AKT activation and inactivation

AKT is subject to a tight auto-inhibition that keeps the kinase in an inactive state.

In the inactive ‘PH-in’ conformation, the 3-phosphoinositide-dependent protein kinase 1 (PDK1) cannot access the activation loop (Figure 6)^[23]. The interaction of the PH domain with the kinase domain displaces the αC-helix, L3 of the PH domain occupies the position of the phosphorylated activation loop and a residue from that

same loop, Trp80, inserts into a cleft in the kinase domain creating a cavity that positions residues from the HM near Trp80[23–25,40,44,45].

The translocation of AKT to membranes containing PI(3,4)P2 and PI(3,4,5)P3,

which is the first step in AKT activation, is mediated by the PH domain (Figure 4 and Figure 5)^[72]. The accumulation of AKT at the PM is transient, peaks at around 5 minutes after GF stimulation and lasts up to 20 minutes^[73,74]. This interaction induces a conformational shift (PH-in to PH-out) that abolishes the inhibitory effect of the PH domain on the kinase domain by exposing the phosphorylation site in the activation loop (Thr308, Thr309 and Thr305 for AKT1, AKT2 and AKT3). This enables phosphorylation of this site by PDK1, another member of the AGC kinase group. This modification enhances AKT activity by 100-fold, which is still only a fraction (~10%) of its maximum catalytic potential[7,13,42,75,76].

Figure 6 AKT conformations. The PH-in or inactive conformation is shown on the left. Here, the activation loop is inaccessible for PDK1 and Trp80 in loop 3 of the PH domain inserts into a cleft in the catalytic domain and blocks the correct orientation of the activation loop. The same residue also interacts with residues from the HM (green) effectively locking AKT in this inactive conformation. The interaction with PIP3 induces a conformational shift to the PH-out conformation. This exposes the activation loop for phosphorylation by PDK1. Active AKT, with both sites phosphorylated and the disengaged PH domain is shown on the right.

Full activation of AKT requires the additional phosphorylation of a Ser residue, located in the HM, that acts in synergy to increase AKT catalytic activity by another 7-10-fold^[43,77]. Mechanistic Target of Rapamycin Complex 2 (mTORC2), a protein complex defined by mechanistic target of rapamycin (mTOR), MAPKAP1/SIN1, RICTOR, MLST8 and PRR5 was identified as the primary kinase complex that phosphorylates the AKT hydrophobic motif (Ser473, Ser474 and Ser472 for AKT1, AKT2 and AKT3) but it has been reported that, in response to DNA damage, AKT1 S473 can also be phosphorylated by DNA-PK^[78]. Although this modification is not

essential for AKT activity, it stabilizes AKT in its active conformation and increases the catalytic activity by up to an order of magnitude^[6,77].

Thus, AKT activation involves interaction of the PH domain with membrane-tethered PIs, a shift in conformation and subsequent phosphorylation of the activation loop and hydrophobic motif. The phosphorylation of the activation loop appears to be dominant as phosphorylation of the HM alone only results in a minor increase in catalytic activity which may not suffice to propagate AKT signalling in cells^[79]. Along this line, the use of non-phosphorylatable AKT mutants has shown that Thr308Ala has no catalytic activity, while mutation of the HM (Ser473Ala) did not completely abolish catalytic activity^[72,80].

As is the case with other kinases of the AGC kinase group, both modifications could be interdependent, but the reports on the order of modification are conflicting.

Phosphorylation of the HM could disrupt its interaction with Trp80 of the PH domain, and would allow the PH domain to disengage from the kinase domain after binding PIP3. These reports indicate that HM phosphorylation stimulates the transition to the PH-out conformation where the activation loop is exposed^[23,24,45]. This is supported by the observation that, in some cases, a phosphomimetic⁶ mutation of the Ser residue in the AKT HM to Aspartic acid (Asp) led to increased phosphorylation of the activation loop. However, this was not observed in all studies that make use of AKT with phosphomimetic mutations[39,40,77,81,82].

On the other hand, the three-dimensional structures of inactive and active AKT show that the hydrophobic groove, where the HM interacts with the kinase domain, only exists in the active state when the activation loop is phosphorylated^[39,83]. Thus, even if phosphorylation of the HM can occur before phosphorylation of the activation loop, there would be no groove where the HM can bind and no active structure to stabilize. This would argue that phosphorylation of the activation loop is a requirement for the HM to exert its stimulating effect on AKT catalytic activity.

Phosphorylation of the activation loop, but not the HM, reduces the affinity of AKT for PI(3,4,5)P3-containing membranes and results in the dissociation of AKT from the PM. This could be due to the re-engagement of the PH domain to a site

6 Phosphomimetics involves the replacement of an amino acid that can be phosphorylated by a naturally occurring amino acid that structurally resembles the phosphorylated original amino acid. For example, an aspartic acid can be used to mimic a phosphorylated serine. There is some controversy on the use of phosphomimetics as the carboxylate of Asp and Glu has less electronegativity, a smaller hydration sphere and volume compared to a phosphate. In the case of AKT the use of phosphomimetics is not recommended as these do not activate the enzyme to the same extent as a phosphorylation

on the phosphorylated AKT kinase domain that is distinct from the interaction site in the PH-in conformation^[23,44]. Consequently, AKT has a far shorter retention time at the PM compared to the presence of its anchors, PI(3,4,5)P3 or PI(4,5)P2[23,84]

.

However, tracking AKT and its activity by using fluorescent reporters in cells, showed that AKT has a slower diffusion speed than can be expected for a protein of that size, indicating that a subpopulation of activated AKT transiently re-engages PIs on the PM or endomembranes after activation^[73,85]. This binding-and-release is rate-limiting for AKT inactivation and explains how AKT phosphorylation is sustained far longer (>60 minutes) than the initial translocation to the PM, as engagement of the PH domain by PI(3,4,5)P3 also protects AKT from dephosphorylation. This closely couples AKT activity to the activity of PI3K and its antagonists such as PTEN^[73].

Phosphorylated AKT in the cytoplasm is rather short-lived. Phosphatases such as protein phosphatase 2A (PP2A) and PH domain leucine-rich repeat protein phosphatase (PHLPP1/2) terminate AKT activity by dephosphorylation of the activation loop (Thr309) and HM (Ser474), respectively^[74,86]. After roughly 30 seconds, the amount of phosphorylated AKT in the cytoplasm is reduced by half^[87–

89]. These phosphatases control the amplitude of AKT signalling, a depletion of PHLPP1/2 causes a drastic increase in both the level and duration of AKT phosphorylation. In Hs578Bst cells, the effect of Epidermal Growth Factor (EGF) stimulation in PHLPP1/2 depleted cells was increased by 30-fold and AKT phosphorylation was sustained for much longer (up to 24 hours instead of 1 hour).

Although it took much longer, AKT phosphorylation did diminish to basal levels while AKT expression remained stable. As AKT associates with Heat-shock protein 90 (HSP90) in the cytoplasm, and has a half-life of up to 36 hours this suggests the existence of additional phosphatases that control AKT activity^[86,90].