Exon Skipping in Mcl-1 Results in a Bcl-2 Homology Domain 3 Only Gene Product That Promotes Cell Death*

Mcl-1 is a member of the Bcl-2 family that is regulated transcriptionally and post-transcriptionally, with expression of the full-length Mcl-1-encoded gene product resulting in enhanced cell survival. As reported here, the human Mcl-1 gene can also undergo differential splicing, which yields an internally deleted, death-inducing gene product, Mcl-1s/Δ TM. Whereas full-length Mcl-1 derives from three coding exons (instead of the two present in Bcl-2 and other anti-apoptotic members of this family), the Mcl-1s/Δ TM splice variant results from the joining of the first and third exons with skipping of the central exon. Because of the skipped exon and a shift in the reading frame downstream, the Bcl-2 homology domain (BH3) remains intact, whereas the BH1-, BH2-, and transmembrane-encoding domains do not. Mcl-1s/Δ TM thus has features similar to BH3 only, pro-apoptotic Bcl-2 family members and, accordingly, was found to promote cell death. In addition to a variety of other types of regulation, the Mcl-1 gene appears ideally designed for the generation of either a Bcl-2-like viability promoting or, as reported here, a BH3 only death-inducing gene product.

Apoptosis is a genetically determined cell death program that is critical for the maintenance of tissue homeostasis in the healthy organism. Alterations in apoptosis contribute to the pathological effects seen in a variety of diseases, including cancer and inflammatory disease (1). Apoptosis is controlled by evolutionarily conserved sets of genes with the Bcl-2 family playing a pivotal role (2,3). The members of this family fall into two groups with opposing functions. Pro-survival family members (e.g. Bcl-2, Bcl-x L , and Mcl-1) inhibit apoptosis, whereas death-inducing members (e.g. Bax, Bid, and Bad) have the opposite effect. The family is characterized by domains of high sequence conservation, termed the Bcl-2 homology (BH) 1 do-mains. Bcl-2 and other pro-survival family members contain BH1, BH2, and BH3 domains, which form a hydrophobic cleft (4) and appear critical for anti-apoptotic function (5). Whereas some death-inducing family members contain all three of these domains (e.g. Bax, Bak, and Bok (6 -8)), others contain BH3 only (e.g. Bid, Bad, and Bim). This suggests that BH3 is a minimum requirement for pro-apoptotic function (9,10), which probably relates to the fact that the exposed BH3 domain can bind to the hydrophobic cleft of pro-survival family members (11). In addition to the BH domains, many family members contain a transmembrane (TM) domain at the C terminus. This anchors the protein to intracellular membranes, such as to the outer surface of mitochondria (12)(13)(14)(15). Although this domain may aid in targeting certain family members for function (16,17), a TM domain is not present in some death-inducing members, namely Bid and Bad (3). In sum, the BH1/BH2/BH3 hydrophobic cleft and the TM anchor appear to contribute to anti-apoptotic activity, whereas the BH3 domain is critical for pro-apoptotic function.
Mcl-1 is a Bcl-2 family member that was originally identified as a gene rapidly up-regulated early in the differentiation of a human myeloid leukemia cell line (ML-1) upon induction with 12-O-tetradecanoylphorbol acetate (PMA) (18). Mcl-1 expression has since been found to exhibit differentiation stage-specific expression in a variety of hematopoietic lineages as well as in epithelial tissues (19,20). In addition, Mcl-1 can be rapidly induced through anti-apoptotic cytokine-mediated pathways, such as those stimulated by hematopoietic cell colony-stimulating factors and interleukins (21)(22)(23)(24)(25). For example, in neutrophils (26), external stimuli that prolong the neutrophil life span (such as lipopolysaccharide and granulocyte-macrophage colony-stimulating factor) cause an increase in Mcl-1 expression (23,27). An increase in Mcl-1 expression can be induced through a mitogen-activated protein kinase-dependent signal transduction pathway acting on SRF/Elk-1 (21,28), as well as through an Akt/ cAMP-response element-binding protein-regulated pathway (29). In addition to regulation at a transcriptional level, the Mcl-1 mRNA can be rapidly turned over (30). Furthermore, the upstream half of Mcl-1 contains PEST sequences, and the protein is likewise subject to rapid turn-over (15). The downstream half of the protein contains the BH1, BH2, and BH3 domains and terminates in a TM domain (15,18). Overall, Mcl-1 is a highly regulated gene product and is frequently expressed during particular stages of cell differentiation and/or in response to specific signals.
Increased expression of the endogenous full-length Mcl-1 protein is associated with the maintenance of cell viability, and decreased expression with cell death (22,23,25,31). Accordingly, transfection with a construct representing the full-length gene product results in enhanced cell survival, as seen both in the FDC-P1 hematopoietic cell line and in Chinese hamster ovary cells (32)(33)(34). Enhancement of cell survival is seen in the presence of a variety of apoptosis-inducing stimuli and at levels of the introduced protein that are readily attainable endogenously. Conversely, loss of the expression of Mcl-1 is associated with cell loss (28,34,35). Similarly, in mice that express full-length Mcl-1 as a transgene in hematopoietic tissues (36), lymphoid (B and T) and myeloid cells exhibit enhanced survival, and the life span of mast cells and monocytes is dramatically prolonged. In sum, expression of the full-length BH1/BH2/ BH3/TM-containing Mcl-1 gene product results in enhanced cell survival.
Alternative splicing is increasingly being recognized to play a significant role in the regulation of proteins involved in cell death, occurring in a variety of death receptors, Bcl-2 family members, and cell death effectors such as caspases (37). Bcl-x can undergo alternative splicing to yield two forms (Bcl-x L and Bcl-x s ) that have opposing biological functions. Bcl-x and Bcl-2 contain two coding exons separated by a conserved intron in the BH2 domain near the C terminus (38 -40). Bcl-x L , which is anti-apoptotic, derives from both of the coding exons in their entirety. Bcl-x S , which is pro-apoptotic, results from the use of an alternate upstream splice donor site that does not lie at an intron/exon border but rather lies within the first coding exon (38 -40). Other splice forms of Bcl-x also exist, and thus alternative splicing to a variety of different isoforms can occur even with a relatively simple gene structure (37,41).
We are interested in the role that the Bcl-2 family plays in controlling cell life span in differentiating tissues, such as in the myeloid lineage (both monocytic and granulocytic branches) (42) and in epithelial tissues. Mcl-1 is of particular interest because of its highly regulated pattern of expression in these tissues (20 -25). In the research presented here, we describe Mcl-1 s/⌬TM , a splice variant of Mcl-1 that derives from exon skipping. Although the full-length human Mcl-1 gene product was found to consist of three coding exons (rather that the two coding exons seen in Bcl-2 and Bcl-x), the Mcl-1 s/⌬TM variant represented joining of the first and third exons. This does not affect the BH3 domain in the first exon but does affect the BH1, BH2, and TM domains downstream because of the skipped second exon and a change in the reading frame downstream. The features of the Mcl-1 s/⌬TM -encoded gene product are thus reminiscent of BH3 only pro-apoptotic family members such Bid and Bad. Accordingly, transfection with Mcl-1 s/ ⌬ TM was found to result in cell death. In addition to regulation at the level of both transcription and turn-over, the Mcl-1 gene appears ideally designed for the generation of either a Bcl-2like viability promoting or a BH3 only death-inducing gene product.

MATERIALS AND METHODS
Isolation and Culture of Human Cells and Cell Lines-Human peripheral blood neutrophils and mononuclear cells were isolated from normal healthy donors by dextran sedimentation and plasma-Percoll gradient centrifugation as described previously (43) and cultured in RPMI and 10% fetal calf serum in Teflon pools. Monocyte-derived macrophages were obtained by purification of peripheral blood neutrophils and mononuclear cells and adherence to tissue culture plasticware in serum-free medium for 1 h followed by culture for 4 -5 days in Iscove's modified Dulbecco's medium and 10% autologous serum (44). Established cell lines, ML-1, BL41-3, K562, Jurkat, A549, H441, HepG2, and MCF-7 were cultured in RPMI with 10% fetal calf serum, 10 mM glutamine, and 50 g/ml streptomycin and penicillin. PMA stimulation of BL41-3 cells was performed using 1-50 ng/ml PMA applied for 2 or 6 h.
RT-PCR Detection of Mcl-1 s/⌬TM in Human Cells and Cell Lines-Total RNA was isolated from peripheral blood cells and cell lines by the RNAeasy system (Qiagen). First strand cDNA was generated by oligo(dT)-primed reverse transcription using avian myeloblastosis virus reverse transcriptase (Promega) and under standard conditions (1 g of RNA in a total volume of 50 l). Aliquots (5 l) were subjected to amplification with one of several sets of primers (Mcl-1-F3, 5Ј-GTT GGT  CGG GGA ATC TGG TA; Mcl-1-F4, 5Ј-ATC TCT CGG TAC CTT CGG  GA; Mcl-1-F5, 5Ј-GTA AGG AGT CGG GGT CTT CCC; Mcl-1-R4, 5Ј-AAA TTA ATG AAT TCG GCG GG; Mcl-1-R7, 5Ј-TCC TCT TGC CAC  TTG CTT TTC) (18)) was used to screen a human leukocyte genomic in the EMBL-3 lambda phage vector (CLONTECH catalog no. HL1006d (Palo Alto, CA)), using standard techniques. The resulting positive clones were characterized by restriction enzyme mapping. Selected restriction enzyme fragments were isolated from agarose gels, subcloned into the pBluescript SKIIϩ vector (Stratagene), and sequenced on both strands.
Identification of Mcl-1 Transcriptional Start Sites-Primer extension was carried out using an oligonucleotide primer complementary to the Mcl-1 coding region in its upstream portion (primer #105C primer, 5Ј-CCCCACAGTAGAGGTTGAGTCCGATTACCG-3Ј). Counting the ATG of the initiator methionine as the first through third nucleotides, the primer used represents nucleotides 23-52 (or nucleotides 77-106 of the p3.2 Mcl-1 cDNA (18,21)). After 5Ј-end-labeling with 32 P using T4 polynucleotide kinase, the primer (ϳ2 ϫ 10 5 cpm) was hybridized to total RNA (20 g) from ML-1 cells that had been treated with PMA to increase expression of the Mcl-1 mRNA (18,21). As a control, the primer was hybridized in parallel to yeast tRNA (20 g). Hybridization was carried out at 30°C for 14 h in 10 mM Tris-Cl buffer, pH 8.3, containing 150 mM KCl and 1 mM EDTA. Reverse transcription with avian myeloblastosis virus reverse transcriptase (Promega) was carried out at 42°C for 90 min in 50 mM Tris-Cl buffer, pH 8.3, containing 10 mM MgCl 2 , 50 mM KCl, 10 mM dithiothreitol, deoxynucleotide triphosphates (0.25 mM each), and 0.15 g/l actinomycin D. After stopping the reaction by the addition of EDTA (to 20 M), samples were treated with RNase A (Sigma) at 37°C for 30 min, extracted with phenol/chloroform, precipitated with ethanol, and subjected to gel electrophoresis.
The S1 nuclease protection assay was carried out using a 32 P-labeled single-stranded DNA probe complementary to Mcl-1 in the upstream portion of the coding sequence and extending into the 5Ј-flanking region immediately upstream. This probe was prepared using a template representing single-stranded Mcl-1 genomic DNA derived from an M13 subclone containing the transcript strand of a 567 base SacI-NotI Mcl-1 genomic subclone from this region (clone SN0.6, which extends from the NotI site within Mcl-1 exon 1 (see Fig. 2) to a SacI site ϳ0.6 kb upstream). The 32 P-labeled #105C oligonucleotide primer used above was annealed to this template, and the extension reaction was carried out. The double-stranded DNA product obtained was then cleaved at a DraI site 153 nucleotides upstream from the initiator methionine (18,21) (see also Fig. 3B). After alkaline-denaturing gel electrophoresis, phenol extraction, and ethanol precipitation, the resultant 5Ј-endlabeled single-stranded probe, 205 bases in length (2 ϫ 10 5 CPM), was hybridized with total RNA (30 g) from ML-1 cells treated with PMA (18). As a control, the probe was hybridized in parallel to yeast tRNA (20 g). Hybridization (20 l/reaction) was carried out at 30°C in 40 mM PIPES, pH 6.4, containing 80% formamide, 400 mM NaCl, and 1 mM EDTA. After 14 h, the hybridization reaction mixture was diluted by the addition of 300 l of S1 nuclease buffer (50 M sodium acetate, pH 4.5, containing 0.28 M NaCl and 4.5 mM ZnSO 4 ) to which denatured calf thymus DNA (0.02 g/l) had been added. Digestion with S1 nuclease (300 units) was carried out at 30°C for 60 min and was terminated by the addition of 80 l of stop buffer (4 M ammonium acetate, pH 8.0, containing 20 mM EDTA and 40 g/ml yeast tRNA). After ethanol precipitation and resuspension in formamide, the size of the protected fragments was assessed by gel electrophoresis on a DNA sequencing gel alongside a sequencing reaction carried out with the 105C primer and the SN0.6 M13 clone as template. The products of the primer extension reaction (above) were analyzed on the same gel.
RNase Protection Analysis-RNase protection analysis was performed using the multiprobe RNase protection analysis system from Pharmingen. A specific probe set was generated which contained probes to Bcl-x, A1, and Mcl-1. The Mcl-1 probe corresponded to positions 961 to 1107 in the human Mcl-1 cDNA sequence; this spans a region of the cDNA containing sequence from exons 2 and 3, part of which is deleted in Mcl-1 s/⌬TM. Full-length wild-type Mcl-1 transcripts would be expected to yield a protected product of 146 bp, and the deleted Mcl-1 s/⌬TM transcript a protected product of 109 bp. For positive controls in the hybridization reactions, mRNA was transcribed from the TOPO pCR II constructs containing wild-type Mcl-1 and Mcl-1 s/⌬TM . tRNA was used as a negative control. As a control for RNA loading, samples were also hybridized with the hAPO-2 probe set (Pharmingen). The RNase protection analysis reactions were carried out as described in the protocol provided by the manufacturer, using 1-3 g of total cell RNA and dilutions of the positive control RNAs. The reaction products were resolved on 5% sequencing gels in 0.5ϫ Tris-borate-EDTA, which were then dried and exposed to film at Ϫ80°C.
Western Blotting and in Vitro Translation-For Western blotting, cell extracts were prepared using a bromphenol blue lysis buffer (containing 80 mM Tris/HCl, pH 7.5, 2% SDS, 20% glycerol, 0.001% bromphenol blue, 0.1 M dithiothreitol, 1 M leupeptin, and 2 mM phenylmethylsulfonyl fluoride). Electrophoresis (12% SDS-polyacrylamide gel electrophoresis) was carried out with the equivalent of 1-3 ϫ 10 5 cells being loaded in each lane, prestained rainbow markers (Amersham Pharmacia Biotech) being run with each gel. Following electrophoresis, samples were blotted onto Trans-blot nitrocellulose membranes (Bio-Rad) using a semi-dry blotter, and protein transfer was confirmed by staining with Ponceau S. A rabbit polyclonal antibody raised against a peptide representing amino acids 121-139 of human Mcl-1 (Santa Cruz, SC-819) was used at a final dilution of 1:200, and detection was with ECL (Amersham Pharmacia Biotech). The specificity of this antibody was confirmed using specific blocking peptides corresponding to the peptide used as the initial immunogen (Santa Cruz, SC-819P). To prepare authentic Mcl-1 proteins as standards for Western blots, in vitro transcription and translation were carried out using the wild-type Mcl-1 and Mcl-1 s/⌬TM constructs in TOPO pCR II. A coupled transcription/translation system (Promega) was used according to the manufacturer's instructions with 2-l aliquots of the reaction products being loaded on the Western blots.
Transient Transfection with Mcl-1 s/⌬TM Expression Constructs to Assess Effects on Cell Death-Two types of expression constructs were prepared for use in transfection into mammalian cells; in one case Mcl-1 or the Mcl-1 s/⌬TM variant were fused to green fluorescent protein and in the other case they were expressed in the pCR-3.1 vector. The Mcl-1 RT-PCR reaction products from the 4F/4R primer pair reactions (cloned in pCR3.1) were digested with KpnI (a KpnI site is present in the 4F primer) and HindIII, as was full-length Mcl-1 (Mcl-1 clone p3.2 in pBluescript II SK (18)), and the resulting 5Ј-region of the cDNA was isolated and ligated into pCR3.1 to yield full-length pCR3.1-Mcl-1 and pCR3.1-Mcl-1 s/⌬TM . C-terminal green fluorescent protein fusion protein expression constructs were made by digesting the two Mcl-1 fragments from the pCR3.1 clones using HindIII and PstI, followed by subcloning in frame into pEGFP-C1 (CLONTECH), to yield pEGFP-C1-Mcl-1 and pEGFP-C1-Mcl-1 s/⌬TM . The resulting clones were sequenced using flanking and internal oligonucleotides.
The pEGFP-based vectors (pEGFP-C1-Mcl-1, pEGFP-C1-Mcl-1 s/⌬TM , and insertless pEGFP-C1; 0.25 g each) were transiently transfected into subconfluent monolayers of the A549 human airway epithelial cell line in chamber slides using LipofectAMINE. Following overnight culture, transfected cells expressing the fusion proteins were examined for the morphological features of cell death using confocal microscopy. The pCR3.1-based Mcl-1 clones were transiently transfected into subconfluent monolayers of either A549 or HeLa cells (in 12 well plates) using Superfect (Qiagen). The constructs to be tested (pCR3.1-Mcl-1, pCR3.1-Mcl-1 s/⌬TM , pcDNABax as a positive control, or insertless pCR3.1 vector as a negative control; 1 g each) were transfected along with pCMV-␤galactosidase (1 g; Promega) as a marker of transfection. The use of a total of 2 g of DNA/well was based on a series of initial transfections. Forty hours after transfection, cells were fixed, stained for ␤-galactosidase activity, and examined as above for the morphologic features of cell death. This analysis was performed in a double blind manner, and cells were scored either as having a normal, viable morphology (flat, healthy cells) or an abnormal morphology characteristic of dying cells (dense, rounded, detaching cells). All blue staining cells in each dish were counted, and the results were expressed as percentages of cells with viable versus dying phenotypes.

A Variety of Cells Contain an Internally Deleted Variant, Mcl-1 s/⌬TM , Along with the Full-length Mcl-1 Transcript-Sev-
eral Mcl-1-encoding cDNA clones in the GenBank TM human EST data base were noted to contain an internal 248-bp deletion corresponding to nucleotides 749 -996 of the full-length Mcl-1 cDNA. This suggested that Mcl-1 might be capable of undergoing some form of alternative splicing, as does Bcl-x. Several such internally deleted Mcl-1 cDNAs were present in three distinct cDNA libraries, lending credence to this possibility (accession AA457098, AA749362, AA521010, and AI435426). While a distinct transcript representing such a deletion had not been distinguished previously on Northern blots, this is not surprising as a 248-bp deletion would represent a decrease in MW of only ϳ7% as compared with the full-length 3.8-kb Mcl-1 transcript (18,30), and probes recognizing the putative deleted transcript would also recognize full-length Mcl-1. We thus tested for the existence of such an internally deleted Mcl-1 by using primers from either side of the putative deletion in RT-PCR. Two distinct cDNA products were obtained, one of which corresponded to the size expected for full-length Mcl-1 and the other of which was ϳ250 bp shorter (Fig. 1A). Two such cDNA products were obtained using three different combinations of primers (F3 and R4, F4 and R4, and F5 and R7). This result was seen using RNA from a variety of different types of cells, including primary hematopoietic cells (e.g. neutrophils) as well as cell lines of both hematopoietic and nonhematopoietic (e.g. epithelial) origin. The smaller of the cDNA products consistently stained less intensely than the longer product, suggesting that it could represent a transcript of lower abundance as is seen with Bcl-x s , which is generally less abundant than Bcl-x L .
Cloning and sequencing of the above cDNA products demonstrated that the longer product matched the full-length Mcl-1 sequence, whereas the shorter product contained the internal deletion of nucleotides 749 -996. This deletion lies within the protein coding sequence, which is encoded by nucleotides 61-1110 of the full-length Mcl-1 cDNA (#L08246). Both the shorter and the longer product contained the Mcl-1 start site of translation (18) as well as a stop codon, as seen upon sequencing of the products obtained with primers (F5 and R7) representing regions upstream (F5) and downstream (R7) of the Mcl-1 coding sequence. In sum, the shorter cDNA appeared to represent a variant Mcl-1 containing a 248-bp deletion within the coding sequence.
Except for the deletion, the nucleotide sequence of the variant Mcl-1 matches that of the full-length cDNA; however, the sequence downstream of the deletion is in a different reading frame in the variant. Therefore, the amino acid sequence predicted for the variant is identical to that of full-length Mcl-1 upstream of the deletion but differs at the C terminus. Specifically, the first 229 (out of 271) amino acid residues of the internally deleted variant are identical to full-length Mcl-1, whereas the C-terminal 42 residues are not (Fig. 1B). The deletion thus does not affect the upstream portion of Mcl-1 encoding the PEST sequences and the BH3 domain but does affect the downstream portion encoding the BH1, BH2, and TM domains (Fig. 1, B and C). Not only does the C terminus of the deleted gene product differ from full-length Mcl-1, but it also does not match other proteins in the data bases. Because the internally deleted variant encodes a shorter protein with an altered C terminus, we term it Mcl-1 s/⌬TM , where s designates short and ⌬TM designates lack of the TM domain. This variant is in some respects similar to Bcl-x s , which retains the BH3 but not the BH1 and BH2 domains. A difference is that the splicing that yields Bcl-x s does not affect the reading frame and therefore does not affect the C-terminal TM domain. However, other pro-apoptotic family members, namely Bid and Bad, contain   (39), and Bcl-x, like other anti-apoptotic family members, contains a single intron within the protein-coding region (near the C terminus) (38 -40, 45, 46). An intron at this position is a highly conserved feature of the Bcl-2 family (37,(47)(48)(49)(50)(51). For clues as to how the Mcl-1 s/⌬TM variant arises, we characterized ϳ30 kb of the Mcl-1 human genomic locus. In contrast to other antiapoptotic family members, Mcl-1 was found to consist of three coding exons (Fig. 2A). Thus, Mcl-1 contains an intron just downstream of BH3, in addition to the conserved intron further downstream in BH2 (Table I). Exon 1 is G/C-rich and encodes the first 229 amino acid residues containing the PEST sequences and BH3; exon 2 encodes BH1 and a portion of BH2, and exon 3 encodes the remainder of BH2 and the C-terminal TM domain. The Mcl-1 s/⌬TM variant represents exon 1 joined to exon 3 with skipping of exon 2 (Fig. 2C). Exons 2 and 3 are not in the same phase (Table I), which accounts for the shift in reading frame at the C terminus. Whereas Bcl-2 and other anti-apoptotic family members do not contain an intron comparable to a Mcl-1 intron 1, an intron is present at this location in pro-apoptotic Bax (Bax intron 3) (8).

Mcl-1 Transcriptional Initiation Sites Lie Directly Upstream of the First Coding Exon without an Intervening 5Ј-Untranslated Exon-Both
Bcl-2 and Bcl-x contain upstream untranslated exons in addition to the two coding exons (38,40). To look for such upstream exons and to identify the start site(s) of transcription in Mcl-1, we carried out primer extension analysis and S1 nuclease mapping. In primer extension, the most abundant of the products obtained extended to about 70 bp upstream of the Mcl-1 translation start site (Fig. 3A, lane 3); additional products extended 10 bp further upstream (these two positions are indicated with arrows to the left of the sequencing ladder in Fig. 3A). Transcription thus appeared to initiate in the region of two tandem initiator sequences (52) present at this location in the immediate 5Ј-flank of Mcl-1 ( Fig.  3A left column of letters and Fig. 3B, double underlines). The primer extension analysis did not specify initiation exclusively at a single nucleotide within the region containing these tandem sequences. Instead, initiation appears to occur at at least two major sites within an ϳ10-bp stretch, possibly reflecting the presence of two initiator sequences. S1 nuclease mapping also yielded a protected fragment indicative of transcriptional initiation in the region of the initiator sequences (Fig. 3, lane  2). The longest fragment obtained upon S1 nuclease mapping corresponded most closely to the more downstream of the sites indicated by primer extension. Additional shorter fragments were also seen, which could reflect incomplete protection from S1 nuclease digestion. The reason that the more upstream site was not detected by S1 nuclease mapping could relate to the fact that it was associated with less abundant primer extension products. It could also relate to the presence of a repeated sequence in the tandem initiators (i.e. CACTTC), which could allow the formation of a loop upon binding to the probe. Whatever the case, both mapping methods suggested that a majority of Mcl-1 transcripts initiated from the more downstream of the two initiator sequences; this location was therefore designated as position ϩ1 (indicated with an asterisk in Fig. 3). scription of the human Mcl-1 gene (underlined in Fig. 3B (21)). Elements similar to these transcription factor binding sites, as well as a potential initiator sequence, are present in the 5Јflank of mouse Mcl-1 (25,29), although, elsewhere, the latter does not demonstrate the extensive identity to the 5Ј-flank of human Mcl-1 (ϳ56% in the region shown in Fig. 3B). Elements in this conserved region appear to function in both species as recent experiments show that mutations affecting this region also have an effect on transcription from the mouse Mcl-1 promoter (29). Further upstream in the human gene lies a GGCCCC repeat region (dot underlined in Fig. 3B) within a G/C-rich upstream area. Thus, the initiator sequences that mark the sites of Mcl-1 transcriptional initiation lie in a break between a G/C-rich, repeat-containing area upstream and the G/C-rich first coding exon downstream ( Fig. 2A and 3B).
One further point is that, with the primer extension method, a very faint product extended beyond the two initiator sequences (approximately another 25 nucleotides beyond the most upstream of the other 2 sites, to ϳ155 bp from the #105C primer; faintly visible in Fig. 3A, lane 3). This may mean that transcription can also initiate further upstream. Consistent with this possibility is the fact that, although several EST clones present in the data bases extend to near the initiator sequences (e.g. accession AI204385, AI202072, AA776756, and AI340205), some clones contain an additional upstream sequence (e.g. AI439001, AA884201, and AA453505). However, in all of the latter, the upstream sequence is co-linear with Mcl-1 genomic DNA, rather than containing a large gap. Thus, although the sequence of upstream portion of Mcl-1 contains regions that could be considered as candidate splice donor and acceptor sequences, there is at present no evidence suggestive of an upstream untranslated exon. Overall, human Mcl-1 can be transcribed from a region containing tandem initiator sequences that lie 69 -82 bp directly upstream of the coding sequence.
Transfection with the Mcl-1 s/⌬TM Variant Results in Cell Death-Because the internally deleted Mcl-1 s/⌬TM variant displays features reminiscent of the BH3 only pro-apoptotic family members Bid and Bad, we wondered whether it might similarly have death-promoting activity. We tested this in epithelial cells, where dead cells are readily detectable as they shrink, round up, and detach from the substratum while viable cells grow as a flat monolayer. We first used expression constructs in which Mcl-1 s/⌬TM was fused to EGFP (in the vector pEGFP-C1), as this provides a rapid assay. With this method, cells express-ing the introduced fusion protein are identified based on green fluorescence and can then be examined for the morphologic features of cell death (53,54). We found that A549 airway epithelial cells expressing pEGFP-C1-Mcl-1 s/⌬TM , as detected by green fluorescence, exhibited cell shrinkage (Fig. 4B) as well as nuclear condensation (Fig. 4C) and blebbing (Fig. 4D), hallmarks of cell death by apoptosis (53,54). In contrast, cells expressing the pEGFP-C1 vector or full-length pEGFP-C1-Mcl-1 ( Fig. 4A and legend) remained morphologically normal and viable, as expected based on previous experiments (32)(33)(34).
To confirm the above observations, we used the pCR3.  (Fig. 4E, lane 4). Endogenous full-length Mcl-1 was also detectable in these cells, but its level of expression was not altered in the presence of Mcl-1 s/⌬TM . Although the level of expression of Mcl-1 s/⌬TM appears lower, the relative levels of the variant and full-length Mcl-1 cannot be determined as the former is present only in transfected cells, whereas the latter represents in the entire cell population. Whatever the relative levels, it was clear that the introduced Mcl-1 s/⌬TM protein was expressed in this system. We therefore performed co-transfection experiments with pCR3.1-Mcl-1 s/⌬TM , along with pCMV-␤galactosidase as a marker for transfected cells, to test the effect of expression on cell death. Transfected (i.e. ␤-galactosidase expressing) cells were found to undergo cell death exactly as described above with pEGFP-C1-Mcl-1 s/⌬TM , an effect that was not seen with the insertless pCR3.1 vector or full-length pCR3.1-Mcl-1 (results not shown). An identical effect was seen in another epithelial cell line, HeLa. Upon transfection with pCR3.1-Mcl-1 s/⌬TM , HeLa cells expressing the ␤-galactosidase marker rounded up and exhibited shrinkage/blebbing (Fig. 4G), as did cells transfected with a Bax expression vector. In contrast, in parallel cultures transfected with the insertless vector or full-length pCR3.1-Mcl-1 (Fig. 4F and legend), the majority of cells remained morphologically normal and viable. Cells expressing ␤-galactosidase were scored as having either a normal, viable morphology (flat, healthy cells) or the altered morphology characteristic of cells dying by apoptosis (dense, rounded, detaching cells). This analysis showed that expression of the Mcl-1 s/⌬TM isoform was nearly as effective as Bax in inducing cell death (Fig. 4H). In two different cell types, in sum,   (lanes A, T, G, and C) was generated using the #105C primer employed in the mapping (see "Materials and Methods"). The RNA used was from PMA-treated ML-1 human myeloid leukemia cells, the cell line from which Mcl-1 was originally cloned (18). Arrows to the left of the sequencing ladder indicate the two major positions where primer extension products were seen (120 and 130 bp upstream from the 5Ј-end of the #105C primer). The more abundant primer extension products were found at the position of the lower arrow, as opposed to the upper arrow 10 bp upstream. The lower position was thus designated ϩ1. This position is indicated with an asterisk on the Mcl-1 coding sequence, which is listed in the column at the left. The sequence as listed includes the two tandem initiator sequences at 69 -82 bp upstream of the start site of Mcl-1 translation (see B below). The sequence of these initiators is CCACTTCTCACTTC, where the consensus for initiator sequences is YYAX(T/A)YY. The sequence in the right column is that of the complementary strand, which was that sequenced. The distance (bp) indicated at the right side of the gel represents the distance from 5Ј-end of the #105C primer.  7 and 8). The alternately spliced variant was indeed present at much lower abundance than the fulllength Mcl-1 transcript, a finding that is perhaps not surprising in view of the fact that the analysis was performed using healthy, viable cells.
Expression of Mcl-1 is markedly induced by PMA in a variety of cells (18,30), and it seemed possible that Mcl-1 s/⌬TM might similarly be increased with this agent. We examined this possibility using a Burkitt lymphoma cell subline, BL41-3, in which the Mcl-1 genomic locus is amplified and overexpressed (55). Our rationale was that the presence of multiple copies of the Mcl-1 gene might optimize the possibility of observing increased expression. The Mcl-1 s/⌬TM transcript, although not detectable in untreated BL41-3 cells, was detectable in cells exposed to PMA (Fig. 5A, lane 2 versus lanes 3-4). The Mcl-1 s/ ⌬ TM -encoded protein was likewise not detected in untreated BL41-3 cells but was detected at low levels in cells treated with PMA (Fig. 5B). The fact that the levels of the alternately spliced variant were lower than those of full-length Mcl-1, as in the other cells tested above, could be the reason that BL41-3 cells remain viable upon PMA treatment. Although the studies performed to date are limited in that they have examined cells that are largely viable and that also express full-length Mcl-1, these studies demonstrate endogenous cellular expression of the Mcl-1 s/⌬TM splice variant. They thus open the way to future studies aimed at monitoring the expression of the two isoforms during the course of cell death and at manipulating the expression of the isoforms as recently described for Bcl-x (56) as an approach to inducing the death of cancer cells. Other members of the Bcl-2 family can also undergo differential splicing (37,39,41,49,(57)(58)(59). In terms of anti-apoptotic family members, the splicing of Bcl-x L to Bcl-x S involves the use of an alternate splice site within the coding sequence (39). Bcl-w can also undergo alternate splicing, and in this case splicing is to an adjacent gene (45). With both Bcl-x and Bcl-2, unspliced transcripts that read through into the intron have also been reported (37). The mechanism involved in the splicing of Mcl-1 differs from that seen with these other anti-apoptotic family members in that it involves exon skipping, where the subsequent exon is placed in an altered reading frame. It is the presence of an additional intron in Mcl-1 downstream of BH3, along with the conserved intron further downstream in BH2,  3 and 4, respectively). Positive control RNA transcribed from plasmids containing full-length wild-type Mcl-1 or Mcl-1 s/⌬TM is in lanes 7 and 8, respectively. Identical samples were hybridized with the hAPO-2 probe set to confirm equivalent loading (not shown). B, BL41-3 cells were either not exposed (lane 2) or exposed (lane 3) to 10 ng/ml PMA and assayed for expression of Mcl-1 by Western blotting. In vitro translated Mcl-1 s/⌬TM was used as a standard (lane 1). The band traveling with an electrophoretic mobility just slightly more rapid than wild-type Mcl-1 is also specifically detected with anti-Mcl-1 antibodies, although its identity has not yet been determined. that allows for this ability to skip an exon and eliminate BH domains critical for anti-apoptotic effects.
Although two coding exons are present in anti-apoptotic family members so far characterized (other than Mcl-1) (38,40,45,46), multiple coding exons (5)(6) are present in pro-apoptotic family members. This is because of the presence of 3-5 introns in the coding region (37,(47)(48)(49)(50)(51), where the terminal intron is that which is conserved throughout the family. For example, additional introns in the Bax gene place BH1, BH2, and BH3 on separate exons. A variety of splice variants have been identified; this includes Bax␦, where exon 4 is spliced to exon 2 (in frame) with skipping of exon 3, as well as Bax␥, where exon 3 is spliced to exon 1 (with a change in reading frame) with skipping of exon 2 (8,37). The activity of these various variants has not been completely elucidated. Differential splicing is also seen in other pro-apoptotic family members; in Bok/Mtd, exon 2 can be skipped (49) and, in Bim, alternate splicing can alter the N terminus of the encoded gene product (10). Although the altered splicing of Bok results in retention of a similar function, the different variants of Bim exhibit differences in pro-apoptotic efficacy. One of the introns in pro-apoptotic Bax lies downstream of BH3 in a position analogous to the additional (first) intron in Mcl-1 (8). Other pro-apoptotic family members, namely Bid and Bak, also contain an intron downstream of BH3 (47,48,51). However, in the latter two genes these introns are not at positions identical to that of intron 1 in Mcl-1 and the analogous intron in Bax. The alternate upstream splice site within the Bcl-x first coding exon that is used in the production of Bcl-x s likewise lies downstream of BH3, although this site is again not located at a position identical to that of the intron in Mcl-1 (39). Overall, the multiple introns in pro-apoptotic family members provide a variety of possibilities in terms of differential splicing. Moreover whereas differential splicing through exon skipping, as seen here with Mcl-1, has not been reported for other anti-apoptotic family members, it is common among pro-apoptotic members.
Whereas these studies demonstrate that the human Mcl-1 gene can undergo differential splicing into two distinct isoforms, it remains to be determined how these isoforms function and are coordinated in vivo. It has been suggested that antiapoptotic members of the Bcl-2 family prevent cell death by inhibiting the action of adaptor proteins involved in the initiation of the caspase cascade, although this is controversial (60). Pro-apoptotic Bcl-2 proteins may then bind to anti-apoptotic family members and neutralize the block on the adaptor proteins to allow the death cascade to develop. The pro-apoptotic activity of several BH3 only Bcl-2 family members has been shown to be regulated in a post-translational manner by protein-protein interactions, which make these proteins unavailable to initiate the death pathway. For example Bad is bound in a phosphorylation-dependent manner to 14 -3-3 proteins, which sequesters the protein in a form that is unable to associate with Bcl-2 and Bcl-x L (61). Upon dephosphorylation, Bad translocates to bind Bcl-2 and Bcl-x L and induces the cell death cascade. Bim, another BH3 only protein, is sequestered in the cytoskeletal associated motor complex bound to LC8, a cytoplasmin dynein light chain protein. Apoptotic stimuli disrupt this interaction, and the freed Bim then binds to Bcl-2, neutralizing its anti-apoptotic activity and promoting cell death (62). A third BH3 only protein, Bid, is sequestered in the cytosol in an inactive form and is cleaved by caspase-8 to yield two distinct fragments (63). The larger fragment can relocalize to the mitochondria where it acts to antagonize anti-apoptotic Bcl-2 proteins and so promotes cell death. Therefore through distinctly different mechanisms it has been shown that BH3 only proteins are normally inactive. Following a pro-apoptotic stimulus, the protein is released and acts as a death inducer by antagonizing the anti-apoptotic activity of Bcl-2-like proteins. In the case of Mcl-1 with two isoforms with opposite functions being present in the same cell, a mechanism may exist for coordinated regulation. By analogy with BH3 only pro-apoptotic family members, the Mcl-1 s/⌬TM variant may be sequestered in an inactive form in viable cells.
Because the conserved intron within the BH2 domain near the C terminus is common to anti-and pro-apoptotic family members, it has been hypothesized that both sides of the family may have arisen from a primordial member, which contained a single intron in the coding region (48) and which underwent gene duplication (46). The addition of introns upstream of the conserved intron could then have occurred during the evolution of pro-apoptotic family members. The fact that the position of the additional upstream introns is not precisely conserved among pro-apoptotic family members led to the hypothesis that these additional introns were added after the divergence of the primordial gene (48). Mcl-1 contains a single additional upstream intron along with the conserved intron near the C terminus. It thus differs from other anti-apototic family members, which do not contain such introns, as well as from proapoptotic family members, which contain multiple upstream introns. Mcl-1 might then represent an intermediate in the development of the pro-and anti-apoptotic family members. This possibility is consistent with evolutionary analysis, which suggests that Mcl-1 and A1 represent a very ancient branch of this family different from the branch containing Bcl-2 and Bcl-x and from that containing Bax (64). This possibility and the findings described here could relate to previous observations concerning the BH3 domain of Mcl-1 (9). Specifically, the core of the BH3 domain of Bcl-2, Bcl-x, and Bcl-w contains an alanine residue at position 4, whereas pro-apoptotic family members contain bulkier domains (e.g. isoleucine or valine). These bulky residues appear important for binding to the hydrophobic cleft of anti-apoptotic family members (11). Mcl-1 contains a valine at this position, which is more typical of the pro-apoptotic family members. In addition, at position Ϫ3 from the BH3 core, the other three anti-apoptic family members have a charged residue (histidine or lysine), whereas pro-apoptotic members have uncharged residues; in this aspect Mcl-1 is also similar to the latter in that this residue is a leucine. The presence in Mcl-1 of residues in BH3 that are more typical of pro-apoptotic family members may relate to the fact that Mcl-1 can be alternately spliced to a pro-apoptotic form. Taking into account these aspects of its structure, Mcl-1 appears to be ideally designed to be capable of undergoing differential splicing to either a viability-promoting or a BH3 only death-promoting gene product.