The insulin signaling story as we know it currently ends at the level of the insulin receptor-substrate proteins (IRS proteins). Last month, we used the IRS proteins as a platform from which to explain the importance of isoforms in partitioning messages to different pathways, leading to unique end effects. As the reader might recall, this specificity is achieved through a combination of characteristics specialized to certain isoforms, including modular subunits, amino acid targeting motifs, and subcellular localization. Additionally, the role of the IRS proteins in amplifying the insulin signal was mentioned; this function in conjunction with isoform-specific motifs was postulated to confer distinctness with respect to signaling proteins recruited by the IRS proteins. We ended the third segment of this series by briefly mentioning some physiologically significant roles for IRS protein isoforms, which appear to be tissue-specific. In this installment, we follow insulin’s signal immediately downstream of the IRSs by reviewing the role of the phosphatidylinositol 3-kinase (PI3K) family of enzymes.
The PI3K family of lipid-protein kinases is a critical group of enzymes not only in insulin signaling, but in a multitude of other pathways, some not even related to receptor tyrosine kinases (1, 42-46). In fact, we see PI3K activation in response to androgens, estrogens, thyroid hormones, leptin, and certain G-protein-coupled receptor agonists (1-3, 42-46), linking these kinases to three diverse signaling families, in addition to being coupled to receptor tyrosine kinases (RTKs). One could easily argue that studying the PI3K family of enzymes will yield greater rewards than looking at almost any other intermediate in understanding how cells regulate a multitude of seemingly intertwined signaling pathways that converge/diverge at a few critical points. So, without further ado, we will look more closely at the roles of phosphoinositides and their enzymes in mediation of cellular signaling.
In regions of the cell where membranes are present, including the plasma membrane, are groups of modified fatty acids called phospholipids. As most readers probably know, triglycerides (AKA triacylglycerols) are a family of compounds that consists of three fatty acids attached to a glycerol backbone. Should one fatty acid be removed, we end up with a diglyceride, more commonly referred to as a diacylglycerol (DAG, which is a primary group of fatty acids in the cooking oil Enova®). Addition of a phosphate and a (relatively) hydrophilic head group results in a phospholipid. Phospholipids can be further modified depending on the head group, and for the phosphatidylinositol 3-kinases, the head group of interest is inositol.
Inositol is a glucose derivative that differs from the other phospholipid head groups in its ability to influence many, many signaling pathways (1-4). When modified through phosphorylation of various positions in its ring, the inositol moiety of phosphatidylinositol serves to recruit proteins with certain domains (Pleckstrin Homology) as well as function as a substrate for generation of unique phosphorylated derivatives. [Furthermore, the phosphorylated inositol can be cleaved from its glycerol backbone and regulate additional cellular effects -mainly having to do with calcium signaling- but this will not be covered in the name of brevity.] What this all means is that the first level of control of signaling through the PI3Ks occurs during the synthesis of differently phosphorylated inositol moieties.
There are multiple positions in the inositol ring capable of accepting a phosphate. As a group, these phosphorylated derivatives are called “phosphoinositides.” Normally, intracellular pools of phosphoinositides are maintained through the efforts of a family of enzymes called the phosphatidylinositol kinases (PIKs), NOT to be confused with phosphatidylinositol 3-kinases (PI3Ks). This means that instead of plain old phosphatidylinositol, depots of phosphatidylinositol-X-phosphates are present (where X = a position on the inositol ring to which a phosphate is attached, usually three, four, or five) (1, 4, 5, 6).
An important point to note is that the activity and cellular localization of PIKs appears to be coupled to receptor activation, and in particular, receptor tyrosine kinases (4, 5). This is significant because PIKs generate the substrates for the PIPKs (phosphoinositide-phosphate-kinases). PIPKs then take the products of PIKs and add additional phosphates to unphosphorylated positions on the inositol ring to make new phosphoinositides that can function as secondary messengers (4, 5, 10). Briefly:
This means that PI3K is a member of the PIPK family of enzymes, with the “3” signifying its effect, which is phosphorylation of PI(X)Ps at the 3-position (where X obviously does not equal 3).
If it appears that this two-layered mechanism (PIK and PIPK) of regulation is overly complex, it is only because phosphoinositides are implicated in diverse functions including, but certainly not limited to nutrient uptake, cellular trafficking, protein synthesis, apoptosis, vesicular secretion, cytoskeleton assembly, and nuclear signaling (1-4, 10, 45). Remember that phosphoinositides are still relatively high up on the signaling hierarchy, and like the multifunctional IRSs (although on a wider scale), phosphoinositides serve to amplify and spread messages in a rapid yet orderly manner. Therefore, having localized, regulated PIKs ensures three things: (a) a diverse group of phosphoinositides (b) where they are needed, (c) and the capacity to make more when necessary. Likewise, putting effort into controlling PIPK activity and localization facilitates second messenger generation in the desired quantities and the proper location. Thus, a model for phosphoinositide signaling that localizes substrate production and utilization through different enzymes is attractive when attempting to explain mechanisms for specificity of signal generation across several classes of signaling pathways.
In insulin signaling, addition of a phosphate at the D3 position of the inositol ring is heavily implicated in signaling downstream of IRS proteins (1-3). The family of enzymes that phosphorylates the 3-position of the inositol moiety is the phosphatidylinositol 3-kinases, or PI3Ks. The PI3Ks are a modular group of protein-lipid kinases (phosphorylate both proteins and lipids, i.e. phosphoinositides) that consist of a regulatory and catalytic subunit (1-3). Normally, both subunits reside in the cytosol, with the regulatory and catalytic subunit bound together (1-3). Upon insulin stimulation of the insulin receptor and recruitment and tyrosine phosphorylation of the IRS proteins, certain phosphotyrosine motifs in the IRS proteins recruit the regulatory subunit of PI3K to the cell membrane (1-3). This is possible because the regulatory (also referred to as “adaptor”) PI3K subunit contains two SH2 domains, which, as one might recall, is a protein domain that is attracted to phosphotyrosines within specific motifs (pYXXM), including those found in properly coupled IRS proteins (1-3). The insulin receptor and IRS proteins are both localized to the plasma membrane, as are phosphoinositides, so bringing the complete PI3K to the membrane allows localization of substrate (phosphatidylinositol-4,5-P2) and catalyst (PI3K) (1-3, 5).
Upon movement to the plasma membrane, PI3K generates PI (3,4,5)P3 (AKA PIP3) from PI(4,5)P2 (1, 9). Clearly, this requires the presence of sufficient PI(4,5)P2 pools, implicating PIKs as potential regulators of PI3K signaling intensity based upon ligand presence and the local concentration of phosphatidylinositol. PIP3 serves as a second messenger for several important downstream mediators of insulin’s metabolic and mitogenic effects by stimulating their recruitment through their PH (Pleckstrin Homology) domains (refer to article III). In this manner, any protein with a PH domain is a potential candidate which responds to PIP3 synthesis by PI3K. As expected, not all PH domains are equal, and among PH domains exist specificity for which phosphatidylinositol-phosphates there is attraction to (8). Also, the local concentration of phospholipids may make a difference, as it has been shown that there are phosphatidylinositol-phosphate-sensitive protein domains that also have a requirement for a certain level of local hydrophobicity (7). PH domain-containing proteins, for future reference, include key transducers of almost all of insulin’s (and other hormones that act through the PI3Ks) terminal physiological responses.
Implications and Speculation
At this point, a reasonable question to ask is why IRS does not just directly recruit these PH domain-containing proteins and avoid the whole mess of phosphoinositides and their many levels of control. One possible explanation is that the advantages of being able to greatly amplify the insulin signal outweigh the hassles of designing an intricate system of regulation. Furthermore, once the phosphoinositide system was established, it could be used by not only the insulin receptor, but many other receptor tyrosine kinases, as well as seemingly unrelated pathways, providing a convenient mechanism to establish cross-talk.
The system is really quite ingenious if one looks closely enough: insulin, being a hormone with many physiological effects, both non-genomic and genomic, must have a way to signal specifically to certain targets (i.e. proteins involved in glucose uptake) yet be able to communicate with other signaling pathways to temper its effects on metabolism and cellular growth in relation to the multitude of other hormonal responses, all in an effort to maintain homeostasis.
Through the use of IRS proteins, insulin is able to spread its signal out in a wider radius, using the large area and phosphorylatable sites of the IRSs to selectively recruit intermediates of both metabolic and mitogenic pathways. Then, by having the IRS proteins pull in specific PI3K family members (mainly type Ia, but also type IIa, discussed below), the cell ensures that proper signal propagation will occur, yet maintains tight control over the activity, as all that is required to shut the signal off is to dephosphorylate the IRS protein’s specific phosphotyrosine residues responsible for PI3K adaptor subunit recruitment (special enzymes –phosphateses- with just this function exist). However, when the PI3K is active, it can generate enormous quantities of second messengers- which also implies that cells can regulate the system at the level of phosphatases that degrade the second messenger (1, 47). What this means in the end is that it is preferable to spread responsibilities out over a wide range of intermediates rather than dumping all of the regulatory roles on a single protein. Control through feedback regulation, not only from the same pathway, but from others as well, is of paramount importance, and this is most readily achieved if there are several points and means for intervention.
Classes of PI3Ks
As is the case with almost every signaling molecule, PI3K has many ever-so-slightly divergent faces. Both the catalytic and the regulatory subunit have several isoforms coded on separate genes, a fact that underscores the effort cells put into ensuring tight control of phosphoinositide signaling (1). Currently identified are three classes of PI3-kinases- I, II, and III (1-3). Categorization is based primarily upon structure, which in turn defines substrate specificity and subunit association (3), although all use phosphoinositides of some type.
With regard to insulin, type IA is the subtype of interest. Even within class I PI3Ks, there is a further division into IA and IB- again, the division is based upon structure and which adaptor and catalytic subunits associate. However, both IA and IB-PI3K share substrate specificity, which is why they are grouped together- namely, they appear to utilize PI(4,5)P2 exclusively in vivo (3, 9). It is interesting to note that a study looking at a related class of PI3K (class II) found that different substrate preferences were seen depending on the local presence of substrate (27). This should be viewed in conjunction with data showing that simple provision of PI(3,4,5)P3 was not sufficient to mimic insulin (with respect to glucose uptake) in the presence of Wortmannin (PIP3 + Wortmannin), a commonly used type-I PI3K inhibitor, yet was able to restore the insulin response under the Wortmannin + Insulin (PIP3 + Wortmannin + Insulin) condition, implying that local regulation of PIP synthesis (which seems to be insulin-dependent) is of absolute importance (11).
Class II and III PI3K have important roles in cell physiology as well, and class II PI3Ks are stimulated by insulin (23-27); the class II alpha-isoform, which is the primary insulin-sensitive class II PI3K, has been shown to be necessary for cell survival in vitro (26). However, class II and III PI3Ks are only mentioned here to make the reader aware of their existence, and we will not go into any further detail at this time.
Subunits of Class Ia PI3K
The remainder of this article will be dedicated to the class Ia PI3Ks; this is not to say that other isoforms or classes are not important- it is simply beyond the scope or relevance of this article to dwell on their intricacies, and a single isoform will do nicely as a tool to draw out general concepts from the molecular soup that is cellular signal transduction.
Type IA PI3Ks are characterized by their catalytic subunits- p110-alpha, beta, and gamma, and their adaptor subunits-p85-alpha, p85-beta, the smaller variants p55-alpha and gamma, and p50 (3, 15). The adaptor isoforms are similar, except the smaller p55 and p50 variants lack a protein domain (SH3) that the p85 isoforms have (3). This may prove important for targeting of the adaptor (and therefore the catalytic subunit) to specific upstream proteins, as the SH3 domain recognizes different amino acid motifs than SH2 domains, which is specific for phosphotyrosines (as present on IRS proteins that have been tyrosine phosphorylated by activated insulin receptors) (3, 21). Interestingly, both SH2 domains appear to be required to bind, at least in the case of IRS-1, as binding of a single SH2 domain reduces PI3K activity by ~50% in vitro (20). Having this requirement is a potential mechanism for controlling intensity of PI3K signaling by negative feedback loops (which will be discussed shortly), for cells might be able to prevent binding of one or both SH2 subunits, with the respective effects of downstream signal dampening or silencing.
The adaptor subunits do far more than simply bind catalytic subunits and upstream phosphotyrosine residues, though. While catalytic subunits appear to be rather promiscuous in their choice of association with adaptors (3), adaptors have rather different effects on catalytic subunit activity (17, 18). In the cytosol, there is normally a set ratio of catalytic subunits to adaptors (18); how this is maintained is unknown, but it was shown by RNA-interference knockdown techniques that a dynamic ratio is somehow established (18). Intriguingly, this paper also found that knocking out both the p110-alpha and beta catalytic subunits, which are believed to be the critical mediators of insulin-stimulated PI3K activities in insulin-sensitive tissues (19, 22), only resulted in a mild pre-diabetic phenotype, much less severe than expected if the model that PI3K activity is essential for insulin-stimulated glucose uptake were true (18). This data in conjunction with previous studies showing that reducing levels of the p85alpha subunits resulted in increased insulin sensitivity (17), which is seen despite reduced total PI3K signaling, implies that the p85alpha subunit may be having non-PI3K-catalytic-subunit-related effects, i.e. activating other intermediates that negatively impact insulin signaling; this model differs from past hypotheses that an excess of p85alpha over p110alpha/beta reduced insulin sensitivity by binding up free sites on IRS proteins (an adaptor bound to IRS without its catalytic subunit also attached would prevent adaptor-catalytic subunit complexes from binding, indirectly inhibiting PI3K activity).
Correlational data supports the conclusion that an increase in the adaptor:catalytic subunit ratio reduces PI3K signaling (29). Based upon this latest finding, it is reasonable to hypothesize that adaptor subunits do far more than bind and transport catalytic subunits- it appears that balance must be achieved between subunits. It also implies that altering the ratios are a potential compensatory mechanism in response to some perturbance (possibly serving as a temporary pressure release valve). In support of a critical regulatory role for adaptors in PI3K-IA signaling, there is evidence that binding to different adaptors results in altered catalytic subunit activity (16); thus, manipulation of adaptor isoform proliferation is an additional mechanism by which cells can regulate PI3K signaling, and as three different genes code for the five different class IA adaptors (3), such regulation is feasible. If the mechanism by which adaptors select catalytic subunits were elucidated, we might have a very interesting model for how insulin-sensitive cells might switch between metabolic and mitogenic effects at the level of differential adaptor-catalytic subunit expression and association.
Not to be completely ignored, the catalytic subunits of PI3K-IA show some interesting differences as well. The alpha isoform has a considerably (25x) greater Km and a higher Vmax for PI(4,5,)P2 than the beta isoform, implying that it is more important under conditions where PI(4,5,)P2 production is greatly increased, although whether levels ever get this high in response to physiological insulin levels is unknown (28). The dominance of the alpha isoform under high concentrations of PI(4,5)P2 is further enhanced by its 70-fold higher kinase activity (28), although the greatly elevated Km of p110-alpha makes the beta isoform considerably more active at lower PI(4,5)P2 levels, as might be seen under normal physiological conditions (19). In vitro data looking at levels of p110 alpha vs. beta with respect to glucose uptake and insulin supports the dominant role of p110-beta in mediating insulin’s effects (19, 19a) at least in an adipocyte and endothelial cell line (19, 19a). However, this by no means rules out p110-alpha’s role in signaling- it simply means that it is probably more active in regions where PI(4,5)P2 levels reach higher concentrations that that normally seen for regions where insulin signaling occurs- this is almost certainly cell-type and/or hormone-type responsive- in support of this, it has been suggested that p110-beta might have a role in the MAPK pathway, which is often implicated in cellular proliferation (19a, 47).
Negative Feedback Loops
It was mentioned in the last article that tyrosine phosphorylation of IRS proteins promotes insulin signal transduction; brief reference was made to the fact that serine phosphorylation sites were also present in IRSs, but nothing more was explained. As it turns out, addition of phosphate groups to serine (and possibly threonine) residues in proteins that are normally activated by tyrosine phosphorylation reduces their signaling capabilities. In the case of IRS proteins, this occurs either by interfering with binding to the insulin receptor via their PTB domain, and/or negatively impacting their recruitment of SH2-domain-containing proteins such as the adaptor subunits of PI3Ks (48). There is additional control by the actions of tyrosine phosphatases that remove phosphates from tyrosine residues, and several have been implicated including the phosphotyrosine-phosphatase-1B and SHP-2 (49, 50-52). However, the actual importance of phosphatases is unknown- the pathway of primary interest has to do with the p70S6-kinase and its potential role as an IRS serine/threonine kinase (50, 53).
The p70S6K is an important serine/threonine protein kinase in anabolic pathways- the fact that it can negatively regulate insulin signaling introduces the possibility that not only can glucose excess downregulate the insulin message, but so too might amino acid excess (53). Unfortunately, regulation is not as simple as serine/threonine phosphorylation = decreased signaling and tyrosine phosphorylation = increased signaling, and evidence exists that serine phosphorylation at certain sites on IRS-1 actually promotes specific branches of insulin signaling (54). Interestingly, White et al. observed positive effects on insulin signal transduction down its anabolic pathway, but not its metabolic pathway in response to phosphorylation of serine 302 in IRS-1; incidentally, this site is phosphorylated by both glucose and amino acids- the actual enzyme that phosphorylates this site is currently unknown (54).
It appears that nutrient partitioning may also be affected at the level of IRS-1/2 serine phosphorylation at the level of the p70S6K. The first convincing evidence of this appeared in 2004 when it was shown that p70S6K-deficient mice were resistant to diet-induced obesity and maintained insulin sensitivity when exposed to a high-fat diet, supposedly because knocking-out the p70S6K prevented phosphorylation of certain serine sites on the IRS proteins known to inhibit downstream signaling (55). A recent finding by Marette et al. showed that obese rats’ tissues demonstrated elevated phosphorylation of one of the same serine sites on IRS-1 as that purportedly phosphorylated by p70S6K (55, 56). Furthermore, inhibition of the mTOR-p70S6K pathway reduced this serine phosphorylation and restored insulin-associated PI3K and downstream signaling (56). This is further supported by a study performed by Saad et al. where persistent hyperinsulinemia blunted insulin signaling in liver and skeletal muscle in rats, but actually increased insulin sensitivity in white adipose tissue- again, the link appears to be at the p70S6K, as rapamycin (an mTOR inhibitor, and therefore, a p70S6-kinase dampener) was able to ameliorate the effects on insulin signaling of prolonged hyperinsulinemia in all three tissues, restoring insulin sensitivity in liver and skeletal muscle, and bringing adipose tissue sensitivity back down to normal (50).
It has been known for some time now that the p110 (at least the alpha isoform) catalytic subunit of PI3K has a protein kinase activity in addition to its well-documented lipid kinase function (12-14). More accurately, it seems that p110 can act as a serine/threonine kinase, and its targets include itself, the p85 subunit, and the IRS proteins (12-15, 30-32). The exact role of autophosphorylation and adaptor subunit phosphorylation by the catalytic subunit is unknown, although the latter may result in reduced activity (13), perhaps by interfering with binding to phosphotyrosine residues; a far more definite role for the protein kinase function of p110 has been shown with respect to the IRS proteins. In the mid-1990s, Lam et al. found that when stimulated with insulin, murine adipocytes’ IRS-1 proteins were serine phosphorylated, whereas only the p85 adaptor subunits were serine phosphorylated in the absence of insulin- both effects were abolished by treatment with the p110 inhibitor Wortmannin. However, this study was unable to definitively map out a role for the serine kinase function of p110 in insulin signaling (14). Later studies utilizing overexpression techniques have made the link. Egawa et al., using a modified p110-alpha subunit that was constitutively active (always targeted to the cell membrane), discovered that overexpression of this modified protein resulted in a reduction in IRS-1/2 protein levels, an increase in serine phosphorylation of IRS-1/2, and a reduction in recruitment of the p85 subunit to IRS-1/2 in response to insulin (30); again, all of these effects were preventable by Wortmannin (pretreatment). This improvement in IRS-protein tyrosine phosphorylation in response to Wortmannin treatment is a consistent one, although, as mentioned above, the fact that signals downstream of PI3K seem to negatively impact upon IRS-signaling muddies the importance of the potential short feedback loop between p110 and IRS (31)- the potential exists nonetheless.
The field of PI3K research is poised for elevation to a new plane with the progression of modern gene manipulation strategies. As seen in the excellent review by Vanhaesebroeck et al. (3), silencing of individual PI3K catalytic and adaptor subunits is an emerging technique that promises to more accurately define the details of adaptor-subunit attraction and localization. This is fortunate news for insulin signaling research, because though the necessity for PI3K is well-established in insulin-sensitive tissues, (34), there still are questions about what downstream mediators are necessary in the many diverse effects of insulin, to say nothing about the multitude of other hormones that signal through various PI3K subtypes. Furthermore, the upstream regulation of PI3K has not yet been clearly mapped out, as there is some discrepancy in the literature regarding the time course of PI3K activation and the limiting factors governing its activity (41, 57). There is then the question about the role of nuclear translocation of PI3K and other signaling components (38-40). And, by no means can we ignore the question regarding IRS isoforms and how they differ in signaling capacities depending on cell type (33, 35-37), with IRS-1 seeming to have a dominant role in insulin-stimulated glucose uptake, and IRS-2 being more important for insulin’s mitogenic effects in skeletal muscle- yet the opposite is commonly observed in the liver.
In this installment, we began to see how partitioning of signaling might be accomplished through subcellular localization of both precursors (i.e. PI(4,5)P2) and enzymes (PIKs, PI3Ks) and differential recruitment of kinases and phosphatases, and their effects on generating specific regulatory feedback loops. While it is unfortunate that elucidation of these phenomena requires much attention to molecular detail at the finest level, there are, as always, general rules and concepts that one can infer by stepping back and appreciating the system as a rational framework. For example, by looking at signaling pathways as strands of a spider’s web, it is obvious that applying pressure on a single strand elicits a response from the entire web- not only that, but the net response is spread out over many threads as opposed to being concentrated on a few particular fibers. The further we pursue insulin’s (or any “web” type signaling molecule- and they are all looking to be webs after all) message, the more we will see this to hold true, especially as cross-talk and regulation come into play; likewise, we may in time be able to predict what happens when critical chunks of the web are broken and understand the deleterious effects of too much pressure being applied onto one pathway.