Identification of centaurin-�2:a phosphatidylinositide-binding protein present infat, heart and skeletal muscle

Paul Whitleya, Alison M. Gibbarda, FranÁoise Koumanova, Susan Oldfielda, Elaine E. Kilgourb,Glenn D. Prestwichc, Geoffrey D. Holman1)aa Department of Biology and Biochemistry, University of Bath, Bath/United Kingdomb Cardiovascular and Gastrointestinal Research Department, AstraZeneca Pharmaceuticals, Macclesfield,Cheshire/United Kingdom

c Department of Medicinal Chemistry, University of Utah, Salt Lake City, Utah/USA

Received November 7, 2001Received in revised version January 17, 2002Accepted January 18, 2002

Centaurin ± PH domains ± IP4 ± phosphatidylinositides ±adipose ± heart ± muscle

We describe here the cloning, expression and characterisationof centaurin-�2 from a rat adipocyte cDNA library. Thecentaurin-�2 cDNA contains an open reading frame, whichcodes for a protein of 376 amino acids with predicted mass of43.5 kDa. Centaurin-�2 shares 51 ± 59% identity with centaur-in-�1 proteins and has the same domain organisation, consist-ing of a predictedN-terminal ArfGAPdomain followed by twosuccessive pleckstrin homology domains. Despite the sequencesimilarity, there are a number of notable differences betweenthe previously characterised centaurin-�1 proteins and thenewly described centaurin-�2: (i) in vitro lipid binding experi-ments with centaurin-�2 do not reveal the same selectivity forphosphatidylinositol 3,4,5-trisphosphate over phosphatidylino-sitol 4,5-bisphosphate that has been shown for centaurin-�1;(ii) unlike centaurin-�1 which is expressed mainly in the brain,centaurin-�2 has a broad tissue distribution, being particularlyabundant in fat, heart and skeletal muscle; (iii) in contrast tocentaurin-�1 which is found in both membrane and cytosolicfractions, endogenous centaurin-�2 is exclusively present in thedense membrane fractions of cell extracts, suggesting aconstitutive membrane association. Insulin stimulation, whichstimulates phosphatidylinositol 3,4,5-trisphosphate produc-tion, does not alter the subcellular distribution of centaurin-�2 between adipocyte membrane fractions. This observation isconsistent with the lack of specificity of centaurin-�2 forphosphatidylinositol 3,4,5-trisphosphate over phosphatidylino-sitol 4,5-bisphosphate.

Introduction

Members of the centaurin family of proteins contain one or twopleckstrin homology (PH) domains and an ArfGAP domain(Randazzo et al., 2001). Centaurin-�1 proteins all have a highbinding specificity for inositol 1,3,4,5-tetrakisphosphate (IP4)and/or phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3).It follows that the membrane association of the centaurin-�1proteins may be regulated by the production of PI(3,4,5)P3(Venkateswarlu et al., 1999). GFP-tagged human centaurin-�1is translocated from a cytoplasmic/nuclear location to theplasma membrane upon production of PI(3,4,5)P3 in transfec-ted cells (Venkateswarlu et al., 1999; Tanaka et al., 1999).Furthermore, other centaurin-�1 family members appear to bepresent in bothmembrane and soluble pools in cells, suggestingthat they may cycle between membrane-bound and unboundstates (Stricker et al., 1997; Aggensteiner et al., 1998; Venka-teswarlu et al., 1999; Hanck et al., 1999). The N-terminal zincfinger motifs of the centaurin-� proteins have sequencesimilarity to the ArfGAP domain of yeast Gcs-1 (a proteinknown to regulate membrane trafficking (Poon et al., 1999).Although evidence for ArfGAP activity from in vitro assays islacking, Venkateswarlu et al. (1999) have shown that humancentaurin-�1 has possible ArfGAP function in vivo, as it canfunctionally complement aGcs-1-deleted yeast strain. Thus, thePI(3,4,5)P3 binding specificity together with the proposedArfGAP function of the centaurin-�1 family of proteinspotentially allows them to link lipid signalling to membranetrafficking.The centaurin-�, � and � family of proteins have been

identified in database searches as having centaurin-�1 like PHdomains andArfGAP domains, although the domain organisa-tion between families is different (Randazzo et al., 2001).Centaurin-�4 (also known as ASAP1) displays ArfGAP

0171-9335/02/81/04-222 $15.00/0

1) Dr. Geoffrey D. Holman, University of Bath, Bath, BA2 7AY/UK,e-mail: g.d.holman@bath.ac.uk, Fax: �441225826779

EJCB222 European Journal of Cell Biology 81, 222 ± 230 (2002, April) ¥ ¹ Urban& Fischer Verlag ¥ Jenahttp://www.urbanfischer.de/journals/ejcb

activity in vitro that is activated by phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) binding and has a proposed functionin actin cytoskeleton regulation (Kam et al., 2000; Randazzoet al., 2000). Centaurin-�3 (also known as PAP) is predicted toact in regulated Arf-mediated vesicular transport. Thus,proteins with identical structural elements appear to beinvolved in a variety of cellular functions.Themajority of the centaurin-�1 proteins have been isolated

and cloned from brain cDNA libraries, and Northern blotanalysis has revealed mRNA expression to be highly abundantin brain. The limited tissue expression of centaurin-�1 led us toinvestigate the possibility of the presence of related proteins inother tissues. In this paper we describe the identification andcloning of aprotein froma rat adipocyte cDNAlibrary.Wehavealso determined the sequence for the human homologue. Thesenew proteins are more closely related to centaurin-�1 than tothe centaurin-� proteins and have therefore been calledcentaurin-�2.As centaurin-�2 has two PH domains, each of which shows

some sequence similarities and some differences to the PHdomains in centaurin-�1 and centaurin-� proteins, we havesought to examine the phospholipid binding specificity of eachof these domains by mutation of critical arginine residues(Fukuda andMikoshiba; 1996; Venkateswarlu et al., 1999). Theligand binding data indicate that centaurin-�2 has broaderphosphoinositide specificity than centaurin-�1. Furthermore,centaurin-�2 appears to be constitutivelymembrane associatedand does not redistribute between subcellular locations in amanner that reflects PI(3,4,5)P3 production.

Materials and methods

MaterialsDiC16PI(3,4)P2, diC16PI(4,5)P2 and diC16PI(3,4,5)P3 were synthesised asdescribed (Thum et al., 1996; Chen et al., 1996) or purchased fromEchelon. [3H]BZDC-phosphatidylinositide triesters of PI(3,4)P2 andPI(3,4,5)P3 were generated by incorporating a benzophenone grouplinked through an aminopropyl ester at the P-1 phosphate of the inositolhead group (Dorman and Prestwich, 1994; Prestwich, 1996; Gu andPrestwich, 1996). D-[3H]Inositol 1,3,4,5-tetrakisphosphate (777 GBq/mmol) and D-myo-inositol 1,3,4,5-tetrakisphosphate were from NENLife Science Products and Alexis Biochemicals, respectively.

Cloning of centaurin-�2A DNA fragment of 1218 base pairs containing the majority of thecoding region of centaurin-�1 was assembled into pBluescript. ThisDNA fragment was excised, labelled and used to probe a rat adipocyte�gt 11 library (a gift from C. Londos, NIH, Bethesda) at low stringency.Most of the strongly hybridising plaques were found to contain identicalinserts of approximately one kilobase when sequenced. The sequencecontained an open reading frame coding for 193 amino acids that had58% identity to amino acids 184 ± 376 of centaurin-�1, plus an additional405 nucleotides which correspond to the 3� untranslated region. Thissequence was used to design a primer (5�-CGGGCTGCACGGAGAG-CATTGAAC-3�) that was used in 5� RACE of an available rat liverMarathon Ready cDNA library. A cDNA fragment of 744 nucleotideswas obtained which overlapped by 143 nucleotides identical to theoriginal 3� product from the �gt 11 library. The sequence contained anopen reading frame coding for 248 amino acids, but the start codon wasassignedat theATGat codon18basedon aneural network predictionofstart codons (Pedersen and Nielsen, 1997) and similarity to centaurin-�1. The 5�RACE PCR fragment was blunt-end cloned into the EcoRVsite of the pT7Blue vector (Novagen) and sequenced. In order toconfirm that the sequencewas theproduct of a singlemRNAspecies and

not a chimera of different mRNA species, a rat adipocyte MarathonReady cDNA library was synthesised using the Marathon cDNAAmplificationKit (Clontech). 3�RACEwas performed using the primer5�-CCGGCCATGGGCGACCGTGAACG-3�. This productwas clonedinto pT7Blue and sequenced. A 1.8-kb DNA fragment was generatedcontaining the full-length coding region and a 3� untranslated regionincluding a poly(A) tail. The adipocyte 3�RACE sequencewas identicalto the sequence assembled from the combination of 5�RACE sequencewith the �gt 11 3� sequence.

Expression and mutagenesis of recombinantcentaurin-�2Asix-histidine [His]6 tagwas added to the amino terminus coding regionof the rat adipocyte centaurin-�2 clone in pET15b (Novagen). Thefollowing oligonucleotideswere used to amplify the cDNAencoding thefull-length centaurin-�2 protein and incorporate NdeI andXhoI sites atthe 5� and 3� ends, respectively: 5�-GAGCATATGGGCGACCGTGAACGCAACAAG-3�; 5�-GAGCTCGAGTCAGCCACTCTTTGCTGATGTGGT-3�. NdeI/XhoI-digested PCR products were gel purified(QIAquick Gel Extraction Kit, Qiagen) and ligated into the NdeI/XhoI-digested pET15b. BL21(DE3)LysS cells were transformed withrecombinant plasmid and expression was induced by the addition of0.5 mM IPTG for 2 h. The cells were harvested and resuspended in lysisbuffer (20 mM Tris-HCl, pH 8.0; 100 mM NaCl) and lysed by freezethawing and sonication. Cleared lysate was bound to the Talon MetalAffinityResin (Clontech) andwashedwith 20 volumes lysis buffer and 5volumes 10 mM imidazole in lysis buffer. [His]6-tagged proteins wereeluted with 100 mM imidazole in lysis buffer and the fractions wereimmediately dialysed against lysis buffer. Approximately 2 mg of eachprotein per litre culture were purified and stored in 50% glycerol at�20 �C at a concentration of 200 �g/ml.Pointmutationswere introduced into thePHdomainsof centaurin-�2

in pET15b, using the QuickChange Site-Directed Mutagenesis method(Clontech) and oligonucleotides; R151C, 5�-CAGTTCCTGAGGTGCAGGTTTGTCCTTCTGTCAAGG-3�, R275C, 5�-CCCTTCAAGAAATGCTGGTTTGCCTTGGATCC-3� and their respective complimen-tary partners. Insertion of mutations was verified by sequencing beforeexpression of the recombinant mutant proteins as described above.

Ligand binding assaysThe binding of recombinant centaurin-�2 to liposomes was assayedfollowing the method previously described (Patki et al., 1997). Briefly,liposomes were prepared by mixing equal amounts of phosphatidylcho-line and phosphatidylethanolamine in the presence or absence of 5%phosphoinositide as required. The mixtures were dried under nitrogenand resuspended to a final concentration of 1 mg total phospholipid/mlin a buffer composed of Hepes at pH 7.2 (50 mM), NaCl (100 mM) andEDTA (0.5 mM). Resuspended lipids were sonicated for 5 min in a bathsonicator and collected by centrifugation at 20000g for 10 min at 4 �C.The supernatant was removed, liposomes were resuspended to 100 �g/100 �l per assay tube in the same buffer and re-sonicated for 5 min.Recombinant centaurin-�2 (1 �g)was added to each tube and incubatedat 18 �C for 1 h. The tubes were cooled on ice and centrifuged at 20 000gfor 10 min at 4 �C. The supernatants were removed and the pellets wereresuspended in SDS-PAGE sample buffer. The samples were thenanalysed for the presence of centaurin-�2 by 12% SDS-PAGE andimmunoblotting with an anti-centaurin-�2 polyclonal antibody (seebelow).Binding of inositol (1,3,4,5)tetrakisphosphate (IP4) to recombinant

centaurin-�2 was assayed using a modification to methods previouslydescribed (Cullen et al., 1995; Hanck et al., 1999). In brief, each assaycontained 500 ng recombinant centaurin-�2 in 500 �l binding buffer(25 mM sodium acetate/25 mM potassium phosphate, pH 5.0, 0.05%Brij58, 1 mM EDTA), 1 nM [3H]IP4 and varying concentrations ofunlabelled IP4 or other phosphoinositides as indicated in the figurelegends. The reaction mixtures were incubated on ice for 15 min, afterwhich the receptor-ligand complexes were precipitated by addition of100 �l 5 mg/ml �-globulin and 1 ml 25%polyethylene glycol (PEG), andthe samples were incubated on ice for a further 5 min. The samples were

223Characterisation of centaurin-�2EJCB

centrifuged at 13000g for 10 min, the supernatants were removed andthepelletswerewashedwith 500 �l 25%PEG.The resultingpelletswereresuspended in 100 �l 0.1MNaOH,and radioactivitywasdeterminedbyliquid scintillation counting.

Photoaffinity labellingRecombinant [His]6-tagged centaurin-�2 (3 �g) was incubated with0.1 �Ci [3H]BZDC-PI (3,4)P2 or [3H]BZDC-PI(3,4,5)P3 in 75 �l 20 mMTris-HCl, 100 mM NaCl, pH 8.0 using a modification of protocolsdescribed indetail elsewhere (Chaudhary et al., 1998; Profit et al., 1998).Samples were equilibrated on ice for 10 min in a 96-well plate and UVirradiated at 350 nM for four 5-min periods with 5-min intervals.Photolabelled protein samples were subjected to electrophoresis on10%SDS-polyacrylamide gels andCoomassie stained.The gelswere cutand the slices placed in scintillation vials and dried at 80 �C for 2 h. Theslices were solubilised by the addition of 0.5 ml alkaline hydrogenperoxide (2% (v/v) ammonium hydroxide in 30% hydrogen peroxide)and by heating at 80 �C for 2 h. The samples were cooled andradioactivity was determined by liquid scintillation counting.

Northern blot analysisThe 5� fragment of centaurin-�2was subcloned into themultiple cloningsite of pSPT18 and linearised at the 3� end of the insert using SmaI.Digoxygenin (DIG)-labelled RNA probes were generated using SP6RNApolymerase andDIG-labelledUTP as described in theDIGRNAlabelling Kit (Roche Molecular Biochemicals). The centaurin-�1fragment used to screen the �gt-11 cDNA library and rat �-actincDNAwere also cloned into pSPT18 and used as a template to produceDIG-labelled RNA probes. Poly(A)� RNA was separated on a 2%formaldehydeagarosegel and then transferredontoapositively chargednylon membrane (Roche Molecular Biochemicals). Prehybridizationwas performed at 68 �C in 50% formamide (v/v), 5�SSC (where SCC is150 mMNaCl, 15 mMsodiumcitrate), 2%blocking reagent (w/v), 0.1%N-lauroyl-sarcosine (w/v), 0.02% SDS (w/v) for 1 ± 2 h before beinghybridised with 10 ± 100 ng/ml DIG-labelled RNA at 68 �C overnight.Post-hybridization washes were performed (2�SSC, 0.1%SDS at roomtemperature and 0.1�SSC, 0.1% SDS at 68 �C) before detection usingDIG Luminescent Detection Kit (Roche Molecular Biochemicals) andHyperfilm ECL (Amersham). The RNA loaded in each lane wasadjusted to obtain approximately equal �-actin hybridization signals forall tissues.

Cell fractionation and Western blot analysisTissues were removed frommaleWistar rats (180 ± 220 g) and placed inice-cold HES buffer (20 mM HEPES, 1 mM EDTA, 255 mM sucrose,pH 7.2, with 100 �MAEBSF, and 1 �g/ml antipain, aprotinin, leupeptinandpepstatinA).The tissueswerehomogenised inHESbuffer (5 ml pergm) andwere centrifugedat 1000g for 2 min at 4 �C.The supernatantwascentrifuged at 541000g at 4 �C for 30 min. The cytosol was removed andthe total membrane pellet resuspended in HES buffer using a 25Gneedle. 20 �g of the total membrane protein in this pellet was separatedon a 10% SDS-PAGE gel, transferred to nitrocellulose and analysed byWestern blotting using the anti-centaurin-�2 antibody followed by ananti-rabbit-HRP antibody (Sigma) and detection by ECL (Amersham-Pharmacia).To more closely examine the cell membrane fraction with which

centaurin-�2 associates a well characterised method for obtainingsubcellular fractions from rat adipocytes was employed (Simpson et al.,1983). Purified plasma membrane, high-density microsomes (mainlyendoplasmic reticulum), low-density microsomes (containing endo-somes and light membrane vesicles) and a dense membrane fraction(containing mitochondria, peroxisomes, lysosomes, secretory granules,etc) were prepared using this method but with the inclusion of anadditional differential centrifugation step to remove nuclei.For preparation of centaurin-�2 antibody, a fourteen-amino-acid

peptide (CPTEKEQREWLENL) located towards the C-terminus ofcentaurin-�2 was coupled to ovalbumin and used to immunize rabbits.The antiserum was used to prepare affinity-purified antibody usingcentaurin-�2 peptide coupled toReactigel (Pierce)matrix. This peptide

sequencewas chosen as itwas predicted to be in ahighly antigenic regionof the protein. Some residues are common with centaurin-�1 proteinsbut data described in theResults section indicate that there is unlikely tobe extensive cross-reactivity with these isoforms.

Results

Cloning of rat and human centaurin-�2The 3� end of centaurin-�2 was obtained by firstly screening arat adipocyte cDNA library with a centaurin-�1 probe. Acombination of 5� RACE and 3� RACE were then sequentiallyused to obtain the full-length sequence from liver and adipocyteMarathon Ready cDNA libraries. The complete nucleotide(1798 bp) andpredicted protein sequence (376 amino acids) areshown in Fig. 1A. The protein has a deduced molecular mass of43.5 kDa. Protein sequence analysis using the ProfileScanserver at ISREC revealed the presence of an N-terminal zincfinger motif (residues 21 ± 82) with the consensusCX2CX16CX2C and two PH domains, PH-N (residues 131 ±232) and PH-C (residues 254 ± 360). The protein shows highestsimilarity to rat p42IP4 (59% identity) and other centaurin-�1-related proteins (Fig. 1B, C). However, as these other proteinsshow around 90% identity to each other it appears that theprotein described here is distinct from the centaurin-�1 family.A search of the EMBL databases for nucleotide sequences

showing similarity to rat centaurin-�1 identified several humanESTs which appeared to fall into two groups. The first had agreater than 80% identity to the centaurin-�1 family andapproximately 65% identity to rat centaurin-�2 whereas thesecond group had the converse relationship. One of the ESTsbelonging to the second group (accession no. AA256750) wasobtained from the UK HGMP Resource Centre, Hinxton,Cambridge. The complete sequencing of this EST clonerevealed a 1.8-kb cDNA sequence coding for a protein with83% identity to rat centaurin-�2 (Fig. 1B, C). The humancDNA clone has an in-frame stop codon 5� of the proposedATG start codon, further strengthening our above assignmentof the start codon in rat centaurin-�2.The nucleotide sequencesfor human and rat centaurin-�2 that are reported in this paperhave been deposited at the GenBankTM/EBI Data Bank withaccession numbersAJ238993 andAJ238994. During the courseof this study, the sequence of human centaurin-�2 was alsoidentified elsewhere as one of 11 products encoded by the genesdeleted in apopulation of neurofibromatosis patientswith largechromosomal deletions (Jenne et al., 2000).

Determination of phosphoinositide bindingspecificity of centaurin-�2The binding specificity of recombinant [His]6-centaurin-�2 wasinitially assessed by liposome binding assays (Fig. 2a). Nocentaurin-�2 binding was detected with phosphatidylcholine/phosphatidylethanolamine liposomes alone or those contain-ing either PI or PI(3,5)P2. However, significant centaurin-�2binding occurred with liposomes containing PI(3,4)P2,PI(4,5)P2 and PI(3,4,5)P3.To quantify interactions of the phosphoinositides with

centaurin-�2 and the R151C, R275C and R153C/R275Cmutants, the ability of non-labelled IP4, PI(3,4)P2, PI(4,5)P2and PI(3,4,5)P3, to compete with [3H]IP4 binding was assessed(Fig. 2b). IP4, PI(3,4,5)P3 and PI(4,5)P2 all showed similardisplacement curves and at approximately 150 nM displaced50% of the bound [3H]IP4. However, PI(3,4)P2 only competed

224 P. Whitley, A. M. Gibbard et al. EJCB

Fig. 1. Comparison of the sequence of centaurin-�2 with relatedproteins. (A) Nucleotide and deduced amino acid sequence of ratcentaurin-�2. (B) Sequence alignment of rat and human centaurin-�2(AJ238993/4 respectively) with rat centaurin-� (U51013), rat p42IP4

(AJ007422), human centaurin-�1 (AJ006422), porcine p42IP4 (U88363),and bovine PIP3BP (D89940). (C) Unrooted neighbour joining tree forthe indicated centaurin-� proteins. The bar indicates the sequencedifference ratio.

225Characterisation of centaurin-�2EJCB

with [3H]IP4 binding at higher concentrations, with 940 nMbeing required to give 50% displacement. PI(3,5)P2 had evenpoorer affinity for the IP4 binding site as indicated by its poorability to displace [3H]IP4 (IC50� 1.9 �M, data not shown).Mutating the conserved arginine residue in thePH-Ndomain toa cysteine (R151C) resulted in an almost complete loss ofbinding to [3H]IP4, as did the double mutation R151C/R275C(Fig. 3b). However, when the conserved arginine in the PH-Cdomain was mutated to a cysteine (R275C) the ability to bind[3H]IP4 was retained. The amount of [3H]IP4 bound to theR275C centaurin-�2 with no competitor present was similar tothat of wild-type centaurin-�2 (Fig. 3b). Displacement of[3H]IP4 binding with non-labelled IP4 as a competitor (Fig. 3c)was essentially identical for wild-type and R275C centaurin-�2proteins. These results suggest that IP4 binds preferentially tothe PH-N domain of centaurin-�2 and that PI(3,4,5)P3 andPI(4,5)P2 bind with similar affinity to the PH-N domain. As[3H]IP4 did not bind to the R151C centaurin-�2 mutant, thebinding specificity of the PH-C domain could not be deter-mined using the IP4 binding approach.Toanalyse further thebinding specificities of thePHdomains

of centaurin-�2 photoaffinity labelling with [3H]BZDC-label-led phosphoinositides was performed. UV irradiation causes

the BZDC groups on the phosphoinositide to covalently cross-link to the protein (Theibert et al., 1997; Chaudhary et al., 1998;Prestwich et al., 1999). [3H]BZDC-PI(3,4,5)P3 could be cross-linked to centaurin-�2 and its two single mutants, R151C andR275C, but not the double R151C/R275C mutant (Fig. 4a).Thus, PI(3,4,5)P3 apparently binds to both PH domains of

Fig. 2. Phosphoinositide interaction with recombinant centaurin-�2.In (a) the specificity of interaction with phosphoinositides in liposomeswas determined by Western blot analysis of the levels of centaurin-�2that sedimented with the liposomes. No PI indicates phosphatidylcho-line/phosphatidylethanolamine (1 :1) liposomes without additionalphosphatidylinositol. The indicated phosphoinositides were added at5% (w/w). The data shown is representative of two similar experiments.In (b) phosphoinositide interaction with recombinant wild-type cen-taurin-�2 was determined from the extent of competition for binding of[3H]IP4 by non-labelled ligands. The approximate IC50 values calculatedfrom these data are: IP4� 55 nM (�), PI(3,4,5)P3� 163 nM (�),PI(4,5)P2� 154 nM (�) and PI(3,4)P2� 942 nM (�). Data are mean-s� S.E.M. of values from at least three experiments.

Fig. 3. Displacement of IP4 from recombinant centaurin-�2 proteinsmutated at PH-N and PH-C. (a) SDS-PAGE analysis of purified [His]6-tagged recombinant centaurin-�2. Each lane contained approximately3 �g protein and has been detected by Coomassie Brilliant Bluestaining. (b) Comparison of total [3H]IP4 binding to the wild-type (WT),the two single mutants (R151C and R275C) and the double mutant(R151C/R275C) centaurin-�2 proteins (expressed as a percentage ofthe IP4 binding to wild-type). Data are mean� S.E.M from fourexperiments. (c) Comparison of the displacement of [3H]IP4 from wild-type (�) and from the PH-C mutant (R275C) (�) by non-labelled IP4.Mean data are from two experiments.

226 P. Whitley, A. M. Gibbard et al. EJCB

centaurin-�2. Interestingly, [3H]BZDC-PI(3,4)P2 could becross-linked to centaurin-�2 and the R151C mutant, but notthe R275C or R151C/R275C mutants (Fig. 4b). These datasuggest that PI(3,4)P2 has low affinity for the PH-N domain(consistentwith its poor ability to displace IP4) but bindswell tothe PH-C domain. In order to further characterise the separatespecificities of the tandem PH domains, attempts to expressPH-N and PH-C as separate GST-fusion protein constructsweremade.However, these constructs persistently precipitatedfrom solution and were unsuitable for use in ligand-bindingassays.

Tissue distribution and subcellular localizationof centaurin-�2Northern blot analysis of multiple tissue mRNAs, normalisedfor �-actin (Fig. 5c), detected the presence of two mRNAtranscripts of 4.0 kb (major) and 5.2 kb (minor) when probedwith an 830-bp DIG-labelled RNA corresponding to the 5� endof centaurin-�2 (Fig. 5a). These transcripts appear to beexpressed in all tissues examined. The mRNA detected inNorthern blots was considerably larger than the cloned cDNA,possibly because of a large amount of untranslated 5� sequence.When a DIG-labelled centaurin-�1 probe was used, twotranscripts of 2.8 kb and 4.0 kb were detected (Fig. 5b), mainlyin brain as has been previously reported (Hammonds-Odieet al., 1996). Affinity-purified centaurin-�2 antibody recog-nised a single protein of the expected apparent molecularweight in the total particulate fractions of adipocytes, heart,skeletal muscle, and brain (Fig. 5d). This protein was alsodetected in membrane fractions of kidney but only low levelswere found in blots of liver and lung membrane fractions (data

Fig. 4. PI(3,4,5)P3 and PI(3,4)P3 photolabels bind differently at cen-taurin-�2 PH-N and PH-C domains. Comparison of photolabel ligandbinding of (a) [3H]BZDC-PI(3,4,5)P3 and (b) [3H]BZDC-PI(3,4)P2 towild-type centaurin-�2 (WT), the PH-N mutant (R151C), the PH-Cmutant (R275C) and the double mutant (R151C/R275C). Data(expressed as a percentage of the total ligand binding to wild-typecentaurin-�2) are means� S.E.M. of values from at least threeexperiments.

Fig. 5. Tissue distribution of centaurin-�2. Amultiple tissue Northernblot was prepared using poly(A)� RNA from the indicated rat tissues.The quantity of poly(A)� RNA was adjusted to obtain approximatelyequal �-actin hybridization signals for all tissues (2.0 kb). Note that the�-actin probe also cross-hybridizes to muscle-type actin (1.8 kb). Inpanels (a) centaurin-�2, (b) centaurin-�1, and (c) �-actin in adiposetissue, brain, heart and skeletal muscle were examined. In (d), 20 �gprotein of total membrane (M) and cytosolic (C) fractions from adiposetissue, brain, heart and skeletal muscle were subjected to SDS-PAGE,blotted and detected using affinity-purified anti-centaurin-�2 antibody.The signal from wild-type recombinant protein (Rec) is also shown.Results are representative of several similar experiments.

227Characterisation of centaurin-�2EJCB

not shown).Centaurin-�2was not detectable in the cytoplasmicfractions of any of these tissues. The low level of immuno-reactive signal in brain (where centaurin-�1 occurs at highlevels (Fig. 5b) and the lack of any immunoreactive signal in thecytoplasmic fraction (where centaurin-�1 is thought to bepresent) suggests that the antibody we have used does notextensively cross-react with centaurin-�1 proteins (althoughthis possibility has not beendirectly tested and cannot be totallyexcluded).Inorder to localise the centaurin-�2 further, subfractionation

of rat adipocytemembraneswas performed (Fig. 6). Centaurin-�2was localised exclusively in the dense granule/mitochondrialmembrane fraction (DG/M; Fig. 6a, upper panel). Further-more, a comparison of membrane fractions from basal andinsulin-stimulated cells indicated that centaurin-�2 was nottranslocated between fractions in response to insulin signalling,PI 3-kinase activation and PI(3,4,5)P3 generation. By contrast,in response to insulin the levels of the glucose transporterprotein GLUT4 in the low-density microsome fraction (LDM)were reduced with a corresponding increase in plasma mem-brane (PM) GLUT4 (Fig. 6a, lower panel) confirming that thecells used were capable of protein translocation.

Wenext examined the type of interaction between centaurin-�2 and the DG/M fraction by treating these membranes with10 �M IP4, 1 M NaCl, or 1% Triton X-100 (Fig. 6b). 10 �M IP4was�50-times the concentration needed to displace 50%of the[3H]IP4 in the binding experiments but did not dissociate anycentaurin-�2 from this fraction. Similarly, a high-salt wash (acondition known to dissociate porcine p42IP4; (Stricker et al.,1999)) was unable to displace centaurin-�2. However, treat-ment of this fraction with 1% Triton X-100 (to solubilise themembranes present) completely removed centaurin-�2 fromthe pellet and led to its recovery in the supernatant. Theseresults suggest that centaurin-�2 is more tightly membraneassociated than centaurin-�1 proteins. In addition, weakeningthe interaction of membranes with the centaurin-�2 PH-Ndomain selectively (by competition with IP4) is insufficient todisplace this protein.

Discussion

The cloning of the cDNA encoding the protein centaurin-�2adds a newmember to the centaurin-� family of proteins, whichhave a rather complicated nomenclature. PIP3BP (Tanakaet al., 1997), human centaurin-�1 (Venkateswarlu et al., 1999)and pig and rat p42IP4 (Stricker et al., 1997; Aggensteiner et al.,1998) all have 90% or greater identity, whereas rat centaurin-�(Hammonds-Odie et al., 1996) is 45 amino acids longer thanthese other family members and differs between amino acids139 ± 161. These differences in the rat centaurin-� sequencemakes it distinct from other family members although it stillshows 75% or greater identity to the shorter sequences.Centaurin-�2 is only 59% identical to the nearest centaurin-

�1 family member. We have also obtained the sequence of ahuman centaurin-�2 that shares 83% identity with rat centaur-in-�2. The sequence differences suggest that the centaurin-�2proteins are distinct from the centaurin-�1 proteins. This isfurther strengthened by the data on the tissue distribution asdetermined by Northern and Western blot analysis. Whereasrat centaurin-�1 is almost exclusively expressed in brain,centaurin-�2 has a more widespread tissue distribution. Wes-tern blot analysis suggests that the expression of the protein ishighest in the insulin-responsive tissues of adipocytes, heart andskeletal muscle but the significance of this finding in terms ofinsulin action is unknown.PH domainmutation experiments indicate that centaurin-�2

only binds [3H]IP4 with high affinity at the PH-N domain.Competitive displacement of [3H]IP4 revealed that PI(3,4,5)P3and PI(4,5)P2 have similar binding affinities for PH-N as�150 nM of each was required to displace 50% of the [3H]IP4.The concentrations of PI(3,4)P2 (940 nM) and PI(3,5)P2(1.9 �M) required to competitively displace 50% of the[3H]IP4 were much greater. The described photolabellingstudies showed that PI(3,4,5)P3 could cross-link to PH-N andPH-C, although the extent of cross-linking to PH-C was lessthan to PH-N, possibly reflecting a higher affinity for PH-N.This is consistent with the higher binding of IP4 to this domain.By contrast, PI(3,4)P2 photolabel could be cross-linked effec-tively to PH-C but not to PH-N. This is consistent with the poorability of this ligand to diplace IP4 fromPH-N.Thus, our bindingdata are consistent with a model in which both of the PHdomains contribute to phosphoinositide binding, but withdifferent specificities.

Fig. 6. Centaurin-�2 is present in a dense subcellular membranefraction of rat adipocytes. Subcellular membrane fractions wereobtained from rat adipocytes either in the basal state or followingtreatment with insulin. Low-density microsome (LDM), high-densitymicrosome (HDM), plasmamembrane (PM), dense-granule/mitochon-drial membranes (DG/M) or cytosol fractions were analysed by SDS-PAGE.Western blot analysis was carried out with either anti-centaurin-�2 antibody (a, upper panel) or anti-GLUT4 (a, lower panel) polyclonalantibodies. In (b), the DG/M fraction was treated with IP4, NaCl orTriton X-100 and the distribution between the pellet and supernatantwas analysed by Western blotting using the anti-centaurin-�2 antibody.Results are representative of several similar experiments.

228 P. Whitley, A. M. Gibbard et al. EJCB

Centaurin-�2 did not show specificity for PI(3,4,5)P3 overPI(4,5)P2 in either liposome binding assays or IP4 bindingassays. This was unexpected since both its PH domains closelyresemble type I PH domains (Rameh et al., 1997; Kavran et al.,1998), and centaurin-�1 proteins all show selectivity forPI(3,4,5)P3 over PI(4,5)P2. It should be noted, however, thatthe PH domain of ASAP1 (centaurin-�4) also most closelyresembles a class 1 PH domain, but binds PI(4,5)P2 andPI(3,4,5)P3 equally well and is actually activated by PI(4,5)P2in preference to PI(3,4,5)P3 (Kam et al., 2000). Furthermore,other proteins containing PH domains closely resembling theputative PI(3,4,5)P3-binding motif Lys-Xaa-Sma-Xaa6 ± 11-Arg/Lys-Xaa-Arg-Hyd-Hyd (where Xaa is any amino acid,Sma is a small amino acid, and Hyd is a hydrophobic aminoacid) (Isakoff et al., 1998) do not necessarily specificallyrecognise this phosphoinositide (Dowler et al., 1999). Themolecular basis for the differences between the bindingspecificities of the centaurin-�1 and centaurin-�2 proteins isnot yet clear, but there are a number of positively chargedresidues present in thePHdomains of rat andhuman centaurin-�2 that are not present in those of centaurin-�1 proteins, e.g.PH-N residues R150 and K236 and PH-C residues R269, R321and K326.The subcellular distribution of centaurin-�2 also differs

considerably from the centaurin-�1 proteins. Whereas thecentaurin-�1 is distributed between both soluble and mem-brane-bound pools and is recruited to membranes uponstimulation of PI(3,4,5)P3 production (Venkateswarlu et al.,1999; Tanaka et al., 1999; Venkateswarlu and Cullen; 1999),centaurin-�2 is found exclusively in themembrane pellets of alltissues tested. Subcellular fractionation of rat adipocytesrevealed that centaurin-�2 is localised to a dense membranefraction. Further resolution of the type of dense membraneparticle with which centaurin-�2 associates will probablyrequire immunocytochemistry at the electronmicroscope level.However, rat adipocytes are generally considered unsuitablefor such studies because of their very large fat deposits.The absence of centaurin-�2 in the cytosol suggests that it is

constitutively associated with membranes. A constitutivemembrane association could be due to the high affinity of thePH-N domain of centaurin-�2 for PI(4,5)P2. This property isnot shared by the centaurin-�1 proteins and therefore thisdifference in specificity may account for the lack of a dynamicredistribution of this protein similar to that observed forcentaurin-�1. However, we cannot rule out additional molec-ular interactions, for example, with other proteins, that conferthe tight membrane association. Although a function for thecentaurin-� proteins remains elusive, one approach that can betaken in order to gain insight into their possible mechanisms ofaction is to identify protein partners. Casein kinase 1 (CK1) hasrecently been identified as binding to centaurin-�1 (Duboiset al., 2001). It remains to be seen whether CK1 interacts withcentaurin-�2, or whether centaurin-�2 has other proteinpartners that will give an indication of function.In conclusion, we have characterised rat centaurin-�2 which

shows awider tissue distribution than the centaurin-�1 proteinspreviously described. The ligand binding and subcellularfractionation properties of centaurin-�2 suggest that thisprotein is constitutively membrane associated.

Acknowledgements. We thank Drs. J. Chen, C. Ferguson, G. Dormanand J. Peng (University ofUtah) for synthesis of photoaffinity analoguesand Chris Boot (AstraZeneca) for help in raising the antiserum to

centaurin-�2. This work is supported by theMedical Research Council,UK and Diabetes UK (to G. D. Holman) and the National Institutes ofHealth (to G. D. Prestwich, grant NS29632).

References

Aggensteiner, M., Stricker, R., Reiser, G. (1998): Identification of ratbrain p42(IP4), a high-affinity inositol(1,3,4,5)tetrakisphosphate/phosphatidylinositol(3,4,5)trisphosphate binding protein. Biochim.Biophys. Acta 1387, 117 ± 128.

Chaudhary, A., Gu, Q. M., Thum, O., Profit, A. A., Qi, Y., Jeyakumar,L., Fleischer, S., Prestwich, G. D. (1998): Specific interaction ofGolgicoatomer protein �-COP with phosphatidylinositol 3,4,5-trisphos-phate. J. Biol. Chem. 273, 8344 ± 8350.

Chen, J., Profit, A. A., Prestwich, G. D. (1996): Synthesis of photoacti-vatable 1,2-O-diacyl-sn-glycerol derivatives of 1-L-phosphatidyl-D-myo-inositol 4,5-bisphosphate (PtdInsP(2)) and 3,4,5-trisphosphate(PtdInsP(3)). J. Org. Chem. 61, 6305 ± 6312.

Cullen, P. J., Dawson, A. P., Irvine, R. F. (1995): Purification andcharacterization of an Ins(1,3,4,5)P4 binding protein from pigplatelets: possible identification of a novel non-neuronalIns(1,3,4,5)P4 receptor. Biochem. J. 305, 139 ± 143.

Dorman, G., Prestwich, G. D. (1994): Benzophenone photophores inbiochemistry. Biochemistry 33, 5661 ± 5673.

Dowler, S., Currie, R. A., Downes, C. P., Alessi, D. R. (1999): DAPP1: adual adaptor for phosphotyrosine and 3-phosphoinositides. Biochem.J. 342, 7 ± 12.

Dubois, T., Kerai, P., Zemlickova, E., Howell, S., Jackson, T. R.,Venkateswarlu, K., Cullen, P. J., Theibert, A. B., Larose, L., Roach,P. J., Aitken, A. (2001): Casein kinase I associates with members ofthe centaurin-� family of phosphatidylinositol-(3,4,5)-trisphosphate-binding proteins. J. Biol. Chem. 276, 18757 ± 18764.

Fukuda, M., Mikoshiba, K. (1996): Structure-function relationships ofthe mouse Gap1m. Determination of the inositol 1,3,4,5-tetrakis-phosphate-binding domain. J. Biol. Chem. 271, 18838 ± 18842.

Gu, Q.-M., Prestwich, G. D. (1996): Synthesis of phosphotriesteranalogues of phosphoinositides PtdIns(4,5)P2 and PtdIns(3,4,5)P3.J. Org. Chem. 61, 8642 ± 8647.

Hammonds-Odie, L. P., Jackson, T. R., Profit, A. A., Blader, I. J., Turck,C. W., Prestwich, G. D., Theibert, A. B. (1996): Identification andcloning of centaurin-�. A novel phosphatidylinositol 3,4,5-trisphos-phate-binding protein from rat brain. J. Biol. Chem. 271, 18859 ±18868.

Hanck, T., Stricker, R., Krishna,U. M., Falck, J. R., Chang, Y. T., Chung,S. K., Reiser, G. (1999): Recombinant p42IP4, a brain-specific 42-kDa high-affinity Ins(1,3,4,5)P4 receptor protein, specifically inter-acts with lipid membranes containing Ptd-Ins(3,4,5)P3. Eur. J.Biochem. 261, 577 ± 584.

Isakoff, S. J., Cardozo, T., Andreev, J., Li, Z., Ferguson, K. M., Abagyan,R., Lemmon,M. A., Aronheim, A., Skolnik, E. Y. (1998): Identifica-tion and analysis of PH domain-containing targets of phosphatidyl-inositol 3-kinase using a novel in vivo assay in yeast. EMBO J. 17,5374 ± 5387.

Jenne, D. E., Tinschert, S., Stegmann, E., Reimann, H., Nurnberg, P.,Horn, D., Naumann, I., Buske, A., Thiel, G. (2000): A common set ofat least 11 functional genes is lost in themajority of NF1 patients withgross deletions. Genomics 66, 93 ± 97.

Kam, J. L.,Miura, K., Jackson, T. R., Gruschus, J., Roller, P., Stauffer, S.,Clark, J., Aneja, R., Randazzo, P. A. (2000): Phosphoinositide-dependent activation of the ADP-ribosylation factor GTPase-activating protein ASAP1. Evidence for the pleckstrin homologydomain functioning as an allosteric site. J. Biol. Chem. 275, 9653 ±9663.

Kavran, J. M., Klein, D. E., Lee, A., Falasca, M., Isakoff, S. J., Skolnik,E. Y., Lemmon, M. A. (1998): Specificity and promiscuity in phos-phoinositide binding by pleckstrin homology domains. J. Biol. Chem.273, 30497 ± 30508.

229Characterisation of centaurin-�2EJCB

Patki, V., Virbasius, J., Lane, W. S., Toh, B. H., Shpetner, H. S., Corvera,S. (1997): Identification of an early endosomal protein regulated byphosphatidylinositol 3-kinase. Proc. Natl. Acad. Sci. USA 94, 7326 ±7330.

Pedersen, A. G., Nielsen, H. (1997): Neural network prediction oftranslation initiation sites in eukaryotes: perspectives for EST andgenome analysis. Proc. Int. Conf. Intell. Syst. Mol. Biol. 5, 226 ± 233.

Poon, P. P., Cassel, D., Spang, A., Rotman, M., Pick, E., Singer, R. A.,Johnston, G. C. (1999): Retrograde transport from the yeast Golgi ismediated by two ARF GAP proteins with overlapping function.EMBO J. 18, 555 ± 564.

Prestwich, G. D. (1996): Touching all the bases: Inositol polyphosphateand phosphoinositide affinity probes from glucose. Acc. Chem. Res.29, 503 ± 513.

Prestwich, G. D., Chaudhary, A., Chen, J., Feng, B., Mehrotra, B., Peng,J. (1999): Probing Phosphoinositide Polyphosphate Binding toProteins. In: K. S. Bruzik (ed.): Phosphoinositides: Chemistry,Biochemistry and Biomedical Applications. American ChemicalSociety, Washington, DC, pp. 24 ± 37.

Profit, A. A., Chen, J., Gu, Q. M., Chaudhary, A., Prasad, K., Lafer,E. M., Prestwich, G. D. (1998): Probing the phosphoinositide bindingsite of the clathrin assembly protein AP-2 with photoaffinity labels.Arch. Biochem. Biophys. 357, 85 ± 94.

Rameh, L. E., Arvidsson, A., Carraway, K. L., III, Couvillon, A. D.,Rathbun, G., Crompton, A., VanRenterghem, B., Czech, M. P.,Ravichandran,K. S., Burakoff, S. J.,Wang,D. S., Chen,C. S., Cantley,L. C. (1997): A comparative analysis of the phosphoinositide bindingspecificity of pleckstrin homology domains. J. Biol. Chem. 272,22059 ± 22066.

Randazzo, P. A., Andrade, J., Miura, K., Brown, M. T., Long, Y. Q.,Stauffer, S., Roller, P., Cooper, J. A. (2000): From the cover: The ArfGTPase-activating protein ASAP1 regulates the actin cytoskeleton.Proc. Natl. Acad. Sci. USA 97, 4011 ± 4016.

Randazzo, P. A., Miura, K., Nie, Z., Orr, A., Theibert, A. B., Kearns,B. G. (2001): Cytohesins and centaurins: mediators of PI 3-kinaseregulated Arf signaling. Trends Biochem. Sci. 26, 220 ± 221.

Simpson, I. A., Yver, D. R., Hissin, P. J., Wardzala, L. J., Karnieli, E.,Salans, L. B., Cushman, S. W. (1983): Insulin-stimulated translocationof glucose transporters in the isolated rat adipose cells. Character-

ization of subcellular fractions. Biochim. Biophys. Acta 763, 393 ±407.

Stricker, R., Adelt, S., Vogel, G., Reiser, G. (1999): Translocationbetweenmembranes and cytosol of p42IP4, a specific inositol 1,3,4,5-tetrakisphosphate/phosphatidylinositol 3,4,5-trisphosphate-receptorprotein from brain, is induced by inositol 1,3,4,5-tetrakisphosphateand regulated by a membrane-associated 5-phosphatase. Eur. J.Biochem. 265, 815 ± 824.

Stricker, R.,Hulser, E., Fischer, J., Jarchau, T.,Walter, U., Lottspeich, F.,Reiser, G. (1997): cDNA cloning of porcine p42IP4, a membrane-associated and cytosolic 42 kDa inositol(1,3,4,5)tetrakisphosphatereceptor from pig brain with similarly high affinity for phosphatidyl-inositol (3,4,5)P3. FEBS Lett. 405, 229 ± 236.

Tanaka, K., Horiguchi, K., Yoshida, T., Takeda, M., Fujisawa, H.,Takeuchi, K., Umeda, M., Kato, S., Ihara, S., Nagata, S., Fukui, Y.(1999): Evidence that a phosphatidylinositol 3,4,5-trisphosphate-binding protein can function in nucleus. J. Biol. Chem. 274, 3919 ±3922.

Tanaka, K., Imajoh-Ohmi, S., Sawada, T., Shirai, R., Hashimoto, Y.,Iwasaki, S., Kaibuchi, K., Kanaho, Y., Shirai, T., Terada, Y., Kimura,K., Nagata, S., Fukui, Y. (1997): A target of phosphatidylinositol3,4,5-trisphosphate with a zinc finger motif similar to that of theADP-ribosylation-factor GTPase-activating protein and two pleck-strin homology domains. Eur. J. Biochem. 245, 512 ± 519.

Theibert, A. B., Prestwich, G. D., Jackson, T. R., Hammonds-Odie, L. P.(1997): The Purification and Assay of Inositide Binding Proteins. In:S. B. Shears (ed.): Signalling by Inositol Lipids and Inositol Phos-phates. Oxford University Press, Oxford, UK, pp. 117 ± 150.

Thum, O., Chen, J., Prestwich, G. D. (1996): Synthesis of a photoaffinityanalogue of phosphatidylinositol 3,4-bisphosphate, an effector in thephosphoinositide 3-kinase signaling pathway. Tetrahedron Lett. 37,9017 ± 9020.

Venkateswarlu, K., Cullen, P. J. (1999): Molecular cloning and func-tional characterization of a human homologue of centaurin-�.Biochem. Biophys. Res. Commun. 262, 237 ± 244.

Venkateswarlu,K.,Oatey, P. B.,Tavare, J. M., Jackson,T. R.,Cullen, P. J.(1999): Identification of centaurin-�1 as a potential in vivo phospha-tidylinositol 3,4,5-trisphosphate-binding protein that is functionallyhomologous to the yeast ADP-ribosylation factor (ARF) GTPase-activating protein, Gcs1. Biochem. J. 340, 359 ± 363.

230 P. Whitley, A. M. Gibbard et al. EJCB