Class IA PI3Ks (phosphoinositide 3-kinases) consist of a p110 catalytic subunit bound to one of five regulatory subunits, known as p85s. Under unstimulated conditions, p85 stabilizes the labile p110 protein, while inhibiting its catalytic activity. Recruitment of the p85–p110 complex to receptors and adaptor proteins via the p85 SH2 (Src homology 2) domains alleviates this inhibition, leading to PI3K activation and production of PIP3 (phosphatidylinositol 3,4,5-trisphosphate). Four independent p85 KO (knockout) mouse lines have been generated. Remarkably, PI3K signalling in insulin-sensitive tissues of these mice is increased. The existence of p110-free p85 in insulin-responsive cells has been invoked to explain this observation. Such a monomeric p85 would compete with heterodimeric p85–p110 for pTyr (phosphotyrosine) recruitment, and thus repress PI3K activity. Reduction in the pool of p110-free p85 in p85 KO mice was thought to allow recruitment of functional heterodimeric p85–p110, leading to increased PI3K activity. However, recent results indicate that monomeric p85, like p110, is unstable in cells. Moreover, overexpressed free p85 does not necessarily compete with heterodimeric p85–p110 for receptor binding. Using a variety of approaches, we have observed a 1:1 ratio between the p85 and p110 subunits in murine cell lines and primary tissues. Alternative models to explain the increase in PI3K signalling in insulin-responsive cells of p85 KO mice, based on possible effects of p85 deletion on phosphatases acting on PIP3, are discussed.
Class I PI3Ks (phosphoinositide 3-kinases) generate PIP3 (phosphatidylinositol 3,4,5-trisphosphate), which interacts with specialized lipid-binding domains present in protein kinases [such as Akt/PKB (protein kinase B)], regulators of small GTPases and scaffolding proteins. This allows PI3Ks to control a wide variety of cell biological responses such as growth, proliferation and migration [1,2].
The class IA subset of PI3Ks signal downstream of receptor tyrosine kinases (including the insulin receptor) and Ras [1,2]. In mammals, these PI3Ks consist of a 110 kDa catalytic subunit (p110α, p110β or p110δ) and a smaller regulatory subunit, of which there are five isoforms (p85α, p85β, p55γ, p55α and p50α; collectively referred to as ‘p85s’; [3,4]). In resting cells, p85 stabilizes the thermally labile catalytic subunit (Figure 1) and conformationally inhibits its lipid kinase activity . The interaction of p85 with p110 is very strong, and can withstand high concentrations of salt, urea or detergent [6,7], indicating that dissociation of p85 and p110 under physiological conditions is unlikely. In activated cells, recruitment of the p85–p110 complex to pTyr (phosphotyrosine) residues in activated receptor or adaptor molecules by means of the SH2 (Src homology 2) domains in p85 brings the p110 subunit in proximity to its lipid substrates and relieves the inhibition of p85 on p110, allowing the latter to generate PIP3 (Figure 1 and ). Moreover, pTyr engagement renders the p85–p110 complex responsive to Ras .
The role of the p85 regulatory subunit in the regulation of class IA PI3K activity
Studies in class IA PI3K KO (knockout) mice revealed that protein expression of p110 and p85 subunits is most often interlinked. Indeed, deletion of p85α leads to a concomitant decrease in the levels of p110α, p110β and p110δ [10–13]. Reciprocally, heterozygous loss of both p110α and p110β, or homozygous loss of p110α or p110δ leads to reduction in the levels of p85/p55/p50 [14–16]. However, in p85β KO mice, p110 levels are unaltered, despite an apparent 20–30% reduction in overall levels of p85 .
The ‘free p85’ hypothesis to explain increased insulin-induced PI3K signalling in p85 KO mice
PI3K catalytic activity positively controls insulin signalling [18–20] and in vitro PI3K lipid kinase activity present in PI3K immunoprecipitates is reduced in insulin-responsive cells of p85 KO mice [11,21,22]. It therefore came as a surprise to find increased insulin-stimulated PIP3 generation and Akt/PKB activation in p85 KO tissues [10,11,17,22].
In order to reconcile these observations, a hypothesis based on the existence of p110-free p85 in insulin-responsive tissues has been put forward (reviewed in ). This was based on the indication of a ≥30% stoichiometric excess of p85 over p110 in wild-type MEFs (mouse embryo fibroblasts)  and liver cells . An excess of class IA regulatory subunits was also reported in nutritionally induced insulin resistance in humans , in women with pregnancy-induced insulin resistance , and in Type 2 diabetic individuals . Subsequent analysis revealed that forced overexpression of class IA PI3K regulatory subunits in skeletal-muscle cells leads to a decrease in Akt/PKB phosphorylation , indicating that an excess of p85 might compete with p85–p110 for pTyr-binding sites.
These observations led to the model that in wild-type cells, a surplus of p110-free p85 exists, which competes with heterodimeric p85–p110 for pTyr-binding sites, thereby dampening PI3K signalling. In p85 KO cells, the levels of free p85 were thought to be preferentially reduced, leading to enhanced access of lipid-kinase-competent, heterodimeric p85–p110 complexes to pTyr-binding sites and thus to increased PI3K signalling.
However, several different mechanisms must exist to regulate PI3K signalling, given that not all cell types respond to p85 deletion in the same manner. Indeed, leucocyte responses (to stimuli other than insulin) are not enhanced in p85α-null mice but are in most cases reduced or unaffected, despite a similar reduction in p85 and p110 protein levels as in insulin-responsive cells (reviewed in ). In p85β KO mice, however, T-cell responses are increased  but this happens without alterations in Akt/PKB activity . This effect may therefore be independent of PI3K lipid kinase activity. It has not been reported whether p85 exists in excess of p110 in leucocytes. It is also interesting to note that deletion of p85α selectively in the liver leads to an increase in Akt/PKB activity and phosphorylation upon insulin stimulation in these cells , whereas selective deletion of p85α in muscle cells does not affect Akt/PKB responses to insulin .
Cracks in the ‘free p85’ hypothesis
Recent experimental evidence argues against the model of ‘surplus’ p85 over p110:
(i) Monomeric p85 is unstable. Pulse–chase experiments of cells transiently overexpressing p85α with or without p110α revealed that the half-life of p110-free p85α is at least three times less than that of p85α bound to p110α .
(ii) Loss of p110α or p110β expression leads to concomitant reduction in p85 expression levels, as is seen upon constitutive deletion of p110α or p110β, or conditional ablation of p110α expression [14,15].
(iii) Introduction of increasing amounts of free p85α in pan-p85α KO adipocytes does not decrease insulin-stimulated PI3K activity in pTyr immunoprecipitates, but decreases Ser473 phosphorylation of Akt/PKB , indicating that p85α can decrease PIP3 levels through mechanisms that are independent of its capacity to compete for pTyr-docking sites (this may be through an effect of p85 on PIP3 phosphatases, see below).
(iv) We have not found biochemical evidence for p110-free p85 (see below).
Analysis of the absolute amounts of class IA PI3K proteins
Quantification of absolute protein amounts in cells is challenging. Evidence for the existence of monomeric p85 has so far largely been inferred from immunodepletion and immunoblotting experiments. Thus the available data are indirect and formal proof of the existence of p110-free p85 is lacking. A possible confounding issue is that the available antibody reagents for the p85 subunits are superior to those against the p110 subunits. In other words, it is easier to detect p85 than p110 using the available reagents.
We have applied ion-exchange chromatography, immuno-precipitation and immunodepletion to monitor the presence of monomeric and dimeric class IA PI3Ks. We also determined the absolute protein amounts of each class IA PI3K isoform by quantitative MS, using an absolute protein quantification method [34–36] that makes use of isotopically labelled internal standards spiked at known concentrations into protein mixtures. This internal standard peptide has the same sequence as a section of the target protein, except that it is enriched in stable isotopes. Analysis of the proteolysed sample by MS results in the detection of both the endogenous and isotope-labelled internal standard peptides. These data are then used to derive absolute quantification of endogenous protein (in molecules per cell or mole per gram of tissue). Using these approaches, no evidence for free p85 or p110 could be obtained in mouse fibroblast and leucocyte cell lines and in a variety of murine tissues (B. Geering, P.R. Cutillas, G. Nock, S.I. Gharbi and B. Vanhaesebroeck, unpublished work). Taken together, these observations indicate that class IA PI3Ks are obligate heterodimers.
Alternative models for increased insulin signalling in p85 KO mice
If an excess of p85 over p110 cannot be invoked to explain increased PI3K signalling in insulin-responsive cells of p85 KO mice, how then might this phenomenon arise? In principle, this could be due to increased activity of the remaining PI3Ks and/or decreased lipid PIP3 activity in these cells.
As mentioned above, lipid kinase activity in PI3K immunoprecipitates is reduced in p85 KO cells [11,21,22], in line with the often reduced levels of p110 subunits. This suggests that the intrinsic lipid kinase activity of the remaining PI3K subunits is not enhanced, for example due to the fact that p110 could now be monomeric (and thus un-inhibited by p85) or because expression of the other (not gene-targeted) p85 subunits is increased and the remaining p110s are now in complex with other p85 isoforms (‘p85 isoform switching’ of p110s). Taken together, these results suggest that insulin signalling is increased in p85 KO cells, independent of alterations in p110 activity.
It is therefore tempting to speculate that the higher PIP3 levels seen upon insulin stimulation of p85 KO cells are due to reduced PIP3 phosphatase activity. Indeed, PTEN (phosphatase and tensin homologue deleted on chromosome 10) activity is decreased in p85α-null liver cells, without alteration in PTEN expression levels . SHIP (SH2-containing inositol phosphatase) activity in p85 KO cells has not been investigated, but its expression level was unaltered in p85α+/− MEFs .
How could loss-of-expression of p85 affect lipid phosphatases? p85 mediates protein–protein interactions [3,29] which could influence lipid phosphatase binding and recruitment, both directly and indirectly, as discussed below.
p85–p110 complexes can bind lipid phosphatases, including SHIP , or might become engaged in signalling complexes that contain lipid phosphatases such as SHIP [38,39] and other 5-phosphatases , the 3-phosphatase PTEN  or 4-phosphatases . Loss-of-expression of p85 may therefore lead to reduced recruitment of phosphatases to receptor complexes.
In addition, proteins that under normal conditions do not bind IRS (insulin receptor substrate) (for example due to lower binding affinities compared with p85) might engage with IRS when p85 is missing (Figure 2). These include SH2 domain-containing proteins such as SHIP (which would admittedly increase PIP3 phosphatase activity), but also molecules like the JAK (Janus kinase) tyrosine kinase [43–45]. JAKs can interact with and phosphorylate IRS-1/IRS-2 [46,47] and SHIP . Such a ‘facilitated’ interaction of JAK with IRS might explain the observed increase in IRS-2 tyrosine phosphorylation in p85 KO mice [17,22]. Moreover, interaction of JAK with SHIP could lead to SHIP phosphorylation and possibly its inactivation, as shown to occur downstream of high-affinity IgE and insulin receptors [49,50].
Hypothetical model explaining increased PI3K signalling in p85 KO cells
An important question is why loss of p85 appears to decrease PIP3 phosphatase activity in insulin-responsive tissues only. Could it be that the latter express ‘p85-responsive’ PIP3 phosphatases, the expression of which is restricted to insulin-responsive cells? Candidates include the 5-phosphatases SKIP (skeletal-muscle- and kidney-enriched 5-inositol phosphatase) and hSac2, which appear to have activity towards PIP3 [51–54]. Furthermore, the expression of SHIP isoforms varies in different cell types. Haemopoietic cells express two isoforms of SHIP, namely SHIP1, which is only found in leucocytes, and SHIP2, which is widely expressed, including in insulin-sensitive tissues. Tissue-specific regulation of SHIP by p85 may contribute to the observed differences between insulin-responsive cells and leucocytes. Phosphorylated IRS molecules also recruit a different set of proteins in leucocytes and insulin-responsive tissues and cells .
The current model of monomeric p85 as a negative regulator in class IA PI3K signalling does not explain all the available data and new findings indicate that this hypothesis may need reconsideration. Although it is possible that the balance of catalytic and regulatory subunits is distinct in different tissues and cells types, several independent studies (including our own) do not support the existence of p110-free p85 in the tissues and cells that have previously been documented to contain free p85. New methodologies, including technologies with enough resolution and sensitivity to determine absolute protein amounts in defined cell populations [56–58], will be needed to resolve these discrepancies. Alternative models to explain the insulin hypersensitivity observed in p85 KO mice invoke alterations in protein–protein interactions as a consequence of diminished p85 expression levels, with as the end result a reduction of PIP3 phosphatase activities. Experiments to test these models are in progress. These will hopefully lead to a better understanding of the role of the regulatory subunits in PI3K signalling.
3rd Focused Meeting on PI3K Signalling and Disease: Biochemical Society Focused Meeting held at Bath Assembly Rooms, U.K., 6–8 November 2006. Organized and Edited by B. Hemmings (Friedrich Miescher Institute for Biomedical Research, Switzerland), B. Vanhaesebroeck (Ludwig Institute for Cancer Research, U.K.), S. Ward (Bath, U.K.) and M. Welham (Bath, U.K.).
insulin receptor substrate
mouse embryo fibroblast
protein kinase B
phosphatase and tensin homologue deleted on chromosome 10
Src homology 2
SH2-containing inositol phosphatase
We thank Katja Björklöf, Lazaros Foukas and Klaus Okkenhaug for critically reading this paper. Personal support was from the Ludwig Institute for Cancer Research (B.G. and B.V.), Roche Research Foundation, Switzerland (B.G.), Overseas Research Scheme U.K. (B.G.), Uarda Frutiger-Fonds, Switzerland (B.G.), Janggen Poehn Stiftung, Switzerland (B.G.) and the Association for International Cancer Research U.K. (P.R.C.).