Expression Cloning of Protein Targets for 3-Phosphorylated Phosphoinositides

The phosphatidylinositol 3-kinase (PI 3 * -K) family of lipid kinases play a critical role in cell proliferation, survival, vesicle trafficking, motility, cytoskeletal rearrangements, and oncogenesis. To identify downstream effectors of PI 3 * -K, we developed a novel screen to isolate proteins that bind to the major products of PI 3 * -K: phosphatidylinositol-3,4-bisphosphate (PtdIns-3,4-P 2 ) and PtdIns-3,4,5-trisphos- phate (PtdIns-3,4,5-P 3 ). This screen uses synthetic biotiny- lated analogs of these lipids in conjunction with libraries of radiolabeled proteins that are produced by coupled in vitro transcription/translation reactions. The feasibility of the screen was initially demonstrated using avidin-coated beads prebound to biotinylated PtdIns-3,4-P 2 and PtdIns- 3,4,5-P 3 to specifically isolate the pleckstrin homology do- main of the serine/threonine kinase Akt. We then demonstrated the utility of this technique in isolating novel 3 * phosphorylated phosphatidylinositol (3 * -PPI)-binding proteins through the preliminary screening of in vitro tran-scribed/translated cDNAs from a small pool expression library derived from mouse spleen. Three proteins were isolated that bound specifically to 3 * PPIs. Two of these proteins have been previously characterized as PIP3BP/ p42 IP4 and

It is now well recognized that dynamic changes in the phosphorylation state of intracellular phosphatidylinositol (Pt-dIns) 1 play critical roles in mediating many cellular events. Phosphatidylinositol 3Ј-kinases (PI 3Ј-Ks) are a subfamily of PtdIns kinases that phosphorylate the 3Ј-OH (D3) position of PtdIns to create four different PtdIns derivatives: PtdIns-3-P, PtdIns-3,4-P 2 , PtdIns-3,5-P 2 , and PtdIns-3,4,5-P 3 (reviewed in Refs. [1][2][3]. Nine different isoforms of PI 3Ј-K have been identified in mammalian cells and they have been grouped into three classes by Domin and Waterfield (4) based on the specific form of PI that is used as a substrate.
Singly phosphorylated PtdIns-3-P is constitutively expressed in cells and is involved in a variety of events associated with membrane protein trafficking (5). Although all classes of PI 3Ј-Ks can phosphorylate PtdIns to generate this lipid, the majority of PtdIns-3-P is probably produced by Class III PI 3Ј-Ks, which is specific for PtdIns (1). PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 are generated following stimulation by a wide variety of extracellular stimuli through many diverse classes of receptors. Class I PI 3Ј-Ks phosphorylate PtdIns-4,5-P 2 to generate PtdIns-3,4,5-P 3 , which can be dephosphorylated to PtdIns-3,4-P 2 by the 5Ј lipid phosphatase SHIP (6). Alternate pathways to PtdIns-3,4-P 2 have also been described involving phosphorylation of the 4-position of PtdIns-3-P by Class II PI 3Ј-K or an unidentified PtdIns-3-P 4-kinase, but the extent to which these enzymes contribute to PI-3,4-P 2 synthesis in not clear (7)(8)(9).
Class I PI 3Ј-Ks play critical roles in many essential cellular processes. Perhaps most importantly, these kinases regulate cell survival. Inhibition of class I PI 3Ј-Ks leads to an induction of programmed cell death or apoptosis, and constitutive unregulated activation of these enzymes or downstream targets of Pt-dIns-3,4-P 2 and PtdIns-3,4,5-P 3 can rescue cells from cell death induced by serum deprivation, loss of matrix attachment, Myc expression, and other apoptotic stimuli (10 -14). These kinases also control the activation of many intracellular signaling pathways that regulate cell proliferation, including extracellular regulated kinase/mitogen-activated protein kinase, protein translation factors (e.g. eIF-4E), and cyclins/cyclin-dependent kinases (15)(16)(17)(18)(19). Also, membrane trafficking events regulated by 3Ј-PPIs control receptor internalization (20). In addition, PI 3Ј-Ks are necessary for glucose transporter recruitment to the plasma membrane and regulation of glycogen synthase kinase 3 and phosphofructokinase, indicating that 3Ј-PPIs are critically involved in insulin-mediated events associated with glucose metabolism (2,(21)(22)(23). Integrin affinity modulation is also blocked by pharmacological inhibitors of PI 3Ј-K, implicating these kinases in critical events associated with leukocyte trafficking and inflammatory responses (24 -26). PI 3Ј-K also plays an important role in regulating cell movement and cytoskeletal rearrangements. For example, 3Ј-PPIs are necessary for controlling receptor-induced changes in actin assembly, the formation of lamellipodial protrusions, and cell migration through the small GTPbinding protein Rac (27)(28)(29)(30)(31).
Because of the central importance of PI 3Ј-Ks in controlling cell proliferation, survival, and motility, it is likely that class I PI 3Ј-Ks and 3Ј-PPI-binding proteins play an important role in the pathogenesis of cancer. Overexpression of PI 3Ј-K in chicken cells is sufficient to induce cellular transformation both in vitro and in vivo (32). PI 3Ј-K has also been implicated in the induction of chronic myelogenous leukemia and acute lymphocytic leukemia by the BCR-ABL oncogene (33,34). As might be expected from its importance in cellular motility, PI 3Ј-K has been shown to play a role in tumor invasion and metastasis in several model systems (31,(35)(36)(37). Recently, a role for PI 3Ј-K in human carcinogenesis was demonstrated by the evidence that the tumor suppressor PTEN is a lipid phosphatase that is specific for 3Ј-phosphate of the inositol head group of and that elimination of the lipid phosphatase activity correlates with the oncogenic potential of PTEN mutants found in human cancers (38 -40).
There are several identified protein motifs that bind to 3Ј-PPIs: PH domains, FYVE domains, SH2 domains, and C2 domains. The primary function of these domains, each of which is approximately 90 -120 amino acids in size, is believed to be localization of the protein to high local concentrations of 3Ј-PPIs found near active signaling complexes at the cell membrane. However, there is evidence indicating that these domains can regulate protein function as well.
The most diverse and best-characterized 3Ј-PPI binding domains are PH domains, which comprise a large family of binding modules that are known to bind proteins, as well as a wide range of lipids. A subset of PH domains binds with a high affinity to PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 (41,42). These PH domains are critical for the function of several signaling proteins, including the serine/threonine kinases Akt and phosphoinositidedependent kinase 1 (PDK1), the tyrosine kinase Brutons tyrosine kinase, and the ARF-guanine nucleotide exchange factor Grp1.
Given the diversity of cellular responses dependent on the products of PI 3Ј-K, it is probable that many as yet unidentified proteins bind to and are regulated by PtdIns-3,4,5-P 3 and Pt-dIns-3,4-P 2 . Identification of these proteins and the elucidation of their role in cellular signaling will be critical to our understanding of these cellular functions as well as of diseases such as cancer. In order to isolate novel 3Ј-PPI-binding proteins, we developed a screen using in vitro coupled transcription/translation technology (47,48) in conjunction with the use of synthetic biotinylated PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 ligands. We show in this report that this screen can isolate 3Ј-PPIbinding proteins with high specificity and selectivity from among a vast excess and diversity of other nonspecific proteins. We report an initial demonstration of the effectiveness of the system using the PH domain of the serine/threonine kinase Akt, a known 3Ј-PPI binder, as a model. Second, we demonstrate the utility of this technique in isolating other 3Ј-PPIbinding proteins by screening a small pool expression library derived from mouse spleen mRNA. Three 3Ј-PPI-binding proteins were identified in this initial screen. Two of these proteins have been previously characterized, PtdIns(3,4,5)P 3 -binding protein/p42 IP4 and PDK1, thereby establishing positive controls that the system operates as predicted. Importantly, the third protein is a novel protein of unknown function that contains both a PH domain and an SH2 domain.

EXPERIMENTAL PROCEDURES
Cloning and Expression of Akt PH Domain-The PH domain of Akt was cloned as a fusion protein to maltose-binding protein (MBP) in the expression vector pMAL2B (New England Biolabs). The N-terminal 130 amino acids of murine Akt1 in the vector pJ3⍀ (a gift of P. Tsichlis, Fox-Chase Cancer Center) was amplified by polymerase chain reaction using Pfu polymerase with the coding strand primer 5Ј-cgatcgggatccatggaacag-3Ј (upstream of the Myc tag in pJ3⍀ and containing a BamHI site) and the noncoding strand primer 5Ј-ccctgaattctcactgggtga-3Ј (from Akt amino acid 130 and encoding an EcoRI site). Both maltose-binding protein alone and the MBP-Akt PH domain fusion construct were Synthesis of Biotinylated and Nonbiotinylated Phosphatidylinositols-All chemically synthesized probes were prepared using a methyl D-glucopyranoside as the chiral starting material for the inositol head group; different phosphate substitution patterns were elaborated using the Ferrier rearrangement of suitably protected glucose-derived precursors (49). Di-C 8 PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 were prepared by modifications (50,51) of the synthesis of the dipalmitoyl analogs (52,53). Biotinylated phosphoinositides were prepared from the sn-2-aminohexanoyl derivatives of PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 by condensation of the active ester of biotin with the water-soluble PIP n analog in the presence of a mild base (Fig. 1). The biotinylated probes, bPtdIns-3,4-P 2 and bPtdIns-3,4,5-P 3 , were purified by ion exchange chromatography and employed as aqueous solutions for attachment to streptavidin-coated surfaces Binding and Isolation of Radiolabeled in Vitro Translated Proteins with Biotinylated Phosphoinositides-Avidin beads (Ultralink-immobilized Neutravidin TM , Pierce) were washed twice with 5 volumes of wash/binding buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40 (nonylphenylpolyethylene glycol), 5 mM dithiothreitol). The beads were then reconstituted in a 500 l of wash/binding buffer, and the biotinylated phosphoinositide was added. Generally, 0.1 l of 100 M biotinylated lipid was bound per 1 l of packed avidin beads. The biotinylated lipid was incubated with the beads for 1-2 h at 4°C with gentle agitation and then washed twice with a 10ϫ bead volume of wash/binding buffer to remove excess ligand. Control beads without biotinylated lipid were prepared in an identical manner but without the addition of the lipid. 5 l of the 35 S-labeling reaction containing the in vitro transcribed/translated protein was then added to the tubes, and the protein was allowed to bind for 2 h at 4°C with gentle agitation. The tubes were then centrifuged briefly, the beads were washed two times with 0.5 ml of wash/binding buffer, and the bound proteins were eluted by boiling the beads in 20 l of Laemmli sample buffer containing 5% 2-mercaptoethanol. The proteins were then separated on a 13.75% SDS-polyacrylamide gel electrophoresis gel. Following electrophoresis the gel was soaked in EnHance (NEN Life Science Products) for 1 h and then in a 7.5% w/v solution of PEG-3350 for 1 h. The gel was then dried and subjected to autoradiography.
Competition Binding of Isolated Proteins for Nonbiotinylated Lipids over Biotinylated Lipids-Competition binding experiments were performed exactly as described above for standard binding of radiolabeled proteins to biotinylated lipids except that avidin beads, which had been prebound with the biotinylated diC 7 PtdIns-3,4,5-P 3 or PtdIns-3,4-P 2 , were incubated with the in vitro translated proteins in the presence of 1 pM to 100 M of nonbiotinylated lipids (diC 8 forms of PtdIns-3,4,5-P 3 , PtdIns-3,4-P 2 , and PtdIns-4,5-P 2 (Echelon Research Laboratories, Salt Lake City, UT)). Binding was quantitated by measuring the intensities of bands from the autoradiogram or by scintillation counting of the eluted proteins in Laemmli sample buffer using Opti-fluor (Packard Instrument Co.). Binding curves were modeled by the equation, % Bound ϭ (IC 50 /(IC 50 ϩ C)) ϫ 100, where % bound represents the quantity of protein bound to the biotinylated lipid coated beads in the presence of a concentration (C) of competing lipid. IC 50 is the molar excess of competitor C required to reduce the percentage bound to 50% of its maximal value in the absence of competitor.
In Vitro Translation Expression Cloning-A murine spleen cDNA library containing 2 ϫ 10 5 independent clones was divided into 1700 individual small pools (54). The cDNAs in this library had been cloned into the in vitro transcription vector pCS2(ϩ) under control of the SP6 promoter. Individual cDNAs from positive pools were isolated from the other cDNAs of the pool using a 96-well format as described previously (47). The individual cDNAs were sequenced (Harvard Biopolymer Facility) and compared with known sequences by searching the NCBI nr data base using the BLAST algorithm.
Northern Blotting-Northern blotting was performed as described in Ref. 71. A 32 P-labeled double stranded DNA probe was made from a polymerase chain reaction fragment from the respective gene using random primed oligonucleotide synthesis (Prime-a-Gene, Promega). Murine tissue RNA was a gift of Dr. W. Swat (Harvard Medical School).

Biotinylated Forms of PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 Specifically Bind to the in Vitro Translated PH Domain of Akt and
Can Be Used for Affinity Isolation-In order to test whether cloning of phosphoinositide-binding proteins through coupled in vitro transcription/translation expression was feasible, we examined whether biotinylated forms of 3Ј-PPIs could be used to affinity isolate a known 3Ј-PPI binding domain that had been produced by a coupled in vitro transcription/translation system. We first prebound avidin-coated beads with diC 7 -analogs of PtdIns-3,4,5-P 3 and PtdIns-3,4-P 2 that had been biotinylated on the -end of the sn-1-aminohexanoyl derivatives (see under "Experimental Procedures" and  1, 4, and 7), labeled proteins bound to avidin beads (lanes 2, 5, and 8), and labeled proteins bound to avidin beads prebound with PtdIns-3,4-P 2 -biotin (lanes 3, 6, and 9). C, PtdIns-3,4,5-P 3 and PtdIns-3,4-P 2 preferentially displace MBP-PH from PtdIns-3,4,5-P 3 -biotin. 35 S-Labeled MBP-PH was bound to avidin beads coated with PtdIns-3,4,5-P 3 -biotin in the presence of the indicated concentrations of PtdIns-4,5-P 2 (squares), PtdIns-3,4-P 2 (circles), or PtdIns-3,4,5-P 3 (triangles) and processed as described under "Experimental Procedures." Data points represent the mean of two independent experiments. cloned into the in vitro transcription vector pCS2(ϩ) under the control of the SP6 promoter. The PH domain of Akt is known to bind 3Ј-PPIs with high affinity and served as a positive control to determine optimal binding conditions and specificity (55)(56)(57)(58). As shown in Fig. 2A, the fusion protein of MBP-Akt PH bound to avidin beads that were prebound with the biotinylated forms of either PtdIns-3,4,5-P 3 or PtdIns-3,4-P 2 . The MBP-PH fusion protein failed to bind to the unmodified avidin beads, and MBP failed to bind to either the avidin beads alone or to the beads prebound with the biotinylated phosphoinositides. These data demonstrate that the biotinylated 3Ј-PPIs can effectively and specifically isolate an in vitro translated form of a phosphoinositide-binding domain.
The goal of this study is to isolate novel 3Ј-PPI-binding proteins from among pools of in vitro transcribed/translated proteins from a small pool expression library. In order to specifically isolate 3Ј-PPI-binding proteins from among other proteins expressed in the library, it is necessary that the system used for screening have a high specificity for 3Ј-PPI-binding proteins and a low background affinity for nonspecific binding proteins. The system must also be sufficiently sensitive to detect small amounts of a 3Ј-PPI-binding protein in a large background of nonspecific binding proteins. To determine whether our system conformed to these criteria, we investigated whether biotinylated PtdIns-3,4,5-P 3 and PtdIns-3,4-P 2 could specifically isolate the MBP-Akt PH domain fusion protein after having been diluted into a pool from the in vitro translation library.
There are approximately 100 independent clones in each pool of cDNA from our pCS2 mouse spleen cDNA library, and the total DNA concentration was approximately 1 g/l for each pool. Therefore, 10 ng of the plasmid encoding MBP-Akt PH fusion protein was mixed with 1 g of DNA of a random pool from the library in order to approximate the amount of 3Ј-PPIbinding protein that would be found in a random pool under the conditions of our screen. As is shown in Fig. 2B, biotinylated PtdIns-3,4-P 2 -beads captured the MBP-Akt PH fusion protein from among the other proteins in the pool. However, some proteins in the library, e.g. the 25-kDa protein, bound to the avidin beads in both the presence and absence of ligand, most likely through a nonspecific hydrophobic interaction. These results provided the proof of concept that this approach is feasible for isolation of 3Ј-PPI binding domains.
Isolation of the Murine Isoforms of PDK1and PIP3BP/p42 IP4 by Small Pool Expression Cloning-After an initial screening of 500 pools of the library using the biotinylated phosphoinositides, we isolated three 3Ј-PPI-binding proteins. We observed a protein of approximately 25 kDa in a single pool that appeared to bind exclusively to biotinylated PtdIns-3,4,5-P 3 (Fig. 3A, top  panel, left). When the cDNA for the protein was isolated from the other cDNAs in the pool, the expressed purified protein was also found to bind to PtdIns-3,4-P 2 but apparently with a lower affinity than for PtdIns-3,4,5-P 3 (Fig. 3A, top panel, right). Sequencing of the cloned cDNA revealed that it was identical to the C-terminal 319 amino acids of murine PDK1 (59). PDK1 is a ubiquitously expressed 559-amino acid (65 kDa) protein that contains an N-terminal serine/threonine kinase domain and a C-terminal PH domain. In agreement with our results, the PH domain of PDK1 has recently been found to bind with a 4-fold higher affinity to PtdIns-3,4,5-P 3 than to PtdIns-3,4-P 2 but with much lower affinity to PtdIns-4,5-P 2 (60). The fragment of PDK1 that we isolated contains the entire C-terminal PH domain and approximately half of the serine/threonine kinase domain (Fig. 3A, bottom panel). However, the first 230 nucleotides of the isolated mRNA are not the sequence of PDK1, and the first in-frame AUG codon appears at the end of the kinase domain such that only the entire PH domain of PDK1 is translated. The initial 230 noncoding nucleotides could either be due to an alternative splicing of PDK1 or be an artifact from the construction of the library.
Another 3Ј-PPI-binding protein identified in our screen was a protein of approximately 25 kDa that bound tightly to both biotinylated PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 but also had some residual binding to the avidin beads (Fig. 3B, top panel). Upon isolation and sequencing of the cDNA from the pool, the gene was found to encode the C-terminal 210 amino acids of a protein that has Ͼ95% amino acid homology to 42-kDa inositol-1,3,4,5-tetrakisphosphate-binding proteins isolated from both porcine brain, called p42 IP4 (61), and bovine brain, called PIP3BP (62). PIP3BP/p42 IP4 is a protein of 374 amino acids that contains an N-terminal zinc-finger domain that has homology to GTPase activating proteins for the ARF family of small G-proteins and two tandem PH domains at the C-terminal end of the protein (Fig. 3B, bottom panel). The fragment of the gene that we isolated contains the entire C-terminal PH domain and approximately half of the N-terminal PH domain but contains none of the putative ARF-GTPase-activating protein domain. The C-terminal PH domain contains a consensus sequence for high affinity binding of 3Ј-PPIs (41,42). The N-terminal PH domain does not contain this sequence, but two studies have shown that it displays some specificity for 3Ј-PPIs (62,63).
The function of PIP3BP/p42 IP4 is currently unknown, although it is highly expressed in the brain and is thought to play a role in vesicle and membrane transport because of the putative ARF-GTPase-activating protein domain. Interestingly, a closely related yeast protein called Gcs1 has been shown to be an ARF-GTPase-activating protein and is important in proper cytoskeletal organization and actin polymerization in yeast (64). A recent study has also shown that PIP3BP is localized to the nucleus but translocates to the plasma membrane upon activation of PI 3Ј-K; however, the functional significance of both the nuclear localization and the PI 3Ј-K-dependent translocation is unknown (65).
The gene fragment containing the sequence for the protein was approximately 1.3 kilobases in length and encoded an open reading frame from the beginning of the fragment to a stop codon at nucleotide position 927 (Fig. 5). A putative Kozak initiator (ATG) codon for methionine is present at position 87, and initiation of translation from this codon is consistent with the observed the size of the translated protein (280 amino acids and 30 kDa). However, because the sequence upstream of this ATG did not contain any in-frame stop codons, it was difficult to determine whether the isolated gene fragment encoded the entire coding sequence of the gene or whether more coding sequences existed upstream.
SH2 domain of the neural adaptor protein Nck (35% identical, 59% homologous at the amino acid level). The PH domain, which encompasses nucleotides 590 -840, is most similar to the PH domain of Akt (39% identical, 63% homologous at the amino acid level) and contains the consensus sequence for high affinity binding of 3Ј-PPIs (42). There is also a tyrosine (Tyr-139) located between the SH2 and PH domains that could be phosphorylated in stimulated cells because the sequence surrounding this tyrosine is a putative consensus motif for phosphorylation by tyrosine kinases (66). The existence of a putative phosphotyrosine-binding SH2 domain, a 3Ј-PPI-binding PH domain, and a sequence for phosphorylation of tyrosine strongly suggests that this protein plays a role in cellular signaling. We have named this protein PHISH for 3Ј-phosphoinositide-interacting SH2-containing protein.
In order to determine the tissue distribution of PHISH, we performed Northern blots on total RNA from several murine tissues, spleen, brain, heart, lung, thymus, and lymph using a probe derived from nucleotides 87-927, the putative coding region of the protein. Two RNA species that are approximately 1.2-1.4 kilobases were detected. These could represent products of alternative splicing of the gene. The larger of the two transcripts was detected in all tissues, and the smaller of the two transcripts was highly expressed in spleen and at lower levels in both heart and lung tissue (Fig. 6). In vitro transcription of the cDNA for PHISH isolated from the expression library also produced two transcripts that were approximately 1.2-1.4 kilobases in size. The major transcript is very close in size to the larger of the two transcripts present in all murine tissues tested, whereas the minor product is close in size to the smaller transcript present in heart, lung, and spleen. This data suggests that the gene fragment isolated from the expression cloning library encodes the complete sequence of the gene.
Upon searching the human EST data base with the nucleotide sequence of PHISH, we found that PHISH had an 87% nucleotide sequence identity with that of a human EST derived from stem cells (locus AF150266). This EST encodes a cDNA that is 1.4-kilobases long and contains entire coding region of PHISH. The sequences of PHISH and the human EST differ significantly outside of the protein coding sequence. Moreover, three codons upstream from the putative start codon in the human EST are in frame stop codons. The high degree of sequence homology between PHISH and the human EST implies that they encode species-specific homologues of the same FIG. 5. Nucleotide and amino acid sequence of murine PHISH. The sequence of PHISH was isolated via expression cloning. The putative SH2 domain is outlined in red, the predicted tyrosine phosphorylation site is outlined in green, and the putative PH domain is outlined in blue.
Cloning of 3Ј-PPI Targets protein and that the gene for PHISH isolated from the expression screen encodes the entire protein.
Because PHISH contains only a PH domain and an SH2 domain, it is possible that PHISH functions as an adaptor protein. The PH domain could dock PHISH to 3Ј-PPI generated at membrane receptor complexes, and the SH2 domain could recruit phosphotyrosine containing protein(s) to such complexes. It is interesting to note that Skolnik and co-workers (42,67) have isolated the PH domain of EST684797 from a homology search of human ESTs for PH domains that would be predicted to bind tightly to 3Ј-PPIs. They have also shown that the PH domain of EST684797, which is the human homolog of PHISH, binds tightly and selectively to PtdIns-3,4-P 2 and Pt-dIns-3,4,5-P 3 but not to PtdIns-4,5-P 2 .
These studies demonstrate that coupled in vitro transcription/ translation libraries can be used in conjunction with affinity isolation technology using synthetic phosphoinositides to isolate 3Ј-PPI binding domains. Recently, several other methods have been described to isolate and clone 3Ј-PPI-binding proteins. One example is the use of resins coupled to phosphorylated inositol phosphates, such as inositol 1,3,4,5-tetrakisphosphate, or inositol phosphates linked to a glycerol moiety, to extract proteins from tissue extracts (61,62,68). Using these techniques, PIP3BP and p42 IP4 were extracted from total brain extracts. However, proteins expressed in low abundance are difficult to isolate by this procedure, and furthermore, cDNA cloning requires multiple steps following affinity isolation.
Expression cloning of genes from cDNA libraries can circumvent some of the limitations of protein affinity isolation techniques. Two expression cloning techniques have been developed recently to identify 3Ј-PPI-binding proteins. The first involved the screening of a gt11 library with a crude mixture of brain phospholipids that had been phosphorylated in vitro by PI 3Ј-K (69). The screening of two murine-derived cDNA libraries (from brain and adipocytes) yielded only one protein that bound tightly to PtdIns-3,4,5-P 3 . This protein, called Grp1 (general receptor for phosphoinositides), contained a PH domain and an ARF-guanine nucleotide exchange factor domain that catalyzes 3Ј-PPI-dependent nucleotide exchange on mammalian ARFs (70). However, no other 3Ј-PPI-dependent binding protein was isolated in the screen, possibly due to improper folding of the proteins.
Another recent expression cloning strategy for isolating 3Ј-PPI-binding proteins is the use of a modified yeast two-hybrid system (42). In this system, genes from a mammalian cDNA library were fused to a constitutively active Ras. The fusion proteins were then tested for their ability to rescue a temperature-sensitive phenotype, due to a defect upstream of Ras, when coexpressed with constitutively active PI 3Ј-K. This system worked well in experiments in which constitutively active Ras was fused to PH domains already known to bind to 3Ј-PPIs. However, when tested with an expression library in yeast, only Akt␥ was isolated.
The screen described here is the first to isolate multiple 3Ј-PPI binding targets from an expression library, including a novel protein. The basis for the better efficiency of this screen may stem from the fact that in vitro translated proteins are more likely to be properly folded. In addition, many internal initiation sites for translation allow for the independent and unoccluded expression of PH and other 3Ј-PPI binding domains. We also used analogs of PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 that are synthetically pure and, as a result of modification in a distal region of the acyl chain, closely resemble the phosphorylated phosphatidylinositols that occur naturally. It has been shown that although the main binding of the protein is to the inositol head group, the contribution of the glycerol chain and the fatty acid side chains of the phosphatidylinositol are essential for specificity and tight binding to the protein (41). Our screening technique is not without drawbacks. For example, there can be nonspecific binding of proteins to the avidin-coated beads, and this nonspecific binding may obscure the binding of other 3Ј-PPI-binding proteins with the same electrophoretic mobility. In addition, we isolated only C-terminal PH domains, possibly because our library was made using oligo(dT) to prime cDNA synthesis from the 3Ј poly(A) tail of mRNAs. In addition, our screen thus far has isolated only strong binders of 3Ј-PPIs, indicating that our binding conditions may be too stringent to accommodate weaker binding proteins.
In summary, we have described a novel and effective way of isolating and cloning 3Ј-PPI-binding proteins from expression libraries. The technique described here has broad applications for the isolation of binding partners for other phosphoinositide polyphosphates or other lipid products.
FIG. 6. Northern blot of mRNA from various murine tissues hybridized to probes derived from PHISH and glyceraldehyde-3-phosphate dehydrogenase. Murine tissue RNA (10 g per lane) was subjected to agarose gel electrophoresis, transferred to nylon membrane, and hybridized to 32 P-labeled probes derived from the coding regions of PHISH or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as described under "Experimental Procedures." Lane 1, brain; lane 2, heart; lane 3, lung; lane 4, lymph node; lane 5, spleen; lane 6, thymus. Arrows represent the mobility of the major (large arrow) and minor (small arrow) products of an in vitro transcription reaction of the PHISH gene isolated from the expression library.