Agonists Cause Nuclear Translocation of Phosphatidylinositol 3-Kinase γ

In hematopoietic cells, the signals initiated by activation of the phosphoinositide 3-kinase (PI3K) family have been implicated in cell proliferation and survival, membrane and cytoskeletal reorganization, chemotaxis, and the neutrophil respiratory burst. Of the four isoforms of human PI3K that phosphorylate phosphatidylinositol 4,5-bisphosphate, only p110γ (or PI3Kγ) is associated with the regulatory subunit, p101, and is stimulated by G protein βγ heterodimers. We performed immunolocalization of transfected p110γ in HepG2 cells and found that, under resting conditions, p110γ was present in a diffuse cytoplasmic pattern, but translocated to the cell nucleus after serum stimulation. Serum-stimulated p110γ translocation was inhibited by pertussis toxin and could also be induced by overexpression of Gβγ in the absence of serum. In addition, we found that deletion of the amino-terminal 33 residues of p110γ had no effect on association with p101 or on its agonist-regulated translocation, but truncation of the amino-terminal 82 residues yielded a p110γ variant that did not associate with p101 and was constitutively localized in the nucleus. This finding implies that the intracellular localization of p110γ is regulated by p101 as well as Gβγ. The effect of PI3Kγ in the nucleus is an area of active investigation.

The four isoforms of human PI3K that phosphorylate PI, PI-4-P, and PI-4,5-P 2 are classified by their catalytic subunits: p110␣, p110␤, p110␥, and p110␦. The most studied human PI3K is a heterodimer composed of p110␣, p110␤, or p110␦ coupled to an adapter protein, p85. These p85-associated PI3K isoforms are tightly linked to signaling mediated by growth factor receptors. After stimulation by extracellular growth factors, the cytoplasmic tail of the growth factor receptor autophosphorylates, enabling it to associate with numerous signaling proteins, including p85, via its two SH2 domains. This recruits p110␣ to the cellular membrane, which appears to be sufficient to activate its lipid kinase activity (3). The binding of activated Ras to p110␣ also appears capable of activating p85-p110, but whether this enhances its membrane association is currently unknown (4). Intracellular localization studies have shown that p85-p110 is present in the cytoplasm with a small component at the extracellular membrane (3); yet two studies using PC12 or human embryonic kidney 293 cells have suggested that p85-associated PI3K can translocate to the nucleus after neuronal growth factor stimulation (5) or H 2 O 2 exposure (6). However, these observations remain controversial.
Several years ago, it was shown that hematopoietic cells possess a PI3K that can be directly stimulated by G␤␥ heterodimers (26). Several groups have demonstrated that this G protein-activated PI3K is a heterodimer composed of a catalytic subunit, p110␥, and an adapter protein, p101 (7,8). In addition to blood cells, Northern blot analysis demonstrated that p110␥ mRNA is also abundant in skeletal and cardiac muscle, liver, and pancreas (9). This PI3K plays a role in the activation of mitogen-activated protein kinase by G protein-coupled receptors and Btk (10,11). In reconstitution assays, the p101-p110␥ complex was inhibited by the pleckstrin homology domaincontaining protein pleckstrin, whereas the p85-p110 complexes were unaffected (12).
The literature suggests that the regulation of p110␣/␤/␦ catalytic subunits is controlled by their intracellular localization, which, in turn, is controlled by their p85-binding partners. In addition, several reports now provide evidence that p85-associated PI3K can translocate to the cell nucleus (5,6). However, no published reports specifically address the intracellular localization of p101-p110␥. Therefore, in this study, we determined, first, the intracellular localization of p110␥ and, second, whether the localization is influenced by binding to p101. Our results show that p110␥, upon serum stimulation, translocates to the cell nucleus. This serum-induced nuclear translocation is pertussis toxin-sensitive and can be mimicked by overexpression of G␤␥ heterodimers, but does not appear to be cell cycleregulated. However, the nuclear localization is regulated by p101 since the ⌬1-82 truncation variant of p110␥, which cannot associate with p101, is constitutively localized in the nucleus.

MATERIALS AND METHODS
Mammalian Expression Vectors-The cDNA clone of human p110␥ was described previously (13). The p110␥ variants were generated by polymerase chain reaction mutagenesis using the techniques of Ho et al. (14) and Landt et al. (15). The deletion variants, p110␥ (⌬1-34) and p110␥ (⌬1-82), also contained the 5Ј-untranslated region from ␤-hemoglobin fused upstream of the initiator methionine. All of these cDNAs were cloned into pCMV5 and contain a carboxyl-terminal additional 9-amino acid hemagglutinin (HA) epitope tag (YPYDVPDYA) recognized by the monoclonal antibody 12CA5. The GFP-p110␥ fusion-expressing plasmid contained the sequence for GFP in place of the stop codon and was generated by the technique of polymerase chain reaction splice overlap extension and cloned into pcDNA3.1 ϩ (Invitrogen, Carlsbad, CA). The sequences of all clones were fully confirmed. The plasmids that direct the expression of an EE epitope (EEEEYMPME)tagged variant of p101 and Myc epitope (EQKLISEEDL)-tagged bovine p110␥ were a generous gift from Dr. Len Stephens (Babraham Institute, Cambridge, United Kingdom). Plasmids that direct the synthesis of G␤ 1 and G␥ 2 were a generous gift from Dr. Janet Robishaw (Geisinger Institute, Danville, PA).
Co-immunoprecipitation Studies-COS-7SH cells were transfected by the calcium phosphate technique as described previously (16) using plasmids that direct the synthesis of HA-p110␥ variants, with and without EE-p101 or G␤␥. Forty-eight hours later, the cells were washed with phosphate-buffered saline and then lysed on ice in 1 ml of 1% Triton X-100, 0.14 M NaCl, 1 mM MgCl 2 , 1 mM EGTA, 20 mM HEPES (pH 7.4), 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 0.1% aprotinin. After clarification at 13,000 ϫ g for 30 min, the supernatant was incubated with Sepharose G coupled to an anti-EE antibody (BAbCO, Berkeley, CA) overnight. The beads and associated proteins were pelleted at 2000 ϫ g for 30 s, washed extensively in the lysis buffer, and boiled in Laemmli loading buffer. The anti-EE immunoprecipitates were then fractionated by 7.5% SDS-polyacrylamide gel electrophoresis and immunoblotted with the anti-HA antibody HA.11 (BAbCO) to detect the co-immunoprecipitation of HA-p110␥ variants along with EE-p101.
Indirect Immunofluorescence-HepG2 cells were transiently transfected by the calcium phosphate technique and stained as described previously (17). Platelet-derived growth factor-transformed porcine aortic endothelial (PAE) cells were electroporated as described previously (18). Staining of cells with ethidium monoazide (Molecular Probes, Inc., Eugene, OR) was performed following the manufacturer's protocol.
Microinjection-Microinjection was performed using an Eppendorf 5171 Micromanipulator with an Eppendorf 5246 Transjector. Plasmid DNA was diluted in 2ϫ injection buffer (100 mM HEPES (pH 7.2), 200 mM KCl, and 10 mM NaPO 4 ) to a final concentration of 25 ng/l. Transjector settings were injection pressure ϭ 60 hectopascals, compensatory pressure ϭ 20 hectopascals, and time of injection ϭ 0.1 s. Cells were incubated for 6 h following injection and were then fixed in 10% neutral buffered Formalin for 30 min prior to image analysis. Two methods were used for image collection and analysis. Conventional fluorescence microscopy was performed using a Nikon Microphot-SA microscope and camera. We also used the resources of the University of Pennsylvania Cancer Center Confocal Microscopy Core Facility. Confocal images were acquired from a TCS 4D upright microscope and processed on an IBM OS9 workstation using Scanware software. All light microscopic figures were shot at ϫ40 magnification.

RESULTS AND DISCUSSION
p110␥ Translocates to the Cell Nucleus of HepG2 Cells after Serum Stimulation-To begin to understand the role of PI3K␥ in vivo, we performed indirect immunofluorescence of transfected p110␥ in human HepG2 hepatoma cells. Liver cells naturally express PI3K␥; therefore, we reasoned that HepG2 cells should contain any accessory proteins needed for proper p101-p110␥ signaling. Since all available antibodies were unable to detect endogenous p110␥ by immunofluorescence, we expressed the HA epitope-tagged p110␥ that was recognized by the anti-HA monoclonal antibody 12CA5. Twenty-four hours after transfection, the cells were placed in medium without serum for 16 h. The cells were then washed, fixed, and stained with an anti-HA antibody.
As shown in Fig. 1 (A and B), when HepG2 cells were transfected with epitope-tagged p110␥ and analyzed under serum-deprived conditions, the cells appeared large and flat. Under these resting conditions, indirect immunofluorescence showed that p110␥ was present in a diffuse cytoplasmic pattern. In contrast, after stimulation of the cells with serum, immunolocalization of p110␥ revealed that it was no longer detected in the cytoplasm, but was found almost exclusively in the nucleus (Fig. 1, C and D). Confocal microscopy verified this finding and revealed that the transported PI3K␥ was diffusely present throughout the nucleus, but most concentrated at the nuclear membrane. Moreover, PI3K␥ staining did not coincide with simultaneous staining of the Golgi apparatus with the antibody G2404, a monoclonal antibody directed against the Golgi 58-kDa protein (data not shown). As has been reported previously, overexpression of PI3K did induce morphologic changes, including shrinking of the cytoplasm and ruffling of the cell membrane (5). Although cells overexpressing this protein had dramatic changes in their appearance, they were still viable. This was demonstrated by the exclusion of the DNA-staining agent ethidium monoazide in the absence of cell permeabilization. In addition, the cells did not show any evidence of apoptosis by the terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling immunofluorescence assay. Therefore, in response to serum stimulation, overexpressed p110␥ induces dramatic morphologic changes and translocates to the nucleus.
Overexpression of p101, which is endogenously present in HepG2 cells, had no influence on p110␥ nuclear translocation and, under either resting or stimulated conditions, was present in both the cytoplasm and nucleus (data not shown). We performed time course experiments to determine how rapidly after cell stimulation p110␥ migrates to the nucleus. In these experiments, cells were transfected with plasmids encoding PI3K␥. Twenty-four hours later, the cells were serum-deprived for an additional 16 h. At this point, all of the PI3K␥ protein was cytoplasmic. The medium was then changed to Dulbecco's modified Eagle's medium containing 10% fetal calf serum, and the time course of nuclear translocation was measured using the addition of serum as the defined time 0. These experiments showed that the translocation of p110␥ began as rapidly as 2 h after serum exposure and was almost complete by 7 h, but FIG. 1. Effect of serum, G␤␥, and pertussis toxin on the nuclear translocation of p110␥. HepG2 cells were transiently transfected with HA epitope-tagged p110␥ with or without G␤␥. Twenty-four hours after transfection, the cells were placed in medium containing the supplements indicated below. After growing on coverslips overnight, the cells were fixed and stained with anti-HA antibody (12CA5). The top row are indirect immunofluorescence images, and the bottom row are phase images. The tips of the arrows are placed at the edge of the cytoplasmic membrane for each paired set. A and B, cells were grown overnight in serum-depleted medium. The p110␥ protein was present in a diffuse cytoplasmic pattern, with no p110␥ visualized in the central dark nucleus. C and D, cells grown in 10% serum overnight demonstrated that p110␥ was exclusively localized in the nucleus. E and F, cells were grown in 10% serum with 200 ng/ml pertussis toxin overnight. Pertussis toxin prevented the usual translocation of p110␥ to the nucleus after serum stimulation. G and H, cells were transfected with p110␥ and G␤␥ and then incubated with serum plus pertussis toxin. Pertussis toxin did not prevent the nuclear translocation of p110␥ initiated by overexpressed G␤␥ heterodimers. maximal after ϳ12 h. This time period is similar to the nuclear translocation seen with p110␣ in PC12 cells after stimulation with neuronal growth factor (5).
We next sought to determine whether G protein-coupled or growth factor receptor-mediated signaling pathways were responsible for the serum-induced nuclear translocation of p110␥. To examine this issue, we preincubated cells with pertussis toxin to inhibit the release of G␤␥ heterodimers from G␣ i -coupled receptors. Cells were then examined in the presence or absence of serum stimulation. In the absence of serum, pertussis toxin did not affect the diffuse cytoplasmic distribution of p110␥ in serum-starved cells (data not shown). However, pertussis-toxin inhibited the serum-induced p110␥ nuclear translocation (Fig. 1, E and F). This suggests that the seruminduced nuclear translocation of PI3K␥ is dependent on a G␤␥mediated signaling pathway. To test this hypothesis, we attempted to mimic serum-mediated nuclear translocation by overexpression of G␤␥ heterodimers. The overexpression of G␤␥ heterodimers resulted in the nuclear translocation of p101 even in the absence of serum stimulation (data not shown). As expected, pertussis toxin failed to alter the nuclear localization of p110␥ induced by overexpression of G␤␥ heterodimers (Fig.  1, G and H). Together, these observations suggest that serum contains a factor that initiates the nuclear translocation of p110␥ by causing release of G␤␥ heterodimers downstream from a G␣ i -coupled receptor.
To confirm this G␤␥-mediated translocation of p110␥, a plasmid directing the expression of a chimeric protein composed of GFP fused to the carboxy terminus of p110␥ was microinjected into HepG2 cells. Cells were analyzed 6 h after microinjection of the plasmid under serum-starved conditions. Similar to our observations with indirect immunofluorescence, under resting conditions, the GFP-p110␥ fusion protein was found in a diffuse cytoplasmic pattern (Fig. 2, A and C). In contrast, when plasmids that direct the expression of G␤␥ heterodimers were simultaneously microinjected with the GFP-p110␥ plasmid, the GFP fusion protein was exclusively localized in the nucleus (Fig. 2, B and D). Thus, in these cells, which endogenously express PI3K␥, there is a G␤␥-mediated transport of this lipid kinase from the cytoplasm to the nucleus.
We questioned whether cell lines derived from tissues that do not endogenously contain PI3K␥ also contain the necessary accessory proteins for PI3K␥ nuclear translocation. As shown in Fig. 3A, when p101 and p110␥ were transfected into stably platelet-derived growth factor-expressing PAE cells, staining for p110␥ demonstrated that it was found in both the cytoplasm and nucleus. In these platelet-derived growth factor-overexpressing cells, this pattern of distribution was independent of serum (data not shown). As shown in Fig. 3 (B and C), the intracellular localization was also independent of cell cycle as demonstrated by thymidine block (arresting cells at the G 1 /S boundary) or by thymidine block and release (arresting cells in S phase). In contrast, transfection of p101 and p110␥ into COS-7SH cells demonstrated an exclusively cytoplasmic distribution for p110␥ (data not shown). Together, this suggests that, like p101 and p110␥, the accessory proteins required for p110␥ nuclear translocation are not ubiquitously expressed.
Studies of the Interaction between p110␥ and p101-There is some controversy in the literature regarding the necessity of the p101 subunit for p110␥ function (7,9,10). We next sought to determine the role of p101 in the nuclear translocation of p110␥ by examining the intracellular localization of a p110␥ mutant unable to associate with p101. Since the interaction between p85 and p110␣ involves the amino terminus of p110␣, we first tested whether p110␥ variants lacking the amino terminus could associate with p101. To perform these experiments, we used a plasmid directing the expression of an EE epitope (EEEEYMPME)-tagged variant of p101, which allowed the immunoprecipitation of p101 along with its associated proteins with an anti-EE antibody. Full-length p110␥ or deletion variants of p110␥ were coexpressed with EE-p101 in COS-7SH cells. After lysis, p101 and its associated proteins were immunoprecipitated with an anti-EE antibody, fractionated by SDSpolyacrylamide gel electrophoresis, and immunoblotted with an anti-HA antibody. The anti-HA antibody recognizes the epitope added to all of the p110␥ variants. As shown in Fig. 4  (left panel), all of the variants coexpressed well with p101. In addition, full-length p110␥ (wild type (WT)) and the 34-amino acid deletion variant (p110␥ (⌬1-34)) both immunoprecipitated efficiently with p101 (Fig. 4, right panel). However, the variant that was missing the first 82 amino-terminal residues completely failed to immunoprecipitate with p101. This implies that a region of p110␥ critical for association with p101 is located between residues 35 and 82 in p110␥.
Having identified a p110␥ variant that did not associate with p101, we next tested whether this mutant translocated toward the nucleus in a fashion similar to wild-type p110␥. In contrast to the wild-type protein, p110␥ (⌬1-82) was found in the nucleus even in the absence of serum (Fig. 5, A and B). Overexpression of G␤␥ heterodimers had no apparent influence on its intracellular localization (data not shown). This implies that p101 plays a role in the regulation of the intracellular localization of p110␥ by retaining it in the cytoplasm in the absence of Plasmids directing the expression of p101 and p110␥ and the cells were analyzed after thymidine block and release. Cells were analyzed after 48 h of 10% serum stimulation alone (A) or following thymidine block to arrest cells at G 1 /S (B) or following thymidine block and release to S phase arrest (C). signals that initiate nuclear translocation. Consistent with this hypothesis, the p110␥-(⌬1-34) deletion variant, which does associate with p101, translocated from the cytoplasm to the nucleus in an agonist-stimulated fashion similar to the wildtype protein (data not shown). CONCLUSION Although PI3K␥ has been well described as a mediator of signaling events at the plasma membrane, this work suggests that PI3K␥ may also play a role at the nuclear membrane. Specifically, we have shown that factors present in serum cause p110␥ to move to the nucleus in transfected HepG2 cells. This response to serum can be inhibited by pertussis toxin and can be mimicked by overexpression of G␤␥ heterodimers. A mutant form of p110␥ that is unable to bind to p101 is constitutively localized in the nucleus.
These observations raise a number of issues, including the mechanism by which PI3K␥ is transported to the nucleus, the impact of PI3K␥ on nuclear signaling, and whether these findings also apply to growth factor-activated p85-p110␣/␤. The mechanism by which PI3K␥ is translocated to the nucleus at this point is unclear. This process certainly can be regulated by the release of G␤␥ heterodimers. Our studies suggest that p101 appears to regulate this translocation since a variant of p110␥ that does not associate with p101 is constitutively found in the nucleus. This is similar to the mechanism by which mitogenactivated protein kinase shuttles between the nucleus and the cytoplasm by alternatively interacting with cytoplasmic and nuclear retention (or anchoring) proteins (20). This model would imply sequences for nuclear import and retention, the existance and perhaps sequences for nuclear export and cytoplasmic retention within p110␥. Examination of the p110␥ sequence reveals two potential nuclear import signals: R 17 RRRR and K 806 KKP. Since the p110␥-(⌬1-34) variant is capable of nuclear translocation, it implies that R 17 RRRR is not critical. Whether K 806 KKP is required for nuclear transport is currently unknown, as is the role of second messenger formation.
The effect of PI3K on mitosis and survival is a topic of recent interest. Evidence derived from the use of inhibitors (or overexpressed effectors) of PI3K implies that a tight regulation of PI3K is critical for both cell growth and cell death. Most likely, this effect of PI3K is the result of an increase in lipid second messengers. Although cytoplasmic p101-p110␥ will phosphorylate PI, PI-4-P, and PI-4,5-P 2 , only cytoplasmic PI-3,4-P 2 and PI-3,4,5-P 3 appear to change significantly after G protein-coupled receptor stimulation. Nuclear membranes, which probably contain the substrate for nuclear lipid kinases, contain abundant quantities of PI, PI-4-P, and PI-4,5-P 2 . But at this point, the substrate for PI3K␥ in the nucleus remains to be determined. It has long been appreciated that individual phospholipid concentrations vary with the cell cycle. For example, Dobos et al. (21) demonstrated that PI-3-P is elevated during the G 2 /M phase of the cell cycle. Recently, it has also been suggested that the cell cycle may actually be influenced by phospholipid content (22)(23)(24). Consistent with this hypothesis, fibroblasts deprived of choline and synchronized in G 1 phase by serum starvation do not efficiently enter S phase after serum stimulation (25). This implies that phospholipid synthesis is required for S phase entry. It is possible that nuclear PI3K␥ influences the cell cycle by transiently elevating the PI-3-P concentration. Recent evidence suggests that p110␥ may also have protein kinase activity (19). This raises the alternative possibility that the substrate for nuclear PI3K␥ is a protein, instead of a lipid.
Although most reports in the literature show a cytoplasmic (or plasma membrane) intracellular distribution for p85, two reports have suggested that it may relocate to the nucleus after neuronal growth factor or H 2 O 2 stimulation (5, 6). Consistent with this observation, preliminary studies in our laboratory have demonstrated that hemagglutinin epitope-tagged p110␣ expressed in HepG2 cells will also translocate to the nuclear membrane after serum stimulation. 2 This implies that the observations of this report will extend to signaling initiated by growth factor as well as G protein-coupled receptors. FIG. 4. Residues 1-82 in p110␥ are required for interaction with p101. COS-7SH cells were transfected with HA epitope-tagged p110␥ variants and EE epitope-tagged p101. Cells were lysed and immunoblotted with the anti-HA antibody 12CA5. As shown on the immunoblot of the total cell lysates, all three p110␥ variants were expressed approximately equally, but only the wild type (WT) and ⌬1-34 variant could co-immunoprecipitate with p101. This implies that the interaction for p101 requires residues 35 through 82.
FIG . 5. Intracellular localization of p110␥ (⌬1-82). HepG2 cells were transiently transfected with HA epitope-tagged p110␥ (⌬1-82) and analyzed by immunofluorescence with the 12CA5 antibody after 16 h of serum deprivation. Shown are the indirect (A) and phase plus (B) immunofluorescence images of two transfected cells. Deletion of the first 82 residues from p110␥ induced it to localize in the cell nucleus in serum-deprived cells. This distribution is in contrast to the diffuse cytoplasmic localization of full-length p110␥ in serum-starved cells (see Fig. 1, A and B).