The Mitotic Regulator Survivin Binds as a Monomer to Its Functional Interactor Borealin*

Survivin is a member of the IAP (inhibitor of apoptosis) protein family, defined in part by the presence of a zinc-binding baculoviral inhibitory repeat (BIR) domain. Most BIR domains bind short sequences beginning with alanine, and in this manner, they recognize and block the action of key targets in apoptotic pathways. However, Survivin binds only very weakly to typical IAP ligands. Unique features of Survivin are the long C-terminal helix following the BIR domain and a short segment (linking the helix and BIR domains) that mediates Survivin homodimerization. Despite this detailed knowledge of the structure of Survivin itself, there is a current lack of understanding about how Survivin recognizes cellular binding partners, and consequently, many questions about Survivin function remain unanswered. We determined two co-crystal structures of Survivin and a minimal binding fragment from the chromosomal passenger protein Borealin, a well validated functional interactor. The interaction between Survivin and Borealin involves extensive packing between the long C-terminal helix of Survivin and a long Borealin helix. Surprisingly, an additional important interaction occurs between the Survivin homodimerization interface and a short segment of Borealin. This segment both structurally mimics and displaces one Survivin monomer. The relevance of this unexpected interaction was tested by mutagenesis of two key Borealin residues. Mutant Borealin introduced into HeLa cells failed to localize properly during mitosis and also caused mislocalization of other chromosomal passenger proteins. This suggests that the mutant is dominant-negative and confirms the functional importance of the interaction surface identified in the crystal structures.

In the 10 years since Survivin was discovered (1), it has seemed at times to be a protein with too many functions. Survivin is a member of the baculoviral inhibitory repeat (BIR) 2 domain family. Originally described as an IAP (inhibitor of apoptosis) protein, it now assumes an increasingly complex role in the intrinsic cell death pathway (2). Survivin has a more clearly understood function in cell division (3). Despite intensive efforts to define Survivin function, little is known about how Survivin recognizes its binding partners. Indeed, the only three-dimensional structures available are for Survivin itself (4 -7). Crystallized Survivin forms a distinctive "bow tieshaped" (4,5) homodimer. Stable solution dimerization occurs through the same interface (7), suggesting that this dimer may contribute to Survivin function.
Throughout mitosis, Survivin is present in the chromosomal passenger complex (CPC), where it is critically important for mitotic progression (8,9). The CPC is currently known to have three members in addition to Survivin: the kinase Aurora B; an adapter protein, INCENP (inner centromere protein); and a recently discovered addition, Borealin (10 -12). A key property of the CPC is that all members are obligate. Depletion of any one member causes mislocalization of all of the others and causes a variety of mitotic abnormalities, consistent with loss of complex function. Early in mitosis, the CPC is found along chromosome arms, but by prophase, it has concentrated at the inner centromeres. It is not clear why the CPC concentrates at the centromeres, as no direct receptor has been identified. One recent candidate is the structural component CENP-C (13), although direct binding to centromeric DNA has also been proposed (14). While at the centromeres, the CPC plays one of its best known roles as part of the spindle assembly checkpoint. Loss of spindle assembly checkpoint function leads to chromosome segregation errors and a failure of the cell cycle to arrest in response to spindle poisons. Once all chromosomes have achieved biorientation and are under tension between the spindle poles, the spindle assembly checkpoint no longer signals, and the cell enters anaphase. At this time, the CPC relocates once more away from the centromeres to the central spindle. As the cell reaches telophase, the CPC concentrates at the midbody, where it is essential for cytokinesis. Loss of CPC function at this stage causes a failure of cells to divide fully, leading to binucleate and, eventually, multinucleate cells.
A major structural role of the CPC is to target Aurora B kinase activity to its substrates at appropriate mitotic stages (3,8,9). Because of the mutual functional interdependence of the subunits at all mitotic stages, it has been difficult to dissect how this is achieved. It has been known for some time that the most important direct interaction of CPC members with Aurora B occurs through the short, highly conserved C terminus of INCENP, called the IN-box. This interaction activates the kinase, and the structure of this subcomplex of the CPC was determined recently (15). In contrast, the extreme N terminus of INCENP is important for Survivin binding. The discovery of the fourth passenger, Borealin, has greatly clarified some organizational aspects of this part of the CPC (10 -12, 14, 16, 17). Significantly, it was observed by researchers working in different organisms that Borealin and Survivin form a robust binary complex in vitro (10,11). It was further determined that this binary complex has enhanced affinity for INCENP compared with either protein alone (14,16).
We decided to map the interacting regions in the binary complex of the human proteins to better understand the ligand recognition properties of Survivin and to gain insight into CPC organization and function. Through sequence analysis and truncation studies, we identified a minimal fragment of Borealin capable of high affinity complex formation with Survivin. Surprisingly, the crystal structures of two different complexes suggest that the accepted Survivin homodimer is not relevant to mitotic function. Instead, mutation of two key Borealin residues present in its interface with Survivin caused extensive mislocalization of CPC proteins, supporting the functional importance of the new binding interface observed in our structures.

Construction of Plasmids for Bacterial Coexpression-
The gene for full-length human Survivin was subcloned in-frame with an N-terminal His 6 tag into the T7 expression vector pET28 (kan R ; EMD Biosciences). The gene for human Borealin was instead subcloned into pET21 (amp R ). A C-terminal Streptag (IBA) was added to the Borealin construct by Kunkel mutagenesis. An alternative N-terminal tag (GB1 domain of protein G) was added to Borealin for some large-scale purifications for crystallography (see below). Truncation mutants were prepared from these constructs by Kunkel mutagenesis. The DNA sequences of all inserts were fully verified.
Coexpression and Copurification Studies-The appropriate Survivin and Borealin constructs were cotransfected into BL21(DE3)pLysS Rosetta 2 cells (EMD Biosciences). Positive colonies were selected on 50 g/ml kanamycin and 50 g/ml carbenicillin. Single colonies were grown in LB broth at 37°C under antibiotic selection until culture density reached A 600 ϭ 0.6. Expression was induced by addition of 0.4 mM isopropyl ␤-D-thiogalactopyranoside, and the cultures were shaken overnight at 22°C. Harvested cells were lysed in 100 ml of buffer A (150 mM Tris (pH 7.5), 250 mM NaCl, and 0.1 mM tris(2-carboxyethyl)phosphine) per liter of culture by passage through a Microfluidizer, and the soluble lysate was loaded onto a nickelnitrilotriacetic acid (Ni-NTA)-agarose column (0.5 ml of resin/ liter of culture; Qiagen Inc.). The column was washed with 20 volumes of buffer A containing 5 mM imidazole, and Survivin was eluted with 5 volumes of buffer A containing 250 mM imidazole. Samples of eluted fractions (10 l) were loaded onto 10 -20% Tris/glycine-polyacrylamide gels (Invitrogen) for analysis of coeluted proteins. Proteins were detected by Coomassie Blue staining. Confirmation of Borealin was done by Western blotting using an anti-Strep-tag primary antibody (Qiagen Inc.) followed by a rabbit anti-mouse secondary antibody conjugated to horseradish peroxidase (Zymed Laboratories Inc.).
Large-scale Purification of Complexes and Crystallization-Two complexes (full-length His 6 -Survivin plus Strep-tag-Borealin-(20 -78) and His 6 -Survivin-(1-120) plus GB1-Borealin-(20 -78)) were expressed and purified on Ni-NTA as described above. GB1 and His 6 tags were removed by thrombin cleavage overnight. The proteins were further purified by sizeexclusion chromatography on Sephacryl S-200 (GE Healthcare) in buffer A and then concentrated to 4 and 17 mg/ml, respectively. Concentrated complexes were dialyzed into buffer B (100 mM NaCl, 20 mM Tris (pH 7.5), and 0.1 mM tris(2-carboxyethyl)phosphine). Protein identities were verified by mass spectrometry and N-terminal sequencing. Note that removal of the GB1 tag was initially incomplete but that further slow cleavage occurred upon concentration, such that there was no tag present in the crystallized complex. The full sequences of expressed proteins and final cleaved chains are shown in supplemental Scheme S1.
Data Collection and Structure Solution-A data set for the Survivin-(1-142)-Borealin crystals was collected at beamline 5.0.1 at the Advanced Light Source (Berkeley Lab) and was processed with HKL software (supplemental Table S1) (18). The crystals belong to space group C2. The data are anisotropic with a functional resolution of ϳ2.4 Å. The structure was solved by molecular replacement using human Survivin (Protein Data Bank code 1F3H) (4) as a search model. Initial electron density maps revealed clear helical density for Borealin. A model was built consisting of Borealin residues 21-76 and Survivin residues 5-141. A 3.3-Å data set for the Survivin-(1-120)-Borealin crystals was acquired at beamline 9-2 at the Stanford Synchronized Radiation Laboratory and was processed with HKL software (supplemental Table S1) (18). The space group is I4 1 22. The structure was solved by molecular replacement using the partially refined crystal structure of the Survivin-(1-142) complex and the program Phaser (CCP4, Daresbury Laboratory), revealing three complexes in the asymmetric unit. Each complex has density for Borealin residues 20 -76 (plus one additional N-terminal residue from the thrombin cleavage site) and Survivin residues 5-119.
Both structures were refined with the program REFMAC5 and included TLS refinement (19). The final models have good stereochemistry: Ramachandran plots show that for the Survivin-(1-142) structure, 98.8% of all residues appear in the most favored or additionally allowed regions, with only two residues (1.2%) in generously allowed or disallowed regions (20). Likewise, for the Survivin-(1-120) complex, 98.2% of all residues appear in the most favored or additionally allowed regions, with only 1.7% of residues in generously allowed or disallowed regions.
Mitotic Localization of Borealin Mutants-Full-length wildtype Borealin and the Y54A/W70A mutant were each subcloned by PCR into pDONOR201 (Invitrogen) and fully sequenced before transfer into pDEST-12 (Invitrogen) additionally modified with an N-terminal hemagglutinin (HA) tag. HeLa cells (CCL-2, American Type Culture Collection) were grown in high glucose Dulbecco's modified Eagle's medium on 18-mm coverslips in 12-well plates for 20 h. Cells were then transfected with plasmid (0.5 g) using FuGENE 6 (Roche Applied Science) for 6 h. The medium was replaced with medium containing 2 mM thymidine, and the cells were incubated for an additional 16 h. Cells were then released from the thymidine block by replacement of the medium. After an additional 8 h, cells were fixed for 20 min in paraformaldehyde.

RESULTS
Analysis of the 280-amino acid Borealin sequence suggests that it may be largely ␣-helical but does not predict any globular domains. However, it has been reported that Borealin-(1-141) is fully competent to bind Survivin (11). Within this N-terminal region are two potential coiled-coil sequences, one predicted strongly and a second predicted only weakly (Fig. 1A and supplemental Fig. S1). A series of truncated Borealin constructs was prepared, focusing on the predicted boundaries of the helical elements. These were coexpressed with fulllength His 6 -tagged Survivin in Escherichia coli. Survivin was purified on Ni-NTA beads, and eluted protein was probed for copurifying Borealin fragments. A fragment as short as Borealin-(20 -60) (encompassing the first predicted coiled coil) could be detected, but Borealin-(20 -78) appeared to be the shortest fragment with full affinity (Fig. 1B). In addition to the strongly predicted coiled coil, this fragment includes at its C terminus a short segment highly conserved in the single human Borealin and all other vertebrate Borealins of the more common Dasra B subtype (supplemental Fig. S2) (12). A subset of Borealin fragments was tested for binding to truncated forms of Survivin (Fig. 1B). Taken together, these results suggest that the strongest interactions occur between Borealin residues 20 -78 and the region of Survivin that follows the BIR domain, including the first part of the long helix. This region also includes the Survivin homodimerization interface (Fig. 1C).
To better understand the interaction of Survivin with Borealin, we determined the x-ray crystal structures of both fulllength Survivin and Survivin-(1-120) bound to Borealin-(20 -78). The most striking aspect of both Survivin-Borealin complex structures is the 1:1 stoichiometry, which was unexpected based on the published structures of Survivin alone ( Fig.  2A). Our complex structures show that Borealin wraps around Survivin, with the N-terminal residues of Borealin (residues 21-60) forming a long helix that packs against the Survivin C-terminal helix. This long Borealin helix is followed by two additional short helices that pack against the juncture of the Survivin BIR domain and its C-terminal helix. The Borealinbinding site overlaps with the Survivin homodimerization interface (4, 5, 7). Thus, Survivin cannot dimerize in this way when bound to Borealin. Closer examination revealed that the backbone conformation of Borealin-(64 -73) is remarkably similar to that of Survivin-(92-101) in the homodimer, suggesting that this region of Borealin structurally mimics the displaced Survivin monomer (Fig. 2B).
The core of the interaction is Borealin-(65-74), which packs tightly against Survivin. In particular, the invariant Borealin Trp 70 side chain enters an enlarged pocket on Survivin, where Ͼ95% of its solvent-accessible surface area is buried (Fig. 2C  and supplemental Fig. S3). This suggests an explanation for the observed importance of this region in the truncation experiments (Fig. 1B). A second significant contact is provided by Borealin Tyr 54 (ϳ70% buried), which is present near the C terminus of the long Borealin helix. Trp 70 and Tyr 54 are the only Borealin residues to bury at least 100 Å 2 of solvent-accessible surface in the complex. The total interface buried by the Sur- vivin-Borealin interaction is large (ϳ2800 Å 2 in the Survivin-(1-142) complex), with Survivin and Borealin contributing equally to the contact surface. In comparison, ϳ1000 Å 2 is buried in the Survivin homodimer interface (5). This suggests that the interaction of Survivin with Borealin is stronger than the interaction of Survivin with itself. Uncomplexed Borealin-(20 -78) was insoluble (precluding equilibrium affinity measurements), but the purification of the heterocomplex from a large excess of expressed Survivin (data not shown) supports a higher affinity interaction. Interestingly, the recognition of Borealin by Survivin is very different from typical IAP-type peptide-binding pockets on other BIR domains (such as XIAP or ML-IAP) (Fig. 2D).
Although the major interaction between Borealin and Survivin appears to be the 1:1 complex described above, we observed a crystal packing interaction for the Survivin-(1-120) complex that suggests the possibility of antiparallel 2:2 complex formation (supplemental Fig. S4). This interaction involves predominantly the long helices of two Borealin subunits that form extensive interhelical packing interactions. Notably, as discussed above, it does not involve homodimerization through the interface present in published Survivin structures. (Indeed, there are no Survivin-Survivin contacts in this crystal packing.) A model of a dimer of the Survivin-(1-142) complex based on the 2:2 packing interaction suggests that it might be possible for the full-length complex to dimerize in this way, but that it would require changes in the degree of kinking or supercoiling of the long helices (supplemental Fig. S4C). We observed a second possible antiparallel 2:2 arrangement in the crystal packing of the Survivin-(1-142) complex. Somewhat surprisingly, this contact does not involve any intimate interhelical packing between the 1:1 complexes, but instead involves BIR-BIR and BIR-helix contacts between Survivin monomers (supplemental Fig. S4D). However, once again, these Survivin-Survivin contacts do not involve any of the previously reported Survivin dimeric crystal contacts (4 -6).
We assessed the solution stoichiometry of the Survivin-(1-142) complex by equilibrium and velocity ultracentrifugation (supplemental Fig. S5). The data are most consistent with a mixed population of 1:1 complex and a smaller amount of aggregate of higher order than 2:2. We see no significant population of 2:2 complex but cannot rule out the possibility that it might form under some conditions. Instead, the observed tendency of the binary complex to aggregate suggested the presence of an additional binding surface and the possibility that our Borealin fragment might be sufficient to recruit INCENP to Survivin. To test this, we coexpressed a previously described minimized INCENP construct, glutathione S-transferase-INCENP-(1-58) (14), and the binary complex. We found that Borealin- (20 -78) and Survivin copurified with the INCENP fragment at apparently stoichiometric levels (supplemental Fig. S6).
To test the biological significance of the major binding interface seen in our crystal structures, we introduced two mutations into Borealin that might be expected to alter its interaction with Survivin (Y54A/W70A) (Fig. 2C). Given that Borealin-(1-60) (corresponding to the long helix) can bind weakly to full-length Survivin (Fig. 1), we tested whether the Y54A/W70A mutations affect binary complex formation between Survivin and Borealin- (20 -78). Interestingly, we found that this binary complex could be purified from bacterial cultures in amounts comparable with those obtained with the wild-type fragment (data not shown), indicating that the mutant fragment retained significant affinity for Survivin.
Full-length tagged wild-type and mutant Borealins were transiently transfected into HeLa cells and were determined to express equally (supplemental Fig. S7A). Immunofluorescence visualization of passenger proteins revealed that in the presence of the Borealin mutant, severe mislocalization occurred during later stages of mitosis ( Fig. 3 and supplemental Fig. S7). In particular, the Borealin mutant was itself incorrectly localized, whereas HA-tagged wild-type Borealin appeared to localize normally, with some background diffuse staining most likely caused by modest overexpression relative to the endogenous FIGURE 3. Mutagenesis of key Borealin residues buried in the interface with Survivin and assessment of chromosomal passenger localization during anaphase and telophase. HA-tagged wild-type Borealin or mutant Y54A/W70A was transfected into HeLa cells as described under "Experimental Procedures." Cells were stained with anti-HA antibody, antibody to Aurora B or Survivin, and Hoechst 33342 to visualize chromosomes. The images shown are for the endogenous passenger protein (Aurora B or Survivin) and for the overlay of passenger protein with DNA and HA staining. A complete set of images is given in supplemental Fig. S7. A, comparison of mutant and wild-type Borealin effects on Aurora B and Survivin localization during anaphase. In the presence of HA-tagged wild-type Borealin, both Aurora B and Survivin localize in punctate regions at the central spindle. When cells are transfected instead with HA-tagged mutant Borealin, both Aurora B and Survivin appear instead to associate predominantly with chromosomes. A minor fraction of Aurora B can be seen in its normal location at the central spindle region, but Survivin appears completely mislocalized. B, comparison of mutant and wild-type Borealin effects on Aurora B and Survivin localization during telophase. In the presence of HA-tagged wild-type Borealin, Aurora B and Survivin are able to translocate to the midbody and are seen in two bright spots flanking the central point. In the presence of HA-tagged mutant Borealin, the localization of Survivin is very strongly affected, with Survivin appearing to be completely absent from the midbody. Again, Survivin appears instead to localize with chromosomal DNA. Aurora B localization is partially affected. Like Survivin, it appears to be significantly associated with the chromosomes. However, it can be seen also in the midbody (with much less intense staining than normal). protein (supplemental Fig. S7). In the presence of transfected wild-type Borealin, Aurora B and Survivin localized in the typical manner of the passenger proteins. In contrast, in cells transfected with the Borealin mutant, Aurora B was partly associated with chromosomes during anaphase and telophase, whereas Survivin was always entirely mislocalized to the chromosomes during these phases (Fig. 3).

DISCUSSION
Purified Survivin forms a dimer in solution (7) and crystallizes as a dimer (4,5). Mouse Survivin forms three distinct dimeric crystal packing contacts (6), one of which involves the same dimer interface seen in the structures of human Survivin. There was much attention paid to the question of which dimer was the "correct dimer," but at the time, mutagenesis studies failed to resolve the issue (21). In addition to our results, recent mutagenesis studies probing the role of a Survivin nuclear export signal indirectly suggest that the accepted Survivin dimer may not be present in the functioning CPC (22). This nuclear export signal spans the Survivin dimer interface, and the mutants studied would not only be expected to affect the function of the nuclear export signal (as intended by the authors), but could also be expected to significantly destabilize the homodimer interface (especially T97E). Nevertheless, the Survivin mutants showed only minor differences in mitotic localization compared with wild-type Survivin (22). Despite this, it has generally been assumed that the relevant form of Survivin in vivo is a dimer. The observation that differently tagged Survivin variants co-immunoprecipitate lends some support to this idea (for example, see Ref. 14). However, it has not yet been established what the overall stoichiometry is of the intact CPC (e.g. it could contain more than one "monomeric" Survivin). In addition, we have observed that even highly purified Survivin has an extreme tendency to form large aggregates, especially in lower ionic strength buffers. We suggest that any report of Survivin co-immunoprecipitating must therefore be viewed with caution.
It would appear that our Borealin mutant (Y54A/W70A) has the capacity to act in a dominant-negative manner and that it therefore interacts with at least one of the other passenger proteins or with a structural receptor for the complex. Our bacterial coexpression assay (with mutant fragment 20 -78) suggests Survivin as a strong candidate. We observed prominent defects in translocation of CPC components to the central spindle and midbody in the presence of mutant Borealin. This is consistent with the recent report that binding of the C terminus of Survivin to Borealin is especially important for targeting to these subcellular structures (17). Intriguingly, the functional defects we observed suggest that some aspect of CPC structure other than Survivin binding per se is perturbed in the presence of the mutant. Possibilities include loss of affinity for INCENP, failure of the CPC to bind to other cellular components, and a conformational change in the CPC involving the interaction surface we have mutated. We conclude that the interaction mode we have identified in our crystal structures is present in the chromosomal passenger complex in human cells and that it is important for the full function of the complex.