Long-term memory traces are often reconsolidated, meaning that they are reactivated and destabilized, updated, and consolidated again. This form of memory updating is thought to be necessary to learn new information and to form more accurate representations of our environment (1). An inability to update and reconsolidate new information from existing memory traces results in cognitive interference, which may form the basis of cognitive impairment.

In the current issue of Biological Psychiatry, Yang et al.

(2) shed new light on the mechanisms and type of plasticity involved in memory updating. In a molecular-driven tour de force series of experiments, Yang et al. identify a new signal transduction cascade that supports memory updating, involving NMDA receptor (NMDAR)-dependent Arc coupling, an uncharacterized p85 class IA PI3K (phosphoinositide 3kinase) adapter/regulatory protein (p55PIK), and mTOR-AKT signaling. They further show that engagement of this signaling pathway is needed to support metaplasticity, often referred to as the plasticity of the plasticity (3).

Traditionally, PI3K regulatory subunit family members dimerize with PI3K catalytic subunits (p110a in this case) to produce the second messenger PIP3 at the cell membrane. The production of PIP3 recruits activators of major signal transduction cascades. These activators include mTORC2 (mechanistic target of rapamycin complex 2), which induces AKT phosphorylation at serine 473, which then phosphorylates GSK3 to decrease its kinase activity. Among its many functions, this pathway regulates neurotransmission and cytoskeletal architecture in neurons. Arc is an immediate early gene that regulates multiple forms of synaptic plasticity (4) and has been extensively studied for its ability to regulate long-term memory and the trafficking of AMPA receptors, which mediate most fast excitatory synaptic transmission in the brain (4). However, Arc was previously shown to interact with the NMDAR subunits NR2A and NR2B (5,6), but the functional significance of these interactions was unclear. The uncharacterized PI3K regulatory subunit family member, p55PIK, was also previously identified as an Arc binding partner (7).

Here, Yang et al.

(2) performed in vitro and in vivo studies to confirm that Arc directly interacts with p55PIK. In a series of follow-up experiments, p55PIK was able to form a complex with the catalytic p110a subunit only when Arc was present, suggesting that Arc serves as a bridge for the formation of a functional PI3K complex. The interaction between Arc, NR2A, and p55PIK was also found to be Ca 21 and CaMKIIa (calcium/ calmodulin-dependent protein kinase II) sensitive.

To address the specificity of complex recruitment and its functional significance, the authors generated a conditional knockout (cKO) mouse to selectively delete p55PIK in the brain. Deletion of p55PIK or Arc in mice prevented p110a interaction with NMDAR subunits. Furthermore, overall PI3K activity was reduced in Arc KO mice. Given the changes in PI3K signaling, the expectation would be decreases in PI3Kdependent recruitment of membrane-targeted signal transduction cascades. Indeed, p55PIK cKO and Arc KO mice had decreased phosphorylated AKT at serine 473 and reductions in GSK3b phosphorylation, which decreased its interactions with the mTORC2 protein mTOR. These changes appeared to depend on NMDAR signaling, as injection of the noncompetitive NMDAR antagonist MK-801 in wild-type (WT) mice led to reduced Arc interactions with NR2A and similar biochemical changes to that of p55PIK cKO and Arc KO mice. Now that a detailed mechanism was established, it was critical to validate this pathway during learning. Here, Yang et al. (2) used a complementary, elegant approach that involved exposure of mice to a novel environment and proximity ligation assays to capture the spatial context of protein complexes in vivo. An increased interaction of Arc with NR2A was observed in the cortex, which persisted up to 2 hours after exploration. Increased complex formation between NMDAR and AKT was also seen using the proximity ligation assay method. As expected, these changes were significantly reduced in Arc KO and p55PIK cKO mice.

The authors next decided to evaluate how p55PIK could support various forms of synaptic plasticity in a similar manner to what was known for Arc. Baseline synaptic transmission, presynaptic alterations and hippocampal-dependent synaptic plasticity that included protein synthesis-dependent forms of long-term depression, and long-term potentiation driven by high-frequency stimulation were measured in p55PIK cKO mice. Surprisingly, however, there was no significant difference even though high-frequency stimulation led to reductions in AKT and GSK3b phosphorylation in p55PIK cKO mice compared to control animals. The lack of plasticity effects also did not appear to disrupt the storage of long-term memories as p55PIK cKO mice did not have defects in working, fear, or read-write reference memories.

If Hebbian forms of plasticity were not altered, then perhaps non-Hebbian plasticity was the culprit. Here, Yang et al. (2) examined changes in metaplasticity in p55PIK cKO mice. Under these conditions, potentiated synapses are resistant to synaptic depression and instead exhibit a potentiation phenotype. Whereas WT mice expressed the expected phenotype, p55PIK cKO mice exhibited a "depotentiation" phenotype. This depotentiation was also found in Arc KO mice. Overall, these findings implied that p55PIK and Arc support the ability of hippocampal potentiated synapses to resist depotentiation. Under this scenario, one may expect that synapses might be vulnerable to de-or repotentiation upon memory recall. The authors tested this in two ways. First, they performed a memory updating task using a radial water maze. The radial water maze contained 6 arms, in which 1 arm had a platform for the mouse to stand on. Mice were tested to find the platform in the same location for 2 consecutive days, and then on subsequent days the platform was changed to a different location to simulate memory updating. If mice had alterations in memory updating, then the expectation would be that they would make more errors during the memory updating component in later trials due to memory instability and would visit the previous platform less frequently during early trials. Indeed, this was the case in p55PIK cKO mice. For the second test, the authors evaluated mice in a 3-chamber social recognition task. Mice were placed in a 3-chamber box and allowed to explore an empty box. Then, an object and mouse were introduced on opposite sides of the chamber and then tested for interaction. As expected, both WT and p55PIK cKO mice had a social preference for the mouse over the object. Next, a familiar mouse and a novel mouse were placed on opposite sides of the chamber to test for social discrimination. Whereas WT mice had a preference for the novel mouse, p55PIK cKO mice did not. Taken together, these behavior studies suggest that p55PIK cKO mice have an impairment to update spatial memories and may even have increased susceptibility to memory interference in multiple behavioral contexts.

Finally, to relate these findings to human-relevant learning impairments, the authors tested for alterations in NMDAR-CaMKII-Arc-PI3K-AKT signaling in early-and late-stage Alzheimer's disease (AD) postmortem brain samples. Using proximity ligation assays, NR2A-AKT interactions were found to be reduced in both early-and late-stage AD brains with no changes in total AKT levels. Moreover, the authors found a reduction in the activation of CaMKII in both early-and late-stage AD. In further evaluating early AD samples, reductions in AKT phosphorylation and decreased NR2A interactions with Arc were observed.

Cumulatively, these findings provide a detailed mechanism for spatial memory updating. NMDAR-dependent increases in Ca 21 activate CaMKII, which then phosphorylates Arc at serine 260 and 278 (8), recruiting it to NMDARs at the cell membrane. The recruitment of Arc serves as a bridge, bringing p55PIK to the p110a catalytic subunit, increasing PI3K activity and activating downstream mTORC2 and AKT transduction pathways. This pathway could serve as a "synaptic tag," which may be relevant for maintaining the history of synapse potentiation for subsequent modifications during memory updating, making them more resistant to depotentiation (Figure 1). Since neural ensemble shifting/activity is associated with memory updating, it would be exciting to map neural ensemble dynamics during memory updating processes between WT and p55PIK cKO mice. These findings further support the notion that NMDAR-CaMKII-Arc-PI3K-AKT signaling may be disrupted in earlystage AD, which may lead to interference vulnerability and facilitate memory failure. Given that deficits in memory updating are associated with mild cognitive impairment and findings of its disruption in a mouse model of AD (9), it is possible that targeting this pathway could be therapeutically relevant. Importantly, disruption of this pathway here demonstrates an increase in GSK3 activity, which is hyperactive in AD (10). One important caveat related to this study is that metaplasticity does not solely rely on NMDAR activity, so there are likely alternative intracellular signaling mechanisms that Disruptions in these signaling pathways may result in deficits in memory updating. CaMKII, calcium/ calmodulin-dependent protein kinase II; mTORC2, mechanistic target of rapamycin complex 2; NR2A, NMDA receptor subunit 2A.

Commentary depend on the type of memory trace that is reactivated. All of these cascades may include Arc as their hub but would engage a context-dependent set of intracellular signaling events. Nevertheless, this study provides one of the most detailed mechanistic pathway maps involved in spatial memory updating.