Molecular Analysis of Survivin Isoforms

Survivin is a protein with proposed roles in cell division and apoptosis. Transcripts encoding splice variants of human survivin have been described and their expression correlated with cancer progression. As survivin forms homodimers in vitro, it has been suggested that these isoforms could interfere with wild type function by forming heterodimers. Here we show that survivin-2β and survivin-δEx3 can interact with wild type survivin but have reduced affinity for the partner protein of survivin, borealin, and thus do not localize with the chromosomal passenger complex in vivo. Furthermore, we demonstrate that overexpression of survivin-2β-green fluorescent protein (GFP) or survivin-δEx3-GFP does not impede cell cycle progression. We also report that wild type survivin, but not survivin-2β-GFP or survivin-δEx3-GFP, can rescue cell proliferation inhibited by small interfering RNA-mediated survivin depletion. These data suggest that, despite their ability to interact with wild type survivin, neither of these isoforms acts as its competitor during mitosis nor has an essential function.

Survivin-2␤ has an additional exon of 23 amino acids, exon 2␤, inserted in its BIR domain, which is predicted to alter its tertiary structure and thus to affect its anti-apoptotic function. Survivin-␦Ex3 omits exon 3 causing a frameshift that results in a COOH terminus with no homology to wild type survivin (10). These differences have led to the suggestion that survivin-2␤ and survivin-␦Ex3 may antagonise wild type survivin by forming heterodimers with reduced capabilities (13). Indeed, when induced to undergo apoptosis, it has been reported that cells transfected with survivin-2␤ have decreased viability over cells transfected with wild type survivin (14). However, to date there is only indirect evidence that heterodimers can form between these isoforms and wild type survivin (14).
The presence of transcripts of survivin isoforms has been correlated with cancer progression. For example, using reverse transcription PCR Mahotka et al. (9) found that wild type survivin and survivin-␦ Ex3 were expressed in cell lines derived from renal carcinomas but not from normal renal cells. Differential expression of survivin-2␤ was also reported in gastric carcinomas and, in this case, negatively correlated with disease progression (13,15). Further reports have linked the expression of survivin isoforms with poor patient prognosis (for review, see Ref. 16). Notably, however, with the exception of one recent study (14), the only description of survivin isoforms at the protein level has come from expression of tagged versions, which have revealed that survivin-2␤-GFP is cytoplasmic throughout the cell cycle, while survivin-␦Ex3-GFP is nuclear during interphase and diffusely localized in mitosis (10,14).
In this study we demonstrate that survivin-2␤ and survivin-␦Ex3 can interact directly with wild type survivin in vitro. Despite these abilities, however, neither survivin-2␤-GFP nor survivin-␦Ex3-GFP co-immunoprecipitate or co-localize with endogenous survivin, or the other chromosomal passengers during mitosis. We provide evidence that their differential localization may be due to their altered affinity for borealin. Furthermore, we report that these proteins are present at very low levels in cycling HeLa and U2OS cells and that their overexpression does not interfere with cell proliferation under our experimental conditions. Finally, we reveal that cDNA to wild type survivin, and only wild type survivin, can compensate for depletion of all forms of survivin by siRNA, indicating that these isoforms do not play a vital role in cell division and cannot compensate for loss of the wild type protein. Thus we conclude that the presence or absence of survivin-2␤ or survivin-␦Ex3 within cultured HeLa or U2OS cells is irrelevant to cell proliferation.

MATERIALS AND METHODS
PCR and Cloning Isoforms-To extract mRNA, exponentially growing cells were lysed in TRIzol (Invitrogen) and treated with chloroform FIGURE 1. Survivin-2␤ and survivin-␦ Ex3 can form heterodimers with wild type survivin in vitro. A, recombinant GST, GST-survivin, and GSTaurora-B were expressed in bacteria, bound to glutathione-Sepharose beads and incubated with radiolabeled in vitro translated wild type survivin, survivin-2␤, or survivin-␦ Ex3 as indicated. Upper panels show autoradiographs of radiolabeled isoforms bound to the GST preparations. Middle panels show Coomassie Blue staining of the corresponding gel to indicate equality of loading. Lower panels show quantitation of autoradiographs for each sample. B, same as for A, showing interaction between GST or GST-borealin and survivin isoforms. A sample of the TNT reactions in which the isoforms were translated is included to demonstrate levels of expression. All isoforms were pulled down by GST-survivin and GST-aurora-B indicating a direct interaction between these proteins. An extremely strong interaction occurred between GST-borealin and wild type survivin (note difference in scale on graphs). The interactions between GST-survivin and survivin-␦ Ex3 and between GST-borealin and survivin-2␤ were comparatively weak, and no interaction was observed between GST-borealin and survivin-␦ Ex3.
(2 min) to remove lipids. After centrifuging (13,000 rpm, 15 min, 4°C), the aqueous phase was removed and nucleic acids precipitated with ice-cold isopropyl alcohol. RNA was pelleted, washed with 70% ethanol, and re-suspended in RNase-free, diethyl pyrocarbonate-treated water (Ambion) before incubation at 60°C (5 min). Samples were then incubated in DNase (Ambion) for 1 h at 37°C. After DNase inactivation, 5 g of RNA sample was combined with 500 ng of oligo(dT) (New England Biolabs) and incubated at 70°C for 15 min. Samples were cooled and then incubated with 100 units of Moloney murine leukemia virus reverse transcriptase (Invitrogen) in 10 mM dithiothreitol, 10 mM dNTP, 1 ϫ 1st strand (42°C, 1 h) and then 70°C for 15 min. cDNA was amplified by PCR using Vent polymerase (New England Biolabs) and cloned into the appropriate vectors.
Real-time PCR was performed using QuantiTect SYBR Green PCR kit (Qiagen), according to the manufacturer's instructions, and analyzed on an ABI Prism 7900HT Sequence detection system (Applied Biosystems). The common forward primer was GAC CAC CGC ATC TCT ACA TTC, and  reverse primers were TGC TTT TTA TGT TCC TCT ATG GG (wild type  survivin), AAG TGC TGG TAT TAC AGG CGT (survivin-2␤), and ATT GTT GGT TTC CTT TGC ATG (survivin-␦ Ex3). Expression of the splice variants was calculated as a value relative to ␤-actin.
GST 6 Pull-down Assays-Expression of GST, GST-aurora-B, GSTsurvivin, and GST-borealin was induced in BL21 cells by the addition of 0.1 mM isopropyl ␤-D-thiogalactopyranoside. After 16 h growth at 18°C bacteria were harvested and lysed as described previously (5). Solubility of GST-borealin was increased by an additional 1 h incubation (4°C) in lysis buffer with 1% Nonidet P-40 and 0.03% SDS. In vitro translation was performed using pBluescript constructs (as indicated) and a TNT T7 kit (Promega), with [ 35 S]methionine (Amersham Biosciences) as tracer as described in Ref. 5.
Tissue Culture, Transient Transfections, and Stable Cell Lines-U2OS and HeLa cells were maintained at 37°C, with 5% CO 2 , in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillin-streptomycin, and fungizome. Transfections were carried out with the indicated pcDNA3.1 constructs using FuGENE 6 (Roche Diagnostics) and stable cell lines generated by transfection followed by selection with G418 (500 g/ml). Prior to analysis, GFPpositive cells were sorted using an LSRII fluorescence-activated cell sorter (BD Bioscience) to ensure a homogeneous population. Tissue culture reagents were from Invitrogen unless otherwise stated.
Immunoprecipitation and Immunoblotting-Whole cell extracts and immunoprecipitations were carried out on colcemid (0.5 ng/ml, 16 h; Sigma)-treated cells arrested in mitosis. Approximately 5 ϫ 10 5 cell equivalents were loaded per lane for whole cell extracts and 3 ϫ 10 6 cell equivalents per lane for immunoprecipitation, except where indicated. SDS-PAGE analysis and immunoblotting were carried out using normal procedures, with secondary horseradish peroxidase-conjugated antibodies (Dako) and chemiluminescent reagents (Amersham Biosciences). Signals were analyzed by exposure to Hyperfilm (Eastman Kodak Co.) or using a PhosphorImager (Storm 480 Phosphorimager, Amersham Biosciences). U2OS cells stably expressing GFP, survivin-GFP, survivin-2␤-GFP, and survivin-␦ Ex3-GFP were transfected 24 h prior to harvesting with either pcDNA-aurora-B (B), pcDNA-borealin (C), or pcDNA-survivin (D), arrested in mitosis, then blotted as indicated. Throughout blots of the initial whole cell extracts were probed with the appropriate antibody (left, upper panel(s)) and anti-GFP to indicate equivalence of expression (left, lower panel(s)). Anti-GFP was also used to indicate efficiency of immunoprecipitation (right, lower panel(s)) in each case. Wild type survivin was able to immunoprecipitate all members of the complex, both endogenous and exogenous. By contrast neither of the isoforms could immunoprecipitate the endogenous forms of these proteins (A). However, when overexpressed, aurora-B, borealin, and wild type survivin co-immunoprecipitated with survivin-2␤-GFP (B-D). The asterisk indicates IgG heavy chain in immunoprecipitates. WCE, whole cell extract; IP, immunoprecipitation.
Growth Curves and FACS Analysis-Cell proliferation was assessed by trypan blue exclusion (Sigma) using a hemocytometer.
For FACS analysis cells were pelleted at 750 rpm for 5 min at room temperature, washed in PBS, then fixed for 1 h in 1% paraformaldehyde at 4°C. Cells were then repelleted, washed in PBS, and resuspended in 500 l of propidium iodide solution (PI; Sigma) containing 40 g/ml PI and 100 g/ml RNase A (MP Biomedicals Ltd. UK) in PBS. Cells were then passed through a 70-m cell strainer (Falcon) into a 2058 tube (Falcon) and incubated at 37°C for 30 min before analysis on an LSRII fluorescence-activated cell sorter (BD Bioscience). Cells were initially analyzed by forward scatter versus side scatter to differentiate between live cells and cell debris. The live cell population was selected for by gate P1. P1-gated cells were plotted as PI-width versus PI-area to establish nuclear size. Gate P2 selected for single cells with increased PI intensity, and cell doublets were excluded. Cells from the P2 population were then plotted versus GFP-A. Gate P3 was selected for GFP-negative cells (established using a GFP-negative cell line), and gate P4 was selected for FIGURE 3. Analysis of survivin isoforms in human cancer cells by PCR. A, mRNA was isolated from the human cancer cell lines indicated and subjected to reverse transcription PCR in the presence or absence of reverse transcriptase. Three bands were revealed in each case, which corresponded to survivin-2␤ (upper), wild type survivin (middle and predominant), and survivin ␦-Ex3 (lower). B, real-time PCR was carried out on mRNA from U2OS and HeLa cells using a common forward primer and a specific reverse primer for each of the isoforms. Consistent with A expression of splice variants was plotted relative to ␤-actin levels indicated that wild type survivin was the predominant form and survivin ␦ Ex3 the least abundant.
GFP-positive cells. To determine the DNA content of the P4 population, PI-A was plotted against the number of cells counted for each cell line.
RNA Interference and Complementation-The RNAi target domain, bases 45-65, within the survivin cDNA sequence was selected as it was common to all human survivin isoforms; hence it should eliminate all splice variants (17). Using site-directed mutagenesis (Stratagene) guanine 54 was substituted for cytosine, thus creating a "wobble" in the RNAi target domain, allowing us to generate isoform specific constructs resistant to "knock down". These constructs were cloned into the expression vector, pcDNA3.1, and introduced into human cells as described above. RNAi was carried out using 60 pmol of doublestranded RNA per reaction and Oligofectamine as described in Ref. 17 or 1.5 l of siPORT NeoFx (Ambion) in 24-well plates with 4 ϫ 10 4 cells/well.

Survivin Isoforms Can Interact Directly with Wild Type Survivin but
Have Reduced Affinity for Borealin-To determine whether survivin-2␤ or survivin-␦Ex3 can interact directly with survivin in vitro, we incu-bated 35 S-labeled, in vitro translated wild type survivin, survivin-2␤ or survivin-␦Ex3, with GST or GST-survivin bound to glutathione-Sepharose beads. In this assay both isoforms bound to GST-survivin (Fig. 1A, middle panel), albeit survivin-␦Ex3 bound far less efficiently (ϳ3-fold) than wild type survivin. These data suggest that survivin isoforms can form heterodimers with wild type survivin. Similar interactions were also detected using yeast two hybrid pairwise interactions. 7 Interactions of similar intensity were also seen between GST-auorora-B and survivin and GST-aurora-B and the isoforms (Fig. 1A, right panel). Next we investigated the affinities of the isoforms for GST-borealin (Fig. 1B). An intense interaction between GST-borealin and wild type survivin was observed, a comparatively weak interaction was seen between GSTborealin and survivin-2␤, and no interaction was observed between GST-borealin and survivin-␦Ex3.
To determine whether these proteins have the potential to interact within the physiological milieu, we generated cell lines stably expressing GFP, survivin-GFP, survivin-2␤-GFP, and survivin-␦Ex3-GFP (see  below) from the cytomegalovirus promoter. These lines were transiently transfected with aurora-B, borealin, or survivin, then subjected to immunoprecipitation with anti-GFP antibodies (Fig. 2). Using this overexpression method we found that wild type survivin and survivin-2␤, but not survivin-␦Ex3, could immunoprecipitate aurora-B, borealin, and wild type survivin.
Expression of Survivin Isoforms-Reverse transcription PCR performed on four human cancer cell lines, HeLa, U2OS, MCF7, and HT29, revealed three mRNA transcripts corresponding to survivin in each cell type (Fig. 3A). Sequence analysis verified that in each case the predominant band was wild type survivin, while the additional bands above and below the major band were survivin-2␤ and survivin-␦Ex3, respectively (data not shown). The signal intensity from the isoform bands varied from line to line. As reverse transcription PCR cannot be used reliably to compare the levels of mRNA expression in a sample, we next used real-time PCR to provide accurate quantitation of mRNA expression of these variants in the cell lines used in this study. As shown in Fig. 3B, real-time PCR revealed that wild type survivin was always the most abundant and survivin-␦Ex3 the least abundant form in HeLa and U2OS cells.
To detect endogenous survivin isoforms at the protein level, antibodies were raised against the survivin-2␤-specific peptide, CNT-STLGGRGGRITR, and a peptide within the COOH terminus of survivin-␦Ex3, COOH-SWLPWIEASGRS. Immunoblot analysis of recombinantly expressed GST-tagged versions of these proteins revealed that the antibodies specifically recognized the isoforms to which they were targetted and did not cross react with the other forms of survivin (Fig. 4, A and B). When tested on whole cell lysates from HeLa cells transiently transfected with GFP-tagged versions of the isoforms, the survivin-2␤ antibody recognized a band of the appropriate size for survivin-2␤-GFP; however, it also highlighted a higher molecular weight band, circa 100 kDa, in both the untransfected and transfected lanes, indicating that it was not completely specific (Fig. 4C, asterisk). Affinity-purified antibodies to survivin-␦Ex3 appeared to be relatively specific revealing a single band in the transfected lane of ϳ40 kDa (Fig. 4D). Despite the reactivity of the antibodies in immunoblots we were unable to detect endogenously expressed protein of either isoform from HeLa or U2OS cells using these antibodies (Fig. 4, E and F, lower panels). Furthermore, we used a commercial polyclonal-survivin antibody that recognizes epitopes present in wild type survivin, survivin-2␤ and survivin-␦Ex3, on whole cell extracts prepared from cells either untransfected or transfected with untagged survivin-2␤ (Fig. 4E) or survivin-␦Ex3 (Fig. 4F). In each case the untagged exogenously expressed isoform was readily detectable in the transfected extracts with this antibody; however, no corresponding band was observed in the untransfected cell lysates. Together these data suggest that the endogenous expression of the survivin isoforms in these cells is extremely low.
Localization of Survivin Isoforms-It is possible that even very low levels of isoforms could interfere with wild type survivin function; therefore we asked whether localization of isoforms would support this. We used GFP tagging and immunofluorescence with the antibodies described above to detect and localize survivin-2␤ and survivin-␦Ex3 (Fig. 5). COOH-terminally tagged cDNA for each variant was transfected into HeLa or U2OS cells and observed 48 h later. Wild type survivin-GFP was found at the centromeres of prometaphase cells, and the midbody of cells executing cytokinesis, as reported previously (Fig.  5A). Using this method, survivin-2␤-GFP was found diffuse throughout the cytoplasm at all stages of the cell cycle and was excluded from the mitotic chromosomes and interphase nuclei. Concentration of sur-vivin-2␤-GFP was noted at the very center of the midbody (Fig. 5B), in a pattern that was distinct from the bipartite midbody localization of wild type survivin-GFP (Fig. 5A). Examination of control cells revealed that GFP alone localizes to the very center of the midbody (Fig. 5D), suggesting that this localization is a GFP artifact rather than a specific localization of survivin-2␤-GFP. Survivin-␦Ex3-GFP was also found diffusely localized in the cytoplasm and was excluded from the chromosomes in mitosis (Fig. 5C, upper panel). In interphase it was present in the nucleus, as indicated by the G 1 cells in Fig. 5C (lower panel). Importantly, at no time did we see localization that corresponded to that of the wild type form as judged by immunostaining for aurora-B (Fig. 5, red  panels). Furthermore, aurora-B localization was not disrupted by the overexpression of any of the survivin isoforms.
The peptide-specific antibodies raised against survivin-2␤ also revealed that it was excluded from the mitotic chromosomes and rarely seen at the midbody. During interphase the signal was diffuse throughout the cytoplasm and absent from nuclei (data not shown). Additionally, using this polyclonal antibody a signal was detected at the centrosome throughout mitosis. However, this signal persisted after RNAi depletion (data not shown), indicating that it was due to a cross-reaction, possibly with the 100-kDa protein in Fig. 4C. Immunolocalization of survivin-␦Ex3 with the antibodies described in Fig. 4 showed a weak signal of diffuse cytoplasmic staining, exclusion from the chromosomes during mitosis, and no concentration at the midbody (data not shown). In interphase cells it was apparent in both the nucleus and cytoplasm. Thus our immunohistological data confirm the localizations with GFPtagged forms above.
Survivin Isoforms Cannot Immunoprecipitate Endogenous Chromosomal Passenger Proteins-As shown in Fig. 2, survivin-2␤-GFP was able to immunoprecipitate aurora-B, borealin, and survivin when they were overexpressed in the cell. Given that the localization of the isoforms is different from the chromosomal passenger proteins (Fig. 5), we next asked whether it could immunoprecipitate the endogenous forms of these proteins from whole cell extracts in vivo. Fig. 6 shows that only wild type survivin-GFP can interact with the endogenous chromosomal passenger complex, as judged by co-immunoprecipitation of aurora-B, borealin, and wild type survivin.

Overexpression of Survivin Isoforms Does Not Impede Cell
Proliferation-Compared with wild type survivin-GFP and survivin-␦Ex3-GFP, we encountered some difficulties in establishing the survivin-2␤-GFP expressing lines, suggesting that it may be slightly cytotoxic. Nevertheless, once established, cell lines stably expressing survivin-GFP, survivin-2␤-GFP, and survivin-␦Ex3-GFP grew with normal kinetics (Fig. 7A). These lines were subjected to FACS analysis to monitor DNA content (Fig. 7B). No change in distribution of 2N and 4N cells was observed, and no increase in ploidy (8N or 16N) was detected in any of the lines, as compared with a control GFP line, indicating that overexpression of these versions of survivin does not interfere with proliferation in these cell lines.
Survivin Isoforms Cannot Complement for Loss of Wild Type Survivin-To assess whether the isoforms have an essential role in mitosis, we carried out RNAi complementation experiments. The RNAi target sequence, bases 45-65, was selected as it is present in all known splice variants of human survivin; hence its application should eliminate all known isoforms (17). The constructs used to generate the stable cell lines described above carried an additional silent mutation (C54G) within this siRNA target region to prevent their elimination, denoted (r) for resistant hereinafter. Immunoblot analysis verified that these constructs were indeed resistant to knock down and that the non-resistant (nr) cytomegalovirus promoted survivin-GFP could be efficiently eliminated from cells by 48 h posttransfection with survivin specific siRNA (Fig. 8A). Using these stable cell lines we asked whether the exogenously expressed proteins could complement for loss of endogenous survivin by RNAi. As indicated in Fig. 8, B and D, in a control population expressing survivin-GFP, cell proliferation ceased 48 h after exposure to survivin specific double-stranded RNA. Introduction of a single cDNA encoding only wild type survivin(r)-GFP was sufficient to restore growth (Fig. 8, C  and D). By contrast neither survivin(r)2␤-GFP nor survivin(r)␦Ex3-GFP was able to complement for loss of survivin expression (Fig. 8D).

DISCUSSION
A number of studies have suggested that the clinical outcome of some cancer sufferers can be correlated to the expression of mRNA of the survivin isoforms, survivin-2␤ and survivin-␦Ex3 (reviewed in Ref. 16). Furthermore, it has been speculated that the reason for this correlation could be due to competition by the isoforms to form dimers with the wild type protein. However, three questions have remained outstanding. First, whether they can form heterodimers; second, whether these forms exist as proteins within the cell; and third, whether they can interfere with the function of the wild type protein. This study addresses all of these questions.
Survivin Isoforms Can Heterodimerise with the Wild Type Protein-Caldas et al. (14) recently reported that survivin-2␤ and survivin-␦Ex3 can co-immunoprecipitate with wild type survivin when both versions are transiently transfected into human cells. Our in vivo data confirm this finding for survivin-2␤, and our in vitro data extend it by demonstrating for the first time that survivin-2␤ and survivin-␦Ex3 can interact directly with wild type survivin to form heterodimers. Two regions of the wild type protein have been implicated in dimerization, residues 6 -10, and the linker region (amino acids 89 -102) between the BIR domain and the COOH-terminal ␣-helix (1, 2). Our in vitro analysis is consistent with this as, although the folding of survivin-2␤ is predicted to be disrupted by the 23-amino acid BIR domain insertion, and the linker region is absent from survivin-␦Ex3, both variants retain amino acids 6 -10, which may be sufficient to permit dimerization. Interestingly, BIR domains are typically present in tandem or multiple repeats, but this domain is singularly present in survivin; thus dimerization may provide a means of duplicating this domain. It should be noted, however, that although crystallographic and NMR studies have decribed survivin as a homodimer (1)(2)(3), it has yet to be determined whether dimerization is critical for its function in vivo.
Survivin Isoforms Do Not Interact with the Chromosomal Passenger Complex in Vivo-During mitosis survivin operates in a complex with aurora-B kinase, borealin, and INCENP, and these "chromosomal passenger proteins" are mutually dependent upon each other for their localization and function (see Ref. 8). Here we report that neither survivin-2␤-GFP nor survivin-␦Ex3-GFP is able to immunoprecipitate endogenous chromosomal passengers from U2OS cells. As they do not co-localize with these proteins at any stage of mitosis this may not be wholly surprising. Thus to determine whether the inability of these isoforms to interact with the chromosomal passenger complex is due to their differences in localization, or vice versa, we performed immunoprecipitations on cells in which the chromosomal passenger of interest was also overexpressed. Using this method we discovered that survivin-2␤-GFP, but not survivin-␦Ex3-GFP, was able to immunoprecipitate aurora-B and borealin, albeit less efficiently than the wild type protein. Consistent with these data, when investigated in vitro, a strong direct interaction occurred between wild type survivin and borealin; a comparatively weak interaction was detected between survivin-2␤ and FIGURE 6. Immunoprecipitation of the chromosomal passenger complex with survivin variants. A, immunoprecipitation (IP) was performed on mitotic U2OS cells stably expressing GFP, survivin-GFP, survivin-2␤-GFP, and survivin-␦-Ex3-GFP using anti-GFP antibodies. Approximately 2 ϫ 10 6 cells were loaded in the immunoprecipitation lanes. One-fifth of the extract was loaded for analysis of whole cell extracts (WCE). Proteins were separated by SDS-PAGE, transferred to nitrocellulose, and incubated with antibodies to aurora-B, borealin, or GFP, as indicated. In contrast to wild type survivin, neither survivin-2␤-GFP nor survivin-␦ Ex3-GFP immunoprecipitated endogenous aurora-B or borealin indicating that they do not associate with these proteins in vivo. B, same as for A, with the exception that 5 ϫ 10 6 cell equivalents were loaded per lane (IP), and the blot was probed with anti-survivin antibodies. These data show that only wild type survivin can immunoprecipitate endogenous wild type survivin.
borealin, and no interaction was seen between survivin-␦Ex3 and borealin. Previous reports have demonstrated that the interaction between wild type survivin and borealin is mediated by the BIR domain of survivin (7,18), that most of survivin is bound to borealin in mitotic cells, and that localization to the mitotic centromeres of survivin is borealindependent (7). As the BIR domain is altered in survivin-2␤ and truncated in survivin-␦Ex3, our data are consistent with this region being responsible for interaction with borealin. Thus we conclude that the differential localization of these proteins within the cell is likely to be a consequence of their inefficiency/inability to bind directly to borealin.
Survivin-2␤ and Survivin-␦Ex3 Are Low Abundance Proteins-To determine whether survivin isoforms are expressed as proteins in cycling HeLa and U2OS cells, we generated isoform-specific peptide antibodies. The level of expression of these versions of survivin was lower than our detection capabilities by immunoblotting; however, we were able to detect the endogenous proteins within the cell by immunofluorescence microscopy. Importantly, the pattern of localization was similar to that observed for GFP tagged versions of these proteins and no co-localization with wild type survivin was observed during mitosis (10). Using the same antibodies Caldas et al. (14) were able to detect survivin-␦Ex3, but not survivin-2␤, in HeLa cell extracts. Whether the low abundance of these proteins is simply a consequence of the low level of mRNA available, or rapid turnover at the protein level, remains to be determined; however, accelerated proteosome-dependent clearance of survivin-␦Ex3, as compared with wild type protein, has been reported (10) and may be due to an additional degradation signal in its COOH terminus (19). Indeed, in our hands we have also found that survivin-␦Ex3 is highly unstable in a number of systems.
If a protein behaves in a dominant or dominant negative fashion, its presence, even at low levels, would be expected to alter cell behavior. To test whether these proteins have a dominant impact on cell cycle progression, we generated stable U2OS cells expressing GFP-tagged forms FIGURE 8. Survivin-2␤ and survivin-␦-Ex3 are non-essential and cannot compensate for loss of wild type survivin. A, immunoblot analysis of U2OS cells stably expressing GFP-tagged versions of survivin (as indicated) 48 h after transfection with control (c) or survivin specific (s) siRNA. Survivin that had not been rendered resistant to siRNA mediated depletion (survivin(nr)-GFP) was eliminated by 48 h, while those with a silent mutation, survivin(r)-GFP, survivin2␤(r)-GFP, and survivin-␦Ex3(r)-GFP, were still expressed. B and C, HeLa cells expressing survivin-GFP or survivin(r)-GFP were exposed to control (triangles) or survivin specific (open circles) siRNA. Cell growth was arrested in the survivin(nr)-GFP culture 48 h post-treatment with survivin specific siRNA (B), but cells expressing the RNAi-resistant survivin(r)-GFP grew similarly to the control population (C). D, cell number in U2OS populations expressing the constructs indicated were analyzed at 48 and 72 h and normalized to the number of cells in the control population. Cell proliferation was inhibited in all populations except survivin(r)-GFP. expressed from a cytomegalovirus promoter. Some initial problems were encountered generating the survivin-2␤-GFP lines that suggested survivin-2␤ may be slightly cytotoxic. Once established, however, all lines expressing these isoforms grew normally and did not accumulate cells of increased ploidy, as compared with cells depleted of survivin, previously demonstrated by Lens et al. (20). Thus these cells are clearly able to tolerate high levels of survivin-2␤ or survivin-␦Ex3 without any deleterious effects indicating that they do not interfere with the essential role of wild type survivin during mitosis.
Only Expression of Wild Type Survivin cDNA Complements RNAimediated Pan-survivin Depletion-Here we have shown that overexpression of survivin isoforms at the protein level is inconsequential to the proliferation of HeLa and U2OS cells in culture. We have also presented complementation experiments to determine whether their expression is required for cell viability. We report that complementation with cDNA to wild type survivin alone is sufficient to restore proliferation indicating that the isoforms are not essential for normal cell cycle progression. In addition, our data reveal that neither survivin-2␤ cDNA nor survivin-␦Ex3 cDNA can rescue cell growth, demonstrating that they cannot substitute for the wild type protein.

CONCLUSIONS
We have shown for the first time that the survivin splice variants, survivin-2␤ and survivin-␦Ex3, can form heterodimers with the wild type protein and have altered affinity for the chromosomal passenger, borealin. We also demonstrate that expression of survivin-2␤ or survivin-␦Ex3 in transformed human cells, even at high levels, is inconsequential to cell proliferation and that neither survivin-2␤ nor survivin-␦Ex3 can substitute for wild type function. We conclude that any correlation between patient prognosis and expression of survivin isoforms is unlikely to be due to interference with mitotic events. It will be interesting to determine whether these correlations can be attributed to interference with the anti-apoptotic activities of survivin.