LL5β Is a Phosphatidylinositol (3,4,5)-Trisphosphate Sensor That Can Bind the Cytoskeletal Adaptor, γ-Filamin

We identified a potential phosphatidylinositol (3,4,5)-trisphosphate (PtdIns(3,4,5)P3) binding pleckstrin homology domain in the data bases and have cloned and expressed its full coding sequence (LL5β). The protein bound PtdIns(3,4,5)P3selectively in vitro. Strikingly, a substantial proportion of LL5β became associated with an unidentified intracellular vesicle population in the context of low PtdIns(3,4,5)P3 levels produced by the addition of wortmannin or LY294002. In addition, expression of platelet-derived growth factor-receptor mutants unable to activate type 1A phosphoinositide 3-kinase (PI3K) or serum starvation in porcine aortic endothelial cells lead to redistribution of LL5β to this vesicle population. Importantly, pleckstrin homology domain mutants of LL5β that could not bind PtdIns(3,4,5)P3 were constitutively localized to this vesicle population. At increased PtdIns(3,4,5)P3 levels, LL5β was redirected to a predominantly cytoplasmic distribution, presumably through a PI3K-dependent block on its targeting to the vesicular compartment. Furthermore, at high, hormone-stimulated PtdIns(3,4,5)P3 levels, it became significantly plasma-membrane localized. The distribution of LL5β is thus dramatically and uniquely sensitive to low levels of PtdIns(3,4,5)P3 indicating it can act as a sensor of both low and hormone-stimulated levels of PtdIns(3,4,5)P3. In addition, LL5β bound to the cytoskeletal adaptor, γ-filamin, tightly and in a PI3K-independent fashion, both in vitroand in vivo. This interaction could co-localize heterologously expressed γ-filamin with GFP-LL5β in the unidentified vesicles.

enzymes seem to act as PtdIns(4,5)P 2 3-kinases in vivo; they can be activated by a variety of close-to-receptor transduction events and drive accumulation of PtdIns(3,4,5)P 3 in the inner leaflet of the plasma membrane. This PtdIns(3,4,5)P 3 serves as signal recruiting proteins from the cytosol that possess modules, typically PH domains, capable of binding its head group (1).
There are a variety of reagents that can be used to inhibit PI3K activity. Most widely used are wortmannin (2) and LY294002 (3); both of which potently inhibit nearly all classes of PI3K and hence cannot generally be used to implicate a particular PI3K in a process. More specific are receptor tyrosine kinase Tyr 3 Phe mutants. A number of receptor tyrosine kinases (relevant here, the PDGF ␤-receptor) are capable of binding type IA PI3Ks at specific tyrosine residues that become phosphorylated following ligand binding (4). Mutation of these tyrosines to phenylalanine blocks type IA PI3K binding and activation but does not affect association of other effectors (1,5). Stable, clonal cell lines have been created overexpressing wild-type PDGF-␤ receptors or (Y740F/Y751F) PDGF-␤ receptors (6) that have allowed the impact of selectivity blocking type IA PI3K activation to be assessed in vivo (7).
There is now a substantial family of PtdIns(3,4,5)P 3 -binding proteins that have been shown to translocate to the plasma membrane in response to receptor stimulation of type I PI3K activity, including PKB (8,9), DAPP-1 (10 -13), PDK-1 (14), ARNO (15), ARAP-3 (16), and GRP1 (17). All of the above PI3K effectors bind 3-phosphorylated lipids via a PH domain. PH domains are protein modules of ϳ100 amino acids that bind a variety of ligands ranging from inositol phosphates and phosphoinositides to possibly G-protein ␤␥-subunits (18,19). Those PH domains that bind phosphoinositides specifically form a subset that can be recognized via a consensus sequence of basic residues implicated in binding. Initially this concept was based solely on a limited number of sequence alignments, however, as more phosphoinositide binding PH domains were characterized the early consensus has been evolved and further validated by work that has described the structure of a number of PH domains, some with phosphoinositide-based ligands bound (20). Many different types of proteins seem to use PH domains as phosphoinositide binding modules including enzymes (e.g. PKB, BTK, Vav, and PDK-1) and adaptor proteins (e.g. DAPP-1).
It is generally thought that phosphoinositide-dependent shifts in signaling proteins from predominantly cytosolic to membrane distributions are, in some way, activating. In the case of PKB this seems to result from co-localization with its upstream regulator PDK-1 combined with increased availability of the site PDK phosphorylates (threonine 308 in PKB␣) as a result of PtdIns(3,4,5)P 3 binding (21). For DAPP-1, which has been claimed to bind PLC-␥ (11), it is presumably the relocation of the PLC-␥, to the cell surface and the location of its phospholipid substrate that could be relevant.
Through the application of PI3K inhibitors, some of which have been described above, it has become clear that type I PI3K signaling regulates a huge variety of cellular responses. One of the most widely important of these is cell survival (22). In essence PI3Ks and PKB are thought to supply a signal from some receptors that block cells from undergoing apoptosis (23). These signals operate constitutively in the presence of relatively low levels of survival factors but their inactivation upon factor withdrawal leads rapidly to apoptosis (1).
Filamins are actin-binding proteins that act to stabilize large three-dimensional actin networks, through their ability to dimerize (24,25). Mammals can make ␣-, ␤-, and ␥-filamins that possess different tissue distributions. The filamins also seem to bind to a variety of membrane-associated structural or signaling proteins, typically via a region of the molecule that does not interfere with the actin-binding or dimerization domains and these include, cytoplasmic tails of integrins and receptors. Hence filamins can be seen as structural proteins contributing directly to the mechanical properties of the cytoskeleton but also as points at which a variety of cell-surface signals can converge on the actin cytoskeleton.
In this article we describe the identification, cloning, and expression of a PH domain-containing protein that binds PtdIns(3,4,5)P 3 and behaves as a PtdIns(3,4,5)P 3 -effector but also associates with ␥-filamin and undergoes a novel redistribution in response to reductions in PtdIns(3,4,5)P 3 levels.

EXPERIMENTAL PROCEDURES
Cloning of Human LL5␤ and Relevant Constructs-The cDNA encoding the full-length open reading frame of LL5␤ was obtained via the I.M.A.G.E. clones 531882 (accession numbers, aa116053), 208876 (accession number, h63748), and 82052 (accession number, t68150), which were all obtained from the I.M.A.G.E Consortium (UK HGMP Resource Centre, Hinxton, United Kingdom). The full-length open reading frame of LL5␤ (3762 bp) was ligated in-frame with an amino-terminal Myc or Glu-Glu tag into pCMV3, pEGFP (Clontech), or the pGEX 4T1 (Amersham Biosciences) bacterial expression vectors. Point mutants at the PtdIns(3,4,5)P 3 binding motif (K1162A and R1163A), LL5␤⌬PH (residues 1-1138), and the isolated PH domain (residues 1138 -1253) were generated by a PCR-based mutagenic strategy and ligated in-frame with an amino-terminal Myc or Glu-Glu (Glu-Glu) tag into pCMV3, pEGFP, and the pGEX4T1 expression vectors. All inserts were verified by sequence analysis (Babraham Technix). Full-length ␥-filamin cDNA was a kind gift from Dominic Chung (University of Washington).
Transient Transfection-PAE and COS-7 cells were transfected by electroporation with 20 g of total plasmid DNA as described previously (10). Transfected PAE cells were plated onto glass coverslips (ϳ8 ϫ 10 4 cells per coverslip) in Ham's F-12 media containing 10% HI-FBS for 12 h, then serum starved for 6 -8 h in F-12 media supplemented with 0.5% fatty acid-free BSA, in the presence of 1 units/ml penicillin and 0.1 mg/ml streptomycin (Invitrogen). Transfected COS-7 cells were plated onto glass coverslips (ϳ8 ϫ 10 4 cells per coverslip) in DMEM containing 10% FBS for 12-16 h, then serum starved for 12 h in DMEM supplemented with 0.5% fatty acid-free BSA, in the presence of 1 unit/ml penicillin and 0.1 mg/ml streptomycin.
Cell Detachment-Twelve hours following transfection of PAE cells with GFP-LL5␤, cells were washed with F-12 media either containing 10% FBS or 0.1% fatty acid-free BSA for 2 h. Cells were trypsinized, treated with trypsin inhibitor (Sigma), and left in suspension (end-on end rotation) for either 30 min or 2 h at 37°C in Hepes-buffered F-12 media (pH 7.2). Cells were then washed, resuspended in PBS, and cytospun (300 rpm for 3 min) onto coverslips, fixed in paraformaldehyde, and examined under fluorescence microscopy.
Immunofluorescence-Transfected cells were grown on coverslips and serum-starved as described above. Cells were then washed and incubated at 37°C for 30 min in F-12 media containing 0.5% fatty acid-free BSA, 30 mM Hepes (pH 7.4), and 1 unit/ml penicillin and 0.1 mg/ml streptomycin prior to stimulation for the indicated times. Cells were treated with human PDGF (B-B10 ng/ml) (Autogen Bioclear) for 5 min, latrunculin B (10 g/ml) (Sigma) for 10 min, or wortmannin (100 nM, Sigma) for 15 min. Following treatment, cells were promptly fixed by incubating with buffer containing 4% paraformaldehyde for 15 min at room temperature followed by 3 washes with 150 mM Tris (pH 7.4). Depending on the requirement of the primary antibody, in some cases, cells were fixed in ice-cold 100% methanol for 5 min and rinsed in dH 2 O. For cells transfected with GFP constructs, coverslips were then rinsed in dH 2 O and mounted on slides using Aqua Polymount (Polysciences Inc.). A series of dyes for detection of mitochondria (mitotracker red, 100 nM, 15 min; Molecular Probes), lysosomes (lysotracker red, 50 nM, 30 min; Molecular Probes), and the transferrin receptor conjugated to Texas Red dye (marker for endosomes, 3 min for early endosomes and 10 min for late endosomes) were added to live cells expressing GFP contructs prior to fixation. For all other immunofluorescence studies, cells were fixed and permeabilized in PBS, 0.1% Triton X-100 for 10 min, washed three times in PBS, then incubated with PBS ϩ 1% BSA (w/v) for 30 min at room temperature before being incubated with anti-Myc monoclonal antibody anti-clathrin polyclonal antibody (Santa Cruz Biotechnology), anti-EEA1 monoclonal antibody (Transduction Laboratories), anti-caveolin-1 antibody (Santa Cruz Biotechnology), TRITC-phalloidin (Sigma), anti-␣-tubulin (Sigma), anti-vinculin (Sigma), or anti-PMP70 antibody (Sigma), as indicated for 1 h at room temperature. Coverslips were washed three times for 5 min in PBS ϩ 0.5% BSA and then incubated with the appropriate secondary fluorescein isothiocyanate/RITC-conjugated antibody (1 h at room temperature). Coverslips were then washed four times in PBS ϩ 0.5% BSA (5 min), PBS (5 min), then rinsed in dH 2 O before being mounted onto slides, allowed to dry, and viewed under a Zeiss Axiophot fluorescence microscope. Images were captured using a SPOT digital camera (Diagnostic Instruments).
Confocal Image Analysis of Live Cells-Cells were transfected, cultured on sterile glass coverslips, and treated as described above. For imaging, coverslips were mounted on the stage of an Olympus 1 ϫ 70 microscope interfaced with an UltraView confocal system. The cells were imaged at 37°C using a thermostatically controlled cell chamber and incubated in PAE salt solution (25 mM Hepes, pH 7.4, 1.8 mM CaCl 2 , 5.37 mM KCl, 0.81 mM MgSO 4 , 112.5 mM NaCl, 25 mM D-glucose, 1 mM NaHCO 3 , and 0.1% (w/v) fatty acid-free BSA).
Time-lapse images of GFP-transfected cells were obtained using an UltraView confocal microscope (PerkinElmer Life Sciences). GFP fluorescence was excited at 488 nm and the emission was collected at wavelengths Ͼ505 nm using a long pass filter. Typically, 12 bit ϳ600 ϫ 400 pixel images were captured every 2-3 s.
Northern Blot Analysis-A 745-bp fragment encoding the unique NH 2 terminus of LL5␤ was used as a probe for Northern blot analysis. The probe was labeled with [␥-32 P]dCTP (Amersham Biosciences) and the Prime-a-Gene labeling system (Promega). The radioactive probe was applied to multiple tissue Northern blots containing RNA from various human tissues obtained from Clontech and carried out according to the recommended protocol.
Purification of LL5␤ Interacting Partners-Recombinant proteins, GST-LL5␤ and GST-LL5␤⌬PH, were expressed in bacteria and purified on GS-Sepharose beads. The glutathione-Sepharose beads bound to GST LL5␤ and GST-LL5␤⌬PH were used to "pull down" interacting proteins from COS-7 cell lysates. Previously seeded COS-7 cells were washed twice in PBS and left in Met-and Cys-free DMEM for 35 min. Then, 0.2 mCi of [ 35 S]methionine and [ 35 S]cysteine (Amersham Pharmacia Biotech) was added 16 h prior to lysis. Lysates were centrifuged for 10 min at 4°C at 13,000 ϫ g av and supernatants were transferred to 2 g of GST LL5␤ and GST-LL5␤⌬PH purified on 30 l of glutathione-Sepharose beads, and allowed to mix for 2 h at 4°C. Four washes were then carried out in lysis buffer (30 mM Hepes (pH 7.4), 10 mM NaF, 5 mM ␤-glycerophosphate, 1 mM MgCl 2 , 1 mM EGTA, 1% Nonidet P-40, 110 mM NaCl, 1 mM dithiothreitol) followed by a final wash in modified lysis buffer containing 0.25% Nonidet P-40 instead of 1% Nonidet P-40. Proteins were eluted in SDS sample buffer, separated by SDS-PAGE, the gel dried down onto Whatman paper, and exposed to x-ray film for 2 days at Ϫ70°C.
In preparation for the trypsin digestion of the interacting protein, the above pull-down method was carried out on a larger scale without 35 S labeling; the SDS-PAGE gel was stained with Coomassie Blue and the relevant band cut out and further cut into gel slices of ϳ1 mm 3 . This was washed in 3ϫ 50% acetonitrile, 25 mM NH 4 bicarbonate (pH 8), then soaked in 100% acetonitrile for 5 min, and dried in a 5200 centrifugal concentrator for 20 min. 10 g of excision grade trypsin (Sigma) in 25 mM NH 4 bicarbonate (pH 8) was added to the dried gel slices and digested at 37°C for 16 h. Extraction of peptides was carried out by soaking the gel slice in 50% acetonitrile, 5% trifluoroacetic acid for 30 min with gentle agitation. A second extraction was carried out as above and the combined extracts were dried as before for 1 h. The dried sample was then reconstituted by adding 4 l of 50% acetonitrile, 0.1% trifluoroacetic acid. The generated peptides were then analyzed at Applied Biosystems by a mass spectrometer (Qstar Pulsar I).

RESULTS
Cloning, Tissue Distribution, and Lipid Binding Properties of LL5␤-Using the consensus Lys-Xaa-Gly/Ser-Xaa(6 -11)-Arg/Lys-Xaa-Arg-Phe/Leu in 1996 (26) we identified a partial human protein sequence in the NCB1/EMBL protein expressed sequence tag data base encoding the COOH-terminal domain of a protein, previously called LL5 (rat) (27), that we predicted would bind PtdIns(3,4,5)P 3 . Sequencing the relevant Image clone (82052) revealed upstream overlaps with further expressed sequence tags and iteration identified a potential upstream start codon with an in-frame stop immediately upstream. Although we began searching for the human orthologue of rat LL5, we ended up with the open reading frame of a paralogue of the human orthologue of rat LL5. The predicted open reading frame was for a 160-kDa protein (see Fig. 1A) and a full-length clone (accession number AJ496194) was created from Image clones 531883, 208876, and 82052. The protein contains a single spectrin repeat and a COOH-terminal PH domain. The same PH domain was also identified as a potential PtdIns(3,4,5)P 3 -binding protein in a screen by Isakoff et al. (26) and subsequently by Dowler et al. (28) (who called the host protein LL5␤ and a closely related molecule LL5␣, which is the human orthologue of the rat LL5). We will retain the nomenclature Dowler et al. (28) applied to the PH domain of this protein and hence will term it LL5␤. LL5␣ and LL5␤ are different proteins that occur at different locations and have less than 70% identity at the protein level. A 745-bp probe from the NH 2 -terminal region of LL5␤, which would not recognize LL5␣, was used to analyze a Northern blot prepared from human tissues. A 6-kb band was detected in a number of tissues with the highest levels found in heart, kidney, and placenta (see Fig.  1B).
Expression plasmids encoding NH 2 -terminal Myc-and GFPtagged LL5␤ (and various constructs, see below) were prepared and transiently transfected into COS-7 and PAE cells. Anti- FIG. 1. Amino acid sequence and tissue distribution of LL5␤. A, amino acid sequence of human LL5␤. The sequence is shown in single-letter code and residue numbers are indicated. The single spectrin repeat (residues 671-778) is in red and the PH domain (residues 1143-1246) is highlighted in gray with the key conserved residues in the PtdIns(3,4,5)P 3 binding motif in pink. B, tissue distribution of LL5␤ by Northern blot analysis. A multiple tissue Northern blot (Clontech) containing polyadenylated RNA from the indicated human tissues was probed for LL5␤ expression using a 32 P-labeled, unique, NH 2 -terminal fragment of LL5␤ (745 bp). The LL5␤ probe was observed to hybridize to a transcript of the predicted size (6 kb).
Distribution of LL5␤ in Cells-We transiently expressed Myc-or GFP-LL5␤ in PAE cells stably overexpressing the PDGF␤ receptor. In the presence of serum or after only short periods of serum starvation (up to 6 h), the LL5␤ constructs appeared predominantly cytosolic in fixed cells. After prolonged serum starvation (8 h or more) a significant proportion of Myc-or GFP-LL5␤ become particulate apparently at the expense of the cytosolic pool in both living or fixed cells. After approximately 6 h of serum starvation, when the protein was predominantly cytosolic, stimulation with PDGF resulted in a partial translocation of both Myc-and GFP-LL5␤ to the edge of the cell (Fig. 3). This event was observed in both living and fixed cells using confocal and standard epifluorescence microscopes. The translocation was apparently more prolonged than that displayed by proteins such as DAPP-1 or PKB studied under similar conditions. Furthermore, it appeared that the peripheral accumulation of LL5␤ constructs correlated with reductions in both its cytosolic and particulate pools (also see below). PAE cells were similarly transfected with a GFP-tagged form of the isolated PH domain of LL5␤. It did not translocate to the edge of the cell in response to PI3K stimulation. However, we note that although full-length GAP1 m translocates to the cell membrane in a PI3K-dependent manner, the isolated PH domain does not translocate when expressed as an independent module. 2 In an attempt to assess the PI3K dependence of this response we preincubated PAE cells transiently expressing Myc-or GFP-LL5␤ with LY294002 or wortmannin, however, there was a profound redistribution of both constructs in response to the PI3K inhibitors alone that was observed in both living (GFP) or fixed cells (both tags). The cytosolic levels of GFP-LL5␤ were very substantially reduced and an intracellu-2 P. J. Cullen, personal communication. Myc-tagged LL5␤ prepared in COS-7 cells were mixed with 50 M dipalmitoyl forms of free, competing phosphoinositides, PtdIns(3,4,5)P 3 , PtdIns(3,4)P 2 , PtdIns-(4,5)P 2 , PtdIns(3,5)P 2 , PtdIns(3)P, Ptd-Ins(4)P, and PtdIns(5)P for 10 min and then transferred to PtdIns(3,4,5)P 3 beads, allowed to mix for 1 h, washed, and proteins were eluted with SDS sample buffer. The first lane was loaded with a sample of Myc-LL5␤ equivalent to 1% of that included in the binding assays. Myc-tagged LL5␤ was detected by immunoblotting with anti-Myc antibodies. C, binding of Myc-LL5␤⌬PH and Myc-LL5␤-K1162A/ R1163A expressed in COS-7 cell lysates to PtdIns(3,4,5)P 3 beads. Proteins were detected by immunoblotting with anti-Myc antibodies.

FIG. 3. The effects of PDGF stimulation on the subcellular localization of GFP-LL5␤ in PAE cells. PAE cells were transiently transfected with GFP-LL5␤.
After recovery and serum starvation, the cells were stimulated with PDGF (10 ng/ ml). A, live cells were viewed with a confocal microscope. Images were captured at the indicated times after PDGF stimulation commenced. B, cells were fixed at the indicated times after PDGF stimulation commenced and viewed under a confocal microscope.
LL5␤, a PIP 3 and ␥-Filamin-binding Protein lar vesicular compartment, apparently identical to that seen in cells after prolonged serum starvation, became decorated (Fig.  4A, see also Supplementary Material for a video showing the effects of wortmannin on the distribution of GFP-LL5␤ in living PAE cells). We have not seen this phenomenon before in the context of similar experiments with PAE cells studying proteins such as PDK-1, PKB, ARAP-3, PRex-1, and DAPP-1 (10,14,16,29).
Interestingly, with PAE cells transiently expressing GFP-LL5␤ in the presence of 30 M LY294002 (where the inhibition of PI3Ks was substantial but not complete), PDGF stimulation caused a significant redistribution of GFP-LL5␤ from the vesicular pool into the cytoplasmic fraction without a clearly detectable accumulation near the plasma membrane. Transient expression of GFP-LL5␤ in a PAE cell line expressing (Y740F/Y751F)-PDGF-␤-receptors (unable to bind and activate type I PI3Ks) revealed that GFP-LL5␤ was constitutively associated with intracellular vesicles in the absence of wortmannin or LY294002. In an attempt to assess whether other procedures, potentially capable of reducing cellular PtdIns(3,4,5)-P 3 levels, could cause this shift of LL5␤ constructs into a particulate compartment, we serum-starved and/or detached PAE cells transiently expressing GFP-LL5␤ and held them in suspension. Both treatments significantly increased the proportion of GFP-LL5␤ in the vesicular compartment, this was seen most clearly in cells that were detached and in the absence of serum (Fig. 4, C and D).
We examined the distribution of GFP-LL5␤ in transiently transfected COS-7 cells to check whether this phenomenon is cell-type specific. We found that the construct was very largely cytosolic in both living and fixed cells and treatment with PI3K inhibitors wortmannin or LY294002 lead to a reduction in cytosolic staining and the decoration of an intracellular vesicular compartment (data not shown).
We examined the distribution of GFP-LL5␤⌬PH and GFP-LL5␤-K1162A/R1163A (double point mutation in the PH domain predicted to abolish PtdIns(3,4,5)P 3 binding) in PAE (Fig.  4B) and COS-7 cells. Both constructs adopted a constitutive vesicular distribution in both PAE and COS-7 cells. These distributions in PAE cells were unaffected by PDGF (not shown).
Together these results suggest that overexpressed LL5␤ can translocate to the plasma membrane in response to PI3K activation but that under conditions of low PtdIns(3,4,5)P 3 , LL5␤ becomes associated with a vesicular compartment. This does not appear to be an artifact of inhibition of PI3K activity as a number of inhibitory strategies are effective nor is the vesicular compartment made up of aggregated protein because LL5␤ enters the compartment rapidly (15 min), can apparently move back into the cytosolic compartment under certain conditions, and the decorated vesicles move actively around the cell in a manner akin to vesicles like endosomes (see Supplementary  Materials). Furthermore, our results with the ⌬PH and LL5␤-K1162A/R1163A constructs suggest that it is likely that this process represents the association of LL5␤ with a pre-existing organelle (unless they are formed specifically in the presence of these constructs) and that the key event is that LL5␤ "perceives" that cellular PtdIns(3,4,5)P 3 is low, rather than the levels of PtdIns(3,4,5)P 3 are actually low.
We consider the best working explanation for these data is that even at relatively low cellular levels of PtdIns(3,4,5)P 3 LL5␤ can cycle on and off the plasma membrane from the cytosol in a PH domain-dependent manner (without substantial accumulation at the plasma membrane) and that this proc- LL5␤, a PIP 3 and ␥-Filamin-binding Protein ess leads to modification of LL5␤ (e.g. phosphorylation or association of a protein) that prevents it becoming localized into the vesicular compartment and has a lifetime of roughly 15 min. The cytosolic pool of LL5␤ can undergo a net translocation to the plasma membrane in response to receptor activation of type I PI3Ks.
The Nature of the Vesicle Compartment That Can be Decorated by LL5␤-The work we described above suggests that the "vesicular" LL5␤ is unlikely to be a aggregated/denatured protein. This view is also supported by the dynamic, jittering movements of GFP-LL5␤-associated structures in the presence of wortmannin or of GFP-LL5␤-K1162A/R1162A decorated structures. We attempted to co-localize GFP-LL5␤ with a variety of markers in wortmannin-treated PAE cells. We used Texas Red-conjugated transferrin to label early (3 min incubation with cells) and late (10 min incubation with cells) endosomes, an antibody against early endosomal autoantigen 1 as an alternative marker for early endosomes, lysotracker as a marker for lysosomes, anti-caveolin antibodies to identify caveolae, anti-clathrin antibodies to decorate clatherin-coated pits, anti-PMP70 to label peroxysomes, mitotracker to label mitochondria, anti-vinculin antibodies to identify focal adhesions, and phalloidin to identify filamentous actin fibers. The GFP-LL5␤ did not co-localize with any of these markers (see Supplementary Material).
We used latrunculin B to disassemble actin fibers in PAE cells. It had no effect on the distribution of GFP-LL5␤-decorated vesicles (see Supplementary Material). We co-transfected cells with Arf-6 and GFP-LL5␤ and localized the Arf-6 with anti-Arf-6 antibodies and RITC secondary antibodies; the constructs did not co-localize (see Supplementary Material). We have previously observed that DAPP-1 becomes localized to an endosomal compartment in the presence of PDGF. We co-transfected PAE cells with GFP-DAPP-1 and Myc-LL5␤-K1162A/ R1163A, stimulated with PDGF and detected the LL5␤ construct via an anti-Myc antibody and a RITC secondary antibody. The internalized GFP-DAPP-1 did not co-localize with the vesicular LL5␤ construct (data not shown). Finally we used a dominant-negative dynamin construct that we have previously used to establish that DAPP-1 is internalized in a dynamin-sensitive fashion and is an effective inhibitor of dynamin-mediated membrane internalization (dynamin and dynamin mutant was a kind gift of H. McMahon). In an experiment where the dynamin point mutant blocked internalization of co-transfected GFP-DAPP-1 in response to PDGF it had no effect on the formation or distribution of GFP-LL5␤-decorated structures in the presence of wortmannin (data not shown). We have not yet positively identified the vesicle compartment that is labeled by LL5␤ constructs although the number markers we have failed to co-localize suggest that it is a tightly defined subpopulation.
LL5␤-binding Proteins-In the context of our hypothesis that the PI3K-dependent redistribution of LL5␤ to a vesicular compartment might be dependent/blocked by an LL5␤-associated protein we attempted to isolate potential binding partners. We prepared GST-LL5␤ and GST-LL5␤⌬PH in bacteria and derived glutathionine-Sepharose beads loaded with them, GST alone, or GST-SHIP-1 as a control. These protein-loaded beads were mixed with aliquots of lysates made from [ 35 S]methionine-labeled COS-7 cells, washed, the bound protein was resolved by SDS-PAGE, and the gel was dried and autoradiographed. A 260-kDa protein was recovered specifically with GST-LL5␤ and GST-⌬PH-LL5␤ (Fig. 5A). Two-dimensional isoelectric focusing and SDS-PAGE could not further resolve the 260-kDa protein band (not shown). The preparation was scaled up without [ 35 S]methionine and the final one-dimensional SDS-PAGE gel was stained with Coomassie Brilliant Blue (Fig.  5B). The 260-kDa protein band was excised, in-gel digested with trypsin, and the resulting peptides were ultimately analyzed by electrospray mass spectrometry (Q star Pulsar i). Four peptides were selected for fragmentation and the patterns of the m/z ratios were used to reconstruct their sequences. Those sequences were used to interrogate the NRDB (nonredundant data base maintained by the European Bioinformatics Institute) and they identified ␥-filamin. This assignment was confirmed by a ␥-filamin-specific monoclonal antibody and a panfilamin antibody preparation in immunoblots of the 260-kDa protein eluted from the GST-LL5␤ affinity supports (Fig. 5C).
We tested whether ␥-filamin could interact with LL5␤ in vivo. (Glu-Glu)-tagged LL5␤ and ␥-filamin were transfected individually and together into COS-7 cells (note, transfection with ␥-filamin did not substantially increase the amount of total ␥-filamin in the cells). Lysates were prepared and immunoprecipitated with anti-(Glu-Glu) monoclonal antibody covalently attached to protein G-Sepharose. ␥-Filamin was only immunoprecipitated in the presence of (Glu-Glu)-LL5␤ (Fig.   FIG. 5. Identification of a LL5␤ interacting protein. A, isolation of an LL5␤-binding protein from 35 S-GST, GST-SHIP-1, GST-LL5␤⌬PH, and GST-LL5␤ were expressed and purified from bacteria on glutathione-Sepharose beads and used to affinity purify any interacting proteins from lysates of 35 S-labeled COS-7 cells. Pull-downs were washed, eluted with SDS sample buffer, resolved by SDS-PAGE, and autoradiographed. The autoradiogram shows the interacting protein, above the 220-kDa marker pulled down by both GST-LL5␤⌬PH and GST-LL5␤ but not GST nor GST-SHIP-1. B, Coomassie detection of the LL5␤ interacting protein affinity purified from COS-7 cell lysate. C, immunodetection of LL5␤ interacting protein with independent antifilamin antibodies. LL5␤ interacting protein purified from COS-7 cell lysate was immunoblotted with RR90, a filamin antibody that recognizes all three filamin isoforms (left panel) and antibody that recognizes specifically ␥-filamin (right panel). 6A). About 5% of the total ␥-filamin was recovered in the washed (Glu-Glu)-LL5␤ immunoprecipitates. This indicates that ␥-filamin and LL5␤ can interact in vivo. We examined the distribution of ␥-filamin and its relationship to LL5␤ in COS-7 (Fig. 6, B and C) and PAE cells (data not shown). We transiently transfected cells with GFP-LL5␤ and/or ␥-filamin (detected by ␥-filamin-specific antibody and a RITC-labeled secondary antibody). In cells co-transfected with GFP-LL5␤ and ␥-filamin in the presence of wortmannin, it was clear that 30 -40% of the ␥-filamin-positive structures were also positive for the GFP-LL5␤ vesicular compartment (Fig. 6B). In cells transfected with ␥-filamin alone, or co-transfected with GFP-LL5␤ and ␥-filamin but not treated with wortmannin, the ␥-filamin adapted a punctate distribution that was insensitive to wortmannin and were clearly smaller than those that contained both GFP-LL5␤ and ␥-filamin (Fig. 6C). In cells cotransfected with ␥-filamin and GFP-LL5␤ and treated with wortmannin, the structures that were only positive for ␥-filamin were the same size as the ␥-filamin-positive structures in cells transfected with ␥-filamin alone. These results suggest that LL5␤ and ␥-filamin can co-localize in both COS-7 and PAE cells in the presence of wortmannin and that ␥-filamin localization is dictated by LL5␤ and wortmannin. This indicates that the interaction between LL5␤ and ␥-filamin (in the presence of wortmannin) leads to the targetting of ␥-filamin into the vesicular compartment by LL5␤ and that ␥-filamin is not responsible for directing or blocking the movement of LL5␤ into a vesicular compartment. DISCUSSION These results imply that LL5␤ has the potential to act as a PH domain-containing PI3K effector that can translocate to the plasma membrane in response to receptor activation of type I PI3Ks. However, at low levels of PtdIns(3,4,5)P 3 or when the PH domain of LL5␤ is unable to bind PtdIns(3,4,5)P 3 , LL5␤ is directed to a vesicular compartment. We consider the simplest explanation for these events, bearing in mind that unstimulated cells contain low levels of PtdIns(3,4,5)P 3 and that PH domain/PtdIns(3,4,5)P 3 -mediated membrane recruitment is probably a dynamic process with turnover times of the order of a maximum of 1-105, that PH domain/PtdIns(3,4,5)P 3 -mediated signaling through LL5␤ blocks targeting of LL5␤ to a vesicular compartment. This signal could be a modification to LL5␤ (e.g. phosphorylation, dephosphorylation, or association of a protein) that is reversed in the absence of reinforcing signals in a time scale of 10 -30 min. The outcome is that LL5␤ shows a dramatic change in distribution as the cellular levels of PtdIns(3,4,5)P 3 alter in the low basal range.
Much literature has shown how many cells require PI3K signals to survive in several different contexts. Notable among these are serum or growth factor starvation and detachment from a substrate. We have noted that LL5␤ redistributes relatively (cf. events like apoptosis) rapidly under these conditions; however, it is completely unclear whether the changes in LL5␤ distribution have a cause/effect relationship with these survival pathways.
LL5␤ can bind ␥-filamin. Our results indicate that this is not a PI3K-regulated interaction and that it appears to serve to redistribute ␥-filamin, although we have not yet established whether LL5␤ can recruit ␥-filamin to the plasma membrane in a PI3K-dependent manner. This contrasts with a number of examples of proteins that bind filamins and as a result are targetted to the actin-containing cytoskeleton e.g. SHIP-1 (30). In the light of the fact that ␥-filamin serves as a stabilizer and organizer of the actin cytoskeleton, this interaction may be important for the role of ␥-filamin. This view is strengthened by the result of a recent study that has suggested that filamin A is effectively down-regulated and as a result cell migration is reduced by interaction with L-FILIP (31). Interestingly, L-FILIP targets filamin A into an undefined punctate intracellular organelle, where the filamin A is degraded. It will be important to investigate the effects of LL5␤ on ␥-filamin degradation and whether FILIP family proteins target filamins into a related intracellular compartment.