Recap From Part 4
So far, when we have looked at any signaling molecule in insulin's path, we have seen the common themes of multiple isoforms and unique tissue distribution, both of which contribute to signaling specificity.
There are at least six isoforms of the IRS-proteins, and PI3K is grouped into three classes and further divided into subclasses. Therefore, it would not be surprising if the next step in the cascade were similarly arranged; such is not the case, however, and when it comes to the primary signal downstream of PI3K, there is only one isoform, and it is ubiquitously expressed.
This month, we will discuss the role of the phosphoinositide-dependent protein kinase (PDK-1) in directing insulin signaling towards several different downstream targets and better understand how structure can be used to analyze signaling.
| Acronym Review!
PIKs = phosphatidylinositol kinases
PI3Ks = phosphatidylinositol 3-kinases
PIPKs = phosphoinositide-phosphate-kinases
PIP3 = PI(3,4,5)P3
MAPK = mitogen-activated protein kinase
PDK-1 = phosphoinositide-dependent protein kinase
IRS = insulin receptor-substrate proteins
It will be useful to briefly review some characteristics of protein kinases. As the reader will recall, phosphorylation events are the driving force behind almost all signaling and is due to the structural changes induced by the charge and steric effects that result from the addition/removal of a phosphate group.
| What Is Phosphorylation?
Phosphorylation refers to introducing phosphates into the cell that function as building blocks needed for ATP formation.
Very often, the phosphorylation of a phosphorylatable amino acid - and these include tyrosine, serine, and threonine - causes a major shift in the shape of the protein. The activity of kinases, which are enzymes that can induce phosphorylation (on themselves and/or other proteins), is regulated by phosphorylation events.
Inactive protein kinase (phosphorylated by a different, active protein kinase) Active protein kinase
While we made mention of phosphorylation as a regulatory mechanism, and saw some examples of phosphorylation of tyrosine residues on the IRS proteins by the insulin receptor (whose kinase activity is stimulated following insulin binding), and even observed the outcome of this (recruitment of proteins with SH2 domain, which are attracted to phosphorylated tyrosine residues in the context of a short series of neighboring amino acids, to the IRS proteins).
We have yet to see an actual example of an enzyme whose activity and structure were changed by phosphorylation. Downstream of the PI3K, this mechanism of regulation is the predominant determinant of signaling strength, specificity, and duration.
The model enzyme we will use to make sense of how this occurs and how it might be manipulated is the phosphoinositide-dependent protein kinase-1 (PDK-1).
The AGC-Family Of Protein Kinases
We will use the AGC protein kinases as a platform for discussion this month. AGC refers to a group of protein kinases, including protein kinase A (PKA), protein kinase G (PKG), and protein kinase C (PKC)8 - PDK-1 is also a member. All AGC family members share certain structural characteristics that make their regulation similar in many ways.
The AGC kinases have two main lobes, one on each end of the protein. In its native conformation, the two lobes of AGC kinases form a central core where actual catalysis occurs.3 However, in order for this core to be fully formed, ATP needs to bind. ATP binding, though, is restricted by the presence of interfering amino acid extensions.
| What Is ATP?
Otherwise known as Adenosine TriPhosphate, ATP is critical to the release of energy. ATP is an adenosine-derived nucleotide that supplies large amounts of energy to cells for various biochemical processes, including muscle contraction and sugar metabolism, through its hydrolysis to ADP.
These interfering residues must be moved before the enzyme has any chance of being activated. Towards this end, there are specific serine and threonine residues nearby which, once phosphorylated, induce the interfering amino acid extensions to move aside, allowing ATP to bind.
This leads to catalytic pocket formation and full catalytic capacity.2 It is the specific serine and threonine residues in AGC kinases that we are most interested in, as they are the keys to AGC protein kinase activation.
Key Serine/Threonine Residues In AGC Kinase Regulation
There is a single serine/threonine residue present in a loop-like structure close to the catalytic core of AGC kinases. This loop is referred to as the T-loop, or activation loop, because phosphorylation of this serine/threonine leads to activation of the enzyme.2
Phosphorylation of the T-loop is considered the crowning step in AGC kinase activation, as activity is increased anywhere from 30-100x by this step alone. The second phosphorylation site increases enzyme activity an additional 7-10x, but only after the activation loop has been phosphorylated.11
There is another critical serine/threonine residue in a more distal region of the AGC kinase called the C-terminal extension. This serine/threonine is always found within a conserved series of amino acids, ordinarily phenylalanine-X-X-phenylalanine (FXXF motif, where X = an amino acid that is non-charged and non-hydrophobic).6
Normally, the serine or threonine comes directly after the last F and a tyrosine or phenylalanine completes the motif, such that the sequence of amino acids to look for is FXXF(S/T) (F/Y).4 Phenylalanine and tyrosine are both hydrophobic amino acids, and thus, the FXXF(S/T) (F/Y) motif is called the hydrophobic motif or HM for short.
With one known exception, all AGC kinase family members contain this motif; a few members do not have a phosphorylatable S/T residue, but rather have an aspartate or glutamate (D or E, respectively) residue.2,4 The significance of this will be discussed later.
Fig. 1. A simplified diagram of a typical AGC protein kinase. For clarity, only the relevant portions of the protein have been drawn. Note particularly the flexible nature of the connector to which the HM is attached.
Another defining structural characteristic of AGC kinases is that they all have a hydrophobic "pocket" in one lobe. This pocket is called either the "PIF pocket" (for phosphoinositide-dependent-protein kinase-interacting factor) or the hydrophobic motif-binding pocket (HM pocket).2 We will use the latter name, as it is far less confusing.
Fig. 2. The diagram to the left depicts an AGC kinase without the HM for clarity. Note the disordered HM-pocket, represented by the white dot in the pocket.
The HM-pocket, as its name implies, binds to hydrophobic motifs. More specifically, the HM-pocket binds to phosphorylated hydrophobic motifs.2 As stated above, most HMs contain an S/T residue which can be phosphorylated - doing so allows the phospho-HM to bind to an HM-pocket.
Figure one shows that the hydrophobic motif is present on a flexible extension which is normally very close to the HM-pocket. As a result, each AGC kinase can actually bind its own phosphorylated HM via its HM-pocket. There are several consequences of this effect, which are discussed below.
In summary, we know thus far that AGC kinases:
1. Have two main lobes that form a catalytic core.
2. Need to bind ATP to form said core.
3. Need to remove interfering amino acids before binding of said ATP.
4. Have two critical serine/threonine residues that govern removal of said interfering amino acids.
- a. One residue is present on a segment of the protein called the "T," or activation-loop.
- b. The other residue is present on a snakelike extension on the C-terminus called the hydrophobic motif
- Phosphorylation of this residue results in a phospho-HM.
- Phospho-HMs can bind to HM-pockets.
5. Have a hydrophobic pocket called the HM-pocket, or PIF-pocket, that can bind phosphorylated hydrophobic motifs.
6. Have a short sequence of amino acids called the hydrophobic motif, which contains the second critical serine/threonine (see #4b).
- The conserved sequence of amino acids of the HM is FXXF(S/T)(F/Y).
- Some members of this family have the following HM: FXXF(D/E)(F/Y).
- D/E = aspartate/glutamate = charged amino acids that can mimic the effects of a phosphate group- essentially, their presence eliminates the need for a separate phosphorylation event, leaving these AGC members with what amounts to an always-phosphorylated HM.
Knowing this, the next series of logical questions to ask are:
- What induces phosphorylation of the T-loop serine/threonine?
- What induces phosphorylation of the HM-serine/threonine?
- Which is phosphorylated first?
- What is the outcome of binding of one's own phosphorylated HM?
- Are there other regulatory mechanisms at work?
In order to answer these questions, we will turn to PDK-1, which is the master regulator of the AGC protein kinases. Somewhat surprisingly, PDK-1 is itself an AGC protein kinase. This brings up an additional question:
- What makes PDK-1 different from other AGC kinases such that it can regulate them?
The phosphoinositide-dependent protein kinase-1 was first discovered eight years ago as the kinase responsible for phosphorylating protein kinase B (usually called PKB and also known as AKT) on its critical T-loop threonine and activating it as a result.1
Subsequently, many other targets of PDK-1 have been discovered, including classical, novel, and atypical PKCs, p70S6K, p90RSK, SGK, and several other AGC kinases.1
PDK-1 is a small enzyme that is a member of the AGC protein kinase family. Like other members, it has a critical serine/threonine in its activation loop that must be (auto) phosphorylated before catalysis may occur. Contrary to most other AGC kinases, this may not be very important to regulation of its activity, as PDK-1 appears constitutively active (no need for a ligand to stimulate activity).1-3
Similar to other AGC kinases, it is bi-lobal in structure, needs to bind to ATP, and has an HM-pocket.2 What makes PDK-1 different, however, is that it does not have a hydrophobic motif.4
All other AGC kinases have some kind of an HM, which when phosphorylated binds to an adjacent (ordered) HM-pocket. PDK-1 has the HM-pocket, but no HM. Rather, PDK-1 uses its substrates' hydrophobic motifs for binding.1-3 This simple difference has allowed PDK-1 to rise above the other AGC kinases as a governing enzyme.
Fig. 3. In order from top to bottom, a series of schematics outlining the recruitment and function of PDK-1 in insulin signaling.
The Universal Importance Of PDK-1
Recall the current model of AGC activation:
Phosphorylation of activation-loop S/T ATP binding Catalytic core formation Phosphorylation of HM S/T binds to own HM-pocket
Right now, there seems to be no point to the latter event. However, we do know that the HM normally hangs off of the C-terminal extension, in plain view, until it is phosphorylated. At that point it buries itself in the HM-pocket.6
Given that HM pockets bind only phosphorylated HMs, this suggests that PDK-1 must be in close vicinity and binds to the newly phosphorylated HM before the enzyme's own HM-pocket sequesters it (how this may happen will be explained later). Sequestration of the phospho-HM eventually occurs, displacing PDK-1. So, why is binding and releasing of PDK-1 important? What does PDK-1 do while bound to its substrate?
PDK-1 is the kinase responsible for phosphorylation of the activation-loop S/T in AGC kinases.
What we are observing here is the importance of the order of phosphorylation, and therefore the order of kinase activation and recruitment, in regulating enzyme activity. While the identity of the kinase that phosphorylates the HM S/T is still cloudy, the bulk of the research suggests that it is not PDK-1; the actual kinase activity has been attributed to autophosphorylation and/or to an elusive set of enzymes tentatively called PDK-2s.1
PDK-1: The Structure Sensor
All AGCs share the same PDK-1 as a necessary factor in phosphorylation of their A-loop S/Ts. However, there is only one isoform of PDK-1, which is pretty much ubiquitously expressed.1,2 Thus, PDK-1 seems a lousy enzyme for signal partitioning specificity.
The many enzymes that seem to function as PDK-2s on the HM S/T should confer a certain degree of specificity, as will localization of different AGC kinases, although the majority seems to prefer the cytosol.1 Knowing this, why even bother with having a step in signal transduction mediated by a promiscuous enzyme such as PDK-1?
We mentioned in the above paragraph that PDK-1 binds phosphorylated HMs. It is perhaps more correct to say that phosphorylation of other AGC's HMs recruits PDK-1.
More and more, it seems that while PDK-1 is unable to do much by way of discrimination, its substrates can change their HM-phosphorylation state and in turn regulate PDK-1 activity. This is essential, because without PDK-1 interaction, AGC kinases cannot become catalytically active.
PKB: Unique among PDK-1 substrates by virtue of its PH domain
Different PDK-1 substrates have evolved their own mechanisms for recruitment of PDK-1. In the case of PKB, there is no requirement for a pre-phosphorylated HM motif in order for PDK-1 to bind.5 This at first seems strange, as HM-pockets are supposed to be specific for phospho-HMs, and binding of the phospho-HM to PDK-1's HM pocket is known to be crucial to AGC kinase activation. How can this discrepancy be resolved?
At first, researchers tried to argue that the PH-domains of PDK-1 and PKB would serve to localize such that PKB's HM would not even need to interact with PDK-1.7 This model is attractive, made more so by the fact that PKB is the sole PDK-1 substrate that has a PH domain.7
However, recent data showed that while the PH domain of PDK-1 performs the classical function of translocation to PIP3-rich membranes, the PKB-PH domain might actually be auto-inhibitory with respect to PKB activity, and binding by PIP3 serves to alter the PKB structure and allow it to be acted upon by PDK-1.10
The same might be the case for PDK-1.9 This is further supported by findings that a PKB mutant without a functional PH domain could still be normally translocated and activated as long as it was expressed with a wild-type (unmutated) PDK-1.10
Recently, X-ray crystallography has revealed that PKB's HM-pocket is actually deformed until the HM is phosphorylated and binds to the HM-pocket (both with respect to PKB).11
The same authors found that provision of a peptide that mimicked a phosphorylated HM to PKB was sufficient to induce order to the HM-pocket and result in binding of the PKB-HM to the PKB-HM-pocket. Thus, it seems that phosphorylation of the HM of PKB causes a structural ordering of the HM-pocket.11 Strangely, the phosphorylation state of PKB's HM seems to have nothing to do at all with PDK-1 binding.
In fact, PKB activity is increased anywhere from 30-100x by virtue solely of activation-loop phosphorylation;3 HM phosphorylation augments activity an additional 7-10x, but activation-loop phosphorylation is clearly the prime mover.
This is also seen in other PDK-1 substrates, with the HM-loop serving more as an aid in recruiting PDK-1 via HM-phosphorylation and/or helping to stabilize the protein's active conformation.1-3
Fig. 4. Once the HM-pocket is ordered by virtue of a structural shift (often induced by PDK-1-mediated phosphorylation of a separate residue), and the HM-serine/threonine is ALSO phosphorylated, the HM of an AGC kinase can bind to its own HM-pocket. Note that both requirements of HM-pocket ordering and HM-serine/threonine organization must be met for this to occur.
So, it may be that PKB, the only PDK-1 substrate that has a PH domain, really does interact simply by virtue of vicinity to PDK-1, making it truly unique among PDK-1 substrates (the only one that does not need a phosphorylated HM to recruit PDK-1).
That the structure of PKB is disordered until HM-phosphorylation, but that ordering requires a phosphorylated activation-loop, suggests that PDK-1 acts first, and then a PDK-2 activity occurs (whether this is a separate enzyme or autophosphorylation is unclear - both may occur depending on the stimulus and cell type).11
This of course would imply that the HM is unphosphorylated when PDK-1 binds; however, this is hardly sufficient evidence to draw the conclusion that PDK-1 actually binds the unphosphorylated HM of PKB. It simply shows that in the case of PKB, activation loop phosphorylation can occur before HM phosphorylation.
Suffice it to say for now that for PKB, the link between the phosphorylation of the activation loop and that of the HM is tenuous at best. This might be one explanation for the rapidity of PKB activation in comparison to other PDK-1 substrates that have an absolute requirement for prior HM phosphorylation by the elusive (but very real) PDK-2.2
P70S6K - HM First
In contrast to the relatively HM-free relationship between PDK-1 and PKB, the p70S6K has a clear need for HM-phosphorylation before PDK-1 can phosphorylate its T-loop.2
Again, it is not clear just how PDK-1 is able to interact with what should be a completely unavailable HM-motif. After all, phosphorylated HMs should immediately bind to their HM-pockets, and the closest pocket resides on the substrate itself.
One hypothesis is that the initial binding of the phosphorylated HM to its cognate HM-pocket does not induce any irreversible structural shifts. PDK-1 can then come along and still interact with the "structurally flexible" phospho-HM. However, upon PDK-1 phosphorylation of the A-loop of p70S6K, a more permanent structure shift occurs and prevents PDK-1 from accessing the phospho-HM anymore, freeing it up to perform more catalysis and leaving the p70S6K fully active.
A simpler explanation might be the presence of an additional PDK-1 interaction site on the p70S6K that is actually hidden/blocked by an unphosphorylated HM, which is freed upon phosphorylation.
Currently, the most popular explanation is based upon X-ray crystallographic evidence showing that PDK-1 is able to sense inactive AGC enzyme conformations, and avoids active AGC enzymes. If this is so, then it suggests that the HM-pockets of AGC kinases that require HM-phosphorylation before PDK-1 interaction, have a disordered pocket reminiscent of the case in PKB.2,11
Binding of PDK-1 to the available phosphorylated HM and subsequent PDK-1-mediated phosphorylation of the kinase activation loop causes a change in the conformation, creating an active HM-pocket, inducing binding of the already phosphorylated HM to it and displacing PDK-1, which has served its function.2
Fig 5. In order from top to bottom, a diagrammatic representation of select intermediary steps in activation of PDK-1 substrates that require HM-phosphorylation first. PDK-1 is represented by purple, and its substrate (p70S6k) is represented by blue.
In the first picture, p70S6K is in an inactive conformation. Next, the serine/threonine residue in its HM-motif is phosphorylated - this is shown as a change in the shape of the serine/threonine from a red box to a red circle. This phosphorylation allows the p70S6k's HM to bind to PDK-1's HM-pocket.
In the third picture. The now-fully active PDK-1 phosphorylates the activation loop serine/threonine residue of p70S6k. This leads to the final picture, where p70S6K is has bound its own (phosphorylated) HM in its now-ordered HM-pocket, simultaneously freeing PDK-1.
Whichever mechanism is active, the initial phosphorylation of the HM of p70S6k is a slow process, taking upwards of fifteen minutes for full activation after ligand binding in vitro versus mere seconds for PKB in vitro.2
This suggests that using a different PDK-2 activity while utilizing the same T-loop kinase allows for separation of signaling temporally without adding unnecessary complexity to the system (the PDK-1 functions as the primary ligand sensor, while the PDK-2 mediates the proper timing of pathway activation). This mechanism is especially plausible for p70S6k, which is activated downstream of PKB.
We now see that the activation of enzymes downstream of PI3K is quite different from those upstream. This is consistent with what we expect, as the initial signal is spread out rapidly, with specificity conferred by several special isoforms and tissue expression levels.
As the signal travels further and enters pathways that will lead to targeted end effects, the mechanism shifts at the level of the PDK-1, which, in conjunction with the yet-undefined PDK-2(s), imparts a mode of regulation that contributes to specificity and to temporal activation.
What PDK-1 lacks in isoforms, it more than makes up for by a clever trick, a single change that makes it different from its peers and elevates it to the status of a master AGC family regulator. That change, of course, is the lack of an HM, leaving its own HM-pocket free to bind to other AGC's HMs.
This mechanism of choosing substrates is perfect, given that all AGC kinases have two regulatory sites that can each (individually) change their structure once phosphorylated. PDK-1 takes full advantage of this by activating only one site - the serine/threonine in the activation loop.
It leaves phosphorylation of the HM-serine/threonine to another activity. Thus, PDK-1 does not interfere with the job of the more enzyme-specific PDK-2s, and vice-versa; when PDK-2s act first (as in the case of the p70S6k), the substrate's HM pocket remains in a disordered configuration but the PDK-2-generated-phosphorylated-HM flags down PDK-1, which completes the job (organizes the enzyme to be in a fully, or at least, primed-to-be-fully-active, state).
In the unique case of PKB, it is the co-localization that accomplishes the task of PDK-1 recruitment, and any phosphorylation of the HM may be enhanced by PDK-1's actions on the activation-loop (via induction of conformational changes that transmit to the HM-pocket to give order to the otherwise disordered HM-pocket), aiding PDK-2 activity.
Understanding the molecular mechanisms of signal transduction is of great benefit to making full sense of simplified signaling diagrams. Now, when one looks at a typical PI3K PDK-1 PKB pathway, there should be questions not only pertaining to what other molecules are involved, cross-talk, etc, but also what kinds of interactions are occurring at the structural level.
It is important to know why PDK-1 activates PKB but not mTOR (they are in different families of kinases), why PKB activation is much more rapid than p70S6k (aside from p70S6k being cytosolic and PKB being recruited to the membrane, the former depends on HM-phosphorylation to occur first, necessitating a second enzyme, while the latter does not need any such second enzyme for at least partial activation).
Now, whenever one encounters another AGC family enzyme, the connection to PDK-1 and PDK-2 activities should immediately be established, and conjectures about how rapidly it is activated, where it will be localized, and what other signaling intermediates it might interact with based upon the first two factors, can be made.
While we did not go into any detail about how different residues spatially organize others to make the necessary interactions for catalysis, this article introduced the concept of kinase-kinase interactions at special phosphorylatable amino acids which act as switches that lead to conformational shifts.
Hopefully, this part included just enough molecular detail to aid in understanding, but not so much superfluous detail as to weigh the reader down. In our next article, we will explore the regulation of PDK-1's substrates, and as we draw closer to the end of the insulin signaling path, we will begin to tie together molecular mechanisms with physiological effects.
- Alessi DR, James SR, Downes CP, Holmes AB, Gaffney PR, Reese CB, and Cohen P. Characterization of a 3-phosphoinositide-dependent protein kinase which phosphorylates and activates protein kinase Balpha. Curr Biol 7: 261-269, 1997.
- Alessi DR, Kozlowski MT, Weng QP, Morrice N, and Avruch J. 3-Phosphoinositide-dependent protein kinase 1 (PDK1) phosphorylates and activates the p70 S6 kinase in vivo and in vitro. Curr Biol 8: 69-81, 1998.
- Biondi RM. Phosphoinositide-dependent protein kinase 1, a sensor of protein conformation. Trends Biochem Sci 29: 136-142, 2004.
- Biondi RM, Cheung PC, Casamayor A, Deak M, Currie RA, and Alessi DR. Identification of a pocket in the PDK1 kinase domain that interacts with PIF and the C-terminal residues of PKA. Embo J 19: 979-988, 2000.
- Biondi RM, Kieloch A, Currie RA, Deak M, and Alessi DR. The PIF-binding pocket in PDK1 is essential for activation of S6K and SGK, but not PKB. Embo J 20: 4380-4390, 2001.
- Biondi RM and Nebreda AR. Signalling specificity of Ser/Thr protein kinases through docking-site-mediated interactions. Biochem J 372: 1-13, 2003.
- Currie RA, Walker KS, Gray A, Deak M, Casamayor A, Downes CP, Cohen P, Alessi DR, and Lucocq J. Role of phosphatidylinositol 3,4,5-trisphosphate in regulating the activity and localization of 3-phosphoinositide-dependent protein kinase-1. Biochem J 337 ( Pt 3): 575-583, 1999.
- Dong LQ and Liu F. PDK2: the missing piece in the receptor tyrosine kinase signaling pathway puzzle. Am J Physiol Endocrinol Metab 289: E187-196, 2005.
- Filippa N, Sable CL, Hemmings BA, and Van Obberghen E. Effect of phosphoinositide-dependent kinase 1 on protein kinase B translocation and its subsequent activation. Mol Cell Biol 20: 5712-5721, 2000.
- Milburn CC, Deak M, Kelly SM, Price NC, Alessi DR, and Van Aalten DM. Binding of phosphatidylinositol 3,4,5-trisphosphate to the pleckstrin homology domain of protein kinase B induces a conformational change. Biochem J 375: 531-538, 2003.
- Yang J, Cron P, Good VM, Thompson V, Hemmings BA, and Barford D. Crystal structure of an activated Akt/protein kinase B ternary complex with GSK3-peptide and AMP-PNP. Nat Struct Biol 9: 940-944, 2002.