Transforming Acidic Coiled-coil-containing Protein 4 Interacts with Centrosomal AKAP350 and the Mitotic Spindle Apparatus*

AKAP350 is a multiply spliced family of 350–450-kDa protein kinase A-anchoring proteins localized to the centrosomes and the Golgi apparatus. Using AKAP350A as bait in a yeast two-hybrid screen of a rabbit parietal cell library, we have identified a novel AKAP350-interacting protein, transforming acidic coiled-coil-containing protein 4 (TACC4). Two-hybrid binary assays demonstrate interaction of both TACC3 and TACC4 with AKAP350A and AKAP350B. Antibodies raised to a TACC4-specific peptide sequence colocalize TACC4 with AKAP350 at the centrosome in interphase Jurkat cells. Mitotic cell staining reveals translocation of TACC4 from the centrosome to the spindle apparatus with the majority of TACC4 at the spindle poles. Truncated TACC4 proteins lacking the AKAP350 minimal binding domain found in the carboxyl coiled-coil region of TACC4 could no longer target to the centrosome. Amino-truncated TACC4 proteins could no longer target to the spindle apparatus. Further, overexpression of TACC4 fusion proteins that retained spindle localization in mitotic cells resulted in an increased proportion of cells present in prometaphase. We propose that AKAP350 is responsible for sequestration of TACC4 to the centrosome in interphase, whereas a separate TACC4 domain results in functional localization of TACC4 to the spindle apparatus in mitotic cells.

Rapid transmission of membrane-initiated signals to appropriate intracellular targets is fostered by signal compartmentalization. Intracellular compartmentalization allows an initial, general second messenger to achieve greater signal specificity. By localizing type II cAMP-dependent protein kinase (protein kinase A) to a particular organelle or cytoskeletal element, protein kinase A-anchoring proteins (AKAPs) 1 effectively pre-position the inactive enzyme at or near its eventual substrate (1). Because there are a number of cAMP-dependent signaling pathways operating simultaneously in the cell, AKAPs facilitate the stimulation of a single unique target at the appropriate time and place (2). Most AKAPs have an amphipathic ␣-helical region that binds the regulatory subunits of the protein kinase A heterotetramer (3). Other domains in AKAPs provide targeting motifs to enable binding of an individual AKAP to its particular intracellular compartment (4) and sequence motifs for the scaffolding of other regulatory proteins. Indeed, many AKAPs also scaffold multiple enzymes that are typically protein kinases, phosphatases, or other second messenger-dependent mediators (5). This complex pattern of protein assembly suggests that protein kinase A-anchoring proteins have the ability to coordinate multiple intracellular signaling events.
Several groups have shown evidence for the functional presence of AKAP scaffolded intracellular signaling complexes. AKAP75, isolated from bovine brain, and its human homologue, AKAP79, localize protein kinase A, protein kinase C, and protein phosphatase 2b to post-synaptic densities (6,7). The specific role for each of these enzymes in this environment remains to be identified, but AKAP79 could functionally organize these multiple effectors in a strategic location for transmission and modulation of post-synaptic signals. Similarly, AKAP250, also known as gravin, localizes both protein kinase A and protein kinase C to the membrane cytoskeleton and filopodia found in human erythroleukemia cells (8). Its submembrane locale places it adjacent to transmembrane receptors, which are likely regulated in part by protein kinase A and protein kinase C. Investigations of the Yotiao scaffolding protein have revealed its essential role in tethering both protein kinase A and the type 1 protein phosphatase to synaptic NMDA receptors (9). Protein phosphatase 1 suppresses and protein kinase A activates NMDA receptor up-regulation, thus defining the role of Yotiao as a signaling system scaffold. Most recently, muscle-selective AKAP has been implicated in bidirectional control of cardiomyocyte protein kinase A activity by anchorage of both protein kinase A and cAMP-specific phosphodiesterase, PDE4D3 (10).
In recent years, our group and others have identified an protein kinase A-anchoring protein, AKAP350, which localizes to the centrosome and the Golgi apparatus (11)(12)(13)(14)(15). AKAP350 is the product of a multiply spliced gene from human chromosome 7q21 that generates numerous isoforms of a large protein scaffold with both centrosomal and non-centrosomal variants. Subsequent to our publication of AKAP350, Ono et al. (16) and Orstavik et al. (12) identified the same open reading frame cDNA as CG-NAP and AKAP450, respectively. In addition to binding type II protein kinase A regulatory subunits, initial studies of AKAP350/AKAP450/CG-NAP show that it interacts with the serine/threonine kinase protein kinase N, protein phosphatase 1, and protein phosphatase 2a (13). Further studies have demonstrated that CG-NAP associates with protein kinase C⑀ in a phosphorylation-dependent manner (16) and that AKAP450 associates with phosphodiesterase 4D3 (17).
Two carboxyl-terminal splice variants from human gastric mucosa (AKAP350A) and human lung (AKAP350B) have been identified (11). AKAP450 and CG-NAP contain carboxyl termini identical to AKAP350A. AKAP350C is a separate readthrough termination that truncates the last 12 exons of AKAP350A and results in a different carboxyl-terminal amino acid sequence (GenBank accession no. AF247727). The NMDA receptor-associated AKAP, Yotiao, is the shortest 3Ј splice variant derived from this locus (11,18). All together, two 5Ј initiation sites, two internal splice sites, and four 3Ј splice variant products of AKAP350/AKAP450/CG-NAP have been reported. For the sake of simplicity, we will refer to this protein family throughout as AKAP350.
Scaffolding proteins could provide a means to organize multiple activities at discrete sites on the centrosome. A recent study of pericentrin, a known integral matrix centrosomal protein (19), showed that it functions as an protein kinase A-anchoring protein, which binds cAMP-dependent protein kinase A regulatory subunits through a novel motif (20). The functional significance of the protein kinase A scaffolding at the pericentrin centrosome is unknown, but it appears to be important in spindle formation. Spindle defects have been observed when protein kinase A anchoring of pericentrin is disrupted (21). Pericentrin interacts with the motor protein dynein (22) and is a substrate for protein kinase A (23). Overexpression of pericentrin results in dynein depletion and a number of spindle defects (22). The apparent scaffolding function of pericentrin may play an important role in the complex function of nucleating microtubule polymers for spindle formation in the centrosome.
In this report, we have utilized a carboxyl-terminal fragment of AKAP350A as bait in a yeast two-hybrid screen of a rabbit parietal cell library. Isolation and subsequent cloning have identified a novel AKAP350-interacting protein that is a member of the transforming acidic coiled-coil-containing (TACC) protein family, TACC4. TACC4 is the first identified AKAP350interacting protein that is not a signal transduction enzyme. Immunolocalization with polyclonal anti-TACC4 antibodies colocalizes TACC4 with AKAP350 at the centrosome in interphase Jurkat cells. Once cells begin mitosis, TACC4 translocates to the spindle apparatus accumulating at the spindle poles, whereas AKAP350 remains at the centrosome. Yeast two-hybrid binary assays and truncated GFP-TACC4 expression studies identify amino acids 247-404 of TACC4 as the region responsible for AKAP350 interaction and corresponding centrosome localization. Further, a separate amino-terminal TACC4 region from amino acid 1 to 380 is responsible for spindle localization. These results indicate that TACC4 is sequestered to the centrosome during interphase by AKAP350 interaction. This interaction is lost in mitosis, and TACC4 translocates to the spindle apparatus, where it may have a functional role in spindle dynamics.

EXPERIMENTAL PROCEDURES
Materials-pEGFP-C2 vector, Advantage Taq, and Marathon cloning kits were purchased from CLONTECH. All DNA sequencing was performed using dye terminator chemistry automated sequencing in the Molecular Biology Core Facility at the Medical College of Georgia. The Molecular Biology Core Facility also synthesized oligonucleotides. A rabbit polyclonal antibody (NE27) was raised against a 16-mer peptide sequence corresponding to amino acids 99 -113 of TACC4 (NH 2 -DEQGTRESPSTPTPRC-COOH) by New England Peptide. Mouse monoclonal antibodies to ␥-tubulin and ␣-tubulin were purchased from Sigma. Mouse GAL4 activation domain and binding domain monoclonal antibodies were purchased from CLONTECH. Mouse monoclonal anti-AKAP350 (14G2) was produced as previously described (11). For production of AKAP350A-specific antibodies, a peptide corresponding to the unique region in the carboxyl terminus of AKAP350A was conjugated to keyhole limpet hemocyanin using an Imject Immunogen EDC kit from Pierce. The AKAP350A peptide (GSTTQFHAGMRR) was synthesized in the Molecular Biology Core Facility. Rabbit polyclonal antibodies were raised in the Antibody Facility at the University of Georgia. The antibody was affinity-purified using Amino-Link system from Pierce with the AKAP350A peptide. 100 mM glycine, pH 2.5, was used to elute the AKAP350A antibody from the Amino-Link column. Species-specific Cy2-, Cy3-, and Cy5-conjugated secondary antibodies were purchased from Jackson Immunoresearch Laboratories (West Grove, PA). Prolong Antifade, 4Ј,6-diamidino-2-phenylindole, and species-specific Alexa 488-conjugated secondary antibodies were purchased from Molecular Probes, Inc. (Eugene, OR). Fail-Safe amplification kits were purchased from Epicentre. [␣-32 P]dCTP was purchased from PerkinElmer Life Sciences. Protease inhibitor mixture and glass beads were purchased from Sigma.
Yeast Two-hybrid Screening-A rabbit parietal cell pADGal4 twohybrid library was screened with the final 2.7 kb of human AKAP350A cDNA in the binding domain vector pBDGal4-Cam (Stratagene). The Y190 yeast strain harboring the HIS3 and ␤-galactosidase reporter genes was used for screening of approximately one million clones as previously described (24). A single positive clone was identified and rescued into XL1-Blue bacteria with selection using ampicillin. Plasmid DNA was prepared using Qiagen miniprep kits, and the clone was sequenced using flanking vector primers and specific derived oligonucleotide sequences by the Medical College of Georgia Molecular Biology Core Sequencing Facility.
5Ј-RACE-The 5Ј end of the isolated clone was highly GC-rich, and resolution of the remaining 5Ј sequence required use of both high temperature cDNA production with rTth polymerase and optimized nested amplification from rabbit spleen cDNA using the Fail-Safe amplification buffer system (Epicentre). Rabbit spleen total RNA was prepared as previously described (25), and cDNA was prepared using rTth polymerase (PerkinElmer Life Sciences) in the presence of manganese ions with oligo(dT) priming. The resulting cDNA was then modified by addition of linkers to prepare a linkered cDNA (Marathon, CLONTECH). Amplification from a rabbit spleen cDNA required Fail-Safe Buffer E (Epicentre) in two rounds of nested amplification. The first round was performed with Adapter Primer 1 (CLONTECH) and CTCCAGGATACACAGGGGTT (5 cycles of 94°C for 10 s, 72°C for 3 min; 5 cycles of 94°C for 10 s, 70°C for 3min; and 30 cycles of 94°C for 10 s, 68°C for 3 min). This was followed by reamplification with Adapter Primer 2 and CCCGAACTGCTCCAGGTAATCGATCTC (35 cycles of 94°C for 10 s, 60°C for 10 s, 68°C for 2 min). The resulting products ranging in size from 500 to 1000 nucleotides were gel-isolated and cloned into pTOPO-T (Invitrogen). Isolated plasmid minipreps were sequenced as above.
Northern Blot Analysis-Northern blots containing 20 g of total RNA from rabbit tissues were hybridized with a random-primed TACC4 cDNA probe (nucleotides 372-765). The probe had been amplified from plasmid using polymerase chain reaction with Advantage Taq polymerase (CLONTECH) (T400 sense, GGCCGAGACCCCAGGACCAG-GAGAC; T400 antisense, GGGGGCCCGACGCGCTCACAGG; 35 cycles of 95°C for 30 s, 60°C for 30 s, and 68°C for 60 s). Blots were washed to high stringency (0.1ϫ SSC, 65°C) and exposed to x-ray film (Eastman Kodak Co.).
Yeast Two-hybrid Binary Assays-Carboxyl-terminal cDNAs for AKAP350B and AKAP350C of similar size to pBD-AKAP350A were cloned into pBD-Gal4. Truncations of AKAP350 including regions common to AKAP350A and AKAP350B were subcloned into pBD-Gal4. Full-length TACC4 cDNA was subcloned into pAD-Gal4. Truncated TACC4 cDNAs were amplified from full-length TACC4 cDNA and subcloned into pAD-Gal4. Assays were performed by dual transfection of the Y190 yeast strain with a binding domain pBD-Gal4 construct and an activation domain pAD-Gal4 construct followed by detection of ␤-galactosidase production as described previously (26). Positive interactions were defined by identification of blue colonies within 2 h of administration of 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal).
Yeast Western Blot Analysis-Yeast two-hybrid binary assays that had negative ␤-galactosidase production were checked for expression of pAD and pBD by Western blot analysis. Cotransformed yeast colonies were inoculated into 50 ml of SD/ϪTrp/ϪLeu and grown overnight at 30°C. 10 ml of the overnight culture were inoculated into 50 ml of YPD and grown to an A 600 ϭ 0.5. Cultures were pelleted at 3000 ϫ g for 10 min, resuspended in 50 ml of sterile water, and pelleted again. The pellets were quickly frozen in liquid nitrogen and kept at Ϫ70°C. Pellets were thawed and put in lysis buffer (50 mM EDTA, 1 M NaCl, 1 mM EGTA, 0.1% Triton X-100, 1 l/ml protease inhibitor mixture). Glass beads were added to the lysis solution, and the yeast were vigorously vortexed for 10 min. Lysis solution was pulled off the top of the beads and spun down at 4°C for 10 min at 20,000 ϫ g. Supernatant was discarded. Pellets were dissolved in 1ϫ SDS, 10% ␤-mercaptoethanol, boiled for 15 min, and run out on a 10% SDS-PAGE gel. Gels were transferred to Immobilon for subsequent Western blotting by anti-pAD-Gal4 and anti-pBD-Gal4. Blots were blocked with 5% nonfat dry milk in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl for 1 h at 25°C. Blots were then probed in 0.5% nonfat dry milk in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl for 2 h at 25°C with a monoclonal antibody against pAD-Gal4 (0.1 g/ml) or pBD-Gal4 (0.1 g/ml). After the primary incubation, the blots were washed three times for 15 min each with 25 mM Tris-HCl, pH 7.5, 150 mM NaCl and then incubated with horseradish peroxidase-conjugated anti-mouse IgG (1:2500) for 1 h at 25°C. The blots were washed three times for 5 min each with 25 mM Tris-HCl, pH 7.5, 150 mM NaCl followed by a 1-min incubation with chemiluminescence substrate (Pierce, Supersignal) and autoradiography.
Immunocytochemistry of NE27-Jurkat cells were pelleted onto glass slides using a Cyto-Spin (Clay-Adams). Cells were fixed in 4% paraformaldehyde for 10 min at 25°C. Fixed cells were washed in PBS, then blocked and permeabilized with 1% milk, 0.3% Triton X-100 in PBS. Immunostaining was subsequently performed with various combinations of anti-TACC4 (1:250), anti-AKAP350 14G2 (1:80), anti-␥tubulin (1:500), and anti-␣-tubulin (1:50) for 2 h at 25°C. The cells were then incubated with Cy3-conjugated anti-mouse IgG and Alexa 488conjugated anti-rabbit IgG for 60 min. Following three 5-min washes in PBS, the cells were incubated in DAPI (1 mM) for 5 min and given a final 5-min wash in 50 mM sodium phosphate, pH 7.4. Slides were mounted with Prolong Antifade (Molecular Probes, Eugene, OR). Cells were examined in the Imaging Core facility of the Institute of Molecular Medicine and Genetics Institute at the Medical College of Georgia on a Zeiss Axiophot microscope equipped with a SPOT digitizing camera or an Amersham Biosciences confocal microscope as described previously (27). Anti-TACC4 (NE27) specificity was confirmed by competitive inhibition of immunostaining with 500 M TACC4 epitope peptide.
Cloning of 3Ј Terminal Sequences from AKAP350 Splice Variants-The 3Ј splice variant sequences, AKAP350A (originally reported as HGAKAP350) and AKAP350B (originally reported as HLAKAP350), were cloned previously (11). Using the Marathon RACE system, a linkered human lung cDNA (CLONTECH) was utilized to perform 3Ј-RACE using a sense primer ATCAATACAATCTCATCTCTAAAG corresponding to a start at nucleotide 8197 of AKAP350. The resulting RACE products were cloned into pBluescript-T, as previously described (25), and selected clones were sequenced. We identified several sequences corresponding to the 3Ј regions of both AKAP350A and AKAP350B. In addition, we identified a novel sequence for a third splice variant, AKAP350C, which is accounted for by a read-through termination of the final 12 exons present in AKAP350A with a short region of unique 3Ј coding sequence (GenBank accession no. AF247727).
Preparation of Recombinant Protein-TACC4 cDNA was cloned into the pGEX5-1 vector (Amersham Biosciences) and transformed into JM109, resulting in a vector coding for GST fused to the amino termi-nus of full-length TACC4. A 1-liter culture of pGEX-TACC4 or empty pGEX5-1 vector was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside in log phase and allowed to grow for 4 h at 37°C. Bacteria were collected with a 5000 ϫ g spin for 15 min, and the pellet was resuspended in 25 ml of lysis buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, 100 mM NaCl, 0.1 mg/ml lysozyme, 5 mM benzamidine, and 0.1 mM AEBSF). The lysate was sonicated on ice with four 10-s bursts. Triton X-100 was then added to a final concentration of 1%, and the mixture was nutated at 4°C for 30 min. The lysate was then centrifuged at 20,000 ϫ g, and the supernatant was filtered through a 0.45-m filter. This filtrate was then added to 2 ml of glutathione-Sepharose beads (Amersham Biosciences) prepared by washing two times in 10 volumes of PBS, one time in PBS plus 1% Triton X-100, and three times in lysis buffer without Triton X-100. The beads and supernatant mixture were mixed on a Nutator (Clay-Adams) at 25°C for 30 min. Nonadherent protein was removed with a 5-min 500 ϫ g centrifugation, and the beads were washed three times with 10 bead volumes of PBS followed by one wash in PBS with 250 mM NaCl. Beads were then transferred to a 1.5-ml centrifuge tube, and protein was eluted with 1 bead volume of elution buffer (10 mM reduced glutathione, 50 mM Tris-HCl, pH 8) for 10 min at 25°C.
In Vitro Binding Assay-One hundred l of GST-TACC4-(1-454)bound glutathione beads (prepared as above without the elution step), GST-bound glutathione beads, or unbound glutathione beads were blocked in PBS, 2 mg/ml bovine serum albumin at 4°C for 2 h. 1 ml of rabbit stomach mucosa 100,000 ϫ g supernatant was then added to each aliquot of beads, and the mixtures were incubated on a Nutator at 4°C for 4 h. Nonadherent proteins were collected by centrifugation at 100 ϫ g, placed in 1% SDS stop solution, and heated at 95°C for 5 min with final addition of 0.1 final volume of ␤-mercaptoethanol. Beads were washed three times with 1 ml of PBS and then incubated in 1.5% SDS stop solution at 95°C for 15 min. The beads were pelleted at 500 ϫ g for 5 min, and 0.1 volume of 2-mercaptoethanol was added to the supernatants. Samples were resolved by SDS-PAGE (3-10% gradient gels) and transferred for 2 h at 750 mA to nitrocellulose (Sarsedt) for subsequent Western blotting with anti-AKAP350 (14G2 1:500).
GFP-TACC4 Cell Quantitation-The cell cycle stage of ϳ100 cells for three separate transfections of GFP-TACC4, GFP-TACC4-(1-380), GFP-TACC4-(247-404), and GFP was quantitated. The number of prometaphase and interphase cells among transfected cells was compared by analysis of variance with post hoc comparison of significant means by Tukey's test.

RESULTS
Yeast Two-hybrid Screen and Cloning of TACC4 -Utilizing the last 2.7 kb of human AKAP350A coding sequence as bait, yeast two-hybrid screening of a rabbit parietal cell library yielded a single partial clone of 1259 base pairs including a 3Ј poly(A) tail and an incomplete 5Ј end. A BLAST search of public data bases with this novel sequence revealed significant identity with members of the TACC protein family including murine TACC3, human TACC3, human TACC2, and human TACC1. As a result, we termed the AKAP350-interacting clone TACC4. Completion of cloning of the 5Ј end of TACC4 identified a 1532-nucleotide cDNA sequence containing an upstream inframe termination codon along with a polyadenylation signal downstream of the termination codon ( Fig. 1; GenBank acces-sion no. AF372837). The 1364-nucleotide open reading frame coded for a 454-amino acid protein with a predicted 49.2-kDa molecular mass and an acidic isoelectric point of 4.6. Structural predictions indicated proline rich regions in the first 255 amino acids and a coiled-coil motif encompassing the carboxyl-terminal 200 amino acids, both of which are characteristics of other TACC family members.
A multiple sequence alignment of TACC4, human TACC3, and murine TACC3 is shown in Fig. 3. The major homology is found within the carboxyl-terminal portion of the TACC4 sequence. As previously described, human TACC3 apparently has a single exon insertion of 996 nucleotides that is not found in murine TACC3 (28). The corresponding region appears to be missing, for the most part, from TACC4 as well. In addition, the coding sequence of TACC4 does not contain the amino-terminal 103 amino acids shared between murine and human TACC3 proteins (Fig. 2, solid box).
Northern Blot Analysis of TACC4 -A Northern blot of rabbit tissue total RNA was probed with a cDNA sequence specific for TACC4 (nucleotides 372-765). Fig. 4 demonstrates a single 1.6-kb RNA species. This message size correlates well with the 1532-nucleotide cloned TACC4 sequence. The message was enriched in spleen, jejunum, and duodenum. It was also detectable in distal colon, ileum, and pancreas.
Yeast Two-hybrid Mapping of the AKAP350 Binding Domain of TACC4 -Binary assays were performed with several pAD-TACC4 truncations against the original pBD-AKAP350A construct utilized in the yeast two-hybrid screen (Fig. 5). Expression of appropriately sized pAD fusion protein was confirmed by Western blot in all negative ␤-galactosidase production assays. Deletion of the carboxyl-terminal 50 amino acids of TACC4 had no effect on AKAP350A interaction. However, further deletion of 24 amino acids abolished this interaction. Although truncation of the 247 amino-terminal residues did not alter AKAP350A interaction, deletion of amino acids 247-284 abolished interaction. Further truncations of pAD-TACC4 were constructed to identify the minimal AKAP350 binding domain of TACC4. pAD-TACC4-(247-380) was not able to interact with AKAP350. However, amino acids 247-404 maintained interaction with AKAP350, identifying the minimal AKAP350 binding domain of TACC4. The coiled-coil domain of TACC4 extends from amino acid 255 to 454. Thus, the binary assay data suggest that two regions of TACC4 at the beginning and end of the coiled-coil domain are important for interaction with AKAP350 (Fig. 5, solid boxes).
Given the importance of the conserved coiled-coil TACC domain to AKAP350A/TACC4 interaction, we evaluated whether other TACC family members with similar coiled-coil TACC domains could interact with AKAP350A. Truncated aminoterminal forms of each human TACC gene were cloned into the pAD-Gal4 activation domain and assayed for interaction with pBD-AKAP350A (Fig. 5). Two amino-terminal truncated fragments of human TACC3 both interacted with AKAP350A, whereas fragments of TACC1 and TACC2 did not interact. These results are apparently because of the high homology between the coiled-coil domains of human TACC3 and TACC4.
Yeast Two-hybrid Binary Assays with AKAP350B, AKAP350C, and Truncated AKAP350A Map the TACC4 Binding Domain of AKAP350 -AKAP350A and AKAP350B differ only in their terminal 23 amino acids. However, AKAP350C truncates a 700-amino acid portion of the AKAP350A carboxyl terminus, eliminating one of the carboxyl leucine zippers and replacing it with a novel 47-amino acid carboxyl-terminal sequence. Each of the splice variant ends were amplified by polymerase chain reaction and cloned into the pBD-Gal4 binding domain vector. Binary assays were performed with pAD-TACC4 and each of the pBD-AKAP350 carboxyl-terminal splice variants (Fig. 6). Expression of appropriately sized pBD protein was confirmed by Western blot in all negative ␤-galactosidase production assays. AKAP350A and AKAP350B, but not AKAP350C, demonstrated an interaction with TACC4. Thus, the carboxyl-terminal 700 amino acid residues truncated in AKAP350C are important for TACC4 interaction, whereas the 23 amino acids differing between AKAP350A and AKAP350B resulted in no change in TACC4 interaction.
To identify the specific region of AKAP350 necessary for TACC4 interaction, amino and carboxyl pBD-AKAP350 truncations were constructed and assayed against pAD-TACC4. A carboxyl truncation from amino acid 3265 to 3520 (the last amino acid common to AKAP350A and AKAP350B) main- tained interaction with TACC4 indicating that the binding domain of AKAP350 is in the region common to AKAP350A and AKAP350B. Further carboxyl truncation of 32 amino acids (amino acids 3265-3488) eliminated interaction with TACC4, indicating that the region of AKAP350 important for TACC4 binding is within the 32 amino acids from 3488 to 3520 (Fig. 6,  solid box). Finally, a minimum TACC4 binding region of AKAP350A (amino acids 3376 -3531) maintained interaction with pAD-TACC4.
In Vitro Association of AKAP350A with TACC4 -In vitro binding assays were performed to confirm the observed yeast two-hybrid interaction of AKAP350 and TACC4. Recombinant GST-TACC4-(1-454) fusion protein was attached to glutathione-Sepharose beads. A 100,000 ϫ g gastric mucosa supernatant was incubated with glutathione beads alone, GST-glutathione beads, or GST-TACC4-glutathione beads. Samples were   FIG. 2. TACC4 is a novel member of the TACC protein family. A BLAST search of available public data bases revealed TACC4 has significant identity to other TACC family members. An amino acid identity calculation of the highly conserved coiled-coil TACC domain (CC, hatched box) and the variable (V) region indicated rabbit TACC4 is most similar to murine TACC3 and human TACC3. Murine TACC3 and human TACC3 share an amino-terminal region of 103 amino acids (solid) with 61% identity that is not present in rabbit TACC4 or other TACC proteins.
FIG. 3. Amino acid sequence alignment of rabbit TACC4, human TACC3, and murine TACC3. A sequence alignment of the TACC family members with highest homology to TACC4, murine TACC3, and human TACC3, was performed. Identical residues are shaded. Rabbit TACC4, murine TACC3, and human TACC3 share highest identity within the carboxyl 200 amino acids in the TACC coiled-coil domain. Human TACC3 and murine TACC3 also share a region of high identity in the first 103 amino acids that is not present in rabbit TACC4. eluted with 1% SDS, and proteins were separated on SDS-PAGE followed by Western blotting with 14G2, an anti-AKAP350 monoclonal antibody (Fig. 7A). AKAP350 immunoreactivity was specifically recovered only from GST-TACC4 beads.
Western Blot Analysis of TACC4 -Anti-TACC4 polyclonal antibody (NE27) was raised against a peptide sequence specific for TACC4 that is absent in human and murine TACC3 (Fig.  8A). Using NE27, Western blot analysis of 3 M urea-treated Jurkat cell lysate detected a band of immunoreactivity ( Fig.  8B) with an approximate molecular mass of 52 kDa as well as a band at ϳ235 kDa. Notably, in samples were prepared without urea, the TACC4 immunoreactive band only migrated at 235 kDa. Immunoreactivity in both cases was not observed when antibody was preincubated with 2.5 M antigen peptide (data not shown). These results suggest that TACC4 may form SDS-insoluble aggregates.
GFP-TACC4 Localizes to the Centrosome and Spindle Apparatus-To analyze further TACC4 targeting, full-length TACC4 was cloned into pEGFP-C2 to express a GFP-TACC4 chimera protein containing GFP on the amino terminus (Fig.  10). The expression level of GFP-TACC4 in transiently trans-  (1-380)) revealed a region of TACC4 from amino acid 380 to 404 (solid box), which is necessary for interaction with AKAP350A. Amino truncations identified that a second region of TACC4 from amino acid 247 to 284 (solid box) is also necessary for interaction with AKAP350A. Using this data, a minimal AKAP350 binding domain of TACC4 from amino acid 247 to 404 (TACC4-(247-404)) was constructed. pAD-Gal4 constructs of the highly homologous coiled-coil TACC regions of human TACC3 (hTACC3), TACC2 (hTACC2), and TACC1 (hTACC1) were constructed and tested for interaction with AKAP350A. Interestingly, only human TACC3 was able to interact with AKAP350A.
FIG. 6. Y2H mapping of the TACC4 binding domain of AKAP350. AKAP350A and AKAP350B differ only in their terminal 23 amino acids (compare hatched box in AKAP350B with checkerboard box in AKAP350A). However, AKAP350C truncates a 700-amino acid portion of the AKAP350A carboxyl terminus, eliminating one carboxyl leucine zipper and replacing it with a novel 47-amino acid carboxylterminal sequence (vertically striped box). The carboxyl end of AKAP350B was amplified from the same start site as the original AKAP350A bait fragment and cloned into pBD-Gal4 binding domain vector as was a carboxyl portion of AKAP350C. Yeast two-hybrid binary assays were performed with full-length pAD-TACC4 and each of the pBD-AKAP350 carboxyl-terminal splice variants. AKAP350A and AKAP350B but not AKAP350C demonstrated interaction with TACC4. Truncation of AKAP350 to amino acid 3488 revealed a single region necessary for TACC4 interaction from amino acid 3488 to 3520 (solid region). Using this data, a TACC4 minimal binding domain of AKAP350 was constructed from amino acid 3376 to 3531 (AKAP350A-(3376 -3531)).
fected Jurkat cells was variable and resulted in different localization patterns. When expressed at low levels, GFP-TACC4 (Fig. 10A) targeted predominantly to the centrosome and caused a slight decrease in endogenous AKAP350 immunostaining. The AKAP350 staining appeared to be dispersed to regions adjacent to the centrosome that also contained GFP-TACC4. High expression of GFP-TACC4 also targeted to the centrosome (Fig. 10B, long arrow) and resulted in an increased number of cells present in prometaphase as determined by DAPI staining (Fig. 10, column 4). In the transfected prometaphase cells, GFP-TACC4 targeted to the spindle apparatus (Fig. 10B, arrowhead) colocalizing with ␣-tubulin (B). Notably, high expression of GFP-TACC4 also resulted in cytosolic accumulations of GFP-TACC4 that were morphologically similar to the pseudo-crystalline aggregates observed by Raff and colleagues (34) in overexpressed TACC coiled-coil domains (34). Their electron microscopic analysis indicated that the aggregates result from a latticework of the coiled-coil TACC domains present in the TACC family members. The aggregations present are probably a result of the highly homologous TACC domain present in TACC4.
In GFP-TACC4-transfected cells, the majority of AKAP350A, localized by anti-AKAP350A staining, remained on the centrosomes with only slight dispersal onto spindle fibers (Fig. 10B,  short arrows). Thus, as with endogenous TACC4 staining (Fig.  9), GFP-TACC4 disassociates from the centrosome during division and relocates to the spindle apparatus, whereas AKAP350A predominantly retains localization to the centrosome. The endogenous TACC4/AKAP350 localization as well as the GFP-TACC4 targeting led us to believe that the AKAP350/ TACC4 association was dependent on the cell cycle. To assess this idea, we constructed two more GFP fusion proteins guided by the yeast two-hybrid binary interaction data.
GFP-TACC4- (1-380), a Non-AKAP350-interacting Protein, Localizes to the Spindle but Is Unable to Target to the Centrosome-We subcloned a carboxyl truncated TACC4-(1-380) into pEGFP-C2 and transiently transfected GFP-TACC4-(1-380) (Fig. 11) into Jurkat cells. TACC4-(1-380) did not interact with AKAP350A in yeast two-hybrid binary assays (Fig. 5). GFP-TACC4-(1-380) was dispersed throughout the cytosol and did not target to any specific organelle in interphase cells (Fig. 11,  A and B). Endogenous AKAP350A localization to the centrosome, as shown by colocalization of ␥-tubulin and anti-AKAP350A, was unaffected by GFP-TACC4-(1-380) expression (Fig. 11A). Importantly, there was no discernible targeting of GFP-TACC4-(1-380) to the centrosome, demonstrating that the region between amino acids 380 and 454 was necessary for TACC4 localization to the centrosome in interphase cells. This same region of TACC4 was required for interaction with AKAP350A in yeast two-hybrid binary assays. This correlation suggested that TACC4 interaction with AKAP350A might be responsible for TACC4 sequestration to the centrosome during interphase.
The net result of GFP-TACC4-(1-380) targeting indicated that this TACC4 truncation differs from endogenous and fulllength GFP-TACC4 by its inability to localize to the interphase centrosome. However, GFP-(1-380) retained the same ability of endogenous TACC4 and GFP-TACC4 to target to the mitotic spindle apparatus (Fig. 11, C and D, arrowheads). AKAP350 staining, on the other hand, remained punctate on the remnant centrosomes (Fig. 11C, long arrows). Notably, GFP-TACC4-(1-380) formed aggregates (Fig. 11C, short arrows) that were morphologically similar to those formed by GFP-TACC4. This expression pattern suggested that the region of TACC4 responsible for interaction with the spindle apparatus was contained within amino acids 1-380 and did not require interaction with AKAP350. Thus, there are distinct regions of TACC4 responsible for centrosome localization and spindle localization.
The AKAP350 Minimal Binding Domain of TACC4, GFP-TACC4-(247-404), Targets the Centrosome but Not the Spindle Apparatus-The second construct linked the AKAP350 minimal binding domain of TACC4 (amino acids 247-404) to GFP forming GFP-TACC4-(247-404). Similar to GFP-TACC4, variable expression of GFP-TACC4-(247-404) resulted in two localization patterns (Fig. 12). Low expression of GFP-TACC4-(247-404) targeted to the centrosome (Fig. 12A). Anti-TACC4 staining of the same cells revealed slight dispersal of endogenous TACC4 from the centrosome with no observed effect in endogenous AKAP350 staining. In yeast two-hybrid binary assays, this region of TACC4 interacted with AKAP350A. These results provide further evidence that interaction with AKAP350 correlates with TACC4 localization to the centrosome.
Overexpression of GFP-TACC4 and GFP-TACC4-(1-380) Results in Prometaphase Arrest-When immunostaining of the chimeric GFP-TACC4 proteins was performed, we observed a general increase in the number of transfected mitotic cells. To quantitate the observed increase, we assessed the cell cycle stage in DAPI staining of ϳ100 transfected cells from three separate transfections. Cell cycle status was evaluated for cells transfected with GFP, GFP-TACC4, GFP-TACC4-(1-380), and GFP-TACC4-(247-404). The most common mitotic morphology observed, condensed chromatin and a dissolved nuclear envelope, corresponded to prometaphase. Therefore, we statistically compared the number of cells present in interphase versus prometaphase (Fig. 13). The number of transfected cells present in prometaphase was significantly increased for GFP- TACC4 and GFP-TACC4-(1-380) compared with GFP and GFP-TACC4-(247-404). These results suggested a link between localization of GFP-TACC4 proteins to the spindle apparatus and prometaphase arrest. Thus, increased number of prometaphase cells required overexpression of GFP-TACC4 proteins that contained the region of TACC4 implicated in spindle apparatus targeting. DISCUSSION The AKAP350 family of proteins interacts with numerous signaling proteins from multiple signal transduction pathways. The shortest 3Ј splice variant, Yotiao protein, associates with the neuronal NMDA receptor and contains a single binding site for the R II subunit of type II cAMP-dependent protein kinase (18,35). The other AKAP350 splice variants contain two binding sites for the R II subunit of type II cAMP-dependent protein kinase (11)(12)(13). In addition, these proteins also contain binding regions for phosphatase 1 and protein kinase N in their Yotiao homology regions, as well as binding regions for protein phosphatase 2a and protein kinase C⑀ in their carboxyl-terminal region (13,16). Most recently, Tasken et al. (17) have shown that a central region of AKAP350 interacts with cAMP-dependent phosphodiesterase 4D3 (PDE4D3). Notably, all of the published AKAP350 interaction partners are enzymatic proteins that regulate signaling pathways by phosphorylation state changes or changes in second messenger levels. Utilizing yeast two-hybrid screening with AKAP350A as bait, we have now identified the first evidence for two AKAP350 interaction partners, TACC3 and TACC4, which have putative non-enzymatic effector functions.
The original rabbit clone identified from the AKAP350A screen contained high homology to the TACC gene family. Thus, we named the putative interacting clone TACC4. Subsequent cloning of the full-length gene for TACC4 identified that TACC4 has significant homology with human and murine TACC3. A polyclonal antibody, NE27, was raised to a 15-mer peptide sequence, which is present only in TACC4 and not in human TACC3. Anti-TACC4 stained the centrosome throughout interphase and the spindle apparatus during mitosis in Jurkat cells. This staining pattern is significantly different from previously published anti-TACC3 localization in which TACC3 only stained a diffuse area around the centrosome during mitosis in HeLa cells (34). Anti-TACC4 also did not recognize a human GFP-TACC3 chimera by immunocytochemistry (data not shown). This distinct staining pattern suggests that TACC4 is a novel gene product. Nevertheless, we cannot rule out the possibility that TACC4 is a splice variant of TACC3 with an alternative 5Ј-start site. Current public data base information does not permit us to identify definitively all of the genomic sequence flanking the human TACC3 gene.
We have confirmed the yeast two-hybrid TACC4 interaction with AKAP350 by both biochemical and immunocytochemical approaches. First, using a biochemical GST-TACC4 pull-down assay, we demonstrated that only GST-TACC4 beads were able to pull down an anti-AKAP350-reactive band. Second, anti-AKAP350 and anti-TACC4 staining colocalized to the centrosome in interphase Jurkat cells. These results support the yeast two-hybrid interaction findings that TACC4 interacts with AKAP350. Further, yeast two-hybrid binary assays revealed that the only mammalian TACC family members that can interact with AKAP350 are TACC3 and TACC4. Binary assays also demonstrated that TACC4 interacts with AKAP350A and AKAP350B, but not AKAP350C. Truncations of AKAP350 localize the TACC4 binding domain of AKAP350 to 31 amino acids from 3488 to 3520 present in the carboxyl ends of both AKAP350A and AKAP350B.
Our data suggest that AKAP350 interaction with TACC4 is necessary for localization of TACC4 to the centrosome in interphase Jurkat cells. Immunocytochemistry with anti-AKAP350 and anti-TACC4 antibodies demonstrated a cell cycle-depend-  1-380). Notably, GFP-TACC4-(1-380) expression did not appear to affect endogenous AKAP350A localization to the centrosome. High expression levels of cytoplasmic GFP-TACC4-(1-380) appeared to perturb microtubules (anti-␣-tubulin), whereas there was no effect on anti-AKAP350A centrosomal staining (B). GFP-TACC4-(1-380) expression in mitotic cells (C) targeted to the spindle apparatus (arrowheads) and formed aggregates (short arrows) in a manner similar to GFP-TACC4. Anti-TACC4 staining was also seen on the spindle apparatus and the aggregates, whereas anti-AKAP350 staining remained (long arrows) at the centrosome structures. Whether or not GFP-TACC4-(1-380) pulled in endogenous TACC4 could not be addressed because GFP-TACC4-(1-380) contains the antigen peptide epitope used to raise anti-TACC4. GFP-TACC4-(1-380) localization to the spindle apparatus was confirmed by colocalization with ␣-tubulin (D). ent variation in TACC4 colocalization with AKAP350. Throughout the cell cycle, AKAP350 remains on the centrosome with only a slight dispersal onto spindle microtubule fibers during mitosis. In interphase cells, anti-AKAP350 staining is located on the centrosome. Anti-TACC4 staining during interphase is also present at the centrosome, suggesting that AKAP350 may sequester TACC4 to the centrosome in nonmitotic cells. Once cells enter mitosis, TACC4 loses its association with the centrosome and translocates to the spindle apparatus where anti-TACC4 staining decorated spindle strands and accumulated at the spindle poles. Thus, endogenous staining patterns suggest that TACC4 only interacts with AKAP350 in interphase.
There are two distinct targeting regions of TACC4. One region is responsible for AKAP350 binding and sequestration to the centrosome during interphase. The other region is responsible for TACC4 localization to the spindle apparatus during mitosis. These two distinct TACC4 regions are deduced from our yeast two-hybrid binaries and GFP-TACC4 expression studies. The TACC4 region responsible for interaction with AKAP350 requires a domain of TACC4 from amino acid 247 to 404. A requirement of the subregion from amino acid 380 to 404 is shown by the inability of pAD-TACC4-(1-380) to interact with pBD-AKAP350A and the loss of AKAP350 colocalization and centrosome targeting by GFP-TACC4-(1-380). Amino-terminal truncation of TACC4 defined a second subregion at the beginning of the coiled-coil TACC domain between amino acids 247 and 284 that is necessary for AKAP350A interaction. A TACC4 construct from amino acid 247 to 404 maintained interaction with AKAP350A. Expression of the same region, GFP-TACC4-(247-404), targeted GFP to the centrosome but did not localize to the spindle apparatus during mitosis. Thus, amino acids 247-404 of TACC4 are responsible for centrosome localization of TACC4 and contain the AKAP350 minimal binding domain. However, this domain is not the region responsible for TACC4 spindle localization. Thus, localization of TACC4 to the centrosome, but not the spindle apparatus, requires interaction with AKAP350A.
TACC4 localization to the spindle apparatus during mitosis requires a distinct region in the amino terminus of the protein. GFP-TACC4 overexpression recapitulated the observed anti-TACC4 staining by localizing to the spindle apparatus in mitotic cells. However, although carboxyl truncation of amino acids 380 -454 of TACC4 in GFP-TACC4-(1-380) resulted in a loss of interphase centrosomal localization, GFP-TACC4-(1-380) still targeted to the spindle apparatus in mitotic cells. These results, along with the observation that GFP-TACC4-(247-404) did not localize to the spindle apparatus, indicate that the region of TACC4 necessary for TACC4 spindle localization is contained in the amino end of the protein between amino acids 1 and 247. Notably, overexpression of GFP-TACC4 constructs with the ability to localize to the spindle apparatus also resulted in an increase in the number of cells present in prometaphase. By contrast, overexpression of GFP-TACC4-(247-404) did not result in increased numbers of prometaphase cells. Thus, the AKAP350 minimal binding region of TACC4 retains neither the region necessary for spindle localization nor the ability to elicit prometaphase arrest.
The explanation for why overexpression of the spindle interaction region of TACC4 causes an increase in the number of prometaphase cells is not immediately apparent. The most likely possibility is that TACC4 overexpression perturbs spindle structure, resulting in an activation of the spindle checkpoint and arrest of the dividing cell at prometaphase (reviewed in Refs. 36 -38). Alternatively, TACC4 may be a protein member of the spindle checkpoint machinery. Overexpression of TACC4 would then result in perturbation of the checkpoint machinery dynamics, causing an increase in prometaphase cells. In either case, the observed prometaphase arrest suggests that interaction of TACC4 with the spindle apparatus is of purposeful importance to normal spindle function.
The dynamics of interaction of TACC4 with the spindle apparatus remain to be elucidated, but recent studies of the Drosophila TACC homologue, D-TACC, suggest it may interact with a conserved microtubule-associated protein family important in microtubule dynamics (39,40). D-TACC was initially identified by copurification with microtubules in a microtubule spin-down experiment of Drosophila embryo extracts (32). D-TACC contains a carboxyl coiled-coil TACC domain homologous to mammalian TACC family members. Overexpression of this TACC domain formed aggregates in HeLa cells, suggesting a conserved function of the TACC domain in TACC proteins (34). Dtacc mutants displayed abnormally short spindle microtubules and eventually died as a result of chromosomal segregation defects.
Recent work by the same group (39) and Cullen and colleagues (40) have implicated interaction of D-TACC with Msps, a Drosophila member of the mammalian XMAP215/ch-TOG microtubule-associated protein family. Vertebrate members of this microtubule-associated protein family stabilize microtubules and antagonize the kinesin XKCM1 in controlling plusend microtubule dynamics in vitro (41). Defective tripolar meiotic acentrosomal spindles were formed in both dtacc and msps mutants (40). D-TACC and Ncd, a minus-end-directed microtubule motor, were involved in efficient localization of Msps to acentrosomal meiotic spindle poles. Similarly, Msps colocalized with D-TACC at mitotic spindle poles in Drosophila embryos and localized to overexpressed D-TACC TACC domain aggregates (39).
Notably, the interaction of these two Drosophila proteins appears to be conserved in human homologues TACC3 and ch-TOG; overexpressed TACC3 aggregates concentrated ch-TOG (39). The ch-TOG gene product, TOGp, localizes to perinuclear cytoplasm in interphase and to the spindle apparatus with concentration at the spindle poles during mitosis (42). It seems plausible that TOGp or a protein complex that contains TOGp may be responsible for TACC4 localization to the spindle apparatus in a similar manner to that seen in Drosophila with D-TACC and Msps.
A remaining question of whether unknown or known AKAP350 interaction partners play a functional role in TACC4 binding to AKAP350 remains to be determined. Notably, Tasken and colleagues (17) have recently reported that AKAP350 has the ability to form functional regulatory complexes with known binding proteins. Targeting of the R II subunit to AKAP350 spatially localizes protein kinase A to the centrosome region whereas temporal control is exerted in the same anchored region by phosphodiesterase 4D3 degradation of cAMP. Therefore, it is reasonable to hypothesize that a similar complex may exist between TACC4 and other AKAP350 regulatory interaction proteins. The existence of such an AKAP350 signaling complex would provide a plausible explanation for the observed cell cycle-dependent translocation of TACC4. Similar functional translocation is seen in many proteins, in particular protein kinases (reviewed in Refs. 1 and 43). Cell cycle phosphorylation-dependent translocation is seen with cyclin B-p34 cdc2 kinase (CDK1) activation. During G 2 phase of the cell cycle, CDK1 is held in an inactive state by phosphorylation at residues Thr-14 and Tyr-15 by WEE1 and MYT1 (44). In late G 2 phase or mitotic prophase, phosphatase CDC25C dephosphorylates and activates CDK1, resulting in the translocation of CDK1 to the nucleus, an event that is required for cell cycle progression (45,46). Notably, recent work by Carlson and colleagues (47) has shown a cell cycle phosphorylation-dependent interaction between R II ␣ and AKAP350. CDK1 is localized to the centrosome at the beginning of mitosis and phosphorylates R II ␣ on Thr-54 (48). This phosphorylation by CDK1 results in dissociation of R II ␣ from AKAP350. A similar event in which TACC4 is phosphorylated or dephosphorylated could explain redistribution of TACC4 from AKAP350 centrosome sequestration to the spindle apparatus. A responsible kinase or phosphatase remains to be determined.
In summary, we have identified a novel TACC family member, TACC4, which interacts with AKAP350 at the centrosome in interphase cells and the spindle apparatus in mitotic cells. We propose that AKAP350 is responsible for sequestering TACC4 to the centrosome in non-dividing cells. AKAP350 anchoring of TACC4 provides for a spatial localization of TACC4 for its functional role in regulating the interaction with the spindle apparatus during mitosis. The numerous interaction partners of AKAP350 suggest that the centrosomal AKAP350 complex may play a role in activation and subsequent translocation of TACC4 to the spindle.