BaxΔ2 Is a Novel Bax Isoform Unique to Microsatellite Unstable Tumors*

Background: Mutations and alternative splicing often silence tumor suppressor gene expression, promoting tumor development. Results: A novel Bax isoform (BaxΔ2) generated from the combination of a microsatellite deletion and unexpected splicing promotes unconventional cell death. Conclusion: BaxΔ2 is an MSI tumor-specific pro-death Bax isoform. Significance: Alternative splicing of mutated gene can restore tumor suppressor function and can be detrimental to the tumors. The pro-death Bcl-2 family protein and tumor suppressor Bax is frequently mutated in tumors with microsatellite instability (MSI). The mutation often results in a “Bax negative” phenotype and therefore is generally thought to be beneficial to the development of the tumor. Here, we report the identification of a novel Bax isoform, BaxΔ2, which is unique to microsatellite unstable tumors. BaxΔ2 is generated by a unique combination of a microsatellite deletion in Bax exon 3 and alternative splicing of Bax exon 2. Consistently, BaxΔ2 is only detected in MSI cell lines and primary tumors. BaxΔ2 is a potent cell death inducer but does not directly target mitochondria. In addition, BaxΔ2 sensitizes certain MSI tumor cells to a subset of chemotherapeutic agents, such as adriamycin. Thus, our data provide evidence that mutation and alternative splicing of tumor suppressors such as Bax are not always beneficial to tumor development but can be detrimental instead.

The instability of genomic tandemly repeated DNA sequences (microsatellites) is caused by a deficiency in the DNA mismatch repair system (1,2). Microsatellite instability (MSI) 3 exists in many types of cancers including colon, gastric, and endometrial cancer (3)(4)(5)(6)(7). MSI mutations usually result in reading frame shifts with subsequent premature termination, producing nonfunctional, truncated proteins, or nonsense-mediated decay of aberrant transcripts (8). Loss-of-function MSI mutations often occur in a number of susceptible tumor suppressors, such as Bax, and they are generally considered to be beneficial to tumorigenesis.
The pro-death Bcl-2 family protein and tumor suppressor Bax plays a critical role in tumorigenesis by regulating programmed cell death and thereby determining the chemo-sensitivity of certain types of tumors (9 -12). The Bax subfamily includes the prototypical Bax␣ and several alternatively spliced isoforms such as Bax␤, , , ␥, , and ␦ (13)(14)(15)(16)(17)(18)(19). Bax␣ consists of six exons encoding a number of functional domains that regulate Bax subcellular localization and pro-death activity (16,20). Bax can form homodimers or heterodimers with Bcl-2 or other Bcl-2 family members (21,22). Upon stimulation by death signals, Bax translocates to mitochondria, where it forms oligomers and leads to the release of cytochrome c (23)(24)(25). The known Bax␣, Bax␤, and Bax isoforms are all capable of directly targeting mitochondria, but the underlying mechanism has yet to be elucidated (17,26,27).
Exon 3 of the Bax gene contains a microsatellite tract comprised of a cluster of eight guanine nucleotides (G8), which is frequently mutated in MSI tumors (7, 28 -30). For instance, approximately half of MSI colon cancers and one-third of gastric and endometrial cancers contain Bax microsatellite tract mutations corresponding to a single nucleotide deletion in the Bax G8 microsatellite tract, i.e., G8 to G7 (3). This causes a reading frameshift and premature termination, usually leading to a "Bax-negative" phenotype (31)(32)(33). Although loss of Bax expression can promote tumor growth and resistance to chemotherapy (9,10), expression levels of Bax do not always correlate with patient prognosis. Intriguingly, a better prognosis is sometimes seen in some patients with Bax-negative MSI tumors (34 -36). The molecular mechanism of this apparent paradox is unknown.
Here we report the identification of Bax⌬2, a novel Bax isoform produced by the combination of a specific microsatellite mutation and an unexpected alternative splicing event. Unlike Bax␣ and Bax␤, Bax⌬2 does not directly target the mitochondria. As an MSI tumor-specific Bax isoform, Bax⌬2 potently induces cell death through an unconventional death pathway and sensitizes MSI tumors to a subset of chemotherapeutic agents.
were purchased from Santa Cruz Biotechnology. The Bax⌬2 monoclonal antibody (2D4) was generated using the peptide (RGGGFHPGSSRAN) by Precision Antibody (Columbia, MD). Adriamycin and tunicamycin were from Sigma, bortezomib was from Selleck Chemicals (Boston, MA), and etoposide and MG-132 were from Calbiochem. Primary tumor RNA samples were purchased from Bioserve. All cell lines were obtained from ATCC unless otherwise specified. Bax Ϫ/Ϫ MEFs were kindly provided by Dr. Xiao-Ming Yin (Indiana University). LNCaP sublines 104-S, 104-R1, and 104-IS cells were kindly provided by Dr. John Kokontis (University of Chicago).
Bax Isoform Analysis and Bax-GFP Mini-gene Constructs-cDNA transcripts of various cell lines and primary tumors were inserted into pcDNA3.1(Ϫ) or PCR-Script vectors, and random clones were selected for analysis. GFP-tagged Bax isoforms were cloned into pcDNA3.1(Ϫ) with GFP at the C terminus of the Bax isoforms. All of the constructs were verified by direct sequencing and analyzed using Geneious software (Biomatters Ltd). To construct Bax-GFP mini-genes, Bax DNA was extracted from 104-R1 (G7) or PC3 (G8) cells, amplified by PCR using primers located in the 5Ј-UTR of Bax and the 3Ј end of Bax exon 4, and ligated into pEGFP-N1. Primers were designed so that Bax genomic sequences were inserted into pEGFP-N1 in-frame with GFP upon appropriate combinations of microsatellite status and splicing events.
Protein Binding Assay-Protein-protein interactions were assayed using a GST fusion protein pulldown assay. 35 S-labeled Bax␣ or Bax⌬2 were generated using the T7 quick coupled translation/transcription system (IVTT) (Promega, Madison, WI) according to the manufacturer's protocol. GST-Bax␣ and GST-Bcl-2 fusion proteins were purified, incubated with 35 Slabeled IVTT proteins, and then extensively washed to reduce nonspecific binding. Protein complexes were resolved on an SDS acrylamide gel and visualized via autoradiography.
Immunofluorescence and Imaging-Cells were transfected using ExGen500 cationic polymer transfection reagent (Fermentas, Burlington, Canada) or Lipofectamine (Invitrogen) according to the manufacturer's protocols. Transfected cells were cultured on coverslips, fixed with 4% paraformaldehyde, permeabilized, blocked, incubated with appropriate antibodies, and mounted onto slides. The cells were photographed using a Nikon TE2000U fluorescence microscope.
Mitochondria Targeting Assay-Mitochondria targeting was analyzed using purified proteins in a cell-free system (37). Briefly, purified recombinant Bax␣ and Bax⌬2 were incubated with freshly isolated murine mitochondria (43). The mixture was then centrifuged to pellet the mitochondria with any bound protein. The pellet was alkaline-treated (0.1 M Na 2 CO 3 , pH 11.5) to strip off peripherally bound protein from the mitochondria surface to distinguish integrally bound proteins in the mitochondria (43).
Cell Death and Cytotoxicity Assays-Percentage of cell death from GFP fusion construct transfected cells was determined by the number of dead GFP-positive cells out of the total GFPpositive cells. Cytotoxicity was determined by measuring mitochondrial activity of treated cells relative to control using MTS, according to the manufacturer's instructions (Promega).

Identification of Bax⌬2 in Bax-negative MSI Cancer Cells-
Expression of Bax proteins in several Bax-negative cancer cell lines including LS174T, LoVo, and DU145 (7, 38) was undetectable when analyzed by immunoblotting with commonly used anti-Bax antibodies (Fig. 1A). Consistent with the clinical observation that some patients with Bax-negative tumor had better prognosis (35,36,39), we noticed that some of the Baxnegative cancer cells are less aggressive than that of Bax-positive cells as analyzed in cell growth and invasion assay (Fig. 1, B-D). Interestingly, all the Bax-negative cells still expressed Bax mRNA ( Fig. 1E) with various sizes of the Bax transcripts (Fig.  1F). Sequencing analyses revealed that the Bax gene isolated from these Bax-negative cells contained a single deletion in a Bax exon 3 guanine microsatellite tract, reducing G8 tracts to G7. If the G7-Bax pre-mRNA undergoes constitutive splicing, this G8 to G7 deletion should cause a reading frameshift and premature termination in the Bax transcript ( Fig. 2A). However, we found that many G7-Bax transcripts actually underwent an unexpected alternative splicing that eliminated almost all of exon 2 and consequently restored the shifted reading frame at the point of the microsatellite deletion ( Fig. 2A). This results in a novel Bax splicing isoform that lacks exon 2 and instead contains a frame-shifted region corresponding to a unique short peptide at the beginning of exon 3 (Fig. 2, A and B). We designated this isoform as Bax⌬2.
To determine whether the Bax⌬2 transcript can be translated as protein in cells, we generated a monoclonal antibody against the unique peptide in Bax⌬2 ( Fig. 2A). We found that the Bax⌬2 antibody specifically recognized Bax⌬2 but not Bax␣ proteins ectopically expressed in Bax null mouse embryonic fibroblasts (MEFs) (Fig. 2, C and D). Conversely, an antibody (N20) that targets the N terminus of Bax␣ recognized ectopically expressed Bax␣ but not Bax⌬2 (Fig. 2, C and D). Thus, Bax⌬2 is a unique isoform distinct from Bax␣.
Bax⌬2 Induces Cell Death without Directly Targeting Mitochondria-Bax can form homodimers or heterodimers with Bcl-2 to regulate cell death (21). To determine whether Bax⌬2 forms dimers with the Bcl-2 family proteins, we employed an in vitro binding assay, in which GST-Bcl-2 or GST-Bax␣ fusion proteins were used to pull down 35 S-labeled Bax⌬2-GFP or Bax␣-GFP in vitro. We found that both GST-Bcl-2 and GST-Bax␣ were able to pull down [ 35 S]Bax␣-GFP and [ 35 S]Bax⌬2-GFP with similar efficiency (Fig. 3A). These data suggest that like Bax␣, Bax⌬2 forms Bax homodimers and can also heterodimerize with Bcl-2.
Bax␣ is sufficient to induce cell death when it is ectopically expressed (16,40,41). To test whether ectopic expression of Bax⌬2 is able to induce cell death, PC3 cells were transfected with Bax⌬2-GFP, Bax␣-GFP, or the control GFP. Analysis of apoptosis in GFP-positive cells revealed that the death of cells expressing Bax⌬2-GFP was significantly higher than that of cells expressing Bax␣-GFP (Fig. 3C), even though the expression level of Bax⌬2 was lower than that of Bax␣-GFP (Fig. 3B). However, PC3 cells have a high basal level of Bax␣ and perhaps other Bax isoforms. It is possible that Bax⌬2 might utilize Bax␣ or other Bax isoforms to induce cell death. To distinguish these possibilities, Bax Ϫ/Ϫ MEFs were transfected with Bax␣-GFP or Bax⌬2-GFP. Consistently, we found that cell death was significantly higher in Bax null MEFs expressing Bax⌬2-GFP than in cells expressing Bax␣-GFP (Fig. 3, D and E). Furthermore, Bax⌬2-induced cell death is mediated by activation of caspase (Fig. 3F), and both caspase 3 and caspase 8 inhibitors, but not caspase 1 inhibitor, can effectively block Bax⌬2-induced cell death (Fig. 3G). Thus, Bax⌬2 itself is sufficient to induce cell death in a caspase-dependent manner and appears to be a more potent death inducer than Bax␣.
Bax␣ induces cell death via directly targeting mitochondria, where it interrupts mitochondrial integrity (42). To determine the cellular localization of Bax⌬2, Bax Ϫ/Ϫ MEFs were transfected with Bax␣ or Bax⌬2. Cell fractionation assays revealed that Bax␣ predominantly distributed in the cytosolic fractions (C) (Fig. 4A), consistent with previous reports (43). By contrast, Bax⌬2 localized in the membrane fractions (M) (Fig. 4A). To determine whether Bax⌬2 targets mitochondria, we used a cellfree mitochondria targeting system. Purified tag-free Bax␣ or Bax⌬2 recombinant proteins were incubated with purified murine mitochondria, followed by centrifugation to separate the mitochondrial pellet from the supernatant (S). The mito-chondrial pellet was further treated with alkaline to distinguish integral membrane (I) from peripheral membrane proteins (P). We found that Bax␣ spontaneously targeted and integrated into the mitochondria, as expected (Fig. 4B, top panel, lane 8) (23,44) By contrast, Bax⌬2 was found almost exclusively in the supernatant (Fig. 4B, bottom panel, lane 3), indicating that Bax⌬2 did not directly target the mitochondria. The notion that Bax⌬2 and Bax␣ have different subcellular localizations was further supported by live imaging analysis of Bax Ϫ/Ϫ MEFs transfected with Bax␣-GFP or Bax⌬2-GFP. Unlike Bax␣-GFP proteins, which were evenly distributed, Bax⌬2-GFP proteins were clustered in the cytosol (Fig. 4C). More importantly, immunostaining analysis of Bax⌬2 revealed that Bax⌬2 did not co-localize with mitochondria under a resting condition (Fig.  4D, top panels) or upon adriamycin stimulation (Fig. 4D, bottom panels).
To determine whether Bax⌬2 may affect mitochondrial functions, we measured the release of cytochrome c from mitochondria in Bax Ϫ/Ϫ MEFs expressing Bax␣ or Bax⌬2. We found that like Bax␣, ectopic expression of Bax⌬2 also triggered the release of cytochrome c (Fig. 4E). Interestingly, the loss of mitochondrial membrane potential in the Bax⌬2 transfected cells was partially blocked by caspase 8 inhibitor but not caspase 3 inhibitor (Fig. 4F). These data suggest that Bax⌬2 affects mitochondrial functions without directly targeting mitochondria, and the underlying mechanism is partly mediated by caspase 8.
Bax⌬2 Is an MSI Tumor-specific Bax Isoform-The generation of Bax⌬2 needs the unique combination of a microsatellite deletion and atypical exon 2 alternative splicing. We used a panel of MSI tumor cell lines to determine the relationship between the production of Bax⌬2 and the Bax gene microsatellite status (G7, G8, or G9). RT-PCR results revealed that all Bax⌬2-positive cells had a single guanine nucleotide deletion at the exon 3 microsatellite site, i.e., a G7 status, although not all G7 MSI tumor cells had detectable Bax⌬2 transcripts (Table 1). Importantly, Bax⌬2 transcripts were also detected in some G7 MSI primary tumors ( Table 1). Analysis of RNA samples isolated from human prostate cancer and colon cancer revealed that two of five colon samples and three of five prostate tumor samples contained a G7 microsatellite tract. Among these sam-ples, one colon and one prostate tumor also had Bax⌬2 transcripts (Table 1). It is possible that in some G7 MSI tumor cells or primary tumors the copy number of Bax⌬2 transcripts is too low to be detected, or the MSI mutation is necessary but not sufficient to produce Bax⌬2. Future studies are needed to distinguish these possibilities. Taken together, these data indicate that Bax⌬2 exists in primary tumors in addition to MSI tumor cell lines.
To determine whether the alternative splicing machinery involved in generation of Bax⌬2 is MSI tumor-specific, we designed a Bax mini-gene GFP assay (Fig. 5A). The mini-gene was constructed by inserting part of the Bax genomic DNA sequence, including exons 1, 2, and 3 and the 5Ј end of exon 4, into the 5Ј end of a GFP reporter vector, with either a G7 or G8 microsatellite tract in exon 3 (G7 or G8 mini-gene) (Fig. 5A). The mini-genes were transfected into two prostate cancer cell lines, microsatellite G7 Bax⌬2-positive 104-R1 cells (Fig. 5B), or G8 Bax⌬2 negative PC3 cells (Fig. 5C). In the case of constitu- tive splicing, the G8 mini-gene will produce an in-frame GFP fusion protein, whereas the G7 mini-gene generates a frameshift and consequent premature termination codon. However, in the case of exon 2 alternative splicing, the reading frame will be restored in the G7 mini-gene but not in the G8 mini-gene (Fig. 5A). We found that in G7 Bax⌬2-positive 104-R1 cells, both constitutive splicing and alternative splicing occurred, but only the G7 mini-gene product was recognized by the anti-Bax⌬2 antibody (Fig. 5B). The same results were also observed in G8 Bax⌬2 negative PC3 cells (Fig. 5C).
Next we examined the propensity of cells to splice endogenous Bax⌬2 transcripts versus Bax␣ transcripts. Transcripts from Bax␣-positive PC3 and Bax⌬2-positive 104-R1 cells were amplified by PCR using primers surrounding exon 1 and the 5Ј end of exon 4 (Fig. 5D). We found that both cell lines produced endogenous transcripts in which exon 2 was spliced, although this splicing event occurred at a higher efficiency in the G7 Bax⌬2-positive 104-R1 cells (Fig. 5D). It is likely that the trans splicing machinery involved in Bax⌬2 generation exists in both Bax G7 and G8 tumor cells with various splicing efficiencies.
However, the splicing itself without a G7 microsatellite mutation is unable to produce Bax⌬2. Thus, Bax⌬2 is an MSI tumorspecific Bax isoform.
Bax⌬2 Selectively Sensitizes MSI Tumor Cells to a Subset of Chemotherapeutic Agents-To determine whether Bax⌬2-positive cells are more sensitive to cell death stimuli, G7 Bax⌬2positive 104-R1 cells and G8 Bax⌬2-negative PC3 cells were treated with adriamycin (topoisomerase II inhibitor), bortezomib (proteasome inhibitor), and tunicamycin (endoplasmic  . Bax⌬2 is localized in membrane fraction but does not target mitochondria. A, Bax Ϫ/Ϫ MEFs transfected with Bax␣ or Bax⌬2 for 16 h, then fractionated, and analyzed for Bax expression by immunoblotting using anti-Bax␣ antibody (N20) and anti-Bax⌬2 antibody (2D4). Lanes C, cytosol fraction; lanes M, membrane fraction. B, mitochondrial targeting assay (41). Purified tag-free Bax␣ or Bax⌬2 proteins were incubated with purified murine mitochondria, followed by alkaline stripping to distinguish integral membrane (I-Memb) from peripheral membrane (P-Memb) proteins. C, live green fluorescence imaging analysis of Bax Ϫ/Ϫ MEFs expressing GFP, Bax␣-GFP, or Bax⌬2-GFP. D, Bax Ϫ/Ϫ MEFs were transfected with Bax⌬2 and then treated without (top panel) or with (bottom panel) adriamycin (4 g/ml). The cells were stained with MitoTracker (red), fixed, and followed by immunostaining with anti-Bax⌬2 antibody (green); nuclei were stained with DAPI. E, Bax Ϫ/Ϫ MEFs were transfected with GFP, Bax␣-GFP, or Bax⌬2-GFP. The cytosol fractions were collected and analyzed by immunoblotting with anti-cytochrome c antibody. reticulum stress inducer). The cytotoxicity assays revealed that G7 Bax⌬2-positive cells were more sensitive to adriamycin but less sensitive to bortezomib or tunicamycin (Fig. 6A), suggesting that Bax⌬2 may be selectively involved in a certain type of cell death pathway. Interestingly, Bax⌬2-positive cells (104-R1 and LS174T) were more sensitive to adriamycin than Bax␣positive cells (PC3 and SW1116) (Fig. 6B) but were not sensitive to another topoisomerase II inhibitor etoposide (Fig. 6B) (45,46). The sensitivity and selectivity to different chemotherapeutic agents were not the result of up-regulation of Bax⌬2 and Bax␣ or Bcl-2 protein levels (Fig. 6C). To exclude the possibility that the cell type differences may count for the different chemosensitivities, Bax⌬2-GFP-transfected PC3 cells were treated with etoposide or adriamycin. We found that Bax⌬2 selectively sensitized the cells to adriamycin-but not etoposide-induced cell death (Fig. 6D). Thus, Bax⌬2 may determine the chemoselectivity through distinct death pathways in the MSI tumor cells.

DISCUSSION
Bax␣ and other Bax isoforms induce apoptosis by targeting mitochondria (17,27,42). In contrast, Bax⌬2 induces apoptosis without directly targeting mitochondria (Figs. 3 and 4). Structurally, Bax⌬2 is almost identical to Bax␣ except for the deletion of amino acids 13-38 encoded by exon 2 and the addition of 10 amino acids generated by the frameshift within this region ( Fig. 2A). The deletion of amino acids 13-38 may be responsible for the inability of Bax⌬2 to target mitochondria, because the amino acids 16 -35 are part of the first ␣-helix of Bax␣ that is required for its mitochondrial targeting (27,(47)(48)(49). However, loss of the first ␣-helix retains Bax␣ in the cytoplasm (27), whereas Bax⌬2 localizes in nonmitochondria membrane fractions, not the cytoplasm (Fig. 4A). It is possible that the addition of the unique short peptide in Bax⌬2 (Fig. 2B) could target it to a yet to be identified membrane compartment. Interestingly, Bax⌬2-induced cell death is accompanied by the release of cytochrome c (Fig. 4E) and loss of mitochondrial membrane potential, which can be partially blocked by caspase 8 inhibitor but not caspase 3 inhibitor (Fig. 4F). Future studies are needed to determine the exact subcellular target of Bax⌬2 and the underlying mechanism by which Bax⌬2 induces cell death. Ϫ 104-S1 Prostate Colon (III) G8 Ϫ  with GFP or Bax⌬2-GFP followed by treatment with or without etoposide (40 M) or adriamycin (4 g/ml). Cell death was measured as described in Fig. 3C. *, p Ͻ 0.05; **, p Ͻ 0.001. Ctrl, control.
Like the prototypic Bax␣, Bax⌬2 is involved in determining the sensitivity of tumor cells to chemotherapeutic agents (Fig.  6). However, Bax⌬2 sensitizes G7 MSI tumor cells to adriamycin but not etoposide, although both are topoisomerase II inhibitors (Fig. 6). It is likely that Bax⌬2 may be involved in an Adriamycin-induced nonmitochondria death pathway rather than promoting the inhibition of topoisomerase II by adriamycin. Interestingly, a side effect of adriamycin treatment is severe cardiotoxicity, which is somewhat attenuated in etoposidetreated patients, suggesting that these topoisomerase II inhibitors do not act in exactly the same manner (50 -53). The selective chemosensitization effect of Bax⌬2 may provide a unique prognostic biomarker for treatment of certain type of MSI tumors.
Bax⌬2 is a G7 MSI tumor-specific Bax isoform ( Fig. 2 and Table 1). Even though G8 MSI tumor cells are capable of carrying out the alternative splicing to delete exon 2 (Fig. 5), without the guanine nucleotide deletion in the microsatellite tract, the frameshift will result in a premature termination codon and presumably nonsense-mediated decay of the transcript ( Fig.  2A). Although expression levels of Bax⌬2 are quite low in MSI tumor cells (Fig. 6), because it is a potent death inducer when overexpressed and therefore is likely detrimental to the tumors, it may be possible to up-regulate Bax⌬2 expression by inducing the transition of G8 to G7 or modulating the splicing machinery to enhance exon 2 splicing, thereby triggering or sensitizing cell death in certain type of MSI tumors. Future studies are needed to explore these possibilities.