Gankyrin Is an Ankyrin-repeat Oncoprotein That Interacts with CDK4 Kinase and the S6 ATPase of the 26 S Proteasome*

A yeast two-hybrid screen with the human S6 (TBP7, RPT3) ATPase of the 26 S proteasome has identified gankyrin, a liver oncoprotein, as an interacting protein. Gankyrin interacts with both free and regulatory complex-associated S6 ATPase and is not stably associated with the 26 S particle. Deletional mutagenesis shows that the C-terminal 78 amino acids of the S6 ATPase are necessary and sufficient to mediate the interaction with gankyrin. Deletion of an orthologous gene in Saccharomyces cerevisiae suggests that it is dispensable for cell growth and viability. Overexpression and precipitation of tagged gankyrin from cultured cells detects a complex containing co-transfected tagged S6 ATPase (or endogenous S6) and endogenous cyclin D-dependent kinase CDK4. The proteasomal ATPases are part of the AAA (ATPases associated with diverse cellular activities) family, members of which are molecular chaperones; gankyrin complexes may therefore influence CDK4 function during oncogenesis.

The 26 S proteasome is an exquisitely regulated protease responsible for most of the non-lysosomal degradation of intracellular proteins (1). The particle is responsible for the degradation of regulatory proteins including tumor suppressors (2,3), transcription factors (4,5), and proteins that regulate the cell cycle (6,7).
The 26 S proteasome consists of a cylindrical catalytic "core" containing 28 subunits (␣ 7 ␤ 7 ␤ 7 ␣ 7 ) with a regulatory complex (RC) 1 attached to each end of the proteolytic core containing at least 15 subunits (1). The 20 S cylinder contains three chambers: two distal antechambers and a central chamber containing the catalytic threonine residues (8,9). The ends of the cylindrical core appear closed in the yeast 20 S proteasome (8); the ends of these antechambers need to be opened for substrate proteins to enter into the catalytic core for proteolysis. The RC can be subdivided into a "base" and a "lid" complex (10). The base contains six ATPases, which belong to the AAA (ATPases associated with diverse cellular activities) superfamily of ATPases together with the non-ATPase RPN1 and RPN2 subunits. The lid subcomplex contains the remaining non-ATPase subunits of the RC. The AAA ATPase superfamily controls events as diverse as 26 S proteasomal functions, peroxisomal biogenesis (11), membrane docking and fusion (12), protein egress from the endoplasmic reticulum (13), nuclear transport (14), and transcription factor regulation (15). The ATPases of the RC may be involved in substrate unfolding for entry into the catalytic core of the 26 S proteasome. It has recently been demonstrated that the base subcomplex of the RC exhibits chaperone-like activity (16); and both the ClpX and ClpA ATPases of Escherichia coli, multisubunit complexes similar to the base subcomplex of the 26 S proteasome, can mediate protein unfolding (17,18) and refolding events (19,20).
The proteasomal regulatory ATPases have in addition to the Walker A and B motifs involved in ATP binding and hydrolysis, sequence patterns shared with DNA/RNA helicases (21). The SUG1 ATPase has 3Ј-5Ј DNA helicase activity (22), and the ATPase activity is stimulated by polynucleotides and polyadenylated mRNAs (23). Several of the ATPases were discovered as modulators of transcription (24). The precise role of the RC ATPases in nucleic acid metabolism is unknown. Recently, core subunits corresponding to human and have been shown to possess nuclease activity (25,26). Formally, the 26 S proteasome may have both proteolytic and nucleolytic activities and is already acknowledged as the cellular counterpart of the ribosome.
In this study we have used a combination of genetic and biochemical approaches relying on protein-protein interactions to discover an ankyrin-repeat protein that specifically associates with the C terminus of the proteasomal S6 ATPase. We have previously shown that the interacting protein, called gankyrin, when overexpressed transforms NIH3T3 cells and is oncogenic in nude mice. Overexpression of gankyrin increases both the phosphorylation and degradation of the retinoblastoma protein (Rb) in vivo (27). These findings have recently been extended to show the importance of gankyrin in a rodent model of hepatocarcinogenesis (28). We have presented a preliminary report on gankyrin (29), and here we show that gankyrin specifically interacts with the S6 ATPase and is present in complexes in cells containing CDK4 and the S6 ATPase.
Yeast Two-hybrid Analysis-Human S6 ATPase cDNA was amplified from total HeLa cell RNA by reverse transcription PCR using Pfu DNA polymerase (Stratagene) and primers containing a 5Ј EcoRI and a 3Ј SalI site.
S6 cDNA was gel-purified, A-tailed by incubation for 30 min at 72°C in 1 ϫ Taq polymerase buffer containing 2 mM ATP and cloned into the T-tailed EcoRV site (35) of pBluescript SK Ϫ (pSK-T Ϫ ) to give plasmid pHSS6. S6 cDNA was excised using EcoRI and SalI restriction endonucleases, gel-purified, and cloned into the corresponding sites of the GAL4 DNA binding domain vector, pGBT9, yielding plasmid pGBS6. S6 cDNA was sequenced to confirm PCR fidelity. An adult human Match-Maker™ brain cDNA library (CLONTECH, Palo Alto, CA) cloned into the GAL4 activating domain vector, pGAD10, was used to screen for S6-interacting proteins. Interacting plasmids were isolated by growing a His ϩ /␤-galactosidase ϩ colony in synthetic SD-leu medium overnight, lysing the cells with acid-washed glass beads (33), and transforming E. coli HB101 with the crude plasmid DNA isolated.
S4, S7, S6Ј, S8 and S10b ATPase cDNAs (36) were PCR-amplified as described for S6 cDNA from total HeLa RNA using Pfu DNA polymerase and the appropriate primers.
Plasmids pHSS4, pHSS6, or pHSS6Ј were utilized for production of the S4, S6, or S6Ј proteasomal ATPases, respectively. All plasmids were prepared by alkaline lysis. Transcription/translation in the presence of 35 S-labeled methionine in the STP3 System (Novagen) was carried out according to the manufacturer's instructions. The 35 S-labeled proteins were kept at 4°C until required. GST Pull-down Analyses-Gankyrin cDNA was excised from the plasmid p161632 with EcoRI, gel-purified, and sub-cloned in frame into the corresponding restriction site of the GST fusion vector, pGEX4T-2 (Amersham Biosciences, Inc.) to give plasmid pGXGankyrin. GST fusion constructs containing N-(GankDelN and GankDelN1) or C-terminal (GankDel56 and GankDel6) deletions have been described previously (27). GST-gankyrin (or the various deletion proteins) was induced in E. coli DH5␣ for 3 h at 25°C by the addition of isopropyl-1-thio-␤-Dgalactopyranoside (100 M final concentration) to a 100-ml bacterial culture (A 600 Ϸ 0.5). Production of the expected GST proteins was confirmed by ␣-GST Western blotting. After induction, bacteria were pelleted for 20 min at 3000 ϫ g and resuspended in 20 ml of ice-cold binding buffer (10 mM Tris-HCl, pH 7.5, 140 mM NaCl, 0.1% Nonidet P-40). Bacteria were lysed by freeze-thawing for 5 min in liquid nitrogen followed by thawing for 10 min at 37°C. The lysis procedure was repeated three times. The freeze-thawed bacteria were then subjected at 4°C to three 20 s rounds of sonication (7 amplitude), and the bacterial debris was pelleted by centrifugation at 15,000 ϫ g for 30 min at 4°C. Supernatants were stored frozen at Ϫ20°C in 100-l aliquots until needed. Free GST lysates were prepared in a similar manner from E. coli DH5␣ transformed with a pGEX4T-2 vector that did not contain the gankyrin cDNA. Free GST and GST fusion proteins were purified on glutathione (GSH)-Sepharose 4B (Amersham Biosciences, Inc.) according to the manufacturer's recommendations and dialyzed against binding buffer. Purified proteins were quantitated by the Bradford assay (37).
For the binding analyses, 6 g of either GST or the various GSTgankyrin constructs diluted to 1 ml with binding buffer were incubated with 50 l bed volumes of GSH-Sepharose (Amersham Biosciences, Inc.) pre-equilibrated in binding buffer for 60 min at 4°C. GSH-Sepharose was then washed four times with 1 ml of ice-cold binding buffer at 4°C. Reticulocyte lysates (40 l) containing 35 S-labeled proteins were diluted in ice-cold binding buffer (or binding buffer containing 5 mM MgATP, 1:25), added to immobilized GST or GST-gankyrin constructs and incubated for 2 h with continuous shaking at 4°C. After binding, each GSH-Sepharose preparation was washed four times in 1 ml of ice-cold binding buffer (or binding buffer containing 5 mM MgATP). All traces of liquid were removed from each Sepharose preparation by aspiration. Each gel sample was resuspended in 50 l of SDS-PAGE (38) loading buffer and boiled for 5 min. After electrophoresis, gels were fixed in 10% acetic acid, 10% methanol and equilibrated in 1 M sodium salicylate, pH 7.5, as a scintillant before drying. Dried gels were subjected to fluorography at room temperature for 24 to 48 h using Kodak BioMax MR film (Kodak).
Deletion of the YGR232w Gene-The YGR232w gene was amplified from 50 ng of total yeast DNA using oligonucleotide primers that spanned a KpnI site 818 bp upstream of the ATG codon (YGR2325) and a HindIII site 417 bp downstream of the stop codon (YGR2323) using a proof-reading thermostable polymerase as recommended by the manufacturer (Pfu Turbo, Stratagene). The 1921-bp product was digested with KpnI and HindIII and cloned into pUC18. The same product was also digested with KpnI and SphI, and the 794-bp 5Ј flank of YGR232w was cloned into pUC18. To amplify the 3Ј end of YGR232w, the YGR2323 primer described above was used in conjunction with another primer designed to create an XmaI site 131 bp upstream of the YGR232w stop codon (YGR232Xma). The resulting 548-bp product was digested with XmaI and HindIII and cloned into pUC18. To create the deletion allele, the URA3 gene was ligated, as a SphI-XmaI fragment, between the SphI site 24 bp upstream of the YGR232w ATG codon and the engineered XmaI site 131 bp upstream of the stop codon, thus removing 81% of the 5Ј end of the open reading frame (ORF). The 3Ј 131bp of the ORF was left intact so as not to disturb the adjacent convergently transcribed PHO81 gene. This plasmid was digested with KpnI and HindIII, and the linear DNA was used to transform the diploid strain DF5 to disrupt the YGR232w gene by homologous recombination (39). Ura ϩ colonies were isolated, and the correct recombination event was confirmed by Southern blot analysis on BglII-digested genomic DNA using the 794-bp KpnI-SphI 5Ј flank of YGR232w as a probe. A ygr232w::URA3/ygr232w::URA3 homozygous diploid strain was constructed by mating haploid ygr232w::URA3 strains of opposite mating types and selecting diploids with a micromanipulator.
Transient and Stable Transfection of Mammalian Tissue Culture Cells-Gankyrin cDNA carrying 5Ј and 3Ј NotI sites was amplified from plasmid p161632 using Pfu polymerase. Amplified cDNA was gel-purified, A-tailed, and cloned into pSK-T Ϫ yielding plasmid pGANKN5N3. cDNA was excised using NotI restriction endonuclease, gel-purified, and ligated into the corresponding site of a modified pcDNA3.1zeo ϩ plasmid to give plasmid pZHFNGANK expressing His 6 -FLAG-gankyrin (HF-gankyrin). Plasmid pCMV5LS6, expressing N-terminally HAtagged S6 ATPase (HA-S6), was constructed by excision of S6 cDNA from plasmid pGBS6 with restriction enzymes EcoRI and SalI, gel purification, and ligation into the vector pCMV5L.
pZHFNGANK was used as a PCR template to amplify, using Pfu polymerase, gankyrin cDNA carrying 5Ј EcoRI and 3Ј SalI sites. The cDNA was TA-cloned as described above and ligated into EcoRI/SalI digested vector pCMV5L yielding plasmid pCMV5LGANK.
Wild-type human CDK4 cDNA was PCR-amplified, using Pfu polymerase, from the plasmid template pKSCDK4 carrying both 5Ј and 3Ј NotI sites. The cDNA was TA-cloned and ligated into the NotI-digested modified pcDNA3.1zeo ϩ plasmid used in the construction of pZHFN-GANK yielding plasmid pZHFNCDK4. All PCR-amplified cDNAs were confirmed by DNA sequencing.
Human HEK293 and U2OS cells were used for both transient and stable transfection experiments. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum in a humidified 5% CO 2 incubator at 37°C. Transfection reactions were carried out using calcium phosphate precipitation. Stable transfectants were selected by culturing transfected cells for 3-4 weeks in the presence of 300 g ml Ϫ1 Zeocin. Western blot analysis indicated that the HEK293 clone 6D1 gave the highest levels of gankyrin expression, and this clone was used in all studies described here involving stable gankyrin-expressing cells.
After glycerol gradient fractionation, shown in Fig. 3A, fractions 5-9 were pooled and loaded on an FPLC ResourceQ column. After extensive washing the remaining proteins were eluted with a linear gradient from 50 mM to 1 M KCl. Fractions 1-60 were collected and checked for 26 S proteasome/19 S regulator by immunoblotting. Fractions 27-44 containing most of the 26 S proteasome/19 S regulator were pooled and loaded on an FPLC MonoQ. Proteins were eluted with a linear gradient from 50 mM to 1 M of KCl. At about 300 mM KCl, a single peak elutes (fractions 17, 18, and 19), which was analyzed (see Fig. 3, B and C). Western blot analysis and nondenaturing electrophoresis were performed as described (41). Affinity-purified ␣-gankyrin antibody was used at a dilution of 1:100. The anti-CSN3, anti-CSN5, anti-S4, and anti-S10a antibodies have been described previously (41).
Immuno-and Affinity-precipitation and Western Blotting-Cells (ϳ 2 ϫ 10 5 ) were lysed in 0.5 ml of "standard" lysis buffer (1% Nonidet P-40, 1% sodium deoxycholate, 150 mM NaCl, 10 mM sodium phosphate, pH 7.2) by freeze-thawing (4 times) and debris removed by centrifugation. Lysates were incubated together with the recommended quantity of antibody for 2 h with agitation at 4°C. Immune complexes were precipitated by the addition of 20 l of protein A/G-agarose beads (Santa Cruz Biotechnology), incubation for 1 h with agitation at 4°C followed by centrifugation. Precipitated immune complexes were washed three times with 1 ml of lysis buffer, excess buffer was removed by aspiration, and the immune complexes were eluted by the addition of 50 l of Laemmli loading buffer.
Nickel chelate precipitation via a poly-His tag was performed using the HisTagCatcher Kit (CytoSignal Research Products) according to the manufacturer's recommendations.
Affinity-purified complexes were analyzed by electrophoresis on polyacrylamide gels followed by blotting onto nitrocellulose membrane and detection with specific antisera and enhanced chemiluminesence.
Anti-CDK4 antibody was obtained from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Rat monoclonal anti-HA antibody was obtained from Roche Molecular Biochemicals. Mouse monoclonal anti-FLAG (M2) antibody was obtained from Sigma-Aldrich.

Isolation of Gankyrin as an S6-interacting Protein-
The gankyrin cDNA was isolated in a yeast two-hybrid screen against an adult human brain cDNA library using the human S6 ATPase as bait. The cDNA sequence is identical to a recently described cDNA encoding a protein called gankyrin (27), found in complexes with the retinoblastoma protein, and will hereafter be referred to by that name. Analysis of the primary amino acid sequence reveals that the protein is composed of tandem elements of the so-called "ankyrin repeat," a 32-amino acid unit first observed in the cytoskeletal protein, ankyrin (42).
Interaction between Gankyrin and the S6 Proteasomal ATPase Occurs in Vitro-To assess the interaction between gankyrin and the S6 ATPase, in vitro binding assays were performed using a GST fusion of gankyrin and in vitro translated 35 S-labeled proteasomal ATPases in the presence and absence of ATP. Radiolabeled ATPases were incubated with GST-gankyrin protein-linked glutathione-Sepharose beads. The radiolabeled S6 ATPase was selectively retained by gankyrin with no interaction observed between S6 and GST alone (Fig. 1). The interaction exhibited specificity toward the S6 ATPase; no binding was detected with the two other ATPases used in the assay (S4 and S6Ј; Fig. 1). No influence of ATP was observed on the binding of S6 ATPase to gankyrin (data not shown).
The S6 ATPase-Gankyrin Interaction Occurs in Mammalian Cells-To assess the interaction in a mammalian cell culture system, transient co-transfection and nickel chelate pull-down analyses of gankyrin-containing complexes was performed. HEK293 cells stably expressing HF-gankyrin were transiently transfected with plasmid pCMV5LS6 expressing HA-tagged S6 (HA-S6) and cells harvested 48 h post-transfection. HFgankyrin-containing complexes were isolated using a nickel chelate pull-down and analyzed by Western blotting using an ␣-HA monoclonal antibody to detect HA-S6. Fig. 2A shows that HA-S6 ATPase is precipitated exclusively in cells that also express HF-gankyrin (lane 4). S6 ATPase transfected into HEK293 cells stably expressing the poly-His-FLAG tag alone or the parent cell line was not detectable in the absence of HF-gankyrin ( Fig. 2A, lanes 2 and 6, respectively). Transfected HF-gankyrin was also able to interact with endogenous S6 ATPase. HEK293 cells stably expressing HFgankyrin were lysed as described previously and analyzed by nickel affinity pull-down followed by Western blotting using a mouse monoclonal antibody raised against human S6 ATPase. Fig. 2B clearly shows that endogenous S6 ATPase can be copurified with HF-gankyrin under the conditions used.
Gankyrin Associates with the RC of the 26 S Proteasome-We followed the fate of gankyrin during the preparation of various protein complexes from human erythrocytes (26 S proteasome, 19 S regulator, 26 S proteasome lid) by Western blotting using an affinity-purified anti-gankyrin antibody. As shown in Fig.  3A (middle panel), analysis of glycerol gradient fractions, an early purification step, shows that gankyrin sediments in fractions containing most of the 20 S/26 S proteasome and/or 19 S regulator (Fig. 3A, upper and lower panels). Following subsequent purification by FPLC MonoQ ion-exchange chromatography, we eluted a single peak with about 300 mM KCl that possessed most of the gankyrin protein (Fig. 3B). Inspection of a Coomassie-stained SDS-polyacrylamide gel containing the 300 mM KCl peak reveals that the fractions of the peak contain large amounts of 20 S proteasome subunits (Fig. 3B, left panel; located between 20 and 30 kDa) and higher molecular mass bands resembling the 19 S regulator subunits. Western analysis of the gankyrin-containing peak (Fig. 3B, right panel) demonstrates the presence of both the 26 S proteasomal S4 ATPase subunit, a component of the base, and S10a, a component of the lid.
We used native PAGE to address the question of which of the observed complexes contains the gankyrin protein. Inspection of the native blots in Fig. 3C reveals that gankyrin is present in a complex that is most likely the 19 S RC. It does not appear to be a component of the 20 S proteasome or the 26 S proteasome lid. Binding to the RC of the 26 S proteasome is to be expected in view of the observed interaction between gankyrin and the S6 ATPase. Gankyrin may also be present in complexes iden-tified as the signalosome (Fig. 3C, * and ****). The low molecular weight gankyrin immunoreactive band seen in Fig. 3C (**) may represent small gankyrin-containing complexes for which the composition is incompletely understood at present (also see Fig. 8). A smaller S4 immunoreactive complex of unknown composition was also observed (Fig. 3C, ***).
Gankyrin Is Not a Stably Associated Subunit of the 26 S Proteasome-We wished to address whether gankyrin is associated with the 26 S proteasome in cells other than terminally differentiated red blood cells. We used glycerol density gradient fractionation of mammalian tissue culture cells to attempt to answer this question. Glycerol gradient fractionation of HEK293 clone 6D1 stably expressing HF-gankyrin was performed, and the fractions obtained were subjected to anti-FLAG Western analysis. Fig. 4E clearly shows that all detectable gankyrin is found within complexes at the top of the gradient after disruption of cells in standard lysis buffer. However, when cells are disrupted in low detergent lysis buffer, although the majority of transfected gankyrin is still found at the top of the gradient, significant amounts are now observed within the same fractions as subparticles derived from the 26 S proteasome (Fig. 4D). Known proteasomal subunits (S6, S8, and ␣6; Fig. 4, A-C, respectively) and the major chymotrypsin activity (not shown) of the proteasome are found toward the bottom of the gradient in both lysis buffers. The presence or absence of ATP during the preparation of cell lysates and gradient analysis had no effect upon the location of gankyrin on the gradients (data not shown).
The S6-Gankyrin Interaction Is Specific for the S6 ATPase-There are six proteasomal ATPases (43), the S4, S6, S6Ј, S7, S8, and S10b subunits of the mammalian 26 S proteasome. The yeast two-hybrid assay was used to assess further the selectivity of the S6/gankyrin interaction. The plasmid constructs pASS4, pASS6, pASS6Ј, pASS7, pASS8, and pASS10b were co-transformed into the yeast strain HF7c together with the gankyrin construct p73Int2 and plated onto synthetic medium lacking leucine and tryptophan (Fig. 5A) to select for cotransformants. Individual yeast colonies were assayed for ATPasegankyrin interactions by re-streaking onto synthetic medium lacking leucine, tryptophan and histidine (Fig. 5B). The interaction between the S6 ATPase and gankyrin is absolutely specific for the S6 ATPase; no yeast growth is seen for the other gankyrin-ATPase combinations (Fig. 5B).
The S6-Gankyrin Interaction Requires the C Terminus of the S6 Protein and the Entire Gankyrin Protein-The regions of the S6 protein required for binding to gankyrin were mapped with both the yeast two-hybrid system and by GST pull-down experiments. The diagram (Fig. 6A) shows the S6 ATPase deletions used in the analysis. For two-hybrid analysis, each cDNA was cloned into a GAL4 DNA binding domain vector, whereas for GST pull-downs, cDNAs were cloned into the plasmid pCITE4a, suitable for in vitro transcription/translation. The p73Int2 plasmid encoding wild-type gankyrin was cotransformed with each of the constructs shown in Fig. 6A into the yeast strain HF7c, and cells were plated onto SD-LWH selective agar plates to assess for specific interactions of gankyrin and the ATPase deletion mutants (Fig. 6B). Growth comparable with wild-type S6 ATPase only occurs for one of the S6 ATPase deletion constructs, ⌬5 (lacking the N-terminal zipper motif). The analysis indicates that the C-terminal portion of the S6 protein is absolutely required for interaction with gankyrin. However, using this assay system expression of the C-terminal portion alone (⌬6) facilitated no interaction with gankyrin. Co-expression of other portions of the S6 protein with gankyrin failed to result in yeast growth.
The interaction of gankyrin and the S6 ATPase was also analyzed using GST pull-down experiments. The same S6 ATPase constructs were [ 35 S]methionine-labeled and subjected to binding experiments with both GST-gankyrin and free GST. Fig. 6C clearly shows that using this experimental system, sequential deletion of the S6 ATPase from the C terminus resulted in an abolition of gankyrin binding (lanes 2-5). However, sequential deletion of the S6 ATPase from the N terminus resulted in little or no decrease in gankyrin binding (lanes 6 -10). Indeed, even expression of the C-terminal 78 amino acids resulted in binding to GST-gankyrin (lane 7). No interaction with any ATPase deletion construct was observed with the free GST protein (not shown). The C terminus of the S6 ATPase is therefore predicted to be both necessary and sufficient to mediate the observed interaction with gankyrin.
In a similar manner, deletional mutagenesis was performed on the gankyrin ORF to ascertain regions of the protein needed for interaction with the S6 ATPase. GST pull-down analyses using full-length gankyrin and various gankyrin deletion mutants (Fig. 6D) was performed with in vitro translated [ 35 S]methionine-labeled S6 ATPase. Fig. 6E shows that interaction between S6 ATPase and the GST-gankyrin constructs occurs only with the full-length GST-gankyrin protein. No interaction with any gankyrin deletion construct was observed with the free GST protein (not shown). The results indicate that a fulllength, presumably correctly folded, ankyrin-repeat structure is essential for interaction with the S6 ATPase.
The Putative S. cerevisiae Homologue of Gankyrin Is a Non-essential Gene-Comparison of the cDNA and predicted protein sequence of gankyrin with the complete sequence of the S. cerevisiae genome suggested that a putative orthologous gene of unknown function (YGR232w; 35% identical, 53% similar) exists within the yeast genome. We carried out a comprehensive analysis of the potential role of this yeast ORF for cell growth and viability under both normal and stressed conditions by constructing both heterozygous and homozygous deletion strains lacking the yeast homologue of gankyrin. Fig. 7A details the strategy used to construct the deletion strains. Tetrad analysis of a YGR232w/ygr232w::URA3 heterozygous diploid (Fig. 7B) revealed that deletion of gene YGR232w had no effect on cell viability. Correct insertion of the URA3 marker cassette and deletion of 81% of YGR232w was verified by Southern blot analysis (Fig. 7D, lanes 2-6). Deletion of YGR232w also had no effect upon susceptibility to a chronic heat shock or to the amino acid analogue, canavanine (data not shown). Tetrad spore analysis of a homozygous diploid strain (Fig. 7C) revealed that spores had wild-type viability, and the diploid strain sporulated at wild-type rates (data not shown). Thus, YGR232w is not required for sporulation or germination. The correct genotype for the homozygous diploid was also verified by Southern blotting (Fig. 7D, lanes 7-11).
Gankyrin Interacts with the Cyclin D-dependent Kinase, CDK4, in Cells-The observation that gankyrin increases the phosphorylation state of Rb (27) [17][18][19] were separated by SDS-PAGE and visualized with Coomassie Blue staining or blotted onto nitrocellulose membrane and tested with anti-gankyrin (1:100), anti-S4 (rabbit polyclonal, 1:10000), anti-S6 (rabbit polyclonal, 1:2500), and anti-S10a (rabbit polyclonal, 1:5000) antibodies. C, aliquots of fractions from the same MonoQ peak of panel B were separated by nondenaturing electrophoresis, after which proteins were blotted onto nitrocellulose and either visualized by Ponceau S staining or tested with anti-gankyrin, anti-S10a, anti-S4, and anti-S6 antibodies. Gankyrin is possibly associated with COP9 signalosome (CSN, *) and smaller complexes (**). A smaller anti-S4 reactive complex of unknown composition is also apparent (***). the p16INK4a-cyclin D/CDK4-Rb pathway. HEK293 or U2OS cells stably expressing HF-gankyrin alone or transiently cotransfected with pCMV5LS6 expressing HA-S6 ATPase were used for analysis. Gankyrin-containing complexes were isolated by nickel chelate pull-down as described previously and subjected to Western analysis for components of the pathway. Fig. 8 shows that CDK4 is selectively isolated as part of a complex containing HF-gankyrin (Fig. 8A, lanes 3 and 4) from 293 cells (similar data from U2OS cells are not shown). Cells containing the parent vector alone showed no ability to precipitate CDK4 (Fig. 8A, lanes 1 and 2). Endogenous S6 ATPase could also be detected in complexes containing HF-gankyrin and CDK4 (Fig. 8B). The presence of complexes containing gankyrin and CDK4 was confirmed by cotransfection of cells with vectors expressing HF-CDK4 and HA-gankyrin. Nickel chelate pull-down followed by detection of HA-gankyrin shows that gankyrin is precipitated only in the presence of HF-CDK4 (Fig. 8C, lane 2). Further Western analyses of these gankyrin complexes did not detect the presence of the proteasomal S8 ATPase, a dimerization partner of S6 (36), subunits of the 20 S proteasome, cyclins D1 and D3, or CDK6 (not shown). These complexes may correspond to the smallest gankyrin-containing complex (**) observed in Fig. 3C. DISCUSSION The yeast two-hybrid screen with the human S6 ATPase of the RC identified gankyrin, which interacts with the S6 ATPase both in vitro (Fig. 1) and in vivo (Fig. 2) but does not bind to the other proteasomal ATPases in the two-hybrid screen (Fig. 5). The C-terminal part of the S6 ATPase is necessary and sufficient for association with gankyrin ( Fig. 6; the C-terminal 78 amino acids are necessary for interaction). Gankyrin contains multiple ankyrin repeats and a putative retinoblastoma-binding motif (27). Many cellular proteins contain ankyrin repeats, which are presumed to be involved in protein-protein interactions. For example, the cytosolic tail of the Notch receptor (44), the cyclin D-dependent kinase inhibitors (INKs) (45) and the C-terminal region of the p105 precursor of the p50 component of NFB1 (46) and IB␣ (47) have multiple ankyrin repeats. The p105 precursor undergoes limited proteolytic cleavage, and IB␣ is destroyed by the 26 S proteasome as part of the activation of NFB1 and the subsequent transcription of genes of the inflammatory and immune systems (46,47). The receptor-induced degradation of IB␣ follows polyubiquitinylation (47). However, the basal continuous degradation of IB␣ is only dependent on the C-terminal ankyrin repeats (48).
Binding of gankyrin to the S6 ATPase can occur within the RC (Fig. 3), which is in agreement with the observations of Saito et al. (49) who showed that p28 (gankyrin) co-purifies through several chromatographic steps with the RC. However, gankyrin was also observed to form lower molecular weight complexes, which may either be derived from 26 S proteasomes because conditions of cell disruption using low detergent lysis buffer are known to cause 26 S proteasome disassembly (Figs. 3C and 4 and Ref. 50) or be formed by the recruitment of additional unknown cellular proteins. Some of these complexes presumably sediment in glycerol gradients at rates equivalent to 13 S or faster and therefore are of considerable size because proteasomal ␣-subunits (see Fig. 4C) have only been described within particles of this size or greater (51). Such complexes may contain the S6 ATPase in the absence of other proteasomal antigens, as Western blot analysis of Ni 2ϩ -purified gankyrincontaining complexes (Fig. 8) showed no evidence for the S8 ATPase (the putative dimerization partner for S6 (36)) or of the 20 S complex (data not shown). This observation suggests that gankyrin, like the HEC protein (highly expressed in cancer cells), may bind to a free ATPase subunit thereby blocking or stimulating the degradation of proteins (52). Significant amounts of free nonproteasomal S7 (MSS1) ATPase have been found in human bladder carcinoma cells (T24 cells) throughout the cell cycle. The free S7 interacts with HEC, which inhibits the degradation of mitotic cyclin B in vitro (52) via modulation of the ATPase activity of the S7 protein. The HEC protein peaks in the M phase of the cell cycle and may interact with the APC/cyclosome (53), which targets mitotic cyclins for degradation. Like HEC/S7, the gankyrin-S6 interactions may occur with free S6 ATPase in the cytosol independent of the 26 S proteasomal particles; a GFP-gankyrin fusion was localized primarily to the cytosol in both HEK293 and NIH3T3 cells (data not shown). There is other evidence for the existence of ATPases in nonproteasomal complexes (54,55).
A recent study detailing the isolation of proteasomes from S. cerevisiae (56) reported that the complement of proteins purified in association with the proteasome is dependent on the presence of ATP throughout the procedure; in the absence of ATP it was reported that YGR232w (NAS6p, the putative yeast orthologue of gankyrin) co-purified with the proteasome. It is as yet unclear what physiological role the presence or absence of ATP satisfies; we were unable to show any difference between the association of HF-gankyrin with proteasome-containing gradients fractions in the presence or absence of ATP.
Cells may contain several complexes involving proteasomal ATPases that control different stages of the cell cycle. S6 ATPase-gankyrin-Rb-E2F1 interactions may control E2F1 stability in relation to the G 1 /S transition and S7 ATPase-HEC-Rb-cyclin B interactions may control cyclin B stability during the mitotic metaphase/anaphase transition. Other ATPasecontaining complexes may control protein degradation during other stages of the cell cycle.
The exact role of the ATPases in these putative complexes is not clear, particularly if free ATPases are involved in the processes. The ATPases may facilitate rapid proteasomal degra-dation of other components of the complexes or may have novel functions in relation to the expression of genes controlling the cell cycle (22).
The full complement of proteins in gankyrin complexes is yet to be determined. However, complexes containing an ankyrin repeat protein, an ATPase, and a kinase were shown many years ago to control gene expression in yeast (57). Furthermore, large complexes containing AAA family proteins are essential for chromatin remodeling, gene expression, and oncogenic transformation (58 -60).
Binding of gankyrin to the S6 ATPase appears to be exclusive to that particular member of the 26 S proteasomal ATPase

FIG. 8. Gankyrin interacts with the cyclin D-dependent kinase, CDK4, in HEK293 cells.
A, HF-tagged proteins were purified from cell lysates on Ni 2ϩ affinity resin and analyzed by polyacrylamide gel electrophoresis and Western blotting using mouse monoclonal anti-CDK4 antibody (1:500). Detection was via the ECL system. Molecular weight (M.wt) markers are shown in kDa. Lane 1 shows HEK293 stably expressing the HF tag alone co-transfected with vector pCMV5L expressing the HA tag alone. Lane 2 shows HEK293 stably expressing the HF tag alone co-transfected with vector pCMV5LS6 expressing the HA-S6 ATPase. Lane 3 shows clone 6D1 stably expressing HF-gankyrin co-transfected with vector pCMV5L expressing the HA tag alone. Lane 4 shows clone 6D1 stably expressing HF-gankyrin co-transfected with vector pCMV5LS6 expressing the HA-S6 ATPase. Lane 5 shows HEK293 cells transiently transfected with vector pCMV5L expressing the HA tag alone. Lane 6 shows HEK293 cells transiently transfected with vector pCMV5LS6 expressing the HA-S6 ATPase. The position of CDK4 protein is marked with an arrow. B, HF-tagged proteins isolated from clone 6D1 stably expressing HF-gankyrin co-transfected with pCMV5L expressing the HA tag alone were isolated on Ni 2ϩ affinity resin and analyzed by PAGE and Western blotting using mouse monoclonal anti-S6 ATPase antibody (1:500). Detection was via ECL. The position of the S6 ATPase is indicated with an arrow. C, HF-tagged proteins were prepared from whole lysates containing HF-CDK4 and HA-tagged gankyrin on Ni 2ϩ affinity resin and analyzed by PAGE and Western blotting using rat monoclonal anti-HA antibody (1:500). Detection was via ECL. Molecular weight markers are shown in kDa. Lane 1 shows HEK293 cells transiently transfected with vectors expressing HF tag alone and HA-gankyrin. Lane 2 shows HEK293 cells transiently transfected with vectors expressing HF-CDK4 and HA-gankyrin. The position of gankyrin is indicated by an arrow. family (Fig. 5), and binding appears to have an absolute requirement for the extreme C terminus of the S6 protein (Fig. 6, A-C). Although expression of this fragment alone did not promote interaction with gankyrin in the yeast two-hybrid assay system, when the same fragment of the S6 ATPase was expressed using a transcription/translation system it was sufficient to mediate interaction with gankyrin in vitro (Fig. 6C). The reasons for this observed discrepancy are unknown, although it may reflect the influence of the differing N-terminal fusion part of the proteins used in the two assay systems. The proteasomal ATPases (with the exception of the S4 protein) contain coiled coil domains in their N termini; these domains have been implicated in a variety of protein-protein interactions (61). This region of the S6 protein might have been predicted as a prime candidate for the mediation of the gankyrin-S6 interaction. It was therefore initially surprising to observe that this region of the S6 ATPase was dispensable for the observed interaction. However, it has been reported that the N-terminal coiled coil regions of the proteasomal ATPases may be involved in pairwise interactions between the ATPases within the RC of the 26 S proteasome (36,62). This may therefore leave the C-terminal domain of the S6 ATPase free to interact with other cellular targets such as gankyrin.
Similar analysis of the domains within the gankyrin protein required for binding to S6 was inconclusive (Fig. 6). Deletion of any part of the gankyrin protein appears to disrupt interaction with the S6 ATPase (Fig. 6, D and E). The gankyrin protein is composed primarily of multiple copies (ϳ6 repeats) of the ankyrin repeat (63) and as such may resemble the structure of the cyclin D-dependent kinase inhibitor family known as INK4s. The most extensively studied member of this family is p16 INK4A . Analysis of tumor-associated mutations within this protein shows that they are distributed throughout the entire molecule from codons 2 to 154 (64) and are not, for example, clustered within the 20 amino acids implicated in CDK binding (65). This observation might be expected if the overall structure of an ankyrin repeat-containing protein is important and if the protein is otherwise thermodynamically unstable. Point mutations scattered throughout the protein are likely to have the same overall effect of destabilizing the three-dimensional structure. It appears that a minimum number of ankyrin repeats are required to stabilize the basic structure (64), and at least in the case of p16 INK4A , interference within any of these repeats is enough to render the protein functionally inactive. In a similar fashion, we propose that the overall tertiary structure of gankyrin is critical for it's interaction with the S6 ATPase.
The completion of the genome sequencing project for the yeast S. cerevisiae allowed us to pursue a systematic search for orthologous genes. Such a search revealed a putative orthologue for gankyrin, ygr232w. PCR cloning of both the ygr232w and rpt3 (the S. cerevisiae orthologue of S6) ORFs has allowed us to show that the two proteins interact in the same manner by transcription/translation/pull-down analysis as their human orthologues (data not shown). Furthermore, the S. cerevisiae RPT3 protein interacted with the human gankyrin protein in a GST pull-down assay (data not shown). A recent report has also shown that the yeast proteins YGR232W and RPT3 interact with one another (66). Construction of both hetero-and homozygous deletion strains suggests that ygr232w is nonessential for vegetative growth, sporulation, and spore germination (Fig. 7, B and C). Subjecting the ⌬ygr232w strain to varying forms of insult, such as chronic heat shock and growth of cells on canavanine-containing media, also resulted in no obvious phenotype different from that of the parent strain (data not shown). We observed no obvious accumulation of polyubiquitinated proteins in extracts taken from the ⌬ygr232w strain.
The observation that gankyrin could induce the hyperphosphorylation of Rb (27) led us to investigate interactions of other proteins involved in the pINK/cyclinD/CDK4/CDK6/pRb axis with gankyrin. We took advantage of Ni 2ϩ affinity chromatography to isolate HF-gankyrin-containing complexes and discovered CDK4 in these complexes (Fig. 8A). In addition to the observed interaction with endogenous CDK4, we were also able to show the presence of endogenous S6 ATPase in the same preparations (Fig. 8B). Surprisingly, we were only able to observe these interactions when cells were lysed using the standard buffer containing deoxycholate; the use of low detergent lysis buffer minus deoxycholate allowed us to detect no interaction with HA-S6, endogenous S6, or CDK4 (data not shown). This observation suggests that the complexes (or some of their components) under investigation are compartmentalized in some way within the cell or may be associated with lipid membranes. A recent report describes such a population of CDK2-cyclin E complexes associated with the endosomes of liver parenchyma (67). To confirm the interaction, we transiently transfected HEK293 cells with vectors expressing HF-CDK4 and HA-gankyrin and looked for the presence of HA-gankyrin after nickel chelate pull-down. Fig. 8C shows that HA-gankyrin could be detected only in the presence of HF-CDK4. It is formally possible that under normal conditions of cellular disruption using low detergent buffer, gankyrin associates with the RC (and presumably the 26 S proteasome) via the observed interaction with the S6 ATPase within the base complex. However, under more stringent conditions of lysis using standard buffer, gankyrin is able to associate with both overexpressed and endogenous S6 ATPase and CDK4.
The presence of CDK4 within HF-gankyrin complexes prompted us to examine the possibility that gankyrin might also interact with the closely related kinase, CDK6, in addition to other proteins that play a role in the regulation of the G 1 /S phase transition in eukaryotic cells. We were unable to detect the presence of CDK6, p21 CIP1 , p27 KIP1 , cyclins D1 and D3, or p16 INK4A in HF-gankyrin-containing complexes. We were also unable to detect either subunits of the 20 S proteasome or other subunits of the RC.
CDK4 exists in proliferating cells within several complexes of differing molecular weight, ϳ450, 150, and 50 kDa. Active CDK4 is associated with the complex of 150 kDa containing, in addition to CDK4, members of the cyclin D and CIP/KIP families of proteins, whereas catalytically inactive kinase is found associated with INK4 family members in the 50-kDa complex (68). The largest CDK4-containing conglomerate of ϳ450 kDa is a cytoplasmic chaperone complex that also contains Hsp90 and p50(CDC37) but not the D-type cyclins (69,70). This complex catalyzes the correct folding of newly synthesized CDK4. It is conceivable that gankyrin functions to disrupt this chaperone complex, thereby releasing more folded CDK4 for association with D-type cyclins and members of the CIP/KIP family of proteins, leading to hyperphosphorylation of Rb and inappropriate progression through the G 1 /S phase of the cell cycle. Such a function for gankyrin may explain the observed oncogenicity (27). The exact composition of gankyrin-containing complexes is currently under investigation.