Perspective

New Responsibilities for the PI3K Regulatory Subunit p85α

See allHide authors and affiliations

Science's STKE  16 Jan 2001:
Vol. 2001, Issue 65, pp. pe1
DOI: 10.1126/stke.2001.65.pe1

Abstract

Class IA phosphoinositide-3 kinases (PI3Ks) are heterodimeric enzymes that regulate many signal transduction pathways. The p85 regulatory subunit recruits the p110 catalytic subunit to the membrane, where p110 phosphorylates inositol lipids. Recent studies present evidence for an additional role for p85α in the regulation of actin cytoskeleton. Okkenhaug and Vanhaesebroeck discuss these results and ask whether experiments describing p85α knockout mice need to be reinterpreted.

In this perspective, we discuss recent evidence suggesting that in addition to being a regulatory subunit for p110 phosphoinositide 3-kinases (PI3Ks), p85α can independently stimulate signaling pathways involved in actin cytoskeletal rearrangements (1, 2).

Cell surface receptors initiate intracellular signaling pathways by recruiting cytosolic signaling enzymes such as PI3Ks to the membrane. Class IA PI3Ks are heterodimers composed of a p110 catalytic subunit and a regulatory subunit. The regulatory subunit consists of tandem SH2 domains (http://www.mshri.on.ca/pawson/sh2.html) and an inter-SH2 region that is tightly bound to p110. The SH2 domains bind specific phosphotyrosines in the cytoplasmic domains of receptor proteins, thus recruiting the p110 catalytic subunit to the membrane, where it phosphorylates phosphoinositides (Fig. 1). These phosphorylated lipids serve as docking sites for a variety of signaling proteins that ultimately regulate cell cycle progression, cell growth, cytoskeletal changes, and vesicular transport (3) [see also the STKE Phosphoinositide 3-Kinase Pathway (http://www.stke.org/cgi/cm/CMP_6557)] (4).

In mammals, complexity arises by the expression of multiple class IA PI3K regulatory and catalytic subunits. There are three genes that encode at least five different regulatory subunit proteins (p85α, p55α, p50α, p85β, and p55γ) (Fig. 2) and three genes encoding three catalytic subunits (p110α, p110β, and p110δ) (5). Together, these can potentially form 15 distinct heterodimers. This begs the following important questions: whether each of the regulatory and catalytic subunit combinations exists in cells, whether they are functionally redundant, and if not, how are they different?

Of the many possible combinations, the p85α/p110α heterodimer has been the most intensely investigated, revealing a complex interplay between the PI3K subunits. When expressed without the regulatory subunit, p110α is catalytically active but unstable at 37°C (6). In a complex with p85α, p110α looses catalytic activity but gains thermal stability (6). Lipid kinase activity of p110α is recovered when the SH2 domains of the regulatory subunit bind to a tyrosine-phosphorylated protein (6). Thus, PI3K recruitment to tyrosine-phosphorylated proteins not only brings p110α near its lipid substrates in the membrane but also releases inhibition of catalytic activity imposed upon it by p85α.

Three of the regulatory subunits, p50α, p55α, and p85α, are derived from the same gene by alternate splicing of the mRNA, which yields distinct start codons (7). The two SH2 domains, the p110-binding inter-SH2 domain, and a proline-rich domain (P2) (Fig. 2) are shared by these isoforms. However, the NH2-terminal domains are divergent. The unique NH2-termini of the p50α and p55α regulatory subunits are 6 and 34 amino acids long, respectively (Fig. 2). Interestingly, after stimulation of cells with insulin, p110α recruited to the receptor by p50α is more active than p110α recruited by p85α (8-10). Whether this is because p50α exerts less inhibition upon p110α or because the unique NH2-terminal sequences in p50α can activate p110α directly is presently unclear (10). The p55α isoform is not recruited to the insulin receptor complex to the same extent as p50α and p85α, and it is therefore difficult to directly compare its effect on p110 catalytic activity during insulin stimulation (10). The regulatory subunits p85β and p55γ are encoded by separate genes and have a similar domain structure to that of p85α and p55α, respectively. Their involvement in insulin signaling has yet to be fully examined.

The fruit fly Drosophila melanogaster (11) and the nematode worm Caenorhabditis elegans each have a single PI3K regulatory subunit, which shares a similar structural organization to the p50α and p55 mammalian regulatory subunits. Thus, the longer p85 subunits in mammals, p85α and p85β, seem to have evolved at a later stage, raising the possibility that they have acquired additional functionality. The unique NH2-terminal regions of p85α and p85β consist of an SH3 domain (http://www.mshri.on.ca/pawson/sh3.html), a unique proline-rich sequence (P1), and a Bcr Homology (BH) domain that is homologous to the Rho guanosine triphosphatase activating protein (RhoGAP) domain of Bcr (Fig. 2). The SH3 domain interacts both with the proline-rich sequences within p85α, as well as with those found in other intracellular signaling proteins, such as SOS and Cbl. The greatest sequence divergence between p85α and p85β is in the BH domains (12), suggesting that these may dictate differential signaling by p85α and p85β.

The BH domain has been implicated in the complex network that exists between PI3Ks and small guanosine triphosphatases (GTPases) (Fig. 1). Small GTPases such as Cdc42, Rac, and Rho control actin cytoskeleton dynamics and also have gene-regulatory roles. Other small GTPases such as Ras and Rab5 are key regulators of cell proliferation and endosomal vesicular traffic, respectively. Small GTPases exist in a guanosine triphosphate (GTP)-bound, active state or a guanosine diphosphate (GDP)-bound, inactive state. Guanosine nucleotide exchange factors (GEFs) catalyze the transition from the GDP- to GTP-bound state. Bound GTP is then slowly hydrolyzed by the GTPase, a reaction sped up by GAPs.

Some evidence has been presented that the BH domain of p85α binds Cdc42 (Fig. 1), an interaction that may contribute to the activation of p110 (13). Because the BH domain of p85 is homologous to Rho family GAPs, it is perhaps not surprising that p85 can interact with Cdc42 and, possibly, other small GTPases. However, unlike bona fide RhoGAPs, the p85 BH domain does not activate the intrinsic GTPase activity of Cdc42 (13). The p110 subunits can also bind members of the small GTPase family, including Ras and Rab5 (14, 15). Ras interaction is thought to lead to p110 activation (16). To make things more complicated, Rac may also interact with PI3K; however, the PI3K subunit responsible for this interaction is unknown (17).

p85 and the Cytoskeleton

PI3Ks have been implicated in the regulation of distinct types of cytoskeletal rearrangements. These include the formation of membrane ruffles (highly dynamic thin veils of membranes protruding from the cell surface) and disassembly of stress fibers (intracellular long bundles of actin filaments).

Addition of platelet-derived growth factor (PDGF) to fibroblasts stimulates the formation of membrane ruffles, most likely through the activation of Rac (18). Upon binding PDGF, the PDGF receptor autophosphorylates on specific tyrosine residues that mediate the association between the PDGF receptor and class IA regulatory subunits, resulting in PI3K recruitment, activation, and lipid phosphorylation (Fig. 1). A mutant PDGF receptor that cannot bind PI3K does not activate Rac to induce ruffles, nor can these structures be induced in presence of the PI3K inhibitor wortmannin (19, 20). By these criteria, both the recruitment and activation of PI3K are required for ruffle formation induced by PDGF, presumably by virtue of the ability of PI3K-phosphorylated lipids to activate a Rac-GEF.

As with ruffle formation, the PDGF mutant that cannot bind class IA regulatory subunits also could not stimulate the disassembly of stress fibers (1). Surprisingly, however, wortmannin did not prevent stress fiber disassembly induced by the wild-type PDGF receptor. Therefore, recruitment of class IA regulatory subunits, but not activation of p110, appears to control stress fiber dynamics. Transfection studies demonstrated that the full-length p85α has the capacity to block stress fiber dissociation, but the shorter p55α and p50α subunits do not, pointing toward a role for the SH3, BH, or P1 domains, or some combination of these domains in actin cytoskeleton regulation. In a similar line of experiments, yet to be published, Hill et al. also determined that the NH2-terminal domains of p85α are required for cytoskeletal changes (2). More specifically, they found that both proline-rich domains and the BH domain but not the SH3 domain are required for cytoskeletal reorganization. All these experiments are consistent with the idea that p85α has functions in addition to the regulation of p110 catalytic activity.

How could the BH and proline-rich regions of p85α regulate cytoskeletal changes? Some evidence has been presented for a role for Cdc42 in this phenomenon. Indeed, disassembly of stress fibers correlated with GTP-loading and activation of Cdc42 and binding of Cdc42 to PI3K (1). It was not determined whether this interaction involved the p85α BH domain or not. If the BH domain were involved in this interaction (13), it is not clear how this would lead to Cdc42 activation. As mentioned above, the BH domain has, thus far, been considered as a domain that allows Cdc42 to activate PI3K, not vice versa (13). One possibility is that binding of p85α to Cdc42 protects this small GTPase from GAP activity. An alternative explanation could be that the proline-rich regions in p85α bind SH3 domain-containing proteins that regulate the activity of Cdc42. Clearly, the molecular basis for the proposed functional interaction between p85α and Cdc42 needs to be further investigated.

Lessons from Knockout Mice

Could the available p85α knockout mice provide more insight into the physiological role of the p85α NH2-terminus? The answer to this question is not straightforward, considering the complexity of the class IA PI3K system and the difficulty in targeting one subunit without affecting the others. Two knockout lines have been generated that target the pik3r1 gene (Fig. 2): one that lacks expression of p85α, p55α, and p50α altogether (21, 22) and one that lacks p85α but retains p50α and p55α expression (the latter is further referred to as p85α-only knockout) (23, 24). The main difference between the two knockouts is that the p85α-p55α-p50α triple knockout is often lethal because of extensive liver necrosis (although some mice survive), whereas the p85α-only knockouts are viable. Surprisingly, knocking out p85α, or all three isoforms, promoted PI3K signaling in some circumstances and reduced PI3K signaling in others.

Other than the liver necrosis, the phenotypes of the two different knockout mice are remarkably similar. They both exhibit reduced B cell development and activation, although T cells appeared normal (21, 23). The impaired B cell phenotype correlated with a reduced capacity to activate PI3K-dependent signaling pathways (21, 23). The action of insulin was also similarly affected in both lines of mice (22, 24). PI3K is considered to be essential for signaling through the insulin receptor tyrosine kinase (25), and the pik3r1 gene knockouts have been investigated in great detail for insulin-related phenotypes. In vitro, insulin-stimulated PI3K activity associated either with tyrosine-phosphorylated proteins, the various p110s, or the remaining regulatory subunits was reduced in both the p85α-only knockout and the p85α-p55α-p50α knockout mice. These observations are compatible with the overall reduced level of class IA regulatory subunits in these mice. The expression levels of p85β and p55γ in the p85α-p55α-p50α triple knockout mice were slightly increased, but not to levels that could compensate for the loss of p85α and its splice variants (22). The expression of p110 subunits was also reduced substantially, consistent with the evidence that PI3K catalytic subunits are unstable when regulatory subunits are limiting (6). On the other hand, the similar reduction of in vitro lipid kinase activity in both the p85α-only knockout and the p85α-p55α-p50α knockout mice is somewhat surprising. On the basis of the higher in vitro PI3K activity associated with p110/p50α compared with p110/p85α (8-10) and an expected lower competition of p50α for access to phosphorylated receptors in p85α-only knockout mice, one might have anticipated enhanced in vitro PI3K activity in the p85α-only knockout mice.

Both lines of mice had reduced blood glucose concentrations (22, 24). This correlated with increased insulin-stimulated glucose uptake into skeletal muscle and adipocytes in the p85α-only knockouts (24). This is unexpected because in vitro studies indicated a positive role for PI3K in the regulation of insulin-stimulated glucose uptake (25) and because of the reduced PI3K activity measured in vitro for the pik3r1 knockouts. Other observations are similarly indicative of unaltered or even increased PI3K action in these mice. Direct measurement of lipid products induced by insulin in adipocytes derived from p85α-only knockout mice showed enhanced and longer lasting PI3K-mediated lipid production than in adipocytes from wild-type mice (24). Insulin-stimulated activation of protein kinase B (PKB), the activity of which is PI3K-dependent (26), was unaffected in the livers and muscle of p85α-p55α-p50α triple knockout mice. Moreover, the p85α-p55α-p50α triple knockout mice had larger muscle fibers, which could also be a consequence of increased PI3K activity (27).

It is at present not clear how reduction of the cellular levels of p85α or all three splice isoforms could lead to enhanced PI3K signaling in some tissues, when immunoprecipitated PI3K activity is reduced. One possibility is that the in vitro lipid kinase assays do not adequately reflect the behavior of PI3K in vivo (28). One also cannot exclude the existence of a small pool of free, unstable, but nonetheless hyperactive p110 that escapes negative regulation by class IA regulatory subunits (6). Such hyperactive p110 subunits could be complexed with Ras, which might substitute for p85 in the stabilization of p110 (14) (Fig. 2). Altered subcellular localization, as a result of the lack of regulatory subunit binding, could also lead to disregulated p110 signaling. A more speculative explanation is that the protein kinase activity (as opposed to the lipid kinase activity) of p110s (29-31) is responsible for the relevant biological activities of insulin signaling in the knockout mice. Last but not least, compensatory or deregulated activities of the class II or class III PI3Ks may occur in vivo (5).

The fact that the p85α-only knockout does have a phenotype despite expression of p55α and p50α could be interpreted as evidence for unique requirements for the NH2-terminal domains of p85α. However, it may also be a quantitative effect, because the amount of p50α and p55α expressed in cells could be insufficient to compensate for the loss of p85α expression in terms of recruitment of p110s to receptors. The similarity in B cell signaling and insulin responsiveness observed between the phenotypes of the p85α-only and the p85α-p55α-p50α knockouts argues in favor of the latter interpretation. So far, therefore, the p85α-only knockout mice have not revealed any functions related to the ability of the p85α NH2-terminal domains to activate signaling independently of p110; however, cytoskeletal rearrangements have not been directly examined.

It would certainly be of interest to determine if the inhibition of stress fiber formation is reduced in fibroblasts from the p85α-only knockouts, as might be predicted by the results of Jimenez et al. (1). The results of such studies may be obscured by compensation from p85β expression. Therefore, crossing the p85α-only knockout mice with p85β knockout mice (32) may at least allow for the production of embryos, and thus fibroblasts, that lack both forms of p85. However, such studies may be difficult to interpret because it is difficult to distinguish the consequences of reduced lipid phosphorylation from those depending only on p85. Another approach to specifically address the function of NH2-terminal regions of p85α in vivo could be to reconstitute the p85α-only knockouts with transgenes encoding full-length p85α, but with mutations that inactivate the SH3, BH, or P1 domains.

Targeting of the p110 genes will be required to completely unravel p85/p110-dependent signal transduction. An attempt was made to knockout the p110α gene (33). Unfortunately, this effort was marred by several difficulties. The mice died as embryos after 10 days of gestation, up to which stage p110α mRNA was unaffected. The lack of good immunoblotting antibodies to p110α combined with the lack of sufficient tissue that early in development did not permit conclusive determination of whether p110α protein production was entirely eliminated or whether truncated forms were still produced. Moreover, p85α was highly overexpressed in the embryos. This suggests that not only is p85α required to prevent degradation of p110α, but that p110α in some way regulates the expression or abundance of p85α. Because of these difficulties, it was unclear whether the effects of the gene targeting were specific to p110α or secondary to dominant-negative effects of p110-free p85α, which could block access to receptors for p110β or p110δ in complex with their regulatory subunits.

The experiments described in this Perspective underline the complex and dynamic interactions between p110 and its regulatory subunits. Because there seems to be interdependence between p110 and the regulatory subunits for expression, we feel that further PI3K gene-targeting efforts should be designed to cause minimal disruptions of the p85/p110 complexes. One way to achieve this is by introducing point mutations (knock-ins) rather than gene deletions.

Fig. 1.

Tyrosine kinase receptors, such as the PDGF receptor, recruit the catalytic subunit of PI3K to the membrane through its associated regulatory subunit. The SH2 domains of the regulatory subunit bind phosphorylated tyrosines in the cytoplasmic domain of the receptor (or phosphorylated tyrosines in associated proteins, in the case of the insulin receptor). The p110 catalytic subunit can also associate directly with activated Ras. The lipids phosphorylated by the p110 catalytic subunit recruit GEFs, which can activate Rac and induce membrane ruffling. Data discussed in this Perspective suggest an alternative pathway in which the NH2-terminal domains of the p85 regulatory subunit control Cdc42 activation. Conversely, Cdc42 may regulate the activity of p110 through its association with p85.

Fig. 2.

Schematic diagram showing the main isoforms of the class IA PI3K regulatory subunits. Note that p85α, p55α, and p50α are derived from the same gene by alternative splicing. P1 and P2, proline-rich regions 1 and 2, respectively; iSH2, the inter-SH2 domain; N-SH2, the NH2-terminal SH2 domain; C-SH2, the COOH-terminal SH2 domain. The names in parenthesis represent other names by which the proteins are known. The p55α and p50α termini of 34 and 6 amino acids are shown in yellow and red, respectively. The NH2 terminus of p55γ (yellow) is most similar to that of p55α.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
View Abstract

Navigate This Article