Pi4KIIα Regulates Unconditioned Stimulus-Retrieval-Induced Fear Memory Reconsolidation through Endosomal Trafficking of AMPA Receptors.

Targeting memory reconsolidation is an effective intervention for treating posttraumatic stress disorder (PTSD). Disrupting unconditioned stimulus (US)-retrieval-induced fear memory reconsolidation has become an effective therapeutic approach to attenuate fear memory, but the underlying molecular mechanisms remain unknown. Here, we report that US-retrieval-dependent increase in phosphatidylinositol 4-kinase IIα (Pi4KIIα) promotes early endosomal trafficking of AMPA receptors, leading to the enhancement of synaptic efficacy in basolateral amygdala (BLA) neurons. The inhibition of Pi4KIIα by an inhibitor or short hairpin RNA impaired contextual fear memory reconsolidation. This disruptive effect persisted for at least 2 weeks, which was restored by Pi4KIIα overexpression with TAT-Pi4KIIα. Furthermore, the blockade of early endosomal trafficking following US retrieval reduced synaptosomal membrane GluA1 levels and decreased subsequent fear expression. These data demonstrate that Pi4KIIα in the BLA is crucial for US-retrieval-induced fear memory reconsolidation, the inhibition of which might be an effective therapeutic strategy for treating PTSD.

Introduction

Posttraumatic stress disorder (PTSD) is a complex neuropsychiatric disorder that is characterized by re-experiencing the trauma, avoidance, emotional numbing, and hyperarousal (Shalev et al., 2017). Abnormalities in the learning and processing of trauma-related fear memory are believed to play a prominent role in the pathophysiology of PTSD (Fanselow and LeDoux, 1999). In laboratory studies, Pavlovian fear conditioning is widely used in both rodents and humans to investigate fear memory. Subjects are typically trained to associate a neutral conditioned stimulus (CS) with an aversive US, such as footshock. After acquisition, fear memory becomes permanent through a process called consolidation (Dudai et al., 2015). Once retrieved or reactivated, the consolidated fear memory is believed to enter a labile state, known as reconsolidation. During reconsolidation, the memory is dynamic and susceptible to modifications, thus providing an opportunity to interfere with seemingly stable memories (Beckers and Kindt, 2017, Lee et al., 2017).

Considerable rodent and human studies have revealed that pharmacological interventions following CS retrieval can disrupt memory reconsolidation and attenuate fear expression (Brunet et al., 2008, Debiec et al., 2010, Nader et al., 2000). Recent studies reported the US-exposure-specific reconsolidation of learned fear in an amygdala-dependent manner (Debiec et al., 2010, Diaz-Mataix et al., 2011). Pharmacological interventions after US retrieval disrupted more CS-US associations than CS retrieval (Huang et al., 2017). US-retrieval-based reconsolidation interventions were also shown to successfully target remote memories (Liu et al., 2014), thus suggesting a powerful noninvasive procedure to treat psychiatric disorders, such as PTSD. However, the practicality of implementing US-retrieval-based manipulations is questionable when considering the ethical aspects of reexposing PTSD patients to traumatic events, such as war, assault, or natural disasters. Perhaps one of the best ways to overcome this problem is to understand the distinct mechanisms that underlie CS-retrieval- and US-retrieval-triggered memory reconsolidation, thus developing a more practical way to clinically rewrite maladaptive memories. Cellular and molecular processes of CS-retrieval-induced memory reconsolidation have been widely studied (Johansen et al., 2011, Tronson and Taylor, 2007), but the mechanisms that underlie US-retrieval-induced memory reconsolidation remain largely unknown.

Accumulating evidence indicates that α-amino-3-hydroxy-5-methyl-isoxazole-4-propionic acid receptors (AMPARs) play an important role in synaptic plasticity, learning, and memory. The exocytosis and endocytosis of AMPARs are regulated by complex molecular and cellular mechanisms, including AMPAR binding proteins, posttranslational modifications, and endosomal trafficking (Anggono and Huganir, 2012, Jacobi and von Engelhardt, 2018). Alterations of AMPAR surface expression have been shown to be involved in fear memory reconsolidation in the amygdala (Clem and Huganir, 2010, Hong et al., 2013, Zhou et al., 2015). Our recent work indicated a possible role for AMPAR trafficking in US-retrieval-induced reconsolidation (Liu et al., 2014, Luo et al., 2015, Yuan et al., 2019). These findings strongly support a role for AMPAR trafficking in fear memory reconsolidation, but the underlying molecular mechanisms remain unexplored.

Phosphatidylinositol 4-kinase IIα (Pi4KIIα) is the dominant lipid kinase that generates Pi4-phosphate. Pi4KIIα is mainly localized in the trans-Golginetwork (TGN) and endosomes (Clayton et al., 2013, Minogue, 2018). Pi4KIIα is involved in many different cell pathways, including membrane trafficking, ion channel regulation, and vesicle trafficking (Guo et al., 2003, Minogue et al., 2006, Pan et al., 2008, Salazar et al., 2005, Wang et al., 2003, Wang et al., 2007). Pi4KIIα regulates the trafficking of cargo proteins, including transferrin, epidermal growth factor receptor, lysosome membrane protein 2, glucocerebrosidase, and vesicle-associated membrane protein 3 (Jovic et al., 2012, Jovic et al., 2014, Minogue et al., 2006). Pi4KIIα was recently reported to play a role in the surface expression of GluA1 in hippocampal neurons (Robinson et al., 2014), suggesting its potential involvement in AMPAR trafficking. Pi4KIIα also regulates receptor sorting at early endosomes, and Pi4KIIα knockdown decreases the efficiency of sorting from early endosomes (Henmi et al., 2016). However, the role of Pi4KIIα in the early endosomal trafficking of AMPARs and US-retrieval-induced fear memory reconsolidation remains unclear.

Based on these previous findings, we hypothesized that Pi4KIIα plays a critical role in US-retrieval-induced fear memory reconsolidation by regulating the early endosomal trafficking of AMPARs. Using a combined behavioral, biochemical, molecular biological, and electrophysiological approach, we found that US retrieval resulted in the rapid and transient upregulation of Pi4KIIα expression, which in turn increased the recruitment of early endosomes, promoted synaptic AMPAR incorporation, and enhanced synaptic transmission in the basolateral amygdala (BLA). These molecular and synaptic alterations ultimately contributed to the US-retrieval-induced reconsolidation of fear memory.

Results

Pi4KIIα Localizes to Early Endosomes in BLA Neurons and Colocalizes with AMPARs

Previous studies have shown that Pi4KIIα is expressed in neurons and astrocytes throughout the brain (Larimore et al., 2011, Simons et al., 2009). In the present study, we used a specific antibody to examine the localization of Pi4KIIα in the rat brain. Pi4KIIα expression was enriched in learning- and memory-related brain areas, including the prelimbic cortex (PrL), infralimbic cortex (IL), nucleus accumbens (NAc) core, NAc shell, dorsal hippocampus (DH), ventral hippocampus (VH), central nucleus of the amygdala (CeA), and BLA (Figure S1A).

We then examined the subcellular distribution of endogenous Pi4KIIα in the BLA by Western blot and immunostaining (Figures S1B–S1E). Consistent with previous findings (Guo et al., 2003), we found that Pi4KIIα was enriched in the cytosolic fraction but not in the synaptosomal membrane fraction and presented no colocalization with postsynaptic density protein 95 (PSD95), a postsynaptic marker (Figures S1C and S1E). Immunostaining revealed the partial colocalization of Pi4KIIα with the somatodendritic marker mitogen-activated protein 2 (Figure S1E). Pi4KIIα was also localized to early endosomes that were labeled with early endosome antigen 1 (EEA1) in BLA neurons (Figure S1D). We also found the colocalization of Pi4KIIα with GluA1 and GluA2 in BLA neurons (Figure S1D).

Given the colocalization of Pi4KIIα, EEA1, and GluA1, we performed co-immunoprecipitation using rat BLA homogenates to detect whether endogenous Pi4KIIα directly binds EEA1 and GluA1. We found that an antibody against Pi4KIIα co-immunoprecipitated EEA1 and GluA1 (Figure S1F). In addition, an antibody against EEA1 co-immunoprecipitated Pi4KIIα and GluA1 (Figure S1G). These results indicate that Pi4KIIα is a binding partner of EEA1 and GluA1 in vivo.

Unconditioned Stimulus Retrieval Transiently Upregulates Pi4KIIα Levels and Promotes Synaptic AMPAR Incorporation in the Rat BLA

We first determined the appropriate US intensity to reactivate fear memory in rats. Six groups of rats were trained for contextual fear conditioning. Twenty-four hours later, the rats underwent different retrieval patterns (no retrieval [NoR], weak US retrieval [USR], or strong USR) immediately followed by bilateral infusions of anisomycin (62.5 μg/μL) or vehicle in the BLA. A freezing test was performed 24 h later (Figure S2A). The two-way analysis of variance (ANOVA) of fear expression revealed a significant retrieval × treatment interaction (F2, 39 = 4.96, p = 0.012). The post hoc analysis showed that fear responses in rats that underwent weak US retrieval and received an anisomycin injection significantly decreased compared with rats that underwent weak US retrieval and received a vehicle injection (p = 0.0093, Figure S2B). These results indicate that exposure to a weak electric shock triggered US-specific memory reconsolidation in rats.

To examine whether Pi4KIIα is activated after US retrieval, four groups of rats were trained for contextual fear conditioning and underwent US retrieval one day later. Brain tissues were then collected 15 min, 1 h, or 4 h later, time points within the reconsolidation window (Figure 1A). Pi4KIIα levels increased 15 min after US retrieval in the BLA but not CeA and returned to baseline levels at 1 h (one-way ANOVA; BLA, F3, 20 = 3.217, p = 0.0448; post hoc, USR 15 min versus NoR, p = 0.0362; Figures 1B and 1C). These results suggest the transient regulation of Pi4KIIα in US-retrieval-induced contextual fear memory reconsolidation.

Figure 1

Unconditioned Stimulus Retrieval Transiently Upregulates Pi4KIIα Levels and Promotes Synaptic AMPAR Incorporation in the Rat BLA

(A) Experimental timeline.

(B and C) (B) Representative Western blots and (C) protein levels of Pi4KIIα and EEA1 in the basolateral amygdala (BLA) and central nucleus of the amygdala after US retrieval. Cytosolic Pi4KIIα levels increased in the BLA but not CeA15 min after US retrieval. Cytosolic EEA1 levels in the BLA decreased 15 min after US retrieval (n = 6 rats/group).

(D and E) (D) Representative Western blots and (E) protein levels of EEA1, GluA1-3, and PSD95 in the BLA after US retrieval. Unconditioned stimulus retrieval increased the synaptosomal membrane expression of EEA1, GluA1, GluA2, and PSD95 but not GluA3 (n = 4–6 rats/group). Data are reported as mean ± SEM. One-way ANOVA followed by Dunnett's multiple-comparison post hoc test. *p < 0.05 and **p < 0.01. NoR, no retrieval; USR, unconditioned stimulus retrieval.

Pi4KIIα is known to play an important role in endosomal trafficking (Minogue, 2018). We examined whether US retrieval alters the levels of endosomal compartments. We found a significant reduction of cytosolic EEA1 levels 15 min after US retrieval but a significant increase in the synaptosomal membrane fraction 1 h after retrieval (one-way ANOVA; cytosol, F3, 20 = 9.666, p = 0.0004; post hoc, USR 15 min versus NoR, p = 0.0227; membrane, F3, 20 = 12.11, p < 0.0001; post hoc, USR 1h versus NoR, p = 0.0026; Figures 1B–1E). This shift suggests the dynamic regulation of EEA1 levels in memory reconsolidation.

Fear memory retrieval induces transient changes in AMPAR surface expression (Hong et al., 2013, Liu et al., 2014, Rao-Ruiz et al., 2011, Zhou et al., 2015). We found that US retrieval significantly increased the levels of GluA1 (one-way ANOVA; F3, 18 = 4.27, p = 0.0192, post hoc, USR 1 h versus NoR, p = 0.0092), GluA2 (one-way ANOVA; F3, 18 = 5.16, p = 0.0095, post hoc, USR 1 h versus NoR, p = 0.005), and PSD95 (one-way ANOVA; F3, 16 = 4.899, p = 0.0133, post hoc, USR 1 h versus NoR, p = 0.0217) in the synaptosomal membrane fraction of the BLA 1 h after US retrieval (Figures 1D and 1E). These data indicate that US retrieval after fear conditioning enhanced AMPAR trafficking and synaptic potentiation.

Unconditioned Stimulus Retrieval Increases Dendritic Spine Density and Synaptic Activity in Rat BLA Neurons

Alterations of synaptic morphology contribute to fear memory acquisition and cocaine relapse (Gipson et al., 2013, Roberts et al., 2010). However, unclear is whether BLA neurons undergo a similar process during memory reconsolidation. We investigated spine density alterations during memory reconsolidation in the BLA using Golgi staining. Four groups of rats underwent contextual fear conditioning. The next day, the rats underwent a US retrieval procedure. Brain tissues were then collected 15 min, 1 h, or 4 h later and subjected to Golgi staining (Figure 2A). The spine densities of BLA neurons in rats in the US retrieval group significantly increased at all three time points compared with the NoR group (one-way ANOVA; F3, 53 = 44.12, p < 0.0001; post hoc, all time points versus NoR, 15 min, p = 0.0001; 1 h, p = 0.0001; 4 h, p = 0.0001; Figure 2B). We then classified dendrite spines into several types, including stubby, thin, mushroom, branched, and filopodia (Harris et al., 1992, Lippman and Dunaevsky, 2005, Zagrebelsky et al., 2005). Golgi staining revealed that US retrieval increased the density of mushroom and thin-shaped spines at the 15 min, 1 h, and 4 h time points, whereas the density of filopodia-shaped spines increased only 4 h after US retrieval (one-way ANOVA; mushroom, F3, 53 = 39.04, p < 0.0001; post hoc, all time points versus NoR, 15 min, p = 0.0017; 1 h, p = 0.0001; 4 h, p = 0.0001; thin, F3, 53 = 14.37, p < 0.0001; post hoc, all time points versus NoR, 15 min, p = 0.0101; 1 h, p = 0.0001; 4 h, p = 0.0006; filopodia, F3, 53 = 5.381, p = 0.0026; post hoc, USR 4 h versus NoR, 4 h, p = 0.0038; Figure 2C). The density of stubby and branched spines remained unchanged (Figure 2C). We also measured miniature excitatory postsynaptic current (mEPSC) frequency and amplitude 1 h after US retrieval, the time point identical to the changes in synaptosomal membrane AMPARs (Figure 2D). Consistent with the increases in spine density and synaptosomal membrane GluA1 and GluA2 levels in the BLA, US retrieval significantly increased mEPSC frequency (unpaired t test; t = 3.032, p = 0.005) but had no effect on mEPSC amplitude in BLA neurons (Figures 2E–2G). These data suggest that US retrieval after fear conditioning enhances neuronal connectivity in the BLA.

Figure 2

Unconditioned Stimulus Retrieval Increases Spine Density and Synaptic Activity in the Rat BLA

(A) Experimental timeline.

(B) Representative images of Golgi-stained sections (left) and quantitative results (right). Total dendritic spine density increased 15 min, 1 h, and 4 h after US retrieval. NoR, n = 17 neurons, 12 slices from 3 rats; USR 15 min, n = 14 neurons, 12 slices from 3 rats; USR 1 h, n = 15 neurons, 12 slices from 3 rats; USR 4 h, n = 11 neurons, 10 slices from 3 rats. Scale bar = 5 μm.

(C) Quantification of five types of dendritic spines. The data are reported as mean ± SEM. One-way ANOVA followed by Dunnett's multiple-comparison post hoc test. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p< 0.0005.

(D) Experimental timeline.

(E) Example mEPSC traces from BLA neurons in the NoR group (n = 18 neurons, 8 slices from 5 rats) and USR 1 h group (n = 14 neurons, 7 slices from 4 rats).

(F and G) Cumulative distribution of mEPSC interevent intervals and average frequency (F), or mEPSC amplitude (G) of neurons in the NoR group and USR group. Data are reported as mean ± SEM. Unpaired t test. **p < 0.01.NoR, no retrieval; USR, unconditioned stimulus retrieval; mEPSC, miniature excitatory postsynaptic current.

Pharmacological Inhibition of Pi4KIIα in the BLA after US Retrieval Impairs Contextual Fear Memory Reconsolidation, and This Effect Is Long Lasting

We next tested whether Pi4KIIα is necessary for the US retrieval-induced reconsolidation process. The intra-BLA infusion of phenylarsine oxide (PAO), an inhibitor of Pi4KIIα (Boura and Nencka, 2015), was performed immediately after retrieval. As shown in Figure 3A, four groups of rats underwent contextual fear conditioning. On the next day, the rats were given different doses of PAO (0, 50, 100, and 200 μM/side) bilaterally in the BLA immediately after US retrieval. A freezing test was conducted 24 h later. Rats that received 200 μM PAO after US retrieval exhibited a significant reduction of fear expression (one-way ANOVA; F3, 28 = 3.56, p = 0.0267; post hoc, vehicle versus 200 μM PAO, p = 0.0139; Figure 3B), suggesting that the PAO injection in the BLA impaired the reconsolidation of contextual fear memory.

Figure 3

Pharmacological Inhibition of Pi4KIIα after US Retrieval Impairs Contextual Fear Memory Reconsolidation and Exerts a Long-Lasting Effect

(A) Experimental timeline.

(B) Tests of fear expression in rats that were injected with PAO (0, 50, 100, and 200 μM/side) in the BLA immediately after US retrieval. USR + Vehicle, n = 7 rats; USR +50 μM PAO, n = 8 rats; USR +100 μM PAO, n = 8 rats; USR +200 μM PAO, n = 9 rats. PAO, phenylarsine oxide. Data represent the mean ± SEM. One-way ANOVA followed by Dunnett's multiple-comparison post hoc test. *p < 0.05.

(C) Experimental timeline.

(D) The inhibitory effect of Pi4KIIα inhibition in the BLA after US retrieval on fear expression lasted at least 2 weeks and was not restored by a reminder footshock. NoR + Vehicle, n = 7 rats; NoR + PAO, n = 7 rats; USR + Vehicle, n = 7 rats; USR + PAO, n = 7 rats; Data represent the mean ± SEM. Repeated measures two-way ANOVA followed by Tukey's multiple -comparison post hoc test. *p < 0.05, **p < 0.01, and ***p < 0.001.

(E and F) (E) Representative Western blots and (F) protein levels of Pi4KIIα, EEA1, GluA1, GluA2, and PSD95 in the BLA in rats that were injected with PAO 1 h after US retrieval (n = 4–6 rats/group). The inhibition of Pi4KIIα activity in the BLA immediately after US retrieval blocked the increases in synaptosomal membrane EEA1, GluA1, and PSD95 levels that were induced by US retrieval. Data are reported as mean ± SEM. Two-way ANOVA followed by Tukey's multiple-comparison post hoc test. *p < 0.05.NoR, no retrieval; USR, unconditioned stimulus retrieval.

We then tested whether PAO affects contextual fear memory acquisition, consolidation, and retrieval. Two groups of rats were microinjected with PAO or vehicle in the BLA before fear conditioning, and a freezing test was performed 1 h after conditioning (Figure S3A). No differences in fear expression were found between rats that received PAO and rats that received vehicle (unpaired t test; t = 0.561, p = 0.586; Figure S3B). We used another two groups of rats to determine the effect of a PAO injection in the BLA on fear memory consolidation. Immediately after contextual fear conditioning, the rats were microinjected with PAO or vehicle in the BLA, and a freezing test was performed 24 h later (Figure S3C). No differences in fear expression were found between rats that received PAO and rats that received vehicle (unpaired t test; t = 0.011, p = 0.991; Figure S3D). Another two groups of rats were used to determine the effect of a PAO injection in the BLA on memory retrieval. Twenty-four hours after contextual fear conditioning, the rats were microinjected with PAO or vehicle in the BLA, and a freezing test was performed 1 h later (Figure S3E). We found no difference in fear expression between PAO- and vehicle-injected rats (unpaired t test; t = 0.397, p = 0.699; Figure S3F). These results indicated that the inhibition of Pi4KIIα with PAO in the BLA had no effect on fear memory acquisition, consolidation, or retrieval. Additionally, the inhibition of Pi4KIIα in the BLA had no effect on CS-retrieval-induced fear memory reconsolidation (Figure S4). The PAO injection in the BLA also had no effect on locomotor activity, anxiety-like behavior, or depression-related behavior (Figure S5).

We next tested the long-lasting effect of Pi4KIIα inhibition after US retrieval on contextual fear memory reconsolidation. As shown in Figure 3C, on day 1, four groups of rats underwent contextual fear conditioning. Twenty-four hours later, two groups of rats were microinjected with PAO or vehicle in the BLA immediately after the US retrieval procedure. Another two groups of rats received PAO or vehicle without retrieval. On day 3, all of the rats underwent a freezing test (Test 1). Fourteen days later, freezing behavior was assessed in all of the rats (Test 2). A strong electric footshock (1.0 mA) was then given immediately after Test 2. Twenty-four hours later, the reinstatement test (Test 3) was performed to evaluate fear expression. The two-way repeated-measures ANOVA of fear expression revealed main effects of test (F2, 48 =10.91, p = 0.0001) and treatment (F3, 24 = 10.73, p = 0.0001). The post hoc analysis revealed that fear expression in the group that underwent US retrieval and received the PAO injection in the BLA significantly decreased compared with the group that underwent US retrieval and received vehicle in Test 1 (p = 0.0108), Test 2 (p = 0.0041), and Test 3 (p = 0.0007; Figure 3D).

We then used four groups of rats to determine the effect of PAO administration in the BLA on the expression of AMPAR subunits. As shown in Figure 3C, tissues were collected 1 h after retrieval for Western blot analysis. The PAO injection in the BLA immediately after US retrieval decreased the levels of EEA1 (two-way ANOVA; group × treatment interaction: F1, 19 = 6.006, p = 0.0241; post hoc, USR-vehicle versus USR-PAO, p = 0.0446), GluA1 (two-way ANOVA; group × treatment interaction: F1, 19 = 6.215, p = 0.0221; post hoc, USR-vehicle versus USR-PAO, p = 0.0456), and PSD95 (two-way ANOVA; group × treatment interaction: F1, 19 = 3.662, p = 0.0709; post hoc, USR-vehicle versus USR-PAO, p = 0.0436), whereas the expression of Pi4KIIα (two-way ANOVA; group × treatment interaction: F1, 12 = 1.376, p = 0.264; post hoc, USR-vehicle versus USR-PAO, p = 0.736) and GluA2 (two-way ANOVA; group × treatment interaction: F1, 12 = 2.374, p = 0.1493; post hoc, USR-vehicle versus USR-PAO, p = 0.616) remained unchanged (Figures 3E and 3F). We also found no difference in Pi4KIIα levels between vehicle- and PAO-injected rats 24 h after US retrieval (unpaired t test; t = 0.731, p = 0.493; Figure S6).

Altogether, these results indicate that the pharmacological inhibition of Pi4KIIα in the BLA after US retrieval disrupted contextual fear memory reconsolidation, which lasted for at least two weeks, and fear memory did not recover after a reminder footshock.

Pharmacological Inhibition of Pi4KIIα in the BLA without Exposure or 6 h after Exposure to a US Has No Effect on the Expression of Contextual Fear Memory

We further tested whether the effect of PAO on subsequent fear expression is retrieval dependent and temporally specific. Two groups of rats were trained for contextual fear conditioning and then received a microinjection of vehicle (1.0 μL/side) or PAO (200 μM/side) bilaterally in the BLA without undergoing a US retrieval trial (Figure S7A). The PAO injection in the BLA had no effect on the subsequent expression of contextual fear memory in the absence of US retrieval (Figure S7B). Another two groups of rats underwent contextual fear conditioning and then received a microinjection of vehicle (1.0 μL/side) or PAO (200 μM/side) bilaterally in the BLA 6 h after US retrieval, and a freezing test was performed 24 h later (Figure S7C). No differences in fear expression were observed between vehicle- and PAO-injected rats (unpaired t test; t = 0.0822, p = 0.936; Figure S7D), indicating that the effect of PAO on contextual fear memory reconsolidation was temporally limited.

Genetic Knockdown of Pi4KIIα Disrupts the Reconsolidation of Contextual Fear Memory and Inhibits Synaptic Transmission in the BLA

The microinjection of PAO in the BLA after US retrieval significantly impaired contextual fear memory reconsolidation. In addition to inhibiting Pi4KIIα, PAO has also been shown to inhibit neurotransmitter release and phosphatase activity (Searl and Silinsky, 2000, Zhang et al., 1992), which may explain the effect on memory reconsolidation that was observed herein. To further reveal the role of Pi4KIIα in the reconsolidation of contextual fear memory, we constructed an adeno-associated virus (AAV) to knockdown endogenous Pi4KIIα. We first evaluated the efficiency of the viral vector and found that it specifically decreased Pi4KIIα levels in the BLA (unpaired t test; t = 8.326, p = 0.0002; Figures 4A–4C).

Figure 4

Knockdown of Pi4KIIα in the BLA Disrupts the Reconsolidation of Contextual Fear Memory

(A) Representative infusion site in the rat BLA.

(B and C) (B) Representative Western blots and (C) Pi4KIIα expression in the BLA 21 days after the injection of AAV-expressing Pi4KIIα shRNA (shPi4KIIα) or scramble shRNA (SCR; n = 4 rats/group). Data represent the mean ± SEM. Unpaired t test. ***p < 0.001.

(D) Experimental timeline.

(E) shPi4KIIα-injected rats and SCR-injected rats exhibited comparable fear expression. 1 h: SCR, n = 7 rats; shPi4KIIα, n = 8 rats; 24 h: SCR, n = 7 rats; shPi4KIIα, n = 7 rats. Data are reported as mean ± SEM. Unpaired t test.

(F) Experimental timeline.

(G) shPi4KIIα-injected rats exhibited impairments in fear reconsolidation, which lasted at least two weeks and were not reinstated by a reminder footshock. The TAT-Pi4KIIα injection restored these behavioral effects. SCR + Saline, n = 9 rats; SCR + TAT-Pi4KIIα, n = 10 rats; shPi4KIIα + Saline, n = 7 rats; shPi4KIIα + TAT-Pi4KIIα, n = 7 rats. Repeated measures two-way ANOVA followed by Tukey's multiple-comparison post hoc test. *p < 0.05, **p < 0.01, and ***p < 0.001.

(H and I) (H) Representative Western blots and (I) protein levels of Pi4KIIα, EEA1, GluA1, and PSD95 in the BLA in shPi4KIIα-injected rats 1 h after US retrieval. Pi4KIIα, EEA1, GluA1, and PSD95 levels decreased in the BLA 1 h after US retrieval. The TAT-Pi4KIIα injection reversed the expression of Pi4KIIα, GluA1, and PSD95 (n = 4–6 rats/group). Data are reported as mean ± SEM. Two-way ANOVA followed by Tukey's multiple-comparison post hoc test. *p < 0.05 and **p < 0.01. USR, unconditioned stimulus retrieval.

To test whether Pi4KIIαdownregulation affects the acquisition and consolidation of contextual fear memory, rats were injected Pi4KIIα short hairpin RNA (shRNA; shPi4KIIα) or scrambled shRNA (SCR) and underwent a training session followed by a freezing test 1 or 24 h later (Figure 4D). No differences in freezing were found between the shPi4KIIα and SCR groups (unpaired t test; 1 h, t = 0.363, p = 0.723; 24 h, t = 1.041, p = 0.318; Figure 4E), suggesting that Pi4KIIα knockdown had no effect on the acquisition or consolidation of contextual fear memory. Rats with Pi4KIIα knockdown were also tested for anxiety- and depression-like behavior, and no significant effects were observed (Figure S8).

To further evaluate the role of Pi4KIIα in the US-retrieval-induced reconsolidation of contextual fear memory, we prepared TAT-Pi4KIIα proteins that were microinjected in the BLA in Pi4KIIα knockdown rats to mimic normal Pi4KIIα expression patterns. We first evaluated the transduction efficacy of TAT-Pi4KIIα by assessing protein levels using Western blot (Figure S9A). Pi4KIIα levels in the BLA significantly increased 1 h after the TAT-Pi4KIIα microinjection (unpaired t test; t = 2.624, p = 0.0394; Figures S9B and S9C).

Four groups of rats received microinjections of shPi4KIIα or SCR in the BLA 21 days before contextual fear conditioning. Twenty-three hours after training, the rats were microinjected with TAT-Pi4KIIα or saline and then underwent US retrieval trials 1 h later. One day later, all of the rats underwent a freezing test (Test 1). Fourteen days later, freezing expression was assessed in all of the rats (Test 2), and then a strong electric footshock (1.0 mA) was given immediately after Test 2. Twenty-four hours later, a reinstatement test (Test 3) was performed to evaluate long-lasting inhibitory effects (Figure 4F). The two-way repeated-measures ANOVA of fear expression revealed main effects of test (F2, 58 = 32.2, p < 0.0001) and treatment (F3, 29 = 13.09, p < 0.0001; Figure 4G). The post hoc analysis revealed that Pi4KIIα knockdown in the BLA significantly reduced fear expression compared with SCR-injected rats (Test 1, p = 0.0144; Test 2, p = 0.0027; Test 3, p = 0.0002; Figure 4G). The TAT-Pi4KIIα microinjection in the BLA restored fear expression in Pi4KIIα knockdown rats (Test 1, p = 0.0230; Test 2, p = 0.0128; Test 3, p = 0.0415; Figure 4G).

We then examined protein expression 1 h after US retrieval in shPi4KIIα- and TAT-Pi4KIIα-injected rats (Figures 4H and 4I). We found that shPi4KIIα microinjections in the BLA significantly decreased the levels of Pi4KIIα (two-way ANOVA; AAV × treatment interaction: F1, 12 = 8.08, p = 0.0148; post hoc, p = 0.0259), EEA1 (two-way ANOVA; main effect of AAV: F1, 20 = 9.823, p = 0.0052; post hoc, p = 0.0432), GluA1 (two-way ANOVA; AAV × treatment interaction: F1, 20 = 5.345, p = 0.0316; post hoc, p = 0.0179), and PSD95 (two-way ANOVA; AAV × treatment interaction: F1, 20 = 6.511, p = 0.0190; post hoc, p = 0.0069) after US retrieval (Figures 4H and 4I). TAT-Pi4KIIα microinjections restored the decrease in the levels of Pi4KIIα (p = 0.0407), GluA1 (p = 0.0279), and PSD95 (p = 0.0058) in shPi4KIIα-injected rats (Figures 4H and 4I).

To investigate the effect of Pi4KIIα knockdown on synaptic activity in the BLA, we performed whole-cell voltage-clamp recordings in SCR- and shPi4KIIα-infected BLA neurons in acute brain slices from untrained rats (Figure 5A). Pi4KIIα knockdown did not significantly alter mEPSC frequency or amplitude (Figures 5B–5D). We then examined whether Pi4KIIα knockdown in the BLA affects the US-retrieval-induced increase in synaptic strength. Pi4KIIα knockdown significantly decreased mEPSC frequency (unpaired t test; t = 2.503, p = 0.0202) 1 h after US retrieval but had no effect on mEPSC amplitude (Figures 5E–5H).

Figure 5

Unconditioned-Stimulus-Retrieval-Dependent Alterations of Synaptic Efficacy Depend on Pi4KIIα

(A) Experimental timeline.

(B) Example mEPSC traces from BLA neurons in the shPi4KIIα group (n = 15 neurons, 7 slices from 4 rats) and SCR group (n = 16 neurons, 5 slices from 4 rats).

(C and D) Cumulative distribution of mEPSC interevent intervals and average frequency (C) or mEPSC amplitude (D) of neurons in the shPi4KIIα group and SCR group.

(E) Experimental timeline.

(F) Example mEPSC traces from BLA neurons in the shPi4KIIα group (n = 12 neurons, 7 slices from 5 rats) and SCR group (n = 12 neurons, 4 slices from 3 rats) 1 h after US retrieval.

(G and H) Cumulative distribution of mEPSC inter-event interval and average frequency (G), or mEPSC amplitude (H) of neurons in the shPi4KIIα group and SCR group 1 h after US retrieval. Data are reported as mean ± SEM. Unpaired t test. *p < 0.05.SCR, scramble; USR, unconditioned stimulus retrieval; mEPSC, miniature excitatory postsynaptic current.

Altogether, these results demonstrate a critical role for Pi4KIIα in US-retrieval-induced fear reconsolidation, likely via the regulation of AMPAR trafficking and synaptic transmission in the BLA.

Preventing Early Endosomal Trafficking in the BLA after US Retrieval Impairs the Reconsolidation of Contextual Fear Memory and Exerts a Long-Lasting Effect

Early endosomal sorting has been shown to be involved in AMPAR trafficking (Parkinson and Hanley, 2018). We found that EEA1 functions as a binding partner of Pi4KIIα (Figures S1F and S1G). Dynasore is used to block early endosome transport (Mesaki et al., 2011). We tested the effect of local dynasore infusion (400 μM/side) in the BLA immediately after US retrieval on the reconsolidation of contextual fear memory. As shown in Figure 6A, four groups of rats underwent contextual fear conditioning. Twenty-three hours later, the rats were microinjected with TAT-Pi4KIIα or saline. One hour later, the rats were microinjected with dynasore or vehicle in the BLA immediately after US retrieval. On day 3, all of the rats underwent a freezing test (Test 1). Fourteen days later, fear expression was assessed in all of the rats (Test 2), and then a strong electric footshock (1.0 mA) was given immediately after Test 2. Twenty-four hours later, the reinstatement test (Test 3) was performed to evaluate fear expression. The repeated-measures ANOVA of the percentage of freezing revealed main effects of test (F2, 58 = 13.09, p < 0.0001) and treatment (F3, 29 = 12.06, p < 0.0001; Figure 6B). The post hoc analysis revealed that the dynasore microinjection in the BLA immediately after US retrieval significantly decreased fear expression in Test 1 (p = 0.0038), Test 2 (p = 0.0126), and Test 3 (p = 0.0160), which was not restored by TAT-Pi4KIIα treatment (Figure 6B).

Figure 6

Preventing Early Endosomal Trafficking in the BLA after US Retrieval Impairs the Reconsolidation of Contextual Fear Memory and Exerts a Long-Lasting Effect

(A) Experimental timeline.

(B) Rats that were injected with dynasore exhibited an impairment in fear expression, which lasted at least two weeks and was not reinstated by a reminder footshock. The TAT-Pi4KIIα injection within the reconsolidation time window did not restore fear expression. Vehicle + Saline, n = 8 rats; Vehicle + TAT-Pi4KIIα, n = 10 rats; Dynasore + Saline, n = 7 rats; Dynasore + TAT-Pi4KIIα, n = 8 rats. Repeated measures two-way ANOVA followed by Tukey's multiple-comparison post hoc test. *p < 0.05 and **p < 0.01.

(C and D) (C) Representative Western blots and (D) protein levels of Pi4KIIα, EEA1, GluA1, and PSD95 in the BLA in rats that were injected with dynasore 1 h after US retrieval. The dynasore injection decreased synaptosomal membrane EEA1 and GluA1 levels in the BLA 1 h after US retrieval. The TAT-Pi4KIIα injection did not reverse the protein expression of EEA1 or GluA1 (n = 6 rats/group). Data are reported as mean ± SEM. Two-way ANOVA followed by Tukey's multiple-comparison post hoc test. *p < 0.05 and ***p < 0.001. USR, unconditioned stimulus retrieval.

Finally, we examined the effect of dynasore and TAT-Pi4KIIα infusions in the BLA on protein expression levels 1 h after US retrieval (Figures 6C and 6D). The dynasore microinjection after US retrieval significantly decreased the levels of EEA1 (two-way ANOVA; main effect of drug [dynasore or vehicle]: F1, 20 = 9.552, p = 0.0058; post hoc, p = 0.0331) and GluA1 (two-way ANOVA; main effect of drug [dynasore or vehicle]: F1, 20 = 47.1, p < 0.0001; post hoc, p = 0.0005; Figures 6C and 6D). TAT-Pi4KIIα treatment did not reverse the decrease in the expression of EEA1 (p = 0.232) or GluA1 (p = 0.232) in dynasore-injected rats (Figures 6C and 6D). These results suggest that Pi4KIIα may regulate US-retrieval-induced fear memory reconsolidation via early endosomal trafficking.

Discussion

In the present study, we found that US retrieval significantly increased cytosolic Pi4KIIα levels and synaptosomal membrane EEA1 and GluA1 levels and decreased cytosolic EEA1 levels in the BLA in rats. The intra-BLA infusion of PAO and knockdown of Pi4KIIα disrupted contextual fear memory reconsolidation and decreased subsequent fear expression. This disruptive effect persisted for at least two weeks and was not reversed by a reminder footshock. We also found that US retrieval increased spine density and synaptic efficacy in BLA neurons. The genetic knockdown of Pi4KIIα abolished the US-retrieval-induced increase in synaptic transmission in the BLA. Furthermore, we found that Pi4KIIα bound to EEA1 and GluA1, and the pharmacological inhibition of early endosomal trafficking after US retrieval impaired the surface expression of GluA1 and subsequent fear expression. Altogether, these results indicate that Pi4KIIα plays a critical role in US-retrieval-induced fear reconsolidation, likely via the regulation of early endosomal sorting and AMPAR trafficking.

Posttraumatic stress disorder causes great suffering and is associated with high social and economic costs. Accumulating evidence suggests that disruption of the reconsolidation of an activated fear memory prevents subsequent fear expression. Our understanding of the cellular and molecular mechanisms of memory reconsolidation is still at an early stage. Several neurotransmitter systems and intracellular signaling cascades have been shown to be required for memory consolidation and reconsolidation (Johansen et al., 2011). However, the nonspecific involvement of some key molecules in multiple memory processes may limit their preclinical to clinical translation. In the present study, we found that Pi4KIIα inhibition did not affect contextual fear memory acquisition, consolidation, or retrieval. Interference with Pi4KIIα outside the time window of reconsolidation had no effect on the subsequent expression of fear, indicating that Pi4KIIα specifically affects the reconsolidation of fear memory.

Unconditioned stimuli serve as powerful reminders to trigger reconsolidation (Debiec et al., 2010). A previous study reported that the second learning occurred independently of dorsal hippocampal N-methyl-D-aspartate (NMDA) receptors in rats that were previously exposed to a similar fear conditioning procedure (Finnie et al., 2018). However, the single exposure to a weak footshock that was used in the present study may not be sufficient to induce the second learning. Indeed, we found that weak US retrieval, followed by an anisomycin injection in the BLA, decreased the fear response, indicating that exposure to the US triggered the reconsolidation process. Additionally, although interventions that target US-retrieval-induced reconsolidation are associated with a better treatment outcome than CS exposure (Huang et al., 2017), the molecular and cellular mechanisms that underlie these differences remain unclear. A recent study showed that US retrieval induced significant CREB activation in almost whole amygdala and hippocampus, whereas CS retrieval only stimulated CREB activation in the lateral amygdala and the CA3 (Huang et al., 2017). In the present study, we found that Pi4KIIα levels in the BLA increased 15 min after US retrieval and decreased to baseline levels 1 h after US retrieval. Additionally, Pi4KIIα inhibition in the BLA had no effect on CS retrieval-induced memory reconsolidation but significantly influenced US-retrieval-induced memory reconsolidation. The dysfunction of Pi4KIIα contributes to Alzheimer disease, Gaucher disease, Hermansky-Pudlak syndrome, and X-linked centronuclear myopathy (Jovic et al., 2012, Kang et al., 2013, Ketel et al., 2016, Salazar et al., 2005, Salazar et al., 2009, Wu et al., 2004). The present study identified a functional role for Pi4KIIα in learning and memory, suggesting that it may be a promising molecular target for erasing aversive emotional memories.

Memory retrieval and reconsolidation are associated with glutamate receptor expression and activity. AMPAR trafficking is essential for the retrieval and subsequent reconsolidation of fear memory in the amygdala (Clem and Huganir, 2010; Hong et al., 2013). In the present study, we found that US retrieval increased GluA1 and GluA2 levels in synaptosomal membrane fractions of the BLA. We also found that US retrieval increased dendritic spine density and synaptic strength in the BLA. The higher expression of PSD95 and GluA1 after US retrieval may be a structural basis for the enhancement of synaptic transmission in the BLA. Consistent with our findings, previous studies revealed that fear memory reactivation promoted the surface expression of GluA1 subunits and excitatory synaptic transmission in the lateral amygdala (Zhou et al., 2015). Fear memory retrieval also induces a transient increase in engram cell excitability (Pignatelli et al., 2019). However, some studies found that fear memory retrieval caused the endocytosis of AMPARs (Hong et al., 2013, Rao-Ruiz et al., 2011). This discrepancy may reflect distinct retrieval paradigms. Indeed, US retrieval has been shown to activate more memory traces and stronger protein alterations than CS retrieval (Liu et al., 2014, Xue et al., 2017, Yuan et al., 2019). Furthermore, we found that the pharmacological inhibition of Pi4KIIα in the BLA decreased GluA1 levels 1 h after US retrieval, and the genetic knockdown of Pi4KIIα impaired the enhancement of mEPSC frequency that was induced by US retrieval, suggesting a regulatory role for Pi4KIIα in GluA1 surface expression and synaptic efficacy in the rat BLA.

Pi4KIIα is extensively expressed in both the central and peripheral nervous systems, especially in the cerebellum, dorsal root ganglion, and spinal cord (Simons et al., 2009). In the present study, we found that Pi4KIIα was also expressed in many brain areas that are related to learning and memory, including the DH and BLA. In neurons, Pi4KIIα localizes to dendrites, endosomes, the Golgi apparatus, and synaptic vehicles (Clayton et al., 2013, Minogue, 2018). Similarly, we found that Pi4KIIα colocalized with EEA1 in rat BLA neurons. Pi4KIIα regulates receptor sorting at early endosomes, and Pi4KIIα depletion by siRNA impairs receptor sorting from early endosomes (Henmi et al., 2016). In the present study, EEA1 levels decreased in the cytosolic fraction 15 min after US retrieval and increased in the synaptosomal membrane fraction 1 h after US retrieval. Such changes suggest an interaction between Pi4KIIα and EEA1. Indeed, the co-immunoprecipitation results showed that EEA1 was a binding partner of Pi4KIIα. The pharmacological blockade of early endosome transport with dynasore inhibited subsequent fear expression, indicating a role for early endosome trafficking in US-retrieval-induced fear memory reconsolidation. The TAT-Pi4KIIα microinjection did not restore the lower expression of EEA1 or GluA1 and did not reverse behavioral inhibition that was induced by the blockade of early endosome trafficking, indicating that Pi4KIIα may serve as an upstream regulator of early endosome sorting and AMPAR trafficking. Interestingly, the TAT-Pi4KIIα microinjection increased Pi4KIIα levels in untrained rats but had no effect on Pi4KIIα expression in trained rats that received the SCR or dynasore injection. This discrepancy may be attributed to the distinct sampling time points and influence of fear conditioning and US retrieval.

In neurons, endosomes have been shown to translocate into spines to promote the insertion of AMPARs (Esteves da Silva et al., 2015, Park et al., 2006). Early endosomes are found throughout the soma and dendrites and play a crucial role in the sorting of AMPARs (Parkinson and Hanley, 2018). AMPARs on early endosomes can recycle back to the dendritic plasma membrane through a mechanism that involves the recruitment of recycling endosomes (Anggono and Huganir, 2012). In the present study, US retrieval decreased cytosolic EEA1 levels and increased EEA1 and GluA1 levels in the synaptosomal membrane fraction of the BLA. The pharmacological inhibition of early endosome trafficking in the BLA after US retrieval decreased GluA1 levels in the synaptosomal membrane fraction. These results support a role for early endosomes in the membrane trafficking of AMPARs.

In summary, the present study demonstrated that Pi4KIIα contributes to the early endosomal trafficking of GluA1-containing AMPARs during US retrieval-induced contextual fear memory reconsolidation. Our findings may contribute to the development of effective therapeutic strategies that target Pi4KIIα to alleviate the return of fear in PTSD.

Limitations of the Study

The present study demonstrated important roles for Pi4KIIα in the early endosomal trafficking of GluA1-containing AMPARs, synaptic enhancement, and US-retrieval-induced contextual fear memory reconsolidation. Further studies are needed to examine whether Pi4KIIα also plays a role in remote memory reconsolidation and whether recycling endosomes also participate in US-retrieval-induced synaptic AMPAR incorporation.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.