N-terminal Domains of the Class IA Phosphoinositide 3-Kinase Regulatory Subunit Play a Role in Cytoskeletal but Not Mitogenic Signaling*

Phosphoinositide (PI) 3-kinases are required for the acute regulation of the cytoskeleton by growth factors. We have shown previously that in the MTLn3 rat adeno-carcinoma cells line, the p85/p110 a PI 3-kinase is required for epidermal growth factor (EGF)-stimulated lamellipod extension and formation of new actin barbed ends at the leading edge of the cell. We have now exam-ined the role of the p85 a regulatory subunit in greater detail. Microinjection of recombinant p85 a into MTLn3 cells blocked both EGF-stimulated mitogenic signaling and lamellipod extension. In contrast, a truncated p85(1–333), which lacks the SH2 and iSH2 domains and does not bind p110, had no effect on EGF-stimulated mitogenesis but still blocked EGF-stimulated lamellipod extension. Additional deletional analysis showed that the SH3 domain was not required for inhibition of lamellipod extension, as a construct containing only the proline-rich and breakpoint cluster region (BCR) homology domains was sufficient for inhibition. Although the BCR domain of p85 binds Rac, the effects of the p85 constructs were not because of a general inhibition of Rac signaling, because sorbitol-induced JNK activation in MTLn3 cells was not inhibited. These data show that the proline-rich and BCR homology domains of p85 are involved in the coupling of p85/p110

The regulation of cellular motility is important in a variety of physiological processes, ranging from wound healing to the metastatic behavior of transformed cells. We have used a metastatic breast cancer cell model, the MTLn3 cell, to study the role of phosphoinositide 3Ј-kinases in EGF 1 -stimulated motility (1). The acute regulation of the actin cytoskeleton by EGF requires the p85/p110␣ isoform of PI 3-kinase. MTLn3 breast cancer cells express similar levels of p85/p110␣ and p85/p110␤. However, EGF-stimulated lamellipod extension is blocked by microinjection of inhibitory antibodies to p110␣ but not p110␤ (2). Inhibition of p110␣ also blocks the production of new barbed ends at the leading edge of EGF-stimulated cells (2). Isoform-specific regulation of the cytoskeleton by class IA PI 3-kinases has also been described in macrophages and fibroblasts (3,4). These findings suggest that different PI 3-kinase isoforms signal differently within the same cell.
Although the mechanism by which PI 3-kinases affect the cytoskeleton are not clear, several potential pathways have recently emerged. Rho family GTPases are central in growth factor-mediated actin reorganization (5), and p85/p110 has been linked to Rac activation in platelet-derived growth factorstimulated cells (6). Furthermore, GTP exchange factors for Rac and CDC42 contain pleckstrin homology or (Fab1, YGL023, Vps27, EEAI) domains and are likely targets of 3-phosphoinositides (7)(8)(9). Potential effectors for Rac/CDC42mediated cytoskeletal signaling include (a) the actin-severing protein cofilin, whose activity is regulated by phosphorylation in a Rac-and PI 3-kinase-dependent manner (10 -13); (b) the N-Wiskott-Aldrich syndrome protein and Wave proteins, which mediate the CDC42/Rac-dependent binding and activation of the Arp2/3 complex of actin-nucleating proteins (14 -18); and (c) the small GTPase Arf6, which is activated by GTP exchange factors that contain pleckstrin homology domains specific for phosphatidyl-inositol (3,4,5)P 3 (19 -21). PI 3-kinase signaling depends on the production of phosphatidylinositol (3,4,5)P 3 at appropriate locations within the cell. This localized signaling is likely to be due in part to the binding of p85 SH2 domains to tyrosine-phosphorylated kinases or substrates. For example, the binding of p85/p110 to activated platelet-derived growth factor receptors leads to the internalization of p85/p110 into endocytic vesicles (22). However, in addition to the SH2 domains located in the C-terminal half of p85 (residues 333-724), p85 contains additional potential targeting domains in its N-terminal half (residues 1-333). These include an SH3 domain, two proline-rich domains (PRDs), and a domain homologous to the breakpoint cluster region (BCR) gene product (23). These domains may target PI 3-kinase to distinct regions of the cell or couple PI 3-kinase to distinct downstream responses. Consistent with this possibility, the short isoforms of p85 (p55␣, p50␣, and p55␥) do not contain the N-terminal domains, and appear to have distinct biological activities in intact cells (24,25).
To assess the roles of the protein-protein interaction domains within p85-(1-333) during EGF-stimulated cytoskeletal regulation, we microinjected recombinant p85 domains into intact cells. We find that a fragment of p85 (residues 82-333), containing just the PRD and BCR homology domains, inhibits lamellipod extension but not mitogenic signaling. These data point to a specific role for residues 82-333 of p85 in signaling to the actin cytoskeleton.

MATERIALS AND METHODS
Recombinant Proteins-GST-N17Rac was purchased from Cytoskeleton, Inc. (Denver, CO). Recombinant p85 constructs used in this study are summarized in Fig. 1. p85␣[R358A/R659A] has been described previously (26). p85⌬BCR (deletion of residues 146 -299) was provided by Dr. Christopher Rudd, Harvard University. Deletion of the PRDs from full-length p85 was accomplished by the method of Kunkel et al. (27) using oligonucleotides that deleted residues 84 -96 (nPRD), 303-314 (cPRD), or both. These constructs were then amplified by polymerase chain reaction using forward primers encompassing bases 1-21 of human p85␣ and reverse primers encompassing bases 1041-1021 and subcloned into pGEX2T (Amersham Pharmacia Biotech). Finally, the 82-333 fragment was amplified by polymerase chain reaction and subcloned into pGEX2t. Recombinant proteins were produced in BL-21 Escherichia coli and purified by affinity chromatography on glutathione-Sepharose (Amersham Pharmacia Biotech). Proteins were extensively dialyzed against phosphate-buffered saline and concentrated to 3 mg/ml using Centricon concentrators (Millipore). Protein purity and final concentration was assayed by SDS polyacrylamide gel electrophoresis and Coomassie Blue staining. The proteins were mixed with rabbit or mouse IgG (3 mg/ml final concentration) prior to injection to facilitate identification of injected cells.
Cell Culture and Microinjection-Culture conditions and microinjection protocols for MTLn3 cells have been described previously (2,28). BrdUrd incorporation was measured after 12 h of EGF stimulation as described (2). EGF-stimulated lamellipod extension in MTLn3 cells was measured as described previously (2), with the recovery time after microinjection as indicated. Cells were scored for lamellipod extension, and the number of cells extending lamellipodia was expressed as a percentage of the number of cells injected with each construct. All data are the mean Ϯ S.E. from at least three experiments.
JNK Activation-Cells were injected as indicated and allowed to recover for 2 h. The cells were incubated in the absence or presence of 1 M sorbitol for 30 min, fixed with 10% paraformaldehyde for 30 min at room temperature, and permeabilized with methanol on dry ice for 10 min. After blocking in 1% bovine serum albumin/5% donkey serum, the cells were stained with anti-active JNK antibodies (Promega) followed by Cy3 anti-donkey antisera, to measure JNK activation, or FITC anti-mouse antibody, to identify injected cells. The data are representative of 2 separate experiments.
Imaging-Images were acquired using a Nikon Eclipse 400 fluorescence microscope with Nikon CFI Plan Apo 60 ϫ 1.4 numerical aperture optics and a Cohu charge-coupled device camera linked to a Scion VG5 frame grabber. Figures were assembled using Adobe PhotoShop.
Statistical Methods-Statistical analysis was performed using ANOVA and Tukey HSD tests, using software from Dr. Richard Lowry, Vassar College.

Inhibition of EGF-stimulated Lamellipod Extension by
Recombinant p85-Overexpression of p85 has been shown previously to inhibit membrane ruffling in Ras-transformed fibroblasts (29). Although some of this inhibition could be due to the titration of intracellular phosphotyrosine residues by the p85 SH2 domains (30), the N-terminal half of p85 also contains numerous protein-protein interaction domains (SH3, prolinerich, and BCR homology domains). Thus, p85 overexpression could also inhibit ruffling by saturating the intracellular targets of these domains.
To study the function of these N-terminal domains of p85, we made GST fusions of wild-type p85 or p85 containing mutations in the conserved FLRV motifs in both p85 SH2 domains. We have shown previously that the R358A/R659A mutations abolish phosphopeptide binding by the p85 SH2 domains (26). We reasoned that even in the absence of functional SH2 domains, these constructs could interfere with interactions between p85 and intracellular targets and block PI 3-kinase-dependent signaling. Purified GST-p85 or GST-p85(R358A/R659A) were microinjected into quiescent MTLn3 cells. After a 2-h recovery, the cells were stimulated with EGF for 3 min, fixed, stained, and scored for lamellipod extension. Both constructs caused a significant inhibition of lamellipod extension; microinjection of the mutant GST-p85(R358A/R659A) inhibited lamellipod extension by ϳ52%, relative to control, whereas GST-p85 inhibited lamellipod extension by 77% (Fig. 2). These data suggest that SH2-independent interactions between p85 and intracellular proteins are required for EGF-stimulated lamellipod extension.
p85-p110 binding is extremely stable (31), and the exchange of p110 between different p85 molecules is presumably slow. Nonetheless, we wanted to rule out the possibility that endogenous p110 was being sequestered by mutant p85 during the 2-h period after injection. We therefore repeated the experiments but stimulated the cells with EGF 10 min after microinjection. Once again, GST-p85(R358A/R659A) inhibited lamellipod extension nearly as well as wild-type p85 (Fig. 2B).
N-terminal Domains of p85 Are Involved in Cytoskeletal but Not Mitogenic Signaling-The data in Fig. 2 show that recombinant p85 inhibited lamellipod extension even in the absence of functional SH2 domains. This suggested that the N-terminal domains of p85 might also be critical for coupling of p85/p110 to the regulation of the actin cytoskeleton. To test this directly, we constructed a truncated GST-p85-(1-333), which lacks the SH2 and iSH2 domains and cannot bind p110 (Fig. 1). When microinjected into MTLn3 cells, GST-p85-(1-333) inhibited EGFstimulated lamellipod extension almost as well as full-length GST-p85. Microinjection of GST had no effect on lamellipod extension (59% of GST-injected cells responded to EGF, as compared with 62% of uninjected cells). However, microinjection of GST-p85 and GST-p85-(1-333) reduced EGF-stimulated lamellipod extension to 25 and 33%, respectively (Fig. 3A).
We next tested whether the N-terminal fragment of p85 could inhibit other PI 3-kinase-dependent signaling pathways. We have shown previously that EGF-stimulated BrdUrd incorporation in MTLn3 cells is dependent on class IA PI 3-kinases (2). Basal BrdUrd incorporation in serum-starved MTLn3 cells was ϳ40% and was unaffected by microinjection of recombinant proteins (data not shown). In contrast, EGF-stimulated BrdUrd incorporation is almost completely blocked by microinjection of recombinant p85 (Fig. 3B). However, microinjection of p85-(1-333) had no significant effect on EGF-stimulated Br-dUrd incorporation.
The BrdUrd incorporation assay requires a longer post-injection incubation (12 h) than is used in the lamellipod extension assays. To rule out the possibility that the BrdUrd results reflected the degradation or inactivation of p85 (1-333), we conducted lamellipod extension experiments using the same protocol as in the BrdUrd assays. Quiescent cells were microinjected with GST, GST-p85-(1-333), or GST-nSH2 domains (a positive control for inhibition of lamellipodia) and then incubated for 12 h prior to stimulation with EGF for 3 min (Fig. 3C). Once again, the p85-(1-333) fragment markedly inhibited lamellipod extension. Thus, our data show that GST-p85-(1-333), which lacks SH2 domains but contains SH3, PRD, and BCR homology domains, can selectively interfere with EGF-stimulated cytoskeletal signaling but not EGF-stimulated DNA synthesis.
Deletional Mapping of N-terminal Domains Involved in Cytoskeletal Signaling-To determine which regions of p85-(1-333) were required for lamellipod extension, we prepared constructs in which we selectively removed the SH3 domain, the nPRD, the cPRD, both PRDs, and the BCR homology domain. The proteins were soluble after expression in BL-21 E. coli, and their electrophoretic mobility on SDS polyacrylamide gel electrophoresis was consistent with their predicted size (data not shown). The purified proteins were injected into MTLn3 cells, and the effects on lamellipod extension were evaluated. GST-p85-(82-333), which lacks the SH3 domain, could still inhibit lamellipod extension (Fig. 5). In con- FIG. 3. A truncated p85-(1-333) inhibits lamellipod extension but not BrdUrd incorporation. A, MTLn3 cells were incubated without serum for 2 h and then injected with GST, GST-p85, or GST-p85-(1-333) as indicated. After an additional 2 h the cells were stimulated for 3 min with 10 nM EGF, and lamellipod extension was determined as described above. Ctl, control. B, MTLn3 cells were rendered quiescent in medium containing 1% fetal bovine serum for 24 h and then injected with GST, GST-p85, or GST-p85-(1-333) as indicated. The cells were then incubated in the absence or presence of 2 nM EGF for 12 h, followed by BrdUrd for an additional 2h. The cells were fixed and stained with anti-BrdUrd antibodies or FITC anti-rabbit antibodies (to identify injected cells), and the percentage of cells that incorporated BrdUrd was counted. C, MTLn3 cells were rendered quiescent in medium containing 1% fetal bovine serum for 24 h and then injected with GST, GST-nSH2 domains, or GST-p85-(1-333) as indicated. The cells were incubated for 12 h and then stimulated with 10 nM EGF for 3 min. Lamellipod extension was determined as described above. All values are the mean Ϯ S.E. of three to five determinations. ANOVA analysis demonstrated significance at p Ͻ 0.01; statistical significance of differences between individual means and GST-injected cells (Tukey HSD test) are indicated.
trast, deletion of either the nPRD or the cPRD, or both PRDs, eliminated the inhibitory effect on lamellipod extension (Fig. 5). These relatively small deletions of 11-12 amino acids were unlikely to cause significant structural perturbation, because their deletion from full-length p85 had no effect on binding to p110 or phosphopeptide activation of p85/p110 dimers (data not shown). Preliminary results also show that a construct lacking the BCR domain was unable to inhibit EGF-stimulated lamellipod extension (data not shown). Although we cannot yet state with certainty that the BCR and PRD domains are necessary for inhibition of lamellipod extension, our data show that the nPRD-BCR-cPRD fragment is sufficient for inhibition of lamellipod extension.
p85 Constructs Do Not Inhibit Rac-dependent Signaling in MTLn3 Cells-EGF-stimulated lamellipod extension is blocked by dominant negative Rac, 2 consistent with data from other systems (32). The BCR homology domain of p85 binds activated Rac and CDC42 (33,34). Because the nPRD-BCR-cPRD domain is sufficient to block lamellipod extension, we considered the possibility that p85-(1-333) was acting as a generalized Rac inhibitor, through the sequestration of activated Rac. We therefore measured activation of the JNK kinase, which is known to be Rac-dependent (35,36). MTLn3 cells were treated with sorbitol for 30 min and then fixed and stained with a phosphospecific anti-JNK antibody. As reported previously in other systems (37), sorbitol treatment lead to the activation of JNK, which could be detected in the nucleus of MTLn3 cells (Fig. 6,  A and B). The nuclear accumulation of active JNK was unaffected by microinjection of GST-p85-(82-333) (Fig. 6B, arrows), whereas it was completely blocked by microinjection of N17-Rac (data not shown). Thus, the Rac-dependent activation of JNK was not inhibited by GST-p85-(82-333), suggesting that the affects of this construct on EGF-stimulated lamellipod extension were not because of a general sequestration of endogenous activated Rac.
In this paper, we have defined a subset of PI 3-kinase-dependent responses that are inhibited by microinjection of Nterminal domains of p85. The p85-(1-333) and p85-(82-333) constructs do not contain the SH2 or iSH2 domains and therefore should not disrupt endogenous p85/p110 binding or activation of endogenous p85/p110 by phosphotyrosine-containing proteins. We therefore presume that these constructs interfere with the targeting of endogenous p85/p110 molecules to sites involved in cytoskeletal regulation. Disruption of this targeting apparently does not interfere with the activation of mitogenic signaling pathways, which may be less spatially organized. Consistent with this latter idea, we have demonstrated previously an increase in DNA synthesis by anti-p85 antibodies that activate p85/p110 dimers but that would be unlikely to cause the targeting of p85/p110 to specific intracellular sites (28).
Although the nPRD-BCR-cPRD fragment of p85 is sufficient to inhibit lamellipod extension, the intracellular targets of these domains of p85 are not yet known. Although the BCR homology domain binds to Rac/CDC42 (33,34), microinjected p85 fragments do not block JNK activation and are therefore not acting as a global Rac inhibitor. However, the p85-derived constructs could disrupt the targeting of a subset of activated Rac to specific regions of the cell.
In addition to Rac and CDC42, a number of cytoskeletal regulatory proteins interact with p85. Cas and focal adhesion kinase bind to the SH2 domains of p85 (38,39) and should not therefore be affected by p85- (82-333). Similarly, cbl binds to the SH3 domain of p85 and should be unaffected by p85-(82-333) (40). Tyrosine-phosphorylated ezrin also binds, to the p85 SH2 domains, but additional binding occurs through the N 2 M. Bailly and J. E. Segall, unpublished results.  6. GST-p85-(82-333) does not inhibit Rac-dependent signaling. Quiescent MTLn3 cells were not injected (A) or injected (B) with GST-p85-(82-333). After 2 h the cells were incubated without (A) or with (B) 1 M sorbitol for 30 min, fixed, and stained with anti-phospho-JNK antibodies followed by Cy3 anti-rabbit antibodies and FITC anti-mouse antibodies (to identify injected cells). Arrows indicated injected cells.
terminus of ezrin to an unknown region of p85. The GTP exchange factor Pak interacting exchange factor and the actinbinding protein profilin also bind to p85 at unknown sites and could be affected by microinjection of p85-(82-333) (41,42). Finally, p85-(82-333) could act by binding to SH3 domains in Src and related kinases (43,44), thereby disrupting their interactions with endogenous p85.
In summary, we have demonstrated that different regions of the p85 regulatory subunit are involved in distinct PI 3-kinasedependent responses. Whereas the SH2 domains of p85 have pleiotropic effects on multiple pathways, the BCR and prolinerich domains of p85 are coupled to cytoskeletal signaling but not DNA synthesis. These data suggest that different isoforms of class IA regulatory subunit, particularly the full-length (p85) versus short forms (p55/p50), are involved in different subsets of PI 3-kinase-dependent signaling events.