Identification and Cloning of Centaurin-α A NOVEL PHOSPHATIDYLINOSITOL 3,4,5-TRISPHOSPHATE-BINDING PROTEIN FROM RAT BRAIN

Using an affinity resin and photoaffinity label based on phospholipid analogs of inositol 1,3,4,5-tetrakisphosphate (InsP4), we have isolated, characterized, and cloned a 46-kDa protein from rat brain, which we have named centaurin-α. Binding specificity was determined using displacement of 1-O-[3H](3-[4-benzoyldihydrocinnamidyl]propyl)-InsP4 photoaffinity labeling. Centaurin-α displayed highest affinity for phosphatidylinositol 3,4,5-trisphosphate (PtdInsP3) (IC50 = 120 nM), whereas InsP4, PtdInsP2, and InsP3 bound with 5-, 12-, and >50-fold lower affinity, respectively. Screening a rat brain cDNA library with a polymerase chain reaction product, generated using partial amino acid sequence from tryptic peptides, yielded a full-length clone. The 2,450-base pair cDNA contained an open reading frame (ORF) encoding a novel protein of 419 amino acids. Northern analysis revealed a 2.5-kilobase transcript that is highly expressed in brain. The deduced sequence contains a novel putative zinc finger motif, 10 ankyrin-like repeats, and shows homology to recently identified yeast and mammalian Arf GTPase-activating proteins. Given the specificity of binding and enrichment in brain, centaurin-α is a candidate PtdInsP3 receptor that may link the activation of phosphoinositide 3-kinase to downstream responses in the brain.

Receptor-stimulated phosphoinositide (PI) 1 metabolism gen-erates numerous inositol polyphosphates (InsP n s) and inositol phospholipids, many of which may function as potential second messengers (1). Of the possible PI metabolites, Ins(1,4,5)P 3 (InsP 3 ) and diacylglycerol (DAG) are the best characterized second messengers. Generated by receptor-stimulated phospholipase C, hydrolysis of PtdIns(4,5)P 2 (2), Ins(1,4,5)P 3 binds to and gates an InsP 3 receptor calcium channel on the endoplasmic reticulum (2,3). The lipid DAG remains in the membrane where it activates several protein kinase C isoforms and may regulate other targets (4,5). In the membrane, DAG is metabolized rapidly to monoacylglycerol and to several phospholipids. In the cytoplasm, Ins(1,4,5)P 3 can be phosphorylated to Ins(1,3,4,5)P 4 by an InsP 3 3-kinase. Other isomers of InsP 4 , InsP 5 , and InsP 6 , some of which are synthesized independently of Ins(1,4,5)P 3 , have been identified (for review, see Ref. 6), and their production may also be regulated by receptors or during cell growth. Information from receptor binding studies, using radioactive InsP 4 and InsP 6 , have demonstrated that a number of important regulatory proteins contain high affinity InsP n binding sites. InsP n s have been implicated in the regulation of clathrin assembly proteins AP-2 (7,8), AP-3 (9), the non-clathrin-associated coatomer proteins (10), synaptotagmin (11), and the regulation of the small GTPases ras and/or rap via a specific GTPase-activating protein (GAP) activity (12).
Inositol phospholipids have also been postulated as messenger molecules. From in vivo, genetic, and permeabilized cell studies, evidence for critical roles for the inositol phospholipids PtdIns(3)P, PtdIns(4)P, and PtdIns(4,5)P 2 as regulators of membrane vesicle trafficking and cytoskeletal rearrangements is accumulating rapidly (5,(13)(14)(15). One inositol phospholipid, PtdIns(3,4,5)P 3 , has emerged as a potential messenger molecule in receptor-stimulated cells (16 -18). Synthesized by receptor-stimulated PI 3-kinase phosphorylation of PtdIns(4,5)P 2 (18), PtdInsP 3 is not a substrate for PI-specific phospholipase * This work was supported in part by National Institutes of Mental Health Grants R29MH50102 and DDRC P50HD32901 (to A. B. T.). Work at Stony Brook was supported by National Institutes of Health Grant NS29632 (to G. D. P.). The first two authors contributed equally to this study. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Although the molecular basis by which PI 3-kinase regulates these cellular activities is still unclear, the existence of specific receptor targets for PtdInsP 3 , analogous to the receptors for InsP 3 and DAG, has been proposed (35,36). In vitro, PtdInsP 3 competes with phosphotyrosine for binding to the SH2 domain of the p85 subunit of PI 3-kinase (37). PtdInsP 3 has been shown to stimulate pleckstrin phosphorylation in permeabilized platelets (38), and there is evidence that 3-phosphorylated inositol phospholipids may interact with Akt kinase (protein kinase B) (32), various protein kinase C isoforms (39 -41), and PRK, a protein kinase C-related kinase (42). The diversity of activities regulated by PtdInsP 3 predicts that numerous cellular effectors for PtdInsP 3 may exist. To isolate and identify specific, high affinity PtdInsP 3 receptors from mammalian brain, we have used affinity chromatography with a tethered aminopropyl-InsP 4 linked to Affi-Gel 10 and a hydrophobic InsP 4 photoaffinity probe. Here, we report the identification, characterization, and cloning of a novel candidate PtdInsP 3 receptor that is highly enriched in rat brain.

Synthesis of 1-O-[ 3 H](3-[4-Benzoyldihydrocinnamidyl]propyl)-InsP 4 -
The precursor 1-O-(3-aminopropyl)-D-myo-Ins(1,3,4,5)P 4 was synthesized in 12 steps from ␣-D-glucose (43) and was coupled with tritiumlabeled 4-benzoyldihydrocinnamoyl N-hydroxysuccinimide ester (BZDC-NHS) (45,46). An ethyl acetate solution of 2.76 mCi (80 nmol) of this reagent was evaporated to dryness under N 2 and dissolved in 200 l of N,N-dimethylformamide. To this solution, 100 l of 0.25 M triethylammonium bicarbonate (TEAB) buffer was added, followed by the addition of 100 l of a 0.56 mg/ml solution of aminopropyl-InsP 4 (0.81 mol) in 0.25 M TEAB buffer. The reaction was kept in the dark and stirred overnight at room temperature. The mixture was then concentrated in vacuo; the residue was dissolved in 500 l of deionized water and applied to a Pasteur pipette (4 ϫ 0.5 cm) column of DEAE-cellulose (HCO 3 Ϫ form). The column was washed with 1 ml of deionized or distilled water and eluted sequentially with the following: 1 ml of 0. Protein Purification-Centaurin 2 -␣ was purified from whole rat brain of 150 -300-g male Sprague-Dawley rats (Charles River) using methods described previously with all procedures at 0 -4°C (44,47). Briefly, whole brains were homogenized in 50 mM Tris/HCl, pH 7.7, containing 1 mM EDTA; 1 mM EGTA; 1 mM ␤-mercaptoethanol; 100 mg/liter phenylmethylsulfonyl fluoride; 5 mg/liter each of chymostatin, antipain, and pepstatin; 10 mg/liter aprotinin and leupeptin; and 250 mg/liter CBZ-phenylalanine (PB). After centrifugation at 45,000 ϫ g for 15 min, the soluble fraction was diluted 3-5-fold with PB containing 250 mM NaCl. Resuspended membranes were solubilized for 45 min with PB plus 1% CHAPS and 250 mM NaCl and centrifuged for 30 min at 45,000 ϫ g. All buffers for membrane-associated fractions contained 1% CHAPS. CHAPS-solubilized membranes and soluble proteins each were incubated with heparin agarose for 1 h. The resin was washed with PB plus 250 mM NaCl and eluted with PB containing 1 M NaCl. The eluates were concentrated in Amicon Centripreps to a final volume of 1-2 ml for each preparation of 10 -15 rat brains for photolabeling experiments. For further purification by InsP 4 affinity chromatography, the concentrated heparin-agarose eluates were diluted with 50 mM Tris/HCl, pH 7.4, containing 1 mM EDTA (ACB) and loaded onto an InsP 4 affinity column (44)  Peptide Sequence Analysis-Approximately 50 g of concentrated centaurin-␣ purified by InsP 4 affinity column chromatography was subjected to SDS-PAGE and transferred to Immobilon-P membrane. The membrane was stained with Coomassie to localize the protein band with an apparent molecular size of 46 kDa, which was then excised and digested with trypsin. The resultant tryptic peptides were fractionated via reverse phase HPLC with a linear gradient of acetonitrile, 0.1% trifluoroacetic acid (90 min). Individual tryptic peptides were then subjected to automated Edman degradation on a Procise 492 protein sequencer (Perkin-Elmer, ABD division). The nine unique peptide sequences that were obtained were compared with sequences in the GenBank data base using the TFASTA and BLAST DNA analysis software package of the University of Wisconsin Genetics Computer Group.
cDNA Cloning-Rat brain template cDNA was prepared by reverse transcription of rat brain poly(A) ϩ RNA (Clontech). Degenerate oligonucleotide primers were designed based on two tryptic peptides having 100% homology to putative peptide sequences at the ends of a humanexpressed sequence tag (GenBank accession number T09325). The sequences of the primers synthesized (DNA International) were as follows: 5Ј-TT(T/C)CA(T/C)TA(T/C)(T/C)TICA(A/G)GTIGCITT(T/C)CC-3Ј (26-mer, 64-fold degeneracy) and 5Ј-C(C/T)TGIGGIAGCATIGG(C/ T)CT(G/A)TCIAC-3Ј (25-mer, 8-fold degeneracy). Using these primers in PCR on the rat brain template cDNA, we were able to obtain a product corresponding to a predicted size of approximately 400 bp. The PCR product was subsequently cloned using the TA cloning system (Invitrogen). A total of 1.6 ϫ 10 6 independent recombinants of a Zap II postnatal day 7 rat brain cDNA library (kindly provided by Dr. Craig Garner, University of Alabama at Birmingham) were screened with a [␣-32 P]dCTP-labeled 400-bp PCR product as a probe (Promega Primea-Gene System). Twenty-five strongly positive plaques were picked and excised with the Stratagene Solr System. The clone containing the longest insert of 2,450 bp was sequenced and called centaurin-␣. Using a nested deletion based strategy, both strands of the centaurin-␣ clone were sequenced as double-stranded plasmids with M13 forward and reverse primers by the dideoxynucleotide chain termination method using deoxyadenosine 5Ј-[␣- 35 S]thiotriphosphate and Sequenase 7.0 (U. S. Biochemical Corp.).
Northern Analysis-The multiple tissue Northern blot containing poly(A) ϩ RNA from the indicated rat tissues was obtained from Clontech. According to the company's catalog, "Multiple tissue Northern blots are premade Northern blots of quality poly(A) ϩ RNA from different (rat) tissues. The poly(A) ϩ RNA is purified by several passages through oligo(dT)-cellulose columns, then run on a denaturing formaldehyde, 1.2% agarose gel, and blotted onto a positively charged nylon membrane. The lanes of the blots each contain approximately 2 g of pure poly(A) ϩ RNA from specific tissues. The RNA loaded in each lane is adjusted so that a visible ␤-actin signal is present in every lane." Prehybridization was performed at 42°C in 6 ϫ SSPE (1.08 M NaCl, 60 mM sodium phosphate, 6 mM EDTA, pH 8.0, 5 ϫ Denhardt's solution, 50% formamide, 100 g/ml sheared salmon sperm DNA, and 0.1% SDS for 4 h. An [␣-32 P]dCTP-labeled AvaI/BamHI fragment (4.5 ϫ 10 9 cpm/ g, 1,240 bp) from the 3Ј end of centaurin-␣ was hybridized to the membrane for 24 h at 42°C (protocol recommended by Clontech). The membrane was washed four times in 1 liter of 2 ϫ SSC (0.9 M NaCl, 99 mM sodium citrate), 0.1% SDS for 15 min at 50°C, followed by two 15-min washes in 1 liter of 1 ϫ SSC, of 0.5 ϫ SSC, and of distilled deionized water at 50°C. Autoradiography was performed at Ϫ80°C with Kodak X-Omat-AR film with intensifying screens for 24 h. Expression of Recombinant Centaurin-␣ in Esherichia coli-A COOH-terminal fusion protein construct (containing ankyrin repeats 7-10) was made by digesting the 2.5-kb pair centaurin-␣ cDNA with BglII and KpnI to generate a 1.6-kb pair fragment that was cloned into the Pinpoint Xa2 expression vector (Promega). The expression of recombinant centaurin-␣ in the bacterial strain DH5␣ was induced by inositol-1-thio-␤-D-galactopyranoside. The COOH-terminal fusion protein was extracted with 8 M urea from insoluble bacterial pellets. The extract was clarified by centrifugation. Then, the supernatant was dialyzed to remove the urea. The COOH-terminal fusion protein was then purified using the Soft-link avidin resin (Promega). Approximately 1 mg of purified fusion protein was obtained from 1 liter of bacterial culture.
Generation of Polyclonal Antisera-A peptide containing the NH 2terminal amino acids 2-19 of centaurin-␣ was synthesized and conjugated to PPDT (purified protein derivative of tuberculin). The bacterially expressed COOH-terminal Pinpoint centaurin-␣ fusion protein and the PPDT-conjugated NH 2 -terminal peptide were injected into rabbits following a standard booster immunization protocol. The NH 2 -terminal antipeptide antiserum was designated J49 and the COOH-terminal fusion protein antiserum, J4.
Expression of Recombinant Centaurin-␣ in COS-7 Cells-COS-7 cells were routinely cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. The EagI/SphI fragment of the centaurin-␣ cDNA was subcloned into the p63d2 vector downstream of the cytomegalovirus promoter (22) to give the centaurin-␣ vector named p63d.2.13. Transient transfections of COS-7 cells (plated at 50 -80% confluence) were performed using 10 -20 g of recombinant centaurin-␣ DNA and 50 l of lipofectamine (Life Technologies, Inc.)/10-cm dish, according to the manufacturer's protocol. Following transfection, the cells were cultured for an additional 48 -60 h. The expressed recombinant centaurin-␣ protein was isolated and partially purified by heparin-agarose and InsP 4 affinity column chromatography under the same conditions as for the native rat brain centaurin-␣ protein. Approximately 20 g of purified recombinant centaurin-␣ protein was obtained from one 10-cm dish of cells.
Immunoblot Analysis-Fractions from rat brain homogenate and cell lysates were separated by SDS-PAGE and transferred to Protean nitrocellulose membranes (Schleicher & Schuell). Membranes were probed with either the rabbit polyclonal NH 2 -terminal J49 antipeptide or the J4 antiserum at 1:500 dilution. Horseradish peroxidaseconjugated anti-rabbit secondary antibodies were detected using 3,3Ј-diaminobenzidine. 3 Receptor-binding Protein-Previously, we have used an InsP 4 affinity column to isolate specific, high affinity InsP n -binding proteins from the particulate fraction of rat cerebella (44,47). This strategy led to the identification of one high affinity InsP n -binding protein, the clathrin-associated protein AP-2 (7). In addition to the proteins characterized for [ 3 H]InsP 4 binding and photolabeling, we noted the presence of a 46-kDa protein (47). We predicted that this 46-kDa protein would have the highest affinity for InsP 4 , since the other InsP n -binding proteins eluted from the resin in rank order of their affinity for InsP 4 , and the 46-kDa protein eluted from the column last. However, under the conditions used for reversible [ 3 H]InsP 4 binding and photoaffinity labeling with an radioiodinated arylazido probe, [ 125 I]ASA-InsP 4 (43), we could not detect binding to this protein (47). Thus, we were unable to characterize its specificity and affinity or determine its localization. As shown in Fig. 1, A and B, a similar heparin-agarose and InsP 4 affinity chromatographic protocol has enabled us to purify this 46-kDa protein (named centaurin-␣) from the soluble fraction of whole rat brain. Because of the presence of multiple high affinity InsP n -binding proteins and the presence of endogenous inhibitors in whole brain supernatant, we could not estimate the specific activity of binding in the crude fraction. Approximately 0.5 mg of 90% homogeneous protein could be purified from the supernatant fraction from 50 whole rat brains. Assuming a recovery of 25% (since this is a two-step, rapid purification procedure), this predicts an abundance of centaurin-␣ protein of about 0.01% of the total brain protein, well within the range of expression of several other PI signaling proteins, such as the InsP 3 receptor (3).

Identification of a PtdInsP
Although in earlier studies we determined that centaurin-␣ did not bind used to identify a specific InsP 3 binding amino acid sequence in the InsP 3 receptor calcium channel (48). The BZDC moiety, which is considerably more hydrophobic than the ASA moiety, is especially useful because of the high efficiency of photocoupling (49). Because the label is covalently attached to the binding protein, this allows for separation by SDS-PAGE and visualization in partially purified fractions as well. Fig. 1C shows that the 46-kDa centaurin-␣ protein (arrowhead) was labeled in the partially purified fractions from rat brain supernatant. Specific photolabeling was blocked by including 10 M unlabeled InsP 4 in the reaction. Photolabeling also displayed a broad pH optimum, which peaked between pH 7.0 and 7.5 (data not shown). Centaurin-␣ partitions in a 2:1 ratio into the soluble and membrane-associated fractions from rat brain. Both fractions behave identically in the InsP 4 affinity column purification and can be specifically photolabeled (Fig. 1D). In addition to brain, a 46-kDa protein is also specifically photolabeled in heparin-agarose fractions from the soluble fraction of rat heart (Fig. 1D), indicating that centaurin-␣, or a related protein, is also expressed in non-neuronal tissues. These results demonstrate that under the conditions used, centaurin-␣ binds to the InsP 4 affinity resin with the highest affinity of the soluble and membrane-associated brain and heart proteins and can be specifically photolabeled by [ 3 H]BZDC-InsP 4 .
Photoaffinity labeling with the [ 3 H]BZDC-InsP 4 was then used to determine the specificity and affinity of centaurin-␣ for several InsP n s and inositol phospholipids. As shown in Fig. 2, increasing concentrations of these unlabeled ligands in the binding reaction causes incremental displacement of [ 3 H]BZDC-InsP 4 from centaurin-␣. The most effective ligand at displacing the label was a synthetic dipalmitoyl PtdIns(3,4,5)P 3 , which had an IC 50 of 120 nM. The next most potent ligand was the deacylated PtdInsP 3 , glycerophosphoinositol 3,4,5-trisphosphate (gPInsP 3 ), which had an IC 50  Given an observed IC 50 for InsP 4 of 620 nM, it is not surprising that in previous reversible binding studies using 5-10 nM [ 3 H]InsP 4 , binding was not observed (47). These data show that several groups are important for binding to centaurin-␣. The long chain fatty acid on the glycerol backbone led to an approx-imate 4-fold increase in binding affinity. The glycerol group esterified to the inositol head group led to a 1.4-fold increase in binding. The 3-phosphate on the D-myo-inositol increased binding between 5-and 13-fold. These results demonstrate that centaurin-␣ has a preference for binding InsP n analogs that are P-1 phosphodiesters and which contain a 3-phosphate on the inositol ring. Importantly, the InsP 4 affinity resin also contains both of these crucial substructures. Together with the data showing that it has extremely high affinity for the InsP 4 resin, this specificity is consistent with the suggestion that centaurin-␣ is a high affinity PtdInsP 3 -binding protein.
Cloning of Full-length Centaurin-␣ cDNA-Centaurin-␣ was purified from whole rat brain supernatant, separated on SDSpolyacrylamide gels, transferred to an Immobilon-P membrane, and digested on the membrane with trypsin. The resultant tryptic peptides were fractionated by reverse phase HPLC. The amino acid sequence of 11 peaks (9 unique) was determined by automated Edman degradation. A GenBank data base search revealed that five of the tryptic fragments of centaurin-␣ had 100% identity to the predicted amino acid sequence of a 375-bp human expressed sequence tag from a human infant brain cDNA library. This identity allowed us to design degenerate oligonucleotide primers and synthesize a PCR product of predicted size (400 bp) from rat brain cDNA template. The nucleotide sequence of the PCR product was 79% identical to the human-expressed sequence tag. The predicted amino acid sequence of the rat brain PCR product and the tryptic fragments of centaurin-␣ revealed that the humanexpressed sequence tag contained a reading error that resulted in a frameshift. The 400-bp cDNA was used to screen a Zap II postnatal day 7 rat brain cDNA library. Of 25 positive clones identified, the largest, of 2,450 bp in length, was sequenced in both directions using a nested deletion strategy. The entire nucleotide sequence of this clone, designated centaurin-␣, is presented in Fig. 3.
Sequence Analysis of Centaurin-␣ cDNA and Proposed Amino Acid Sequence-Centaurin-␣ contains a 112-bp 5Ј-untranslated region lacking any in-frame stop codons, with a putative Kozak's consensus sequence upstream of the first ATG, an open reading frame of approximately 1.4 kb pairs, and a large 3Ј-untranslated region with a 3Ј-poly(A) ϩ tract. The predicted amino acid sequence of centaurin-␣ (Fig. 3) encodes a novel protein of 419 amino acids with a calculated molecular mass of 48 kDa, which also contains all of the tryptic peptides (underlined). The calculated molecular mass of centaurin-␣ protein is close to the molecular mass of 46 kDa estimated from SDS-PAGE of the native protein. Information from the calculated isoelectric point (8.82) predicted that overall, centaurin-␣ is likely to be a basic protein, which would be consistent with its ability to bind to the highly negatively charged inositol phospholipid. In addition, information from hydropathy analysis indicated that centaurin-␣ is a soluble protein lacking any putative membrane-spanning regions. However, a region at the NH 2 terminus (see below) shows a stretch of 30 amino acids with considerable hydrophobic character.
Centaurin-␣ Contains a Putative Zinc Coordination Site and Ankyrin-like Repeats-Searching protein and nucleic acid sequence data bases indicated that centaurin-␣ shows significant homology to a number of other proteins, putative ORFs, and expressed sequence tags (more than 30 distinct sequences) found in animals, plants, and fungi. The most highly conserved region across all of these putative proteins lies in a region of approximately 50 residues, which may form a novel zinc coordination site around two pairs of cysteine residues with the consensus CX 2 CX 16 -17 CX 2 C (Fig. 4, A and B). Within this region, centaurin-␣ shows greatest identity (56%) with the putative YIE4 protein of Saccharomyces cerevisiae and next (52%) to both a human protein RIP1 (a cofactor for retroviral Rev proteins) and a human ORF, KIAA0041, expressed in myeloid cells. Further, this region of centaurin-␣ has 42% identity with an Arf-GAP purified recently from rat liver (50). Outside of this region, centaurin-␣ shows considerably less homology to these proteins with overall only 25.9% conservation of identity with S. cerevisiae YIE4, 25.3% identity with the human ORF KIAA0041, and only 20% identity with the rat Arf-GAP (Fig. 4). However, the human EST (EST07218, Gen-Bank accession number T09325), the sequence that aided in the design of PCR primers, shows 91% identity to centaurin-␣ in a region outside of the putative zinc binding site. This suggests that none of these proteins is a direct homolog of the rat brain centaurin-␣ protein, but instead that each represents a different member of a large family of centaurin-␣-related proteins.
Our initial searches also indicated that centaurin-␣ showed homology to ankyrin and ankyrin-related proteins from a wide variety of species (for review, see Ref. 51). Subsequent analysis has indicated that centaurin-␣ itself contains 10 divergent ankyrin-like repeats (Fig. 5A). A variable number of such repeats, ranging from 4 to 12 repeats, is found in each of the other centaurin-␣-related proteins. The sequences of the first and second repeats are the best conserved between different centaurin family members; this may reflect a conservation of function. Notably in centaurin-␣, as well as in a number of the other proteins, the second ankyrin repeat (Fig. 5B) contains a se-  Fig. 5C). The percentage identity between each protein and centaurin-␣ (produced by pairwise alignment using the GAP program of GCG8) is presented on the right side of the figure. Panel B, comparison of sequences in putative Zn 2ϩ coordination region of centaurin-␣ and related proteins. The amino acid residue number marking the beginning of the sequence in each protein is given on the left side of the alignment. The percentage identity between each protein and centaurin-␣ in this region is presented on the right side of the alignment. Residues conforming to a consensus for the members of this family are marked above the sequence and within the alignment are indicated by bold type. In addition to rat centaurin-␣, the sequences represented are: SpAC26a3 10, a putative S. pombe ORF (GenBank accession number Z69240); ScYIE4, the putative product of the S. cerevisiae YIE4 gene (accession number P40529); HsKIAA0041 and HsKIAA0050, two human putative ORFs (accession numbers D26069 and D30758); ScGCS1, the product of the S. cerevisiae GCS1 gene (59); ScGTS1/LSR1, the predicted product of the S. cerevisiae gene cloned both as GTS1 (61) and LSR1 (accession number X91489); ScSPS18, the putative product of the S. cerevisiae SPS18 sporulation-specific gene (60); HsRIP1/RabCCF, the cellular cofactor of retroviral Rev and Rex RNA regulatory proteins (68,69); RnArf-GAP, the rat liver Arf-GAP (50); ScD971727, the predicted product of the S. cerevisiae putative ORF D71727 (accession number U33057); ScGLO3, the product of the S. cerevisiae GLO3 gene (59).
quence similar to one present in two InsP n -binding proteins, GAP InsP4-BP (12) and synaptotagmin II (11), including the presence of a conserved positive residue (asterisk) in the position of a lysine known to be critical for InsP 4 binding to synaptotagmin.
In the centaurin-␣ sequence, there is a region of homology with some of the TERM family proteins (talin, ezrin, radixin, moesin, and merlin), of which band 4.1, identified in erythrocytes, is a prototype (for review, see Ref. 53). The region of homology with the TERMs, present in the sixth repeat region of  HsInsP4-BP; 12). The suggested consensus sequence for inositol polyphosphate binding is marked; ϩ indicates a conserved positively charged residue, and residues conforming to this sequence are printed in bold in the alignment. The lysine residue demonstrated to be essential for InsP 4 binding to synaptotagmin is marked with an * above the sequence. The sequence of repeat 16 of human ankyrin (54) and the consensus sequence for centaurin repeats derived in panel A are also presented. Panel C, alignment of a region including centaurin repeat 6 with homologous sequences in members of the TERM protein family. The repeat 6 region of centaurin-␣ was compared with some of the TERM family proteins (53): MmMerlin, L28176; HsEzrin, P15311; and HsMoesin, P26038. The consensus sequence for the centaurin-␣ repeats is shown above the alignment. The residues showing conservation within Dayhoff groups in all proteins are in bold, and a consensus sequence for this homology is presented below the alignment. centaurin-␣, is shown in Fig. 5C. The highest degree of similarity is with merlin, a tumor suppressor protein identified in neurofibromatosis type 2. The TERM family proteins have been proposed to link the actin cytoskeleton to membrane proteins in a variety of substructures, including microvilli, filopodia, and cell-cell junctions (53).
The Centaurin-␣ mRNA Transcript Is Highly Expressed in Rat Brain-A 3Ј fragment (1,244 bp) was used to probe a multiple-tissue Northern blot to determine the tissue-specific distribution of centaurin-␣. The Northern blot analysis revealed that a 2.5-kb mRNA is highly enriched in brain, with substantially lower but detectable levels in lung, kidney, and spleen (Fig. 6). The 2.5-kb transcript size on the Northern is in good agreement with the size of the isolated centaurin-␣ clone of 2,450 bp. In rat brain, other minor transcripts, one of which was 4.0 kb, were also detected with the centaurin-␣ probe, but at a considerably lower level of expression than the 2.5-kb transcript.

COS-7 Cell-expressed Centaurin-␣ Binds to InsP 4 Affinity
Resin-To ascertain whether the centaurin-␣ cDNA encodes for the PtdInsP 3 -binding protein that we purified, the coding sequence of the centaurin-␣ was subcloned into a mammalian expression vector and expressed in transfected COS-7 cells. The COS-7 cell-expressed centaurin-␣ protein was isolated from lysed cells by both heparin-agarose (not shown) and InsP 4 affinity resin (Fig. 7A). The Coomassie-stained SDS-polyacrylamide gel revealed the specific enrichment of a protein of 46 kDa (arrowhead), which corresponds, in molecular size, to centaurin-␣ from rat brain. The binding of the COS-7-expressed protein to the affinity column demonstrates that centaurin-␣ encodes a protein that contains a high affinity InsP 4 resin binding site, similar to the endogenous 46-kDa protein. A bacterial fusion protein also bound to the heparin-agarose column under conditions used for the brain centaurin-␣ (not shown), which suggests that auxiliary proteins are probably not required to form the PtdInsP 3 binding site in centaurin-␣, and further, that this binding site is determined by the primary amino acid sequence.
Centaurin-␣ Antibodies Recognize the Rat Brain 46-kDa Protein-Antibodies generated against either a synthetic NH 2terminal peptide of centaurin-␣ (J49) or a COOH-terminal fusion protein of centaurin-␣ (J4) recognized a 46-kDa protein band in immunoblot analysis (Fig. 7B) of fractions from rat brain homogenate, heparin-agarose column eluate, or InsP 4 affinity column eluate (lanes 1-3). The J49 and J4 antibodies were also immunoreactive to the expressed centaurin-␣ protein in transfected COS-7 cell lysate (lane 5), but not in mocktransfected cell lysate (lane 4). Thus, antibodies raised against sequences generated from the cloned centaurin-␣ protein recognized the authentic 46-kDa rat brain centaurin-␣ protein and COS-7-expressed protein. DISCUSSION We have shown here the isolation and cloning of a 46-kDa protein, which we have named centaurin-␣. Centaurin-␣ appeared to bind to the InsP 4 resin with the highest affinity of any InsP 4 binding protein identified. Paradoxically, although it interacted very tightly with the InsP 4 affinity resin, initial binding studies with [ 3 H]InsP 4 and photolabeling with [ 125 I]ASA-InsP 4 indicated that centaurin-␣ did not bind InsP 4 with high affinity (44,47). We have demonstrated here that a more hydrophobic photoprobe, [ 3 H]BZDC-InsP 4 , can efficiently label the centaurin-␣ in fractionated rat brain extracts. Important to this study is the observation that the hydrophobic benzophenone photophore, appended to an aminopropyl chain and attached via a phosphodiester linkage to Ins(1,3,4,5)P 4 , makes this photoaffinity label a good structural surrogate for the inositol phospholipid PtdInsP 3 (Fig. 8). Both the InsP 4 affinity column and the BZDC-InsP 4 photolabel have three key substructures that mimic PtdIns(3,4,5)P 3 : (i) the correct 3,4,5trisphosphate on a D-myo-inositol ring; (ii) linkage to a threecarbon chain through a P-1 phosphodiester linkage; and (iii) substantial hydrophobic character in the acyl or amidyl groups appended to the three-carbon chain. Thus, these two materials resemble PtdInsP 3 more closely than they mimic the watersoluble InsP 4 . In an analogous fashion, [ 3 H]BZDC-InsP 3 was used as a mimic of PtdIns(4,5)P 2 to accomplish the isozymeand domain-selective photocovalent modification of the pleckstrin homology binding domain of phospholipase C␦1. 3 The benzophenone group has largely replaced the arylazide as the photoprobe of choice for many biochemical studies (49). It is chemically more stable, does not decompose in ambient light, and selectively modifies hydrophobic binding regions in target macromolecules. Thus, [ 3 H]BZDC-InsP 3 was used to map the InsP 3 binding site in the rat brain InsP 3 receptor, but insufficient labeled protein was obtained with an [ 125 I]ASAmodified photoprobe (48). An entire family of phosphoinositide photoaffinity probes (55) has been developed based on the efficacy of the BZDC-InsP n probes for selective covalent modification of target proteins in signaling pathways.
Displacement of the [ 3 H]BZDC-InsP 4 photoprobe with PI pathway intermediates indicated that centaurin-␣ had the highest affinity for synthetic dipalmitoyl PtdIns(3,4,5)P 3 with an apparent IC 50 of 120 nM. Since the photolabeling assay occurs under nonequilibrium conditions, the absolute values for binding constants are only approximate. Nonetheless, the rank order of displacement accurately reflects the relative affinity. The other inositol-containing compounds tested were less potent, with dipalmitoyl PtdInsP 3 Ͼ gPInsP 3 Ͼ daPtdInsP 3 Ͼ InsP 4 Ͼ PtdInsP 2 Ͼ gPInsP 2 Ͼ InsP 3 , indicating that the 3-phosphate group on the inositol ring and the long chain fatty acids on the glycerol backbone were both important in ligand binding. Based on this specificity comparison, we propose that centaurin-␣ is an excellent candidate PtdIns(3,4,5)P 3 receptor. As arachidonate is normally esterified at the C-2 position in endogenous PtdInsP 3 (Fig. 8), it is possible that centaurin-␣ may bind the authentic PtdInsP 3 with even higher affinity than synthetic PtdInsP 3 .
Both the bacterial and mammalian cell-expressed centaurin-␣ protein could be partially purified by heparin-agarose chromatography under the same conditions used to isolate the native rat brain protein. Moreover, the COS-7 cell expressed protein bound to the InsP 4 affinity resin. Polyclonal antisera against a bacterial COOH-terminal fusion protein and an NH 2terminal peptide recognize both the native rat brain protein and COS-7 cell-expressed protein. Together with the tryptic peptide data and similar molecular size of the cloned protein, this expression and immunological data are evidence that the centaurin-␣ clone is identical to or very highly homologous to the 46-kDa PtdInsP 3 -binding protein isolated by InsP 4 affinity chromatography.
The observation that both soluble and membrane-associated centaurin-␣ isolated from rat brain behave similarly in binding to the InsP 4 affinity resin and the [ 3 H]BZDC-InsP 4 photoprobe suggests that centaurin-␣ can partition to the lipid bilayer. Receptor-stimulated membrane association is a common theme for many signaling proteins, including protein kinase C, PI 3-kinase, phospholipase A 2 , and phospholipase C (for review, see Ref. 20). Centaurin-␣ may be related to a 42-kDa InsP 4binding protein recently isolated from pig cerebella (56,57) since two peptides from the 42-kDa protein show 100% identity to centaurin-␣. However, the porcine 42-kDa protein is smaller than the rat 46-kDa protein described here, and a third peptide is only 70% identical to an NH 2 -terminal sequence in centaurin-␣ (56). Antibodies to the 42-kDa pig cerebellar protein recognize proteins in soluble and particulate fractions from the brain of several animal species, including rat (57). This may, therefore, represent a very closely related isoform of centaurin-␣, perhaps arising through the use of alternative mRNA splicing. Consistent with this, we have observed that in addition to the major 2.5-kb centaurin-␣ transcript, brain also contains several minor mRNA species that could represent alternatively spliced centaurin mRNAs.
Data base searches revealed that centaurin-␣ is one of a family of related proteins represented in fungi, plants, and animals. The region of highest homology between each of the centaurin relatives is in the putative zinc coordination site. A wide variety of signaling proteins, including protein kinase Cs, chimaerins, rafs, and DAG kinases, contain zinc coordination sites. In these proteins, a pair of zinc binding sites composed of both cysteine and histidine residues cooperate to form an individual zinc "butterfly" domain. In protein kinase C, this zinc butterfly domain has been shown to form a hydrophobic cleft that provides the site of phorbol ester and DAG binding (58). In centaurin-␣ and its relatives, there is only a single potential zinc coordination site composed of four cysteine residues, which is distinct from that found in these other signaling proteins. However, hydropathy plots of centaurin(s) predict that this zinc finger domain comprises a highly hydrophobic region that could provide a surface for interaction with membrane lipids.

FIG. 8. Comparison of [ 3 H]BZDC-InsP 4 and PtdInsP 3 structures.
Both molecules have the 3,4,5-trisphosphate recognition element of D-myo-Ins(1,3,4,5)-P 4 , as well as a P-1 phosphodiester linked to a 3-carbon chain with a hydrophobic appendage. mutation in one of these putative zinc-binding cysteines which gives rise to a mutant phenotype, similar to a null GCS1 mutant, which causes failure to resume growth from stationary phase at restrictive temperatures, indicating an important function for this region in cell regulation (59). Another centaurin relative has been identified recently as a GAP for the small molecular weight GTPase Arf (50). Again, the zinc coordination site appears to be critical for this activity as an equivalent cysteine substitution abolishes Arf-GAP activity in this protein in vitro. GCS1 has also been reported to show Arf-GAP activity. Less is known about the biological activities of other centaurin relatives. In yeast, SPS 18 has been implicated in the process of sporulation (60), whereas mutations affecting the GTS1/LSR1 gene lead to alterations in timing of budding and cell size determination (61). Although the zinc coordination region is highly conserved in centaurin-␣ and its relatives, it remains to be seen whether they also possess GAP activity toward Arf or other small GTPases. Further, it is not yet clear whether other members of the centaurin family are capable of binding PtdInsP 3 or what may be the consequences of this binding for their function.
Recently, potential PtdIns(4,5)P 2 binding motifs have been identified in numerous proteins. In the pleckstrin homology domain of ␣-actinin, phospholipase C␦, ␤-spectrin, and Grb-7, a subdomain RX 7 (H/R/K)X 2-3 W(R/K) has been identified as a PtdInsP 2 binding site consensus sequence (62). Centaurin-␣ does not have homology with the pleckstrin homology domain or this pleckstrin homology subdomain, nor does it have homology with a distinct PtdInsP 2 binding region identified in cofilin (63). We have, however, located a basic sequence, PX 13-14 (R/ K)X 4 (R/K)XKX 5 FX 6 E, in centaurin-␣ which is also found in the C2B domains of the InsP 4 -binding protein GAP InsP4-BP (12) and InsP 4 /InsP 6 -binding protein synaptotagmin II (11), which may be involved in the binding of the 3,4,5-trisphosphate moiety of PtdInsP 3 . This region is located within the second ankyrin repeat in centaurin-␣ and is partially conserved in other family members.
All the members of the centaurin family also contain a variable number of repetitive domains which show homology to those found in the cytoskeletal protein ankyrin and a wide variety of other proteins (51). The potential importance of these repeats in the centaurin family is provided by the observation that truncations of GCS1, in which the protein terminates within the fourth or fifth ankyrin repeats, produce phenotypes similar to that in the mutant cysteine (zinc coordination) allele. The fact that these sequences are divergent in centaurin family members may indicate that these regions are responsible for providing specific interaction information that contributes distinct activities and specialization to each member. Specific cytoskeletal association may be predicted as centaurin-␣ also has a short region of homology with the TERM proteins (53), which have been shown to be involved in linking membrane proteins to the actin cytoskeleton.
Data from many reports have led to the prediction that PtdInsP 3 is involved in regulating actin cytoskeletal rearrangements and/or vesicle trafficking events (15,(22)(23)(24)(27)(28)(29)(30). The finding here that a PtdInsP 3 -binding protein has homology to a family of proteins which may regulate Arf-GTPase activity supports the suggestion that PI 3-kinase-coupled receptors might act via centaurin-␣ to regulate Arf-dependent postendosomal membrane trafficking events (64) which transport lipid and protein components to the plasma membrane during processes such as ruffling, glucose transporter translocation, or neurite outgrowth. Further, centaurin(s) may provide a link between PtdInsP 3 production and receptor regulation of phospholipase D activation, a process shown to be sensitive to inhibition of PI 3-kinase (65) and to involve the regulation of Arf proteins (66,67).
In conclusion, we propose that centaurin-␣ is a member of a new class of PtdInsP 3 -binding proteins which may have a role in mediating downstream events stimulated by growth/trophic factor activation of PI 3-kinase in the brain. The predicted homologies of centaurin-␣ with other proteins, its affinity for PtdInsP 3 , and its partitioning between soluble and membrane compartments are consistent with it playing a pivotal role linking regulation of the actin cytoskeleton and/or specific vesicle movements to changes in PtdInsP 3 production in the lipid bilayer. Whether a cell signaling pathway linking PtdInsP 3 production to centaurin-␣ and the regulation of Arf-GTPase activities is important in the brain awaits future investigation.