The Wnt signaling pathway plays critical roles in development. However, to date, the role of Wnts in learning and memory in adults is still not well understood. Here, we aimed to investigate the roles and mechanisms of Wnts in hippocampal-dependent contextual fear conditioning (CFC) memory formation in adult mice. CFC training induced the secretion and expression of Wnt3a and the activation of its downstream Wnt/Ca2+ and Wnt/β-catenin signaling pathways in the dorsal hippocampus (DH). Intrahippocampal infusion of Wnt3a antibody impaired CFC acquisition and consolidation, but not expression. Using the Wnt antagonist sFRP1 or the canonical Wnt inhibitor Dkk1, we found that Wnt/Ca2+ and Wnt/β-catenin signaling pathways were involved in acquisition and consolidation, respectively. Moreover, we found Wnt3a signaling is not only necessary but also sufficient for CFC memory. Intrahippocampal infusion of exogenous Wnt3a could enhance acquisition and consolidation of CFC. Overexpression of constitutively active β-catenin in the DH could rescue the deficit in CFC memory consolidation, but not acquisition induced by Wnt3a antibody injection, which suggests β-catenin signaling pathway acts downstream of Wnt3a to mediate CFC memory consolidation. Our study may help further the understanding of the precise regulation of Wnt3a in differential memory phases depending on divergent signaling pathways.
Wnts constitute a family of evolutionarily conserved and secreted lipid-modified glycoproteins that act as ligands to activate different signaling pathways. To date, 19 Wnt proteins have been identified in higher vertebrates (Miller 2002; Kawano and Kypta 2003). Three Wnt signaling pathways have been well characterized, including the canonical Wnt/β-catenin, the noncanonical Wnt/Ca2+ and the Wnt/planar cell polarity (Wnt/PCP) pathways (Behrens and Kuhl 2003). In the Wnt/β-catenin pathway, Wnts bind to the Frizzled receptors and the coreceptors LRP5/6, which leads to the phosphorylation of Disheveled (Dvl) (Bhanot et al. 1996). The activation of Dvl leads to the inhibition of GSK3, which allows β-catenin to be stabilized and accumulated in the cytoplasm and translocated to the nucleus where it activates the transcription of T-cell factor/lymphoid enhancer factor (TCF/LEF) target genes (Logan and Nusse 2004). In the Wnt/Ca2+ pathway, the binding of Wnts to Frizzled could increase the intracellular Ca2+, leading to the activation of CaMKII and PKC (Kuhl et al. 2000). Finally, in the Wnt/PCP pathway, the binding of Wnts to Frizzled could induce the activation of Rho and Rac small GTPases, which activate ROCK and JNK, respectively (Simons and Mlodzik 2008; Oliva et al. 2013).
Wnts are expressed in the brain. Previous studies have shown that Wnts played critical roles in the development of the nervous system, including hippocampal formation, axon pathway finding, dendritic morphogenesis and synapse formation (Hall et al. 2000; Lee et al. 2000; Krylova et al. 2002; Packard et al. 2002; Yu and Malenka 2003). Recently, increasing evidence has shown that Wnts also exert functions in the synaptic plasticity in adults (Ahmad-Annuar et al. 2006; Chen et al. 2006; Cerpa et al. 2008, 2010, 2011; Oliva et al. 2013). Moreover, the roles of Wnts in amygdala or hippocampal-dependent memory were also reported recently. Maguschak and Ressler (2011) found that Wnt1 mRNA in the amygdala decreased immediately following cued fear conditioning, and the infusion of Wnt1 protein prior to training impaired long-term memory (LTM) in cued fear conditioning. In the hippocampal-dependent memory paradigm, it was found that spatial learning in a hidden platform water maze induced the elevation of Wnt7 and Wnt5a, but not Wnt3 expression in the hippocampus (Tabatadze et al. 2012). However, whether the increased Wnt7 and Wnt5a expression was functionally necessary for memory formation was not determined. In the object recognition task, the infusion of the canonical Wnt antagonist DKK1 into the dorsal hippocampus (DH) immediately after training impaired the memory consolidation, which suggests canonical Wnt signaling is required for hippocampal-dependent memory consolidation (Fortress, Schram et al. 2013). However, it was not determined whether Wnt/β-catenin signaling was sufficient for memory consolidation or whether Wnt/β-catenin signaling was involved in other memory phases, such as acquisition or expression in the hippocampus. It also remains puzzling which Wnt molecule mediates the activation of the Wnt/β-catenin signaling pathway in hippocampal-dependent memory formation.
Contextual fear conditioning (CFC) is a well-established model to study hippocampal-dependent memory processes, where the subject associates an aversive stimulus, such as a footshock (termed the unconditioned stimulus [US]), with a distinctive context (termed the conditioned stimulus [CS]). The involvement of the DH in CFC is well documented (Rudy et al. 2002). In the present study, using the CFC paradigm and a variety of molecular and pharmacological approaches, we aimed to investigate which Wnts in the DH played roles in CFC memory formation and its underlying mechanism.
Materials and Methods
Male C57BL/6 mice (8-week-old weighing 23–25 g) were housed at 22°C ± 2°C on a 12 h light/dark cycle. Food and water were available ad libitum. All procedures were conducted on the basis of the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the institutional animal care and use committee of Shandong University.
Reagents and Antibodies
Recombinant mouse Wnt3a, DKK1, and recombinant human sFRP1 were purchased from R&D Systems. KN-62 and control rat IgG were purchased from Sigma Aldrich. The primary antibodies used in our experiments were as follows: rabbit anti-CaMKIIα, rabbit anti-phospho-CaMKIIα, and rabbit anti-phospho-GSK3β antibodies were purchased from Cell Signaling Technology. Mouse anti-active-β-catenin and rabbit anti-Wnt3a antibodies were purchased from Millipore. Rat anti-Wnt3a antibody was purchased from R&D; rabbit anti-actin antibody was purchased from Sigma Aldrich; mouse anti-β-catenin antibody was purchased from BD; goat anti-lamin B and goat anti-Wnt5a antibodies were purchased from Santa Cruz Biotechnology; and rabbit anti-GFP antibody was purchased from Invitrogen. Secondary antibodies: Alexa Fluor 488-conjugated donkey anti-rabbit IgG was purchased from Invitrogen; horseradish peroxidase (HRP)-conjugated goat anti-mouse, rat or rabbit IgG and HRP-conjugated rabbit anti-goat IgG antibodies were purchased from Calbiochem.
Contextual Fear Conditioning
The mice were put into the conditioning chamber (25 × 25 × 25 cm) and allowed to habituate for 120 s without any stimulation (habituation); they then received 3 consecutive footshocks (0.4 mA or 0.7 mA, 2 s duration each) through a stainless steel grid floor (Panlab, Barcelona, Spain). Each footshock was separated by a 60 s time period. After an additional 60 s following the last shock, the mice were placed back to their home cage. The footshock US was generated by a programmable animal shocker, and the CS was the experimental context.
Short-term memory (STM) and LTM were tested 1 and 24 h after training, respectively. Two separate groups of mice were used for the STM and LTM tests. The animals were returned to the previous chamber in which the training occurred and tested for 5 min without footshock; memory was assessed by measuring freezing behavior.
In addition, to separate the impact of the “context” and the “shock” on Wnt expression in the DH, 2 experiments were performed. To determine the effect of the context exposure (“context”), the mice were allowed to freely explore for 5 min in the training chamber without receiving footshock. To determine the effect of the shock and minimize the context exposure (“immediate shock”), the mice were given a 2 s footshock (0.7 mA) immediately after being placed in the training chamber and were quickly removed and returned to their home cage (Huff et al. 2006; Lopez-Fernandez et al. 2007).
Surgery and Microinjection
The mice, anesthetized with 5% chloral hydrate (0.6 mL per 100 g, i.p.), were implanted bilaterally with 26-gauge guide cannulas to the DH or amygdala. The coordinates were as follows: DH: anteroposterior (AP), −1.7 mm; lateral (L), ±1.5 mm; dorsoventral (V), −2.3 mm (Fortress, Fan et al. 2013). Basolateral amygdala: AP, −1.4 mm; L, ±3.5 mm; V, −5.1 mm (Ogden et al. 2014). To prevent clogging, a stylus was placed in the guide cannula. After surgery, the animals were allowed to recover for 1 week before training.
For microinjection, the stylus was removed, and an infusion cannula that extended 0.5 mm beyond the tip of the guide cannula was inserted. The infusion cannula was connected to a 10 µL microsyringe via PE20 tubing driven by a microinjection pump (KDS200, KD Scientific). Wnt3a antibodies (dissolved in 0.1 M PBS, 1 µg/µL, 0.5 µL/side), a control IgG (dissolved in 0.1 M PBS, 1 µg/µL, 0.5 µL/side), Dkk1 (dissolved in 0.1 M PBS, 200 ng/µL, 0.5 µL/side), sFPR1 (dissolved in 0.1 M PBS, 250 ng/µL, 0.5 µL/side), Wnt3a (dissolved in 0.1 M PBS, 5 ng/µL, 0.5 µL/side), KN-62 (dissolved in 0.2% dimethyl sulfoxide, 32 ng/µL, 0.5 µL/side) or their respective vehicles were microinfused bilaterally. Lentivirus carrying an active-β-catenin expressing cassette (referred to as “lenti-active-β-catenin”, 1 µL/side) or a green-fluorescent protein (referred to as “lenti-GFP”, 1 µL/side) with a final infectious unit titer of 109 TU/mL was used for microinfusion or stereotaxic injections.
The mice were deeply anesthetized with 5% chloral hydrate and perfused with 0.9% normal saline followed by 4% paraformaldehyde (PFA). Then, the brains were acquired for postfixation in 30% sucrose in PBS. The brains were frozen and sliced (30 µm, coronal) using a freezing microtome. The slices were blocked with a blocking solution (0.3% Triton X-100 and 10% normal donkey serum in TBS) for 1 h and incubated in the primary antibodies for Wnt3a or GFP overnight at 4°C. The slices were incubated with Donkey anti-mouse Alexa 488 or Donkey anti-rabbit Alexa 488 for 1 h at room temperature. The microphotographs were captured using confocal fluorescence microscopy with a Carl Zeiss LSM-780 microscope or a spinning disk microscope (Microstructural Platform of Shandong University). The images were analyzed using ImageJ.
Quantitative Real-Time PCR
Total RNA was isolated using TRNzol-A+ RNA isolation reagent (TIANGEN) following the manufacturer's instructions. Purified total RNA (500 ng) was then reversely transcribed using the RevertAid First Strand cDNA Synthesis Kit (Fermentas) according to the manufacturer's instructions.
Quantitative Real-time PCR was performed in a Cycler (Bio-Rad) with the use of SYBR Green (Roche). The primer sequences used are listed in Supplementary Table 1. Each sample was assayed in duplicate and the relative levels of mRNA were normalized for each sample with the expression levels of the β-actin mRNA using the 2−ΔΔCT method.
Tissue Preparation and Western Blot
The brains were removed immediately after the animals were killed at the desired time points. The DH was dissected on ice and stored at −80°C until used. The samples were homogenized in ice-cold RIPA buffer, which contained 25 mM Tris–HCl, pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate and 0.1% sodium dodecyl sulfate (SDS) with protease and phosphatase inhibitors. After grinding into homogenate and incubating for 15 min on ice, the extracts were centrifuged at 14 000×g for 15 min at 4°C, and the supernatants were collected as total proteins. The nuclear proteins were isolated by NE-PER nuclear and cytoplasmic extraction reagents (Pierce). The concentration of protein was detected using the bicinchoninic acid protein assay. Densitometry analysis on the bands was performed using Quantity one.
Synaptoneurosomes (SNS) Preparation
The DH was obtained, and SNS were extracted as described previously (Villasana et al. 2006; Chen et al. 2011). In brief, the tissue was homogenized in ice-cold homogenization buffer (10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.5 mM dithiothreitol, 1.0 mM ethylenediaminetetraacetic acid, 2.0 mM ethylene glycol tetraacetic acid) with protease and phosphatase inhibitor cocktails (Roche). The homogenates were then filtered through 2 100-µm-pore filters and two 5-µm-pore filters sequentially. SNS were acquired by centrifuging the resulting filtrates at 1000×g for 15 min at 4°C. The SNS was dissolved in 1% SDS, boiled for 10 min and stored at −20°C.
Data were analyzed by 2-tailed t-test or one-way analysis of variance (ANOVA), followed by least significant difference post hoc comparisons. Significance was set to 0.05 for all statistical analyses, and the results represent the means ± SEM. Data analyses were performed using the SPSS statistical program, version 18.0.
Selective Induction of Wnt3a Expression in the DH Following CFC Training
CFC is a well-known hippocampal-dependent learning task. We initially examined whether the expression of various Wnts was altered in the DH after CFC training. We focused on 6 Wnt molecules (Wnt1, Wnt3, Wnt3a, Wnt4, Wnt5a, and Wnt7a), which have been reported to be highly expressed in the hippocampus. Figure 1A illustrates that among the 6 different Wnts, only Wnt3a mRNA levels were increased significantly at 2 h after CFC training using real-time PCR (P = 0.003, 2-tailed t-test). The 2 h time point was chosen based on a previous report that 2 h after the fear conditioning was the optimal time to observe changes in expression for most genes (Ressler et al. 2002).
To determine the time course that Wnt3a was regulated in the DH, the mice were sacrificed at 15 min, 30 min, 1 h, 2 h, 3 h, or 4 h after CFC training. The real-time PCR results showed that Wnt3a mRNA levels peaked at 2 h (P < 0.001, one-way ANOVA) and returned to baseline at 4 h (Fig. 1B). Interestingly, Wnt3a mRNA did not change significantly in the amygdala, which is a brain region also involved in CFC memory formation (Fig. 1B; ANOVA: F6,38 = 0.463, P = 0.831). We then investigated whether the CFC training-induced Wnt3a mRNA change was specific to associative fear learning rather than only exposure to either context or shock alone. The mice were separated into the following 4 groups: naive mice as the control group, the context exposed group, the immediate shock group, and the CFC group (Fig. 1C). The mice could not learn to associate the context with the shock when the mice were shocked immediately after being placed in the conditioning context and removed quickly afterward (“immediate shock”) (Fanselow 2000; Rao-Ruiz et al. 2011). The results revealed that the levels of Wnt3a mRNA were only elevated in the associative fear learning in the 4 groups (Fig. 1C; ANOVA: F3,30 = 8.245, P < 0.001; post hoc: P < 0.001), which suggests that the increased Wnt3a gene expression during the contextual fear memory formation is specific to the association of the CS plus the US, rather than the CS or US modality.
We next investigated whether the increased Wnt3a mRNA levels would lead to Wnt3a protein level elevation after CFC training. As shown in Figure 1D, compared with the time course of Wnt3a mRNA level changes, Wnt3a protein in the DH showed a delayed increase, which peaked at 3 h (P = 0.002, one-way ANOVA) and returned to baseline at 4 h after CFC training. The specific changes in Wnt3a expression in the DH following CFC training suggested that Wnt3a in the DH might play a role in the contextual fear memory formation.
Wnt3a-Dependent Wnt/β-catenin and Wnt/Ca2+ Signaling Pathways Activation After CFC Training
Wnt3a is reported to activate both the canonical Wnt/β-catenin and noncanonical Wnt/Ca2+ signaling pathways. Whether the Wnt/β-catenin and/or noncanonical Wnt/Ca2+ pathways downstream of Wnt3a are activated during CFC memory formation remains unknown. Previous studies showed that Wnt3a-induced phosphorylation of GSK3β at an amino acid residue Ser9 (Cook et al. 1996; Yokoyama et al. 2007; Gelebart et al. 2008). Thus, we first evaluated the levels of p-GSK3β (Ser9) and GSK3β at various time points after CFC training by immunoblotting. Figure 2A shows that the total GSK3β levels in the DH after CFC training had no significant change (ANOVA: F6,22 = 0.623, P = 0.710). However, the p-GSK3β (Ser9) levels in the DH were increased starting at 1 h (P = 0.011) and peaked at 3 h (P = 0.001, one-way ANOVA) after CFC training (Fig. 2A). The increased p-GSK3β (Ser9) levels were specific to associative fear learning because the context exposed group or the immediate shock group did not differ from the naive control group (see Supplementary Fig. 1A). The phosphorylation of GSK3β at Ser9 inhibits its kinase activity, which leads to increased active β-catenin (unphosphorylated β-catenin protein at Ser 33, 37 or Thr 41) accumulation and translocation to the nucleus, where it activates the transcription of TCF/LEF target genes (Chen and Bodles 2007; Salcedo-Tello et al. 2011). Therefore, we then measured the levels of total β-catenin, active β-catenin and nuclear β-catenin at various time points after CFC training. We observed that although the total β-catenin levels in the DH after CFC training had no significant change (Fig. 2B; ANOVA: F6,24 = 0.663, P = 0.680), the levels of active β-catenin and nuclear β-catenin were elevated at 2 and 3 h after training (Fig. 2B, active β-catenin: 2 h, P = 0.002; 3 h, P = 0.011; Fig. 2C, nuclear β-catenin: 2 h, P = 0.004; 3 h, P = 0.049; one-way ANOVA), peaked at 2 h and returned to baseline at 4 h, which suggests that CFC training induced the activation of the Wnt/β-catenin signaling pathway. The expression of active β-catenin and nuclear β-catenin in the context and immediate shock groups did not differ from the naive control group (see Supplementary Fig. 1B,C). Furthermore, we determined whether the expression of the downstream targets of β-catenin was affected after CFC training. The real-time PCR results showed that the mRNA levels of axin2 and tcfe2a were significantly increased at 2 h after CFC training (Fig. 2D; axin2, P = 0.005; tcfe2a, P = 0.031, 2-tailed t-test), which suggests that CFC training leads to the activation of the downstream targets of the Wnt/β-catenin pathway.
In addition to the canonical Wnt/β-catenin pathway, Wnt3a has also been reported to activate the noncanonical Wnt/Ca2+–CaMKII pathway. We next investigated whether the p-CaMKII levels were altered in the DH after CFC training. In a crude homogenized DH fraction, the total CaMKIIα (ANOVA: F4,19 = 0.587, P = 0.676) and p-CaMKIIα levels (ANOVA: F4,18 = 0.398, P = 0.808) did not change after CFC training (Fig. 2E). Interestingly, the synaptic p-CaMKIIα levels were significantly increased at 15 min (P < 0.001, one-way ANOVA) and returned to baseline at 30 min after CFC training, whereas the levels of CaMKIIα in the SNS were not significantly changed (ANOVA: F4,15 = 0.666, P = 0.625) (Fig. 2F). Moreover, the synaptic p-CaMKIIα levels in the context and immediate shock groups were not significantly different compared with the control group (see Supplementary Fig. 1E). It should be noted that the synaptic p-CaMKIIα levels were elevated at 15 min after CFC training when the Wnt3a protein levels did not change. It is possible that CFC training may induce rapid activity-dependent Wnt3a release, which contributes to the increased synaptic p-CaMKIIα levels. We then examined whether CFC training could induce rapid activity-dependent Wnt3a release in advance of the increased Wnt3a synthesis. According to a previous report (Chen et al. 2006), we performed fluorescence immunostaining of Wnt3a in the DH immediately or 15 min after CFC training and measured the fluorescence signaling intensity for Wnt3a-positive cells in the DH to detect the release of Wnt3a. Compared with the naive group, the values of the signaling intensity for Wnt3a-positive cells in the DH were significantly reduced at 0 min (P = 0.001) and 15 min (P = 0.004, one-way ANOVA) after CFC training (Fig. 2G). Because the decreased Wnt3a mRNA expression was not observed, the decrease in Wnt3a immunostaining intensity was most likely due to a rapid Wnt3a release rather than an inhibition of synthesis following the CFC training. A previous study showed that Wnt3a was localized at the postsynaptic region in the hippocampus (Chen et al. 2006). Thus, rapid activity-dependent Wnt3a release could induce quick activation of synaptic CaMKIIα.
The previous results showed that the Wnt/β-catenin and Wnt/Ca2+–CaMKII signaling pathways were activated after CFC training; however, whether these changes were dependent on Wnt3a remained unclear. To address this question, the mice were injected with function-blocking anti-Wnt3a antibodies (Wnt3a ab-T) (Chen et al. 2006) or vehicles (veh-T) 15 min before CFC training (Fig. 3A). When compared with the veh-T group, Wnt3a antibody microinfusion significantly decreased the synaptic p-CaMKIIα levels at 15 min (Fig. 3B; P = 0.015, one-way ANOVA), which suggests that Wnt3a contributes to the activation of the Wnt/Ca2+ signaling pathway in the DH upon CFC training. In addition, the increased p-GSK3β (Ser9) levels in the veh-T group at 1 h were completely blocked by Wnt3a antibody microinfusion (Fig. 3C; P = 0.016, one-way ANOVA). Wnt3a antibody microinjection also blocked the increased nuclear β-catenin levels and the increased mRNA levels of axin2 and tcfe2a at 2 h after CFC training compared with the veh-T group (Fig. 3D, nuclear β-catenin, P = 0.003; Fig. 3E, axin2, P < 0.001; tcfe2a, P = 0.001; one-way ANOVA). These results suggested that Wnt3a was responsible for the activation of the Wnt/β-catenin signaling pathway induced by CFC training.
Wnt3a is Required for the Acquisition and Consolidation of Contextual Fear Memory
Because the above results showed that CFC training could induce the secretion and expression of Wnt3a and activate the Wnt3a mediated downstream signaling pathways, we next investigated whether endogenous Wnt3a is functionally necessary for CFC memory formation. Given that CFC training could induce rapid activity-dependent Wnt3a release in advance of the increased Wnt3a synthesis, Wnt3a antibodies, Wnt5a antibodies (Bilkovski et al. 2010) or vehicles were bilaterally injected into the DH 15 min before CFC training. The cannula placements in different brain regions are shown in Supplementary Figure 2. The mice injected with Wnt3a antibodies, but not Wnt5a antibodies, in the DH showed a significant decrease in the freezing time during fear training (see Supplementary Fig. 3), which suggests that Wnt3a in the DH was specifically required for the acquisition of CFC memory formation. Furthermore, we found that STM and LTM were also impaired by Wnt3a antibody microinjection (Fig. 3F; STM, P = 0.001; LTM, P = 0.001; one-way ANOVA), which confirms that Wnt3a was involved in the CFC acquisition. Intrahippocampal infusion of a control IgG had no effect on CFC STM or LTM (see Supplementary Fig. 4), which indicates that the impairment of CFC memory was not due to a nonspeciﬁc protein-loading effect. In addition, no effect on the STM or LTM was found when Wnt3a antibodies were injected into the amygdala (Fig. 3G), which suggests that Wnt3a in the amygdala was not involved in the CFC memory formation. To exclude the possibility that Wnt3a in the DH affects the locomotion and anxiety-like behaviors, Wnt3a antibodies were administered 15 min before the open field and elevated plus-maze test. No difference was found between the Wnt3a antibody and vehicle groups on locomotion or anxiety-like behaviors (see Supplementary Fig. 5), which suggested that the effect of Wnt3a on CFC memory in the DH is not due to alterations in locomotion or anxiety-like behaviors.
The above results showed that Wnt3a in the DH was involved in the CFC acquisition, we next assessed whether Wnt3a was involved in other memory processes such as consolidation or expression by varying the time point of Wnt3a antibody microinfusion into the DH (Fig. 4A). To examine the effect of Wnt3a on CFC consolidation, Wnt3a antibodies or vehicles were infused into the DH immediately after CFC training, and the STM and LTM were tested. Our results showed an intact STM and an impaired LTM (P < 0.001, 2-tailed t-test) after the microinfusion of Wnt3a antibody immediately after CFC training compared with the vehicle group (Fig. 4B), which suggests that Wnt3a was also involved in the CFC consolidation. No effect on the STM or LTM was observed when a control IgG was injected immediately after CFC training (see Supplementary Fig. 6). To examine the effect of Wnt3a on the expression of conditioned fear, Wnt3a antibodies or vehicles were injected 15 min before the LTM test. Compared with the vehicle group, the mice injected with Wnt3a antibodies showed similar freezing time, which suggests that Wnt3a was not necessary for the expression of CFC (Fig. 4B).
Wnt/Ca2+–CaMKII and Wnt/β-catenin Signaling Pathways Downstream of Wnt3a are Respectively Involved in the CFC Acquisition and Consolidation
To determine the signaling pathways coupled to Wnt3a mediated CFC consolidation, Wnt3a antibodies were microinjected immediately after CFC training, and the activation of the Wnt/Ca2+ and Wnt/β-catenin pathways was examined. Compared with the veh-T group, Wnt3a antibodies microinfusion immediately after CFC training had no effect on the rapid increased levels of synaptic p-CaMKIIα (Fig. 4C), which suggests that the Wnt/Ca2+–CaMKII signaling pathway may not be involved in the Wnt3a mediated CFC consolidation. However, the increased p-GSK3β (Ser9) levels in the veh-T group were completely blocked by the microinfusion of Wnt3a antibodies (Fig. 4D; P = 0.004, one-way ANOVA). Post training microinfusion of Wnt3a antibodies also blocked the increased nuclear β-catenin levels and the increased mRNA levels of axin2 and tcfe2a at 2 h after CFC training compared with the veh-T group (Fig. 4E, nuclear β-catenin, P = 0.002; Fig. 4F, axin2, P < 0.001; tcfe2a, P = 0.002; one-way ANOVA). These results suggested that the Wnt/β-catenin signaling pathway might be involved in Wnt3a mediated consolidation of CFC.
To further distinguish the roles of the Wnt/β-catenin and Wnt/Ca2+–CaMKII signaling pathways in the different CFC memory processes, DKK1 and sFRP1 were applied. DKK1 has been reported to specifically inhibit the canonical Wnt signaling pathway by binding to the LRP coreceptor, whereas sFRP1 is a soluble antagonist of Wnt signaling by binding directly to Wnts and preventing their interactions with Frizzled receptors (Diep et al. 2004; Moon et al. 2004). DKK1 or sFRP1 was microinfused into the DH 15 min before CFC training (Fig. 5A). Compared with the veh-T group, an immunoblot analysis showed that the DKK1 injection had no effect on the increased synaptic p-CaMKIIα levels in the DH at 15 min post conditioning (Fig. 5B). However, DKK1 microinfusion blocked the increased p-GSK3β (Ser9) levels at 1 h post conditioning compared with the veh-T group (Fig. 5C; P = 0.018; one-way ANOVA). These data suggested that DKK1 treatment selectively blocked the activation of the canonical Wnt/β-catenin signaling pathway, but had no effect on the noncanonical Wnt/Ca2+–CaMKII signaling pathway induced by CFC training. In contrast, the administration of sFRP1 into the DH blocked both the increased synaptic p-CaMKIIα levels (Fig. 5B; P = 0.012) and the increased p-GSK3β (Ser9) levels compared with the veh-T group (Fig. 5C; P = 0.023, one-way ANOVA). These data indicated that sFRP1 could block both the canonical Wnt/β-catenin and noncanonical Wnt/Ca2+–CaMKII signaling pathways activated by CFC training.
We further examined the effects of sFRP1 or DKK1 administration on fear memory formation. The acquisition curve showed that the mice injected with sFRP1, but not DKK1, spent less time freezing during fear training when compared with the control group (see Supplementary Fig. 7). Furthermore, we found that DKKI administration selectively impaired the LTM (P < 0.001, one-way ANOVA), but not the STM (Fig. 5D). However, the infusion of sFRP1 in the DH disrupted both the STM (P = 0.001) and the LTM (P < 0.001, one-way ANOVA) of CFC (Fig. 5D). These data suggested that the Wnt/Ca2+–CaMKII signaling pathway was involved in the acquisition of CFC, while the Wnt/β-catenin signaling was involved in the consolidation of CFC. Overall, our findings suggested that Wnt3a may regulate CFC acquisition and consolidation through a differential signaling pathway.
Wnt3a is Sufficient for CFC Memory Formation
The previous results showed that Wnt3a loss-of-function in the DH impaired the CFC memory, which suggests that Wnt3a was necessary for CFC learning. Next, we determined whether exogenous Wnt3a administration was sufficient for CFC memory formation. To prevent a ceiling effect, we used a weak CFC training protocol in which the intensity of the US was lowered to 0.4 mA (Fukushima et al. 2008; Suzuki et al. 2011). Exogenous Wnt3a or vehicle was microinjected into the DH immediately after weak CFC training (Fig. 6A). Immunoblot analysis showed that the microinfusion of Wnt3a significantly increased the levels of synaptic p-CaMKIIα, p-GSK3β (Ser9), and nuclear β-catenin compared with the vehicle injected mice after weak CFC training (synaptic p-CaMKIIα, P = 0.044; p-GSK3β (Ser9), P = 0.038; nuclear β-catenin, P = 0.025; one-way ANOVA) (Fig. 6B–D). These data suggested that exogenous Wnt3a administration could enhance the activation of the Wnt/Ca2+–CaMKII and Wnt/β-catenin signaling pathways.
We next examined the effect of exogenous Wnt3a on CFC memory formation. The results showed that exogenous Wnt3a infusion could enhance both the STM (P = 0.023) and the LTM (P = 0.020, one-way ANOVA) induced by weak training to normal levels compared with the vehicle control (Fig. 6E). These data indicated that Wnt3a in the DH is not only necessary but also sufficient for CFC memory formation.
To confirm that Wnt3a mediates CFC memory acquisition through the Wnt/Ca2+–CaMKII signaling pathway, the mice were injected with the CaMKII inhibitor KN-62 or vehicle 15 min before training followed by exogenous Wnt3a infusion. Our results showed that KN-62 blocked the Wnt3a-induced enhancement of CFC STM and LTM (see Supplementary Fig. 8), which suggests that Ca2+/CaMKII acts downstream of Wnt3a to mediate CFC memory acquisition.
β-catenin Acts Downstream of Wnt3a to Enhance CFC Memory Consolidation
Previous studies and our experiments demonstrated that the specific deletion or the inhibition of β-catenin disrupted the memory consolidation, which suggests β-catenin was necessary for memory formation (Maguschak and Ressler 2008, 2011; Fortress, Schram et al. 2013). However, it remains unclear whether β-catenin was sufficient for the memory consolidation in adults. In the present study, we constructed a lentiviral vector that expressed constitutively active β-catenin in which the first 90 amino acids (ΔN90) that contained the GSK3β phosphorylation site were deleted to mimic active Wnt signaling. The mice were infected with lenti-active-β-catenin (lenti-abc) or lenti-GFP and killed 4 weeks later (Fig. 7A). A large number of GFP-positive cells in the dentate gyrus (DG) and CA3 regions of the hippocampus were detected in the lentiviral infected mice (Fig. 7B), which suggests that the lentivirus successfully infected the cells in vivo. Immunoblotting showed increased active β-catenin ΔN90 expression in the nuclear protein extraction in the lenti-abc injected mice (Fig. 7C). When compared with the lenti-GFP group, the lenti-abc virus injection could increase the mRNA levels of the axin2 (P = 0.003) and tcfe2a (P = 0.015, 2-tailed t-test) (Fig. 7D), which indicates that constitutively active β-catenin can induce the expression of the downstream targets of β-catenin.
We next investigated the effect of β-catenin gain-of-function on the acquisition and consolidation after weak CFC training. As shown in Figure 7E, the overexpression of constitutively active β-catenin in the DH had no effect on the CFC STM, which confirms that β-catenin was not involved in the acquisition of CFC. However, the overexpression of constitutively active β-catenin in the DH enhanced the CFC LTM (Fig. 7E; P = 0.026, 2-tailed t-test). These data indicated that β-catenin in the DH was not only necessary but also sufficient for the consolidation of CFC.
To confirm that β-catenin downstream of Wnt3a selectively mediates CFC memory consolidation but not acquisition, an additional experiment was performed. Four weeks after the infusion of the lenti-abc or lenti-GFP virus, Wnt3a antibodies or vehicles were bilaterally injected into the DH 15 min before CFC training (Fig. 7F). In the Wnt3a antibody or vehicle injection groups, the overexpression of constitutively active β-catenin in the DH had no effect on the CFC STM compared with the lenti-GFP virus injected mice, which further confirmed that β-catenin had no effect on the acquisition of CFC memory. However, in both the Wnt3a antibody or vehicle injection groups, the lenti-abc virus injection could enhance LTM compared with the lenti-GFP group (P < 0.001, one-way ANOVA), which further confirmed that β-catenin acted downstream of Wnt3a to mediate CFC memory consolidation.
The purpose of this study was to investigate the involvement of hippocampal Wnts in CFC memory formation. In the present study, we provided the first evidence for the selective expression changes of Wnt3a in the CFC memory process. CFC training induced the activation of the Wnt/Ca2+ and Wnt/β-catenin signaling pathways, which depended on Wnt3a. Moreover, Wnt3a regulated CFC memory acquisition and consolidation through the Wnt/Ca2+ and Wnt/β-catenin signaling pathways, respectively. Finally, we demonstrated that the infusion of exogenous Wnt3a could enhance CFC memory formation. The overexpression of constitutively active β-catenin selectively rescued the deficit in CFC consolidation, but not acquisition, induced by Wnt3a antibody injection, which supports the notion that Wnt3a regulated the CFC memory consolidation by activating the Wnt/β-catenin pathway.
Our data provided several new insights into the role of Wnt3a in CFC memory formation. First, we found that CFC training could specifically induce the increased Wnt3a expression and activate its downstream Wnt/β-catenin and Wnt/Ca2+ signaling pathways. In our study, a significantly increased expression of Wnt3a was observed in the DH after CFC training, whereas the levels of other Wnts (Wnt1, Wnt3, Wnt4, Wnt5a, and Wnt7a) did not change. However, Tabatadze et al. (2012) reported that the levels of Wnt7 and Wnt5a, but not Wnt3, were elevated in the hippocampus after spatial learning in the water maze (WZ). The differential expression changes of Wnts in the hippocampus after learning may be attributed to the different learning models used. Although CFC and WZ are both hippocampal-dependent tasks, there are many differences between the tasks. CFC involves the association of footshock within a context and is quickly acquired. WZ is used to investigate spatial learning and memory, and several days of training are needed to acquire the memory. Indeed, the increased expression of Wnt7 and Wnt5a in the WZ was observed 7 days later after 5 days of training. This finding may suggest that the training-induced increased expression of Wnt7 and Wnt5a occurs at a later time point, while the increased Wnt3a expression occurs at an early time point in the hippocampus. Increased Wnt3a expression levels in the DH following contextual fear training have been selectively found after associative CS–US pairings, which suggests that an interaction of the representations of the CS and US is obligatory for Wnt3a involvement in CFC memory formation. The hippocampus has been reported to be critical for forming contextual representations (Rudy et al. 2002). Previous studies showed that the context alone could induce molecular changes as CFC did in the hippocampus (Huff et al. 2006; Lopez-Fernandez et al. 2007). However, in our study, the context alone did not induce increased Wnt3a expression in the DH, which suggested that Wnt3a did not contribute to the hippocampal representation of the context. Consistent with our report, a previous report showed that the context alone could not induce the expression of NGFI-B and SGK3 in the hippocampus, whereas CFC could induce the expression (von Hertzen and Giese 2005). To date, there is a debate concerning whether the hippocampus is necessarily important for associating a context with a shock. Our results supported the notion that the hippocampus was involved in forming a context–shock memory. It was reported that the amygdala is also involved in the CFC memory formation (Huff and Rudy 2004; McGaugh 2004). In our study, the increased Wnt3a expression was only observed in the DH, but not in the amygdala after CFC training, which implies that Wnt3a in the DH is specifically involved in CFC memory formation. Whether Wnt3a in the amygdala was involved in other learning tasks needs to be further investigated. Moreover, we found that CFC induced the activation of the Wnt/β-catenin and the Wnt/Ca2+ signaling pathways, which were dependent on Wnt3a. Neutralizing Wnt3a blocked the CFC induced activation of the Wnt/β-catenin and Wnt/Ca2+ signaling pathways. In addition, we found CFC training could induce rapid activity-dependent Wnt3a release in advance of the increased Wnt3a synthesis. A previous brain slice study showed that N-methyl-d-aspartate receptor-dependent synaptic Wnt3a release was induced by tetanic stimulations (Chen et al. 2006). To our knowledge, our study is the first to show in vivo that associative learning could rapidly trigger Wnt3a release, which may regulate the synaptic plasticity underlying learning and memory in adults.
Second, we found that Wnt3a in the DH was necessary for the CFC memory formation, and the Wnt3a activated Wnt/Ca2+–CaMKII and Wnt/β-catenin signaling pathways were involved in CFC memory acquisition and consolidation, respectively. The infusion of Wnt3a antibody before CFC training impaired the acquisition of CFC and blocked the activation of both the Wnt/Ca2+–CaMKII and Wnt/β-catenin signaling pathways, while the infusion of Wnt3a antibody immediately after CFC training impaired the consolidation of CFC and blocked the activation of the Wnt/β-catenin signaling pathway, but not the Wnt/Ca2+–CaMKII signaling pathway. These findings demonstrated the functional importance of Wnt3a in CFC memory formation and suggested that differential signaling pathways downstream of Wnt3a may be involved in CFC acquisition and consolidation. Moreover, we found that the specific canonical Wnt/β-catenin signaling inhibitor DKK1 impaired the LTM, but not the STM, which suggests that Wnt/β-catenin signaling was involved in the consolidation, but not acquisition, of CFC memory. Consistent with our findings, a recent study showed that canonical Wnt signaling was involved in object recognition memory consolidation (Fortress, Schram et al. 2013). However, whether canonical Wnt signaling in the DH was involved in other phases of memory was not determined. We found that the Wnt/β-catenin signaling pathway was not involved in the acquisition of hippocampal memory. Moreover, intrahippocampal injection of another Wnt inhibitor sFRP1 impaired CFC acquisition. Since our study showed that sFRP1 could block the activation of both the Wnt/β-catenin and Wnt/Ca2+–CaMKII signaling pathways, it suggested that noncanonical Wnt/Ca2+ signaling may be involved in the CFC acquisition. Indeed, the effects of p-CaMKIIα in the CFC acquisition have been well demonstrated. αCaMKII-T286A mutant mice showed impairment in the acquisition of CFC (Irvine et al. 2005, 2011). The local infusion of the CaMKII inhibitor KN-62 into the amygdala also impaired the CFC acquisition (Rodrigues et al. 2004). Our study provides the first evidence that the noncanonical Wnt/Ca2+ signaling pathway is necessary for hippocampal memory acquisition. Together with the biochemical and behavioral changes induced by varying the delivering time point of Wnt3a antibodies, it is suggested that Wnt3a may differentially regulate CFC acquisition and consolidation through the Wnt-Ca2+/CaMKII and Wnt/β-catenin signaling pathways, respectively.
Finally, we found that Wnt3a was sufficient for CFC memory formation. Exogenous Wnt3a administration in the DH could enhance CFC memory, which indicates that Wnt3a in the DH is not only necessary but also sufficient for CFC learning. Previous studies demonstrated that β-catenin loss-of-function impaired memory consolidation (Maguschak and Ressler 2008, 2011; Fortress, Schram et al. 2013), which suggests that β-catenin functions as a positive regulator for memory formation. More recently, Vargas et al. (2014) reported that chronic infusion of WASP-1, a potentiator of the canonical signaling, enhanced the episodic memory in mice. However, WASP-1 is not a specific pharmacological agent targeting β-catenin. It may increase the levels of β-catenin through the inhibition of GSK3β (Vargas et al. 2014). To specifically identify the effects of β-catenin gain-of-function on the memory formation, we overexpressed the constitutively active β-catenin via lentivirus infection. We found that the overexpression of constitutively active β-catenin in the DH enhanced CFC consolidation, but not acquisition. Furthermore, we found that β-catenin gain-of-function rescued the Wnt3a antibody microinjection induced impairment in CFC consolidation, but not acquisition, which suggests that β-catenin acts downstream of Wnt3a to mediate CFC memory consolidation.
In conclusion, using the CFC memory paradigm, we determined the specific roles of Wnt3a in CFC memory formation. We found that contextual fear training induced a specific elevation in Wnt3a expression, which activated both the Wnt/β-catenin and Wnt/Ca2+–CaMKII signaling pathways. One novel and important finding in our study was that the Wnt/Ca2+–CaMKII and Wnt/β-catenin signaling pathways were involved in Wnt3a mediated CFC acquisition and consolidation, respectively. Moreover, we found that Wnt3a signaling was sufficient for CFC memory. The overexpression of constitutively active β-catenin in the DH could selectively rescue the Wnt3a loss-of-function induced deficit in CFC memory consolidation. Overall, our study facilitates understanding the role of Wnt3a in differential CFC memory formation processes on the basis of its divergent downstream signaling pathways.
This work was supported by the National 973 Basic Research Program of China (no. 2012CB911000), National Natural Science Foundation of China (no. 31371138, 31130026), the State Program of National Natural Science Foundation of China for Innovative Research Group (no. 81321061), and the Independent Innovation Foundation of Shandong University (IIFSDU).
Conﬂict of Interest: None declared.