Aurora-A kinase interacting protein (AIP), a novel negative regulator of human Aurora-A kinase.

Aurora kinases have evolved as a new family of mitotic centrosome- and microtubule-associated kinases that regulate the structure and function of centrosomes and spindle. One of its members, Aurora-A, is a potential oncogene. Overexpression of Aurora-A is also implicated in defective centrosome duplication and segregation, leading to aneuploidy and tumorigenesis in various cancer cell types. However, the regulatory pathways for mammalian Aurora-A are not well understood. Exploiting the lethal phenotype associated with the overexpression of Aurora-A in yeast, we performed a dosage suppressor screen in yeast and report here the identification of a novel negative regulator of Aurora-A, named AIP (Aurora-A kinase Interacting Protein). AIP is a ubiquitously expressed nuclear protein that interacts specifically with human Aurora-A in vivo. Ectopic expression of AIP with Aurora-A in NIH 3T3 and COS cells results in the down-regulation of ectopically expressed Aurora-A protein levels, and this down-regulation is demonstrated to be the result of destabilization of Aurora-A through a proteasome-dependent protein degradation pathway. A noninteracting deletion mutant of AIP does not down-regulate Aurora-A protein, suggesting that the interaction is important for the protein degradation. AIP could therefore be a potential useful target gene for anti-tumor drugs.

20q-13.2 (30), which is frequently amplified in several human tumors (29, 34 -38). Ectopic overexpression of Aurora-A in near diploid normal breast epithelial cells causes centrosome hyperamplification and aneuploidy (31). Also, overexpression of Aurora-A in rodent cells displayed cellular transforming activity, suggesting that when overexpressed, Aurora-A could function as a potential oncogene (29,31). The extensive research done on the yeast homolog Ipl1, Xenopus homolog Eg2, and Drosophila aurora had shed some light into the functional role of Aurora-A kinase in mitosis. Kinesin-related proteins CIN8 (39), Pav (40) and Eg5 (41) were found to interact directly with the Aurora kinase homologs in yeast (Ipl1), Drosophila (aurora), and Xenopus (Eg2), respectively. Also, yeast Ipl1 is also found to interact with the kinetochore protein Ndc10, implying the possible role of Aurora kinase in the establishment of the mitotic checkpoint via monitoring the capture of the chromosome kinetochores by the spindle microtubule (24,42,43). Human Aurora-A kinase is shown to interact with Cdc20 (44), which is involved in the mitotic activation of anaphase-promoting complex APC/C (45).
Presently, only very limited knowledge is available on the function(s) of Aurora-A kinase in mammals. Understanding the functions of Aurora-A kinase and delineation of the Aurora-A kinase signaling pathway would definitely help us to have a clearer understanding of the role of the kinase in chromosome segregation and neoplastic transformation. Hence, in an effort to identify any interacting proteins as well as the negative regulators of Aurora-A kinase, a dosage suppressor screen in which HeLa cell cDNAs that can alleviate Aurora-A-mediated cytotoxicity in yeast has been carried out. In this paper, we report the identification of AIP, one such potential negative regulator of Aurora-A kinase.

EXPERIMENTAL PROCEDURES
Yeast Dosage Suppressor Screening-Yeast strain EGY188 (MATa trp1his3ura3leu2::2 LexAop-LEU2) was maintained in the rich YPD medium. Yeast transformation, plasmid isolation, and protein extracts were prepared as described (46). For cDNA library screening, EGY188 cells were grown to log phase in YPD and cotransformed with plasmids containing 150 g of Aurora-A cDNA in pEG202 and 150 g HeLa cell cDNAs in pJG4-5 using the LiOAc method (47). The resulting transformants were selected on galactose containing synthetic dropout media lacking histidine and tryptophan (SDϪHisϪTrp). Yeast clones, which survived the Aurora-A-mediated cytotoxicity, were reconfirmed by streaking onto glucose-and galactose-containing synthetic dropout media, and the clones, which grew only on the galactose-containing plates, were characterized further by sequencing.
Cloning of AIP and Plasmid Constructs-To clone a full-length AIP cDNA, a PCR-based approach was employed. Two primers, GG8 (5Ј-CGC TGC CGA TCG GGG CCG ACT-3Ј) and GG10 (5Ј-ACT ACG GAT CAC AGC AGC AAC-3Ј), were designed for PCR cloning of AIP from the HeLa cell cDNA library. All Aurora-A and AIP constructs were made in the mammalian expression vector pCDNA3 (Invitrogen). The cyclin B1 expression plasmid pAPuro-CyclinB1 was a kind gift from Dr. Prochownik, Pittsburgh, PA. To trace the transfected AIP, a FLAG epitope was introduced at the N terminus of both truncated AIP-TR (87-600 bp) and full-length AIP (1-600 bp) constructs by PCR as described previously (28).
Northern Blot Analysis-Pre-made blots containing poly(A) RNA isolated from adult human tissues and a human cancer cell line panel were purchased from Clontech and used for hybridization with AIPspecific probe. Blots were hybridized according to Church and Gilbert (48) with a 477-bp AIP 3Ј-end fragment labeled using a random prime labeling kit. Blots were then stripped and reprobed with ␤-actin to quantitate RNA loading.
Cell Culture, Transfection, and Drug Treatment-NIH 3T3 and COS cells were maintained in Dulbecco's modified Eagle's medium, and HeLa cells were maintained in RPMI 1640 supplemented with 10% fetal bovine serum. Transfections of cultured cell lines have been carried out using LipofectAMINE (Invitrogen) according to the manufacturer's recommended protocol. Typically, 8 ϫ 10 5 HeLa cells were seeded in a 60-mm dish 24 h prior to the transfection and transfected with different plasmids at a total concentration of 3 g. For the in vivo interaction assay, equal amounts of either HsAurora-A or MmAurora-B plasmids were cotransfected with different combinations of control or AIP-expressing constructs using the LipofectAMINE PLUS reagent (15 l of LipofectAMINE and 8 l of PLUS reagent) for 5 h. Similarly, 7 ϫ 10 5 NIH 3T3 cells or 1.8 ϫ 10 6 COS cells were plated in a 60-mm dish 24 h prior to the transfection. For Aurora-A degradation study, COS7 cells were cotransfected with HsAurora-A and FLAG-tagged AIP at different ratios while maintaining the total amount of DNA transfected to 3 g. The same optimized transfection conditions were used. For immunofluorescence staining, 3 ϫ 10 5 cells were seeded on the coverslip placed in the 35-mm dish 1 day prior to the transfection. A total of 1 g of plasmid DNA and 6 l of LipofectAMINE/6 l of PLUS reagent were used for transfection. To inhibit 26 S proteasome-mediated protein degradation, COS cells were treated with 20 M N-Cbz-Leu-Leu-Leu-AL (MG132; Sigma), 25 M ALLM (Calbiochem), 25 M lactocystin ␤-lactone (Calbiochem), and 150 M ALLN (Calbiochem) for 12 h.
Cell Lysis, Immunoprecipitation, and Immunoblotting-The cells were lysed for 15 min on ice in lysis buffer (1ϫ TBS, 10% glycerol, 1% Nonidet P-40) containing protease inhibitors mixture (Roche Molecular Biochemicals). The lysates were cleared by centrifugation at 13,000 rpm for 10 min at 4°C. The protein concentration of the lysates was measured by the Bio-Rad Protein Assay (Pierce). Prior to the immunoprecipitation, 1 mg of lysates was precleared by incubation with 80 l of 50% slurry of protein G-agarose (Sigma) for 1 h at 4°C. For antibody coupling to the protein G-agarose, 20 l of rabbit anti-HsAurora-A serum (44) or 6 g of FLAG M2 mouse monoclonal antibody (Stratagene) was incubated with 80 l of 50% slurry of protein G-agarose (Sigma) for 1 h at room temperature. For immunoprecipitation, the precleared lysate and antibody-coupled protein G-agarose were mixed and rotated for 2 h at 4°C. Immune complexes were washed twice with wash buffer I (1ϫ TBS, 10% glycerol, 0.5% Nonidet P-40, 1% bovine serum albumin) and twice with wash buffer II (1ϫ TBS, 10% glycerol, 0.5% Nonidet P-40). The immune complexes were solubilized by boiling with SDS sample buffer and resolved by SDS-PAGE. The proteins were subsequently transferred to Hybond Cϩ nylon membrane (Amersham Biosciences). After blocking with 5% nonfat milk in TBS, the blots were incubated with rabbit anti-HsAurora-A (1:1,500) or mouse monoclonal anti IAK1 (Transduction Laboratories) at a dilution of 1:1,000 or FLAG M2 mouse monoclonal antibody (Stratagene) at 1:2,000 overnight at 4°C. The horseradish peroxidase-conjugated secondary antibodies were also diluted accordingly in blocking buffer (goat anti-rabbit horseradish peroxidase (Bio-Rad), 1:5,000; goat anti-mouse horseradish peroxidase (Pierce), 1:8,000) and incubated with the blot for 1 h at room temperature. The secondary antibodies were detected by enhanced chemiluminescence (ECL; Amersham Biosciences) and exposed to Kodak Biomax MR film.
Construction of AIP Deletion Mutants-Four AIP deletion mutants were created by PCR-based deletion mutagenesis. A 99-bp and a 198-bp deletion, each separately from the N and C terminus of AIP, were synthesized using four pairs of primers flanking the desired domain. The forward primers were designed to add the 8-amino acid FLAG tag to the N terminus of each mutant protein. The amplified fragments spanning different regions of AIP were cloned into pCDNA3 for expression purposes. The expected sequences of the deletion mutants were confirmed by sequencing.
In vivo interaction and degradation assays were carried out with these AIP mutants as described previously.
Immunofluorescence Staining-Cells grown on coverslips were fixed in Ϫ20°C methanol for 5 min at room temperature. After blocking for 30 min in blocking buffer (1ϫ TBS, 1% bovine serum albumin, 0.1% Triton X-100, 10% goat serum, 0.02% sodium azide), cells were incubated with the primary antibody, mouse anti-FLAG (Stratagene; 1:800), for 1 h at room temperature. The cells were washed thoroughly in 1ϫ TBS and incubated further with the respective secondary antibodies. Alexa Fluor 488-conjugated goat anti-mouse IgG (Molecular Probes) was used as the secondary antibody. For propidium iodide staining, cells were incubated with 0.05 g/ml propidium iodide. Cells were analyzed by using a Leica epifluorescence microscope (Bio-Rad) equipped with a multiband filter set and/or confocal microscopy. (29). By exploiting the lethal phenotype of Aurora-A kinase, we attempted to isolate mammalian proteins that can suppress the lethal phenotype when cotransformed and rescue the yeast from Aurora-A-mediated death.

Molecular Cloning of AIP-Overexpression of Aurora-A kinase is lethal in yeast
For this purpose, a plasmid construct was made in yeast expression vector pEG202 where constitutive expression of the full-length Aurora-A kinase in yeast is achieved using the alcohol dehydrogenase promoter. Yeast strain EGY188 was cotransformed with this Aurora-A plasmid and a HeLa cell cDNA library in pJG4-5 where cDNAs were expressed under a galactose-inducible GAL1 promoter. 0.5 ϫ 10 6 cotransformants were screened, and the resulting positive clones were selected for galactose-dependent reversal of Aurora-A-mediated cell death and characterized further by sequencing. Interestingly, sequence analysis of a total of 141 positive clones revealed that a 477-bp cDNA fragment, which we designate as AIP, containing the 3Ј-end of the mRNA, was represented 17 times. The authenticity of these 17 clones was verified by the galactosedependent rescue from Aurora-A-mediated cell death. The high frequency (12% of the total) and the reproducible rescue from the Aurora-A-mediated lethality by AIP led us to characterize this cDNA fragment further. The predicted translation product of the cDNA fragment isolated by the suppressor screen in yeast is presented in Fig. 1a. Comparison of the protein and nucleotide sequence of AIP with the sequences in the GenBank data base revealed that it is identical to the sequence corresponding to an uncharacterized protein with the accession number AK000615, a sequence submitted to the data base as a part of the human genome sequence project. Sequences similar to human AIP were found in mouse and rat EST data bases also. Fig. 1b compares the deduced amino acid sequence of AIP with the homologous sequences available in the data bases. Human AIP shares 72 and 73% identity at the amino acid level over its entire length with mouse and rat AIP, respectively. However, AIP-related sequences were not found in the lower eukaryotic genomes such as yeast, Drosophila, and C. elegans. Based on the above information, we concluded that AIP is a novel gene and cloned the putative full-length AIP cDNA by 5Ј-rapid amplification of cDNA ends. The full-length AIP cDNA contains a 597-bp open reading frame that encodes a 199amino acid polypeptide with a predicted molecular mass of 22 kDa. RNA blot analysis of human tissues and cancer cell lines indicated that AIP is ubiquitously expressed in a wide variety of tissues, especially high in heart, skeletal muscles, and testis (Fig. 2a). Computer-assisted search for the motifs presented in AIP protein found a tandem bipartite nuclear localization signal, suggesting AIP could be a nuclear protein. Indeed, ectopically expressed FLAG epitope-tagged AIP was localized to the nuclear compartment of the cell (Fig. 2b).
AIP Interacts with HsAurora-A in Vivo-The dosage suppressor screen employed here to isolate AIP is capable of identifying both the direct and indirect regulator(s) of Aurora-A kinase. Preliminary information that AIP might interact directly with Aurora-A kinase came from the yeast two-hybrid in vivo interaction assay where the partial AIP cDNA interacted with Aurora-A to activate the LEU reporter in yeast (data not shown). To verify whether a similar interaction between AIP and Aurora-A occurs in mammalian cell context, we overexpressed the FLAG-tagged AIP cDNA into HeLa cells and attempted to coimmunoprecipitate the Aurora-A with the transfected AIP protein. The results presented in Fig. 3 indicate that AIP associates with Aurora-A in vivo and that AIP can be coimmunoprecipitated with Aurora-A, and conversely, Aurora-A can be coimmunoprecipitated with FLAG-tagged AIP using FLAG antibody. However, it is noted that the interaction of transfected AIP with the endogenous Aurora-A in vivo was difficult to demonstrate. We presumed that the difficulty in demonstrating the coimmunoprecipitation could be caused by the lower amounts of Aurora-A available in AIP-transfected cells. The result presented in Fig. 3a is the best that is achieved under the given experimental conditions. To explore the interaction further, HeLa cells that are otherwise contained in comparatively lower levels of Aurora-A protein (data not shown) were transfected with HsAurora-A together with FLAG-tagged AIP, and coimmunoprecipitation followed by Western blot analysis were carried out. The results presented in Fig. 3b demonstrate that the transfected AIP protein and Aurora-A protein can be coimmunoprecipitated independently of whether FIG. 1. AIP amino acid sequence alignment. a, the deduced amino acid sequence of AIP isolated by dosage suppressor screen in yeast is presented. This sequence lacks the 87 amino acids from the N terminus of the full-length AIP protein. The location of the tandem bipartite nuclear localization signal is highlighted with italics and underlining. b, amino acid sequence alignment of human AIP (hAIP) with those of mouse (mAIP) and rat (rAIP). The mouse and rat sequences were derived from the EST data base. The mouse AIP sequence was derived from EST clones AI425574 and AA545527, and the corresponding rat AIP sequence was derived from EST clone AI104388. Identical amino acids in the sequences are presented in bold face.
FLAG M2-or Aurora-A-specific antibodies were used.
Overexpression of AIP Down-regulates Aurora-A Protein-Because AIP has been isolated as the negative regulator of Aurora-A kinase, we presumed that direct interaction of AIP with Aurora-A kinase should result in the down-regulation of either the stability and/or activity of Aurora-A kinase. In an attempt to study the impact of AIP-Aurora-A interaction on the Aurora-A protein, the levels of Aurora-A protein in AIP-transfected cells were investigated. Initial attempts to study the effect of AIP overexpression on the levels of Aurora-A protein were unsuccessful because of lower transfection efficiency, which was not sufficient to demonstrate the effect of AIP overexpression on endogenous Aurora-A protein. Hence, dividing NIH 3T3 or COS cells were cotransfected with FLAG-tagged AIP and HsAurora-A expression constructs at different ratios, and the levels of HsAurora-A protein were followed by Western blot analysis. A human Aurora-A-specific peptide antiserum was used to detect the transfected human Aurora-A in the background of the endogenous mouse and monkey Aurora-A protein. Ectopic expression of Aurora-A protein in human or monkey cell lines resulted in multiple Aurora-A-specific bands. These protein bands were verified to be Aurora-A-specific by Western blot analysis with two different Aurora-A-specific antibodies (data not shown). In COS cells, ectopic expression of Aurora-A results in two Aurora-A-specific bands of which the top band comigrated with the 46,000 endogenous Aurora-A protein from HeLa cells (data not shown). The nature of these other fragment(s), at present, is not clear. However, the results presented demonstrate that AIP, when overexpressed, could down-regulate the Aurora-A protein-specific bands in both a dose-dependent (Fig. 4a) and time-dependent manner (Fig. 4b).
Both full-length AIP as well as the N-terminal truncated form of AIP (AIP-TR) were able to down-regulate Aurora-A protein (data not shown) although the truncated AIP was more efficient in that it could completely deplete the ectopic expressed Aurora-A protein in COS cells (Fig. 4, a and b).
AIP Interaction with Aurora-A Is Important for the Downregulation of Aurora-A-To address the question of whether the interaction between AIP and Aurora-A is a necessary step for the down-regulation of Aurora-A, we attempted to isolate a deletion mutant of AIP protein, which does not interact with Aurora-A protein. A total of four deletion mutants lacking regions from either the N or C terminus of AIP protein were constructed and used for Aurora-A interaction studies as described previously. The size and location of these deletions in the different deletion mutants in relation to the wild type AIP protein are given in Fig. 5a. Expression of these deletion mutants in HeLa as well as COS cells showed that these mutant proteins have comparable stability (data not shown) except the FIG. 2. AIP mRNA expression and nuclear localization. a, Northern blot analysis of AIP mRNA in adult human tissues and cancer cell lines was carried out with pre-made Northern blots purchased from Clontech. The blots were hybridized with the 477-bp AIP cDNA derived from the yeast dosage suppressor screen. The blot was stripped and reprobed with ␤-actin. b, HeLa cells were transiently transfected with a FLAG epitope-tagged AIP cDNA, and the subcellular localization of the transfected AIP protein (panel 1) was detected by staining with FLAG M2 monoclonal antibody (Stratagene) followed by confocal microscopy. Counterstaining of DNA was carried out with propidium iodide (panel 2). Panel 3 represents the merged image to show the nuclear localization of the transfected AIP protein.

FIG. 3. AIP interacts with HsAurora-A kinase in vivo.
a, HeLa cells were transfected with empty vector or FLAG-tagged AIP as described under "Experimental Procedures." Cell lysates equivalent to 1 mg of protein were used for immunoprecipitation (IP) with rabbit antiserum against HsAurora-A protein as well as mouse FLAG M2 monoclonal antibody. The cell lysates and the corresponding immunoprecipitates were separated by PAGE and blotted onto nitrocellulose filters. The blots were probed reciprocally with HsAurora-A (lanes 1-4) and FLAG M2 (lanes 5-8) antibodies. b, HeLa cells were transfected with HsAurora-A and FLAG-tagged AIP constructs at a 1:1 ratio as described before. Cell lysates equivalent to 1 mg of protein were used for immunoprecipitation with rabbit antiserum against HsAurora-A as well as FLAG M2 monoclonal antibody. The immunoprecipitates and the corresponding lysates were separated by SDS-PAGE and blotted onto nitrocellulose filters. The blots were probed reciprocally with HsAurora-A (lanes 1-6) antibodies and FLAG M2 (lanes 7-12). mutant ⌬C198-AIP, which showed lower stability. This mutant lacks the bipartite nuclear localization signal, and it remains to be shown whether the lack of nuclear localization signal is responsible for the lower stability and function of this mutant. However, all of the deletion mutants can be expressed successfully in both HeLa and COS cells. To compensate for the lower stability of the ⌬C198-AIP mutant, a higher ratio of ⌬C198-AIP to Aurora-A (9:1) was used instead of the usual 1:1 in the interaction assay without compromising the levels of Aurora-A protein. Under these conditions, the expression of ⌬C198-AIP protein was comparable with that of other mutant AIP proteins. To rule out further the possibility that the lower levels of ⌬C198-AIP protein in the lysate are responsible for the absence of detectable protein in the interaction assay, the coimmunoprecipitation was carried out from 1 mg as well as 2 mg of ⌬C198-AIP protein lysates. The data presented in Fig. 5b suggest that the mutants ⌬N99-AIP, ⌬N198-AIP, and ⌬C99-AIP can interact with Aurora-A protein efficiently (lanes 7-9), like the wild type protein (Fig. 3b). However, the C-terminal deletion mutant ⌬C198-AIP did not show any interaction with Aurora-A protein irrespective of the levels of the mutant protein (lanes 10 and 11). This suggests that amino acids 127-166 of AIP contain elements that are necessary for the interaction with Aurora-A protein. To investigate further the efficacy of the noninteracting ⌬C198-AIP mutant in degrading Aurora-A protein, an in vivo degradation assay as described previously (Fig. 4) was performed with the wild type AIP as well as the AIP mutants. The results presented in Fig. 5c demonstrate that the noninteracting ⌬C198-AIP mutant was less efficient in degrading Aurora-A protein compared with the wild type and other deletion mutants. This suggests that the AIP/Aurora-A interaction is important for the degradation of Aurora-A protein.
AIP Specifically Down-regulates Aurora-A-To verify the specificity of the effect of AIP overexpression on the down-regulation of Aurora-A protein, the effect of AIP overexpression on MmAurora-B, another member of the Aurora kinase family, as well as cyclin B1 was investigated. The rationale for selecting cyclin B1 is that, like Aurora-A protein, the proteasome-dependent pathway (50,51) also degrades it. COS cells were transfected with FLAG-tagged AIP-TR together with either MmAurora-B or human cyclin B1 at ratio of 9:1, respectively, and the effects of AIP-TR overexpression on the levels of these proteins were analyzed. The data presented in Fig. 6, a and b, indicate that the overexpression of AIP-TR does not affect the down-regulation of either MmAurora-B or human cyclin B1 and support the notion that AIP downregulates Aurora-A protein specifically. Also, the failure to down-regulate cyclin B1 suggests that the effect of AIP is not mediated by the generalized activation of the proteolytic machinery.
Proteasome Inhibitors Reverse the AIP-mediated Down-regulation of Aurora-A-It has been shown the proteasome plays a major role in the regulation of Aurora-A stability (50). Hence, it is possible that the effect of AIP overexpression on the downregulation of Aurora-A could be mediated through the potentiation of proteasome-dependent degradation of Aurora-A. To address this question, COS cells were transfected with FLAGtagged AIP-TR together with empty vector or HsAurora-A expression constructs, and the effect of AIP overexpression on the down-regulation of Aurora-A was followed in the presence and absence of proteasome inhibitors such as MG132, ALLN, and clasto-lactacystin ␤-lactone. As shown in Fig. 7, proteasome inhibitors could reverse the AIP-mediated down-regulation of Aurora-A protein to different levels depending on their potencies to inhibit the proteasome machinery. Calpain inhibitor ALLM could not reverse the AIP-mediated degradation of Aurora-A protein, suggesting that the cysteine protease calpain is unlikely to play a role in the AIP-mediated down-regulation of Aurora-A. Taken together, these results indicate that the proteasome plays a major role in AIP-mediated down-regulation of Aurora-A protein. DISCUSSION Aurora-A kinase is a member of a serine/threonine kinase family implicated in equal segregation of chromosomes between daughter cells. Aurora-A kinase is suggested to play a role also in tumorigenesis (29). Overexpression of Aurora-A kinase transforms cultured rodent cells and causes aneuploidy in near diploid mammary epithelial cells (31). Regulation of Aurora-A kinase expression and activity occurs at multiple levels such as gene amplification, transcription, phosphorylation, and degradation through the proteasome-dependent pathway (29,31,50). Currently, attempts are being made to understand the functions of Aurora-A kinase at the molecular level. In this paper, using a dosage suppressor screen in yeast, we have isolated and investigated AIP, a novel negative regulator of Aurora-A kinase. We have shown that AIP interacts specifically and down-regulates Aurora-A kinase by potentiating its degradation through the proteasome-dependent pathway. We demonstrated that both the full-length and N-terminal truncated AIP could interact with Aurora-A kinase, suggesting that the C-terminal portion of AIP alone is sufficient for the interaction. However, the interaction of endogenous Aurora-A kinase with AIP in cells overexpressing AIP was difficult to demonstrate probably because of the degradation of endogenous protein by AIP. This inference was supported by the observation that when Aurora-A protein levels were increased by the coexpression of Aurora-A and AIP, AIP and Aurora-A kinase can be coimmunoprecipitated readily (Fig. 3, a and b). Similarly, both full-length and truncated AIP were effective in the down-regulation of Aurora-A kinase (data not shown).

FIG. 4. AIP down-regulates HsAurora-A level in vivo. a, NIH
3T3 cells were transfected with HsAurora-A and FLAG-tagged, truncated AIP construct at different ratios starting from 1:0 to 1:9, respectively, for 36 h. Cell extracts were prepared, and HsAurora-A and AIP proteins were detected with the corresponding antibodies described before. The blot was reprobed with goat anti-actin to verify loading. HeLa cell extract was used as the positive control on the Western blot. b, COS cells were transfected with HsAurora-A and FLAG-tagged, truncated AIP constructs at a 1:9 ratio, respectively, for different time points until 48 h, and the cell extracts were analyzed for HsAurora-A and AIP proteins using monoclonal IAK1 and FLAG M2 antibodies. Extract from COS7 cells was used as the negative control. The blot was reprobed with mouse anti ␤-tubulin to verify loading.
However, the truncated AIP was more efficient in the downregulation probably because of either the higher levels of the truncated protein accumulated inside the cell or the better binding to Aurora-A kinase. The increased level of truncated AIP was evident from all of our Western blot analysis that the cells accumulated more of the truncated AIP than the fulllength AIP (Fig 3b). Analysis of the mutant AIP proteins for the interaction/degradation of Aurora-A protein demonstrated that the mutant ⌬C198-AIP lacks the elements All of the AIP variants contain a FLAG tag at the N terminus. The numbers within parentheses denote the nucleotides of AIP cDNA, and number 1 corresponds to the nucleotide A of the translational start ATG. b, HeLa cells were transfected with Aurora-A and FLAG-tagged AIP mutant constructs at 1:1 ratio, respectively, as described before, except that the ⌬C198-AIP was cotransfected at a higher ratio of 1:9. For all samples, cell lysates equivalent to 1 mg of protein were used for immunoprecipitation (IP) with rabbit antiserum against human Aurora-A. In the case of ⌬C198-AIP, coimmunoprecipitation was carried out from 1 mg as well as 2 mg of protein lysates. The immunoprecipitates and the corresponding lysates were separated by SDS-PAGE and blotted onto nitrocellulose filters. The blots were probed FLAG M2 antibody to detect the AIP proteins. c, COS cells were cotransfected with HsAurora-A and either pCDNA3 or any of the AIP mutant constructs at a 1:9 ratio, respectively, and the effect of overexpression of the AIP proteins on the degradation of Aurora-A protein was assessed at 36 h after transfection as described previously. Cell extracts were analyzed for HsAurora-A and AIP proteins using monoclonal IAK1 and FLAG M2 antibodies, respectively. The blot was reprobed with mouse anti ␤-tubulin to verify loading.

FIG. 6. Overexpression of AIP does not down-regulate either
MmAurora-B or cyclin B1. HsAurora-A or MmAurora-B or human cyclin B1 was cotransfected with FLAG-tagged AIP-TR expression construct at a ratio of 1:9, respectively, into COS cells, and the effect of overexpression of AIP-TR on the levels of HsAurora-A, MmAurora-B, and cyclin B1 was assessed. AIP-TR, HsAurora-A, and MmAurora-B (a) and AIP-TR and cyclin B1 (b) levels were analyzed by Western blot analysis using the respective antibodies. Extracts from COS7 cells were used as the negative control. The blot was reprobed with mouse anti-␤-tubulin to verify loading. essential for the AIP/Aurora-A interaction, and the interaction is important for the degradation of Aurora-A kinase (Fig.  5, a and b). Interestingly, this mutant lacks the nuclear localization signal, suggesting that the targeting of AIP to the nucleus may be necessary for the interaction and degradation of Aurora-A kinase.
We have shown that AIP down-regulates Aurora-A kinase possibly through proteasome-dependent degradation. AIP is not unique in that there are other examples of proteins involved in instigating the degradation of cell cycle-related interacting partners through the proteasome pathway. Jab1 has been shown to promote the degradation of the cell cycle regulator p27kip1 in a proteasome-dependent manner (52). However, the exact role of Jab1 in the degradation is still unclear. The WD repeat-containing protein cdc20 interacts with and targets the budding yeast anaphase regulator Pds1 (securin) for degradation through the APC/C (53). Similarly, it has been well documented that MDM2 can facilitate the degradation of p53 (54). In this case, it is evident now that MDM2 itself can act as the ubiquitin ligase facilitating the ubiquitination of p53 (55). The observation that AIP also could destabilize Aurora-A kinase specifically through 26 S proteasome raises an interesting question as to what proteasome-targeting mechanism is employed for AIP-mediated Aurora-A degradation. AIP sequence analyses do not reveal any similarity to either F box proteins (56) or U box proteins (57), which play crucial roles in targeting and ubiquitination, respectively. On the other hand, the failure of AIP-dependent cyclin B1 degradation in cells overexpressing AIP confirms the notion that AIP does not activate the 26 S proteasome machinery in a generic way. The specific interaction of AIP with Aurora-A kinase as well as the essential nature of the AIP/Aurora-A interaction for the degradation of Aurora-A supports interesting possibilities such as AIP directly modifying and/or targeting Aurora-A kinase for destabilization or AIP/Aurora-A kinase interaction being the key rate-limiting step in the Aurora-A kinase degradation pathway. Although it is known that Aurora-A kinase is polyubiquitnated before degradation by APC/ cyclosome (50), it still remains to be shown that AIP plays a role in the ubiquitination of Aurora-A kinase. It has been shown that cdc20/p55cdc, which is capable of activating APC, interacts with human Aurora-A (44). However, the question of whether cdc20 targets Aurora-A for degradation also remains to be answered.
Apart from its role in destabilization of Aurora-A kinase when overexpressed, the normal function of AIP is yet to be established. The results obtained so far point to the fact that AIP is a ubiquitously expressed nuclear protein. It will be interesting to investigate whether AIP is the normal physiological trigger for Aurora-A degradation. Other pertinent questions to be answered will include the subcellular location where interaction between AIP and Aurora-A occurs and at which stage of cell cycle AIP-mediated Aurora-A degradation occurs. Being a negative regulator of Aurora-A, a potential oncogene, there is a possibility that down-regulation of AIP could play a major role in tumorigenesis. Currently, experiments are being carried out to address these issues.
In summary, the findings reported in this paper identify a novel negative regulator of Aurora-A kinase. Understanding the normal function of AIP as well as the characterization of the molecular mechanisms involved in the AIP-mediated destabilization of Aurora-A will be the next chapter in this investigation. Moreover, the targeted degradation of Aurora-A by AIP provides us with the handle to manipulate the endogenous level of the oncogenic Aurora-A kinase. Hence, AIP could therefore be a potential target gene for anti-cancer drugs in the future.