Regulation of Aurora B Kinase by the Lipid Raft Protein Flotillin-1*

The lipid raft protein Flotillin-1 was previously shown to be required for cell proliferation. Here we show that it is critical for the maintenance of the levels of the mitotic regulator Aurora B. Knockdown of Flotillin-1 induced aberrant mitotic events similar to those produced by Aurora B depletion and led to a marked decline in Aurora B levels and activity. Transfection of wild-type full-length Flotillin-1 or forms directed to the nucleus increased Aurora B levels and activity. Flotillin-1 interacted with Aurora B directly through its SPFH domain in a complex distinct from the chromosomal passenger protein complex, and the two proteins co-purified in nuclear, non-raft fractions. These observations are the first evidence for a function of Flotillin-1 outside of lipid rafts and suggest its critical role in the maintenance of a pool of active Aurora B.

in-frame with the hemagglutinin tag into the EcoRI and BglII sites, and the BglII and KpnI sites of pCMV-HA, to yield plasmids HA-FLOT and HA-SPFH. pGST-Flotillin-1 was generated by cloning the EcoRI-XhoI fragment containing the Flotillin-1 full-length cDNA into pGEX-4T. The nuclear localization sequence from the SV40 T antigen was inserted in-frame downstream of the GFP sequence into pGFP-Flot1 (10).
Real-time Reverse Transcription-PCR-RNA was isolated from cells with the RNeasy Mini Kit (Qiagen). After reverse transcription using random primers (Invitrogen) and Moloney murine leukemia virus reverse transcriptase, the reaction products were analyzed by PCR with SYBR Green incorporation. The sequences of the primers used are shown in Table 2. Two independent experiments with triplicate determinations each were performed. The ⌬⌬C t method was applied to estimate relative transcript levels, normalized for ribosomal RPS14 transcript amplification.
Immunofluorescence-Immunofluorescence was performed as described before (10). DNA was stained with Hoechst 33258 (Sigma). Fluorescent images were captured with a confocalspectral microscope (FV1000, Olympus, Tokyo, Japan) and quantified with Olympus Fluorview software. In all quantifications, at least 300 cells were scored in duplicate experiments.
Flow Cytometry-Cells were fixed and washed in 70% ethanol in phosphate-buffered saline at Ϫ20°C for 1 h, washed several times with cold phosphate-buffered saline, treated with RNase A (50 g/ml) and propidium iodide (25 g/ml), and analyzed on an Epics XL flow cytometer (Coulter, Miami, FL) for DNA content (filter set at 675 nm). Ten thousand events were analyzed. Cells were assigned to specific cell cycle phases by applying the Multicycle cell cycle analysis software (Phoenix Flow Systems, San Diego, CA).
Immunoprecipitation and Pulldown Assays-Co-immunoprecipitation experiments were performed as described (10). For pulldown assays, GST-tagged proteins were affinitypurified from bacteria with glutathione-Sepharose 4B beads (Sigma), eluted with 10 mM reduced glutathione and frozen until use. Cell lysates were incubated with glutathione-Sepharose-immobilized GST-tagged proteins, washed, eluted with Laemmli loading buffer, and subjected to immunoblotting.
TUNEL Assays-TUNEL assays were performed using the In Situ Cell Detection Kit, TMR Red (Roche Applied Science). Samples were analyzed by immunofluorescence.
Sucrose Density Gradient Centrifugation-Cells were disrupted with a Dounce homogenizer, and nuclear and non-nuclear fractions were separated by centrifugation as described (10). The non-nuclear fractions were further fractionated on discontinuous sucrose gradients. Fractions were collected and analyzed by Western blotting.

RESULTS
We had previously observed that Flotillin-1 depletion causes a significant inhibition of cell proliferation (10). To further investigate this effect, we tested siRNA duplexes (siRNA-92, -93, and -94) or shRNAs (shRNA-1 and -2) that target five distinct sequences on the Flotillin-1 mRNA and deplete protein levels with different efficiencies (Fig. 1A). Of these, siRNA-94 and shRNA-2 reproducibly inhibited Flotillin-1 expression at significant levels and were used in subsequent experiments (Figs. 1 and 2). The ensuing proliferative inhibition showed a good correlation with the degree of Flotillin-1 depletion, obtained with two different shRNA sequences (shRNA-1 and -2, Fig. 1B). Growth inhibition was accompanied with the accumulation of a significant fraction of cells with aberrant mitotic events, including uncongressed and lagging chromosomes, multispindle cells ( Fig. 1, C-E), multinucleated and apoptotic cells (Fig. 1D, and supplemental Fig. S1). Cells treated with control siRNA or shRNA duplexes presented normal mitotic morphologies and correct chromosome congression at the metaphase plate, with Ͻ5% showing aberrant mitoses in all cases. Knockdown of Flotillin-1 caused an accumulation of cells in prometaphase and a decrease in the proportion of cells distributed in metaphase, anaphase, and telophase ( Fig. 1F), without affecting the mitotic index (Fig. 1G) or the cell cycle phase distribution (supplemental Fig. S2).
These effects are reminiscent of those caused by depletion of the chromosomal passenger complex (CPC) protein Aurora B (11). Indeed, RNA interference depletion of Aurora B in HeLa cells led to a similar failure of chromosomes to congress, failure to complete cytokinesis, and the accumulation of multinucleated cells ( Fig. 1D; see also Fig. 2D).
Because of the above phenotype, we hypothesized the occurrence of a functional association between Flotillin-1 and Aurora B. Depletion of Flotillin-1 by siRNA-94 or shRNA-2 was followed by a significant decline in Aurora B protein levels (Fig. 2, A and B). The decrease in Aurora B was paralleled by a marked decrease in the phosphorylation of histone H3 at serine 10, a major substrate of Aurora B activity (12,13) (Fig. 2 (A and B) and supplemental Fig. S3). Likewise, depletion of Flotillin-1 caused a reduction in the levels of INCENP, a

Flotillin-1 Regulates Aurora B
protein tightly associated with Aurora B (Fig. 2, C and E). The stability and relative abundance of Aurora B and INCENP are strongly interdependent (14), and therefore the down-regulation of INCENP after Flotillin-1 knockdown could be an indi-rect consequence of the depletion of Aurora B. Because INCENP is an activator of Aurora B (11), its concomitant decrease after Flotillin-1 knockdown might also contribute to the significant reduction of histone H3 (Ser-10) phosphoryla-

Flotillin-1 Regulates Aurora B
JULY 2, 2010 • VOLUME 285 • NUMBER 27 tion observed. Transfection of cells depleted of endogenous Flotillin-1 with HA-Flotillin-1, resistant to siRNA-94 and specifically targeting the endogenous protein, re-established the levels of Aurora B and the levels of phosphorylation of histone H3 at serine 10 ( Fig. 2C). Similarly, expression in these cells of an Myctagged form of Aurora B led to the complete correction of the chromosomal abnormalities caused by Flotillin-1 depletion (Fig. 2D). These results suggest that the mitotic phenotypes caused by knockdown of Flotillin-1 are due to a down-regulation of Aurora B. Supportive of these results, the depletion of Flotillin-1 and the concomitant downregulation of Aurora B and phospho-histone H3 (Ser-10) levels were also detected in PC3 prostate cancer cells (supplemental Fig. S4).
The best described function of Aurora B is in mitosis. As expected, in mitotically enriched nocodazoletreated cells, depletion of Flotillin-1 clearly diminished the levels of Aurora B and phosphorylated histone H3 (Ser-10) (Fig. 2E). In addition, INCENP levels also diminished in both mitotic and asynchronous cells depleted of Flotillin-1, although changes in other CPC components in mitotic cells were not significant. The decrease in Aurora B levels that follows knockdown of Flotillin-1 might reflect a decrease in the amount of holo-CPC formed in mitosis, as also suggested by the altered mitotic phenotypes observed. To explore this possibility, equal amounts of proteins from cells depleted of Flotillin-1 and treated, or not, with nocodazole were immunoprecipitated with Aurora B antibody (Fig. 2F). The amount of Aurora B and co-immunoprecipitated INCENP and survivin was decreased in Flotillin-1-depleted cells, suggesting a depletion of both the Aurora B/INCENP subcomplex and the holo-CPC in cells depleted of Flotillin-1 (11,14,15).
Knockdown of Flotillin-1 by shRNAs did not significantly affect the levels of Aurora B or INCENP

Flotillin-1 Regulates Aurora B
transcripts, as shown by quantitative real-time reverse transcription-PCR (Fig. 2G), ruling out a direct effect on their mRNA levels and suggesting that the down-regulation of Aurora B and INCENP proteins in Flotillin-1-depleted cells is a post-transcriptional effect. Treatment of cells with the proteasome inhibitors lactacystin or epoxomicin restored to control levels the Aurora B protein levels that were down-regulated upon Flotillin-1 depletion, suggesting that Flotillin-1 depletion might affect the stability of Aurora B (supplemental Fig. S5). However, the half-life of the kinase was not significantly different in Flotillin-1-depleted from control cells (supplemental Fig. S6), suggesting that mechanisms other than protein stability might be involved in the regulation of Aurora B levels by Flotillin-1.
Next, we ascertained if the functional association observed between Flotillin-1 and Aurora B involved physical interactions between these proteins. Both endogenous and transfected Flotillin-1 co-immunoprecipitated with endogenous Aurora B and INCENP but not with borealin or survivin (Fig. 3 (A and B) and supplemental Fig. S7). In reciprocal experiments, anti-Aurora B brought down Flotillin-1 and, as expected, also survivin, borealin and INCENP ( Fig. 3B and supplemental Fig. S7). Moreover, GST-Aurora B, but not GST-survivin, pulled down endogenous and exogenous Flotillin-1 (Figs. 3C and 4A). As controls, both GST-Aurora B and GST-survivin pulled down the CPC protein borealin (Fig. 3C). Conversely, GST-Flotillin-1 pulled down endogenous Aurora B and INCENP from whole cell extracts (Fig. 3D). In vitro translated Flotillin-1 specifically interacted with immobilized GST-Aurora B, evidencing a direct interaction between the two proteins (Fig. 3E). These results support the occurrence of a complex between Flotillin-1 and Aurora B distinct from the canonical CPC. Although Aurora B forms a tight tetrameric complex with the CPC proteins in mitotic cells (15), a substantial pool of Aurora B is found in other complexes (16 -18), including an Aurora B/INCENP subcomplex devoid of survivin or borealin (16), which does not target to any defined structure during mitosis and might have functions distinct from the CPC (15). This subcomplex has been suggested to be directly responsible for the phosphorylation of histone H3 at Ser-10 (13,19). Consistent with its association with this non-CPC Aurora B/INCENP complex, Flotillin-1 did not localize to kinetochores during mitosis (supplemental Fig. S8), and, in mitotic cells depleted of Flotillin-1, no changes were seen in the expected localization of the CPC proteins Aurora B, survivin, and INCENP (supplemental Fig. S9).
Flotillin-1 contains two major functional and structural domains, the SPFH domain, also present in stomatin, prohibitin, podocin, and erlin, and the coiled-coil FLOTILLIN domain, shared only with the closely related paralog Flotillin-2 (20). Pulldown experiments with lysates from cells transfected with HA-Flot1-SPFH, expressing only the SPFH domain, or with HA-Flot1-FLOT, expressing only the FLOTILLIN coiledcoil domain, showed that the SPFH domain, but not the FLOTILLIN domain of Flotillin-1, was sufficient to associate with Aurora B (Fig. 4A). Immunofluorescent analysis of cells transfected with full-length or deletion variants showed that the SPFH domain of Flotillin-1, but not its coiled-coil domain, retained the capacity for nuclear localization, comparable to that of the full-length protein (Fig. 4B). Furthermore, subcellular fractionation and sucrose gradient centrifugation showed that, unlike Flotillin-1, Aurora B did not sediment in the lipid raft fraction, whereas both Flotillin-1 and Aurora B co-purified in the nuclear fraction (Fig. 4C). These observations suggest that Flotillin-1 and Aurora B interact outside of lipid rafts and mainly in the nucleus, through the SPFH domain of Flotillin-1.
To study the effects of overexpression and nuclear localization of Flotillin-1 on Aurora B protein levels, HeLa cells

Flotillin-1 Regulates Aurora B
were transfected either with HA-Flotillin-1 or a form forcefully directed to the nucleus through the inclusion of a nuclear localization signal at the N terminus of the protein, GFP-NLS-Flot1. Transfection of increasing amounts of the latter plasmid, but not a control plasmid expressing GFP alone, resulted in corresponding increases in the levels of Aurora B (Fig. 4, D and E). Transfection of GFP-NLS-Flot1 induced a large enhancement of histone H3 (Ser-10) phosphorylation that was significantly superior to the increase in Aurora B levels observed in the same experiments (Fig. 4, D and E), suggesting that Flotillin-1 regulates not only the protein levels but also the enzymatic activity of the kinase. In contrast, transfection of Flotillin-1-HA or HA-Flotillin-1⌬C, forms that are deficient in nuclear translocation and are unable to induce cell proliferation (10), did not induce increased levels of Aurora B (supplemental Fig. S10). In these experiments, the levels of endogenous Flotillin-1 were significantly increased after transfection of HA-Flot1 or GFP-NLS-Flot1 (Fig. 4, D and E; see also Fig. 2C), suggesting that overexpression of exogenous Flotillin-1 leads to the accumulation of the endogenous protein. It has been shown that the formation of homo-and heterooligomers stabilizes Flotillin-1 (25). These observations, together with the significant decline in Aurora B protein levels upon depletion of Flotillin-1, reinforce the hypothesis of a direct correlation between levels and nuclear localization of Flotillin-1 with levels and activity of Aurora B.

DISCUSSION
Aurora B kinase activity peaks in mitosis, when it promotes the destabilization and release of microtubules that are incorrectly attached to kinetochores (11). After metaphase, the kinase either associates with the central spindle in anaphase or is inactivated by PP2A (21). Termination of Aurora B is also achieved by APC/C ubiquitination and proteasome degradation (22,23). Inhibition of Aurora B leads to mono-orientated and mal-orientated chromosomes (24,25) and impairs cytokinesis with appearance of multinucleated cells (11,12,14). Aurora B also has functions outside of mitosis, as exemplified by its regulation of mammalian target of rapamycin activity whereby it is required for the G 1 -to S-phase transition (18).
Our observations suggest that Aurora B protein levels and activity are dependent on the levels of Flotillin-1. In our experiments, knockdown of Flotillin-1 caused a decline in cell proliferation that was associated with a decrease in the levels and function of Aurora B. Conversely, overexpression of Flotillin-1 was associated with increased Aurora B levels and cell proliferation and significantly stimulated phosphorylation of its major substrate histone H3 (Ser-10).
Our evidence also suggests that Flotillin-1 forms a complex with Aurora B and INCENP that does not contain survivin or borealin, and it is therefore different from the chromosomal passenger complex (14,17). In agreement with its association with a non-CPC Aurora B/INCENP complex, Flotillin-1 failed to localize to kinetochores in mitosis. Flotillin-1 interacted with Aurora B directly through its SPFH domain and, when it was directed to the nucleus, greatly induced the activity of the kinase on its substrate histone H3 (Ser-10). Our observation that only nuclear Flotillin-1 participates in the regulation of Aurora B is in agreement with our previous findings showing that only forms of Flotillin-1 capable of nuclear translocation are able to stimulate cell proliferation (10). Although Flotillin-1 does not interact with the CPC at mitosis, it might affect its function by determining the amount of Aurora B available to enter the complex, as suggested by our observation that depletion of Flotillin-1 was accompanied with a concomitant depletion of holo-CPC.
The requirement for Flotillin-1 to maintain Aurora B function and holo-CPC abundance represents a new function for Flotillin-1, possibly related to the chaperone functions found for other SPFH domain-containing proteins (26,27). Prohibitins show a partial homology with GroEL/Hsp60 that has suggested a chaperone function for oligomerized prohibitins, defined as "holdases," for the assembly of the respiratory chain complexes in mitochondria (28,29). Similarly, the notion that flotillins form scaffolds for protein assemblies that are involved in signaling across the plasma membrane is supported by several evidences (1,30). Flotillin-1 and Flotillin-2 associate in hetero-or homo-oligomers that stabilize the oligomerized proteins, such that the proteasome-dependent degradation of Flotillin-1 is accelerated in the absence of Flotillin-2 (26). Our observation that the overexpression of exogenous HA-Flot1 and GFP-NLS-Flot1 significantly increased the levels of endogenous Flotillin-1 may be related to the formation of stable homo-oligomers between endogenous and exogenous Flotillin-1.
Currently we do not know the mechanism by which Flotillin-1 depletion causes a down-regulation of Aurora B. Our observations, that Flotillin-1 depletion induces a down-regulation of Aurora B but does not affect its transcript levels or its protein half-life, might suggest regulation at levels other than transcription or protein degradation, possibly by modulating the rate of translation of Aurora B. On the other hand, overexpression of forms of Flotillin-1 that can translocate to the nucleus increases the levels of Aurora B and strongly enhances its activity, as deduced from hyperphosphorylation of histone H3 (Ser-10). Flotillin-2, a closely related Flotillin-1 paralog, does not translocate to the nucleus (10), and therefore it would not be expected to participate in the nuclear Flotillin-1/Aurora B complex observed here.
To our knowledge, our observations are the first evidence that Flotillin-1 plays relevant physiological roles beyond its previously known functions in cell and organelle membranes, namely the regulation of the overall abundance of Aurora B for a correct and timely cell cycle progression. These results are compatible with a model in which Flotillin-1 is released from its membrane localization in response to extracellular cues, such as serum growth factor stimulation (10), followed by migration to the nucleus where it enhances Aurora B kinase activity and facilitates CPC function. It remains to be explored if the nuclear function of Flotillin-1 observed in the present study is specific for a subset of proteins with which it associates, such as Aurora B, or if it involves more general mechanisms that may regulate additional proteins involved in cell cycle control. Because increased levels of Aurora B are frequently associated with the development of cancer (31), targeting Flotillin-1 might represent a novel approach to modulate Aurora B activity and a potentially useful avenue for investigation in cancer therapy.