Structure-function relationships of the mouse Gap1m. Determination of the inositol 1,3,4,5-tetrakisphosphate-binding domain.

Gap1IP4BP, one of a member of Ras GTPase-activating proteins, has been identified as a specific inositol 1,3,4,5-tetrakisphosphate (IP4)-binding protein (Cullen, P. J., Hsuan, J. J., Truong, O., Letcher, A. J., Jackson, T. R., Dawson, A. P., and Irvine, R. F. (1995) Nature 386, 527-530). In this paper we describe Gap1m, which is closely related to Gap1IP4BP, to also be an IP4-binding protein and show that the pleckstrin homology domain (PH) is the central IP4-binding domain by expressing fragments of the mouse Gap1m in Escherichia coli as fusion proteins and examining their activities. However, in addition to the PH domain, an adjacent GAP-related domain and carboxyl terminus are required for high affinity specific IP4 binding. The PH domain is highly conserved in the Gap1 family and also has striking homology to the amino-terminal region of Bruton's tyrosine kinase. Substitution of Cys for Arg at position 628 in the PH domain corresponding to the mutation of Bruton's tyrosine kinase observed in X-linked immunodeficiency mice results in a dramatic reduction of IP4 binding activity as well as phospholipid binding capacity of Gap1m. This mutant also showed the GAP activity against Ha-Ras to be similar to that of the wild type Gap1m. Our results suggest that the PH domain of Gap1m functions as a modulatory domain of GAP activity by binding IP4 and phospholipids.

Inositol 1,4,5-trisphosphate (IP 3 ) 1 is one of the well characterized second messengers that trigger IP 3 receptor-mediated Ca 2ϩ release from intracellular pools (1,2). IP 3 is rapidly metabolized by various phosphatases and kinases. Among the metabolites of IP 3 , inositol 1,3,4,5-tetrakisphosphate (IP 4 ), produced by the specific IP 3 -3 kinase (3), has been suggested to have a physiological function potentially as a second messenger regulating intracellular Ca 2ϩ concentration (4 -11). However, the actual function of IP 4 is still a matter of debate because the specific receptor protein involving intracellular Ca 2ϩ homeostasis has not been identified.
Recently Cullen et al. (12) established that Gap1 IP4BP , one member of a family of Ras GTPase-activating proteins (GAP), is a specific IP 4 -binding protein and showed that IP 4 specifically restored the Ras GAP activity after inhibition by phospholipids in vitro. To determine the physiological function of IP 4 binding to Gap1 IP4BP in vivo, it is necessary to determine its binding site. A knowledge of the binding site allows the effects of specific antibodies against or peptides from the binding domain in functioning cells. Using this strategy, the roles of inositol high polyphosphates, including IP 4 , inositol 1,3,4,5,6pentakisphosphate (IP 5 ), and inositol 1,2,3,4,5,6-hexakisphosphate (IP 6 ) have been determined in neurotransmitter release (13)(14)(15)(16)(17)(18)(19). We have determined that the C2B domain of synaptotagmin II is an inositol high polyphosphate-binding domain by structure-function analysis (13). The two C2 domains of synaptotagmin were shown to have different functions in synaptic vesicle trafficking by injecting the domain-selective antibodies against C2A or C2B domain into the squid giant presynapse (15,17), superior cervical ganglion cells (18), and chromaffin cells (19). Specifically, these results indicated that the C2A domain functions as a Ca 2ϩ sensor in exocytosis, the inositol high polyphosphate blocks synaptic transmission by binding to the C2B domain, and the C2B domain is also involved in endocytosis.
Gap1 m was originally identified as a mammalian homologue of Drosophila Gap1 (20,21), and at least two forms of Gap1 have been shown to be present in mammals (Gap1 m , see Ref. 21; and Gap1 IP4BP /GapIII, see Refs. 12 and 22). Both proteins are almost identical in size and share the same structural features, two C2 domains homologous to the C2 regulatory region of protein kinase C, a GAP-related domain (GRD), and a pleckstrin homology (PH) domain (12,20,21). The structures of these domains are highly conserved in the Gap1 family from Drosophila to human. However, the functional properties of the C2 domains and the PH domain have not been elucidated.
In the present study, we report studies on bacterially expressed fusion proteins of the mouse Gap1 m and show that the central IP 4 -binding domain is the PH domain, not the C2 domains. The PH domain is an approximately 100-residue protein module that is found in many proteins involved in signal transduction (23)(24)(25). Although these domains are divergent, PH domains of the Gap1 family and Bruton's tyrosine kinase (Btk), a cytoplasmic tyrosine kinase, are highly conserved (21,26). Btk is crucial for B cell development and a mutation in the PH domain of Btk in the X-linked immunodeficiency (Xid) mice interferes with normal B cell signaling (26). We introduced a mutation into the PH domain of Gap1 m similar to that of Btk in Xid mice and found that the mutant Gap1 m had a dramatically reduced IP 4 binding activity but retained normal Ras GAP activity. On the basis of these results, we discuss this novel function of the PH domain of Gap1 m and Btk in signal transduction.

MATERIALS AND METHODS
Chemicals-Phosphatidylserine (PS), phosphatidylinositol (PI), phosphatidylcholine (PC), and phosphatidylethanolamine (PE) were obtained from Sigma. IP 4 and IP 3 were purchased from Boehringer Mannheim. IP 5 , IP 6 , and Ha-Ras were from Calbiochem. All other chemicals were commercial products of reagent grade. Solutions were prepared in deionized water.

Measurement of [ 3 H]IP 4 Binding to GST Fusion
Proteins-GST fusion proteins (5 g) were incubated with 9.6 nM [ 3 H]IP 4 (DuPont NEN) in 50 l of 50 mM HEPES-KOH, pH 7.2, for 10 min at 4°C. The sample was then mixed with 1 l of 50 mg/ml ␥-globulin and 51 l of a solution containing 30% polyethylene glycol 6000 and 50 mM HEPES-KOH, pH 7.2, and placed on ice for 5 min. The precipitate obtained by centrifugation at 10,000 ϫ g for 5 min was solubilized in 500 l of Solvable (DuPont NEN) and radioactivity was measured in Aquasol 2 (DuPont NEN) with a liquid scintillation counter (13). Inhibition of specific  Table I). The data are means Ϯ S.E. of three measurements normalized to 100% for binding to GST-E or GST-F in the absence of cold inositol phosphates.
[ 3 H]IP 4 binding to GST fusion proteins was also performed in the above reaction mixture containing various inositol phosphates.
Phospholipid Binding Assay-Four kinds of liposomes (PS, PI, PC, and PE) were prepared in 50 mM HEPES-KOH, pH 7.2, and 100 mM NaCl by sonication, collected by centrifugation, and equilibrated with 50 mM HEPES-KOH, pH 7.2, and 100 mM NaCl in the presence of 2 mM EGTA or 0.5 mM Ca 2ϩ . GST fusion proteins (5-10 g) were incubated with liposomes corresponding to 160 g of phospholipid for 15 min at room temperature (27). After centrifugation at 12,000 ϫ g for 10 min at room temperature, the phospholipid pellets were washed in 500 l of the above equilibration buffer and then extracted with 300 l of acetone at Ϫ20°C for 30 min to remove excess lipid. The pellets obtained by centrifugation at 12,000 ϫ g for 15 min at 4°C were dissolved in SDS sample buffer. The proteins in the supernatants were precipitated by adding an equal volume of 20% trichloroacetic acid. After a 15-min incubation on ice, the samples were centrifuged at 12,000 ϫ g for 15 min at 4°C, and the precipitates were mixed with SDS sample buffer. Equal proportions of the supernatants and pellets were analyzed by 10% SDS-polyacrylamide gel electrophoresis followed by Coomassie Brilliant Blue R-250 staining. The protein concentrations were determined by the Bio-Rad protein assay kit using bovine serum albumin as a reference.
Ras GAP Assay-GAP assays were performed according to Calés et al. (28)  KOH, pH 7.2, 5 mM MgCl 2 , and 5 mM dithiothreitol in the presence or absence of 53 M PS (and IP 4 ) for 10 min at 37°C. After addition of 5 l of 10 mM EDTA and 0.5% SDS, the samples were incubated for 5 min at 70°C to release bound nucleotide from Ha-Ras. After centrifugation, the supernatant (0.5 l) was loaded onto a TLC plate (PEI-cellulose, Merck) and chromatographed in 0.5 M lithium chloride and 1 M formic acid. The amounts of GTP and GDP were determined by a Fuji Bioimage Analyzer (Fuji Photo Film Co.) using an imaging plate.

Determination of the IP 4 -binding Domain of the Mouse
Gap1 m -To delineate the IP 4 -binding domain of Gap1 m , we produced seven GST fusion proteins, which almost covered the entire Gap1 m sequence (amino acid residues 22-847) (Fig. 1A). Although the C2B domain of Gap1 m was expected to be an IP 4 -binding domain like that of synaptotagmin II (12,13), the IP 4 binding activity was mapped to the PH domain as shown in Fig. 1B. However, competition experiments indicated that the PH domain alone is not sufficient for high affinity IP 4 binding and inositol phosphate binding specificity (Fig. 2). In the case of GST-E, which contained only the PH domain, the order of competitive potencies was IP 6 Ͼ IP 5 Ͼ IP 4 Ͼ IP 3 , and the apparent dissociation constant (K d ) for IP 4 was 350 nM ( Fig.  2A). Thus the PH domain alone behaves as an inositol high polyphosphate-binding domain like the C2B domain of synaptotagmin II (13). In contrast, GST-F, which contained the GRD, PH domain, and C (carboxyl) terminus, bound IP 4 with high affinity (K d ϭ 40 nM) and showed an inositol phosphate binding specificity that was similar to that of the purified porcine platelet Gap1 IP4BP (IP 4 Ͼ IP 5 Ն IP 6 Ͼ IP 3 ; see Refs. 29 and 30). Table I summarizes the inositol phosphate binding properties of GST-C to -G. With the exception of GST-F, all deletion mutants showed low affinity IP 4 binding (K d ϭ 300 -500 nM) and nonspecific inositol phosphate binding. These results indicated that high affinity IP 4 binding to Gap1 m requires at least the GRD, PH domain, and adjacent C terminus.
Mutational Analysis of the PH Domain of Gap1 m -PH domains were found in many proteins involved in signal transduction and showed relatively low similarities (23,24). However, between the Gap1 family and Btk, these domains are highly conserved beyond species difference as shown in Fig. 3A.

binding properties of deletion mutants of Gap1 m
The inositol phosphate binding specificity of GST-C to -G was analyzed by competition experiments as shown in Fig. 2 The N-terminal region of the PH domain contained a cluster of conserved positively charged amino acids (KrxqgR/KKR/ KxgxknFKkR; subdomain II). These basic residues between subdomains I and II (Fig. 3A) are likely to be essential for IP 4 binding to Gap1 m by analogy with the IP 3 binding to ␤-spectrin (31) and phospholipase C-␦1 (32,33). To test whether this region is important for the binding of negatively charged IP 4 , we introduced mutations into this region and analyzed the mutant GST-F in IP 4 binding assays (Fig. 3B). GST-F (R628C) carried a conversion of Arg into Cys at position 628, which corresponded to the mutation of Btk in the Xid mice (Ref. 26; Fig. 3A, asterisk). GST-F (K626Q,K627Q,R628Q) had triple mutations of serial positive amino acids. IP 4 binding activity of GST-F (R628C) was reduced to about 12% of that of GST-F, and GST-F (K626Q,K627Q,R628Q) showed no significant IP 4 binding (Fig. 3C). These results indicated that Arg at position 628 is one of the essential residues for IP 4 binding to Gap1 m . Analysis of Phospholipid-binding Domain of Gap1 m -Maekawa et al. (34) reported that the Gap1 m could be isolated from both the cytoplasmic and particulate fractions of rat brain. To determine the membrane associated domain of Gap1 m , we first examined the phospholipid binding capacity of two C2 domains (C2A ϩ C2B) and PH domain (Fig. 4, top left  and right). GST-A mostly bound PS liposomes in a Ca 2ϩ -dependent manner like the C2A domain of synaptotagmins (27, 35, 36) but did not bind PI, PC, and PE liposomes. Since the C2A domain of Gap1 m (amino acids 22-114) showed Ca 2ϩ / phospholipid binding activity (data not shown) but the C2B domain did not (Fig. 4, bottom left), the C2A domain alone is sufficient for Ca 2ϩ /phospholipid binding. In contrast, most of the GST-G bound PS liposomes and interacted weakly with PI liposomes irrespective of the presence of Ca 2ϩ . The phospholipid binding properties of the mutant PH domain, GST-G (R628C) and GST-F (K626Q,K627Q,R628Q), were also tested. Both mutants showed a reduced phospholipid binding capacity (Fig. 4, bottom right). In addition, IP 4 binding to GST-F was inhibited by PS in a dose-dependent manner (data not shown). These results suggest that the PH domain is involved in association with membranes and that the IP 4 -binding site overlaps with the phospholipid-binding site of the the PH domain.
Ras GAP Activity of Wild and Mutant Gap1 m -To further examine whether the mutant lacking IP 4 binding capacity showed Ras GAP activity, the GAP assay was carried out using Ha-Ras⅐[␣-32 P]GTP as a substrate (Fig. 5). GST-F (R628C) also activated the GTPase activity of Ha-Ras but was less potent as an activator than wild type GST-F (Fig. 5). These results are consistent with the fact that the mutant Btk protein (R28C) in Xid mice shows normal kinase activity (26). In the presence of 53 M PS liposomes corresponding to the concentration in the inner plasma membrane leaflet, the Ras GAP activity of both GST-F and GST-F (R628C) were slightly suppressed. This weak suppression was meaningful because it was restored by 10 M IP 4 in the case of GST-F but not GST-F (R628C), probably due to the lack of IP 4 binding capacity in this mutant protein.

DISCUSSION
In this study we have demonstrated the structure-function relationships of the mouse Gap1 m in terms of Ras GAP activity, IP 4 , and phospholipid binding. From a study of deletion mutants of Gap1 m , the PH domain was determined to be a effector domain for GAP activity and functions as a central IP 4 -and phospholipid (PS)-binding domain.
In a previous study (13,14), we have reported that the C2B domain of synaptotagmins I, II, and IV are essential for inositol high polyphosphate binding and determined the putative IP 4 binding motifs (IHLMQNGKRLKKKKTTVKKNTLNPYFNE-SFSF; amino acid residues 315-346 of synaptotagmin II). However, it is not easy to predict whether other proteins that have the C2B domain bind IP 4 because there are some exceptions. For example, synaptotagmin III has putative IP 4 binding sequence in the C2B domain but did not bind IP 4 , probably due to the steric hindrance of the C-terminal flanking region (14). We infer that the C2B domain of Gap1 m is such a case because the peptide from the C2B domain of Gap1 m (amino acid residues 194 -228; ATVSLVGPSRNDQKKTKVKKKTSNPQFNEVFYF-EV), corresponding to the putative IP 4 -binding domain of synaptotagmin II, bound IP 4 (K d Ն 1 M; data not shown).
The PH domain is a new divergent protein module of approximately 100 amino acids found in many proteins involved in signal transduction (23,24). Although the consensus functional properties have not been elucidated, many PH domains bind to the ␤␥ subunits of heterotrimeric G proteins (37), and phosphatidylinositol 4,5-bisphosphate (38) and the PH domain of phospholipase C-␦1 bind IP 3 (32,33). The PH domain was divided into six subdomains (25), and subdomain VI, which is the most conserved region of the PH domain, is responsible for ␤␥ subunits of heterotrimeric G proteins binding to ␤-adrenergic receptor kinase (37). In contrast, the IP 3 binding sequences of the PH domain of phospholipase C-␦1 were determined to be KVKSSSWRRERFYK (amino acid residues 30 -43) in subdomains I and II (32,33). Although these sequences were not found in the Gap1 family (Fig. 3A), we have shown that the corresponding region (the cluster of positively charged amino acids at position 626 -628 in subdomain II) is essential for IP 4 binding to Gap1 m by mutational analysis. In addition to the PH domain, we have demonstrated that inositol phosphate binding specificity and high affinity IP 4 binding to Gap1 m also requires the adjacent GRD and C terminus.
We also showed that Gap1 m has two phospholipid binding sites, double C2 domains and the PH domain. The C2A domain but not the C2B domain of Gap1 m showed Ca 2ϩ -dependent phospholipid binding, whereas the PH domain of Gap1 m bound phospholipid irrespective of the presence of Ca 2ϩ . Thus it is likely that the PH domain is required for association with the membranes. Furthermore, the Ras GAP activity of recombinant Gap1 m without two C2 domains was slightly suppressed by PS and restored by IP 4 . Although we did not examine whether PS and other phospholipids (such as phosphatidylinositol 4,5-bisphosphate) directly affect the GRD of Gap1 m like that of neurofibromin (39), our results suggest that the PH domain binds IP 4 and phospholipids competitively to regulate subcellular localization of Gap1 m and GAP activity.
Our findings also suggest that the PH domain is unlikely to be a functionally independent domain. As pointed out by Musacchio et al. (25), many PH domains are found adjacent to other functional domains. The PH domain of Akt and ␤-adrenergic receptor kinase is adjacent to the kinase domain. In Xid mice, the mutant Btk protein has normal kinase activity despite the mutation in the PH domain (R28C) (26). Because of the high similarity of the PH domain between the Btk and Gap1 families, we infer that the PH domain of Btk also binds phospholipids and/or inositol high polyphosphate. Loss of these properties in the mutant Btk protein (R28C) may cause false activation or inactivation of kinase activity or protein-protein interaction via the SH2 (Src homology domain 2) or the SH3 domain. On the basis of these results, together with our own observations, we propose that some PH domains function as modulatory domains of kinases.