high levels that we have examined results in reduced apoptosis activity compared with comparable levels of wild type.