If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
NMDAR-dependent synaptic potentiation via APPL1 signaling is required for the accessibility of a prefrontal neuronal assembly in retrieving fear extinction
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
Liangzhu Laboratory, MOE Frontier Science Center for Brain Science and Brain-machine Integration, State Key Laboratory of Brain-machine Intelligence, Zhejiang University, Hangzhou, 311121, ChinaNHC and CAMS Key Laboratory of Medical Neurobiology,ZhejiangUniversity ,Hangzhou ,310058 ,China
To whom correspondence should be addressed: Dr. Yi-Hui Cui, Hangzhou, Zhejiang 310058, China, Phone: +86-18868431588, , or Dr. Shuang Qiu, Hangzhou, 310058, China, Phone: +86-13588413699, .
Affiliations
Department of Neurobiology and Department of Neurology of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, China
To whom correspondence should be addressed: Dr. Yi-Hui Cui, Hangzhou, Zhejiang 310058, China, Phone: +86-18868431588, , or Dr. Shuang Qiu, Hangzhou, 310058, China, Phone: +86-13588413699, .
Affiliations
Department of Neurobiology and Department of Anesthesiology of Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310058, ChinaLiangzhu Laboratory, MOE Frontier Science Center for Brain Science and Brain-machine Integration, State Key Laboratory of Brain-machine Intelligence, Zhejiang University, Hangzhou, 311121, ChinaNHC and CAMS Key Laboratory of Medical Neurobiology,ZhejiangUniversity ,Hangzhou ,310058 ,China
The ventromedial prefrontal cortex (vmPFC) has been viewed as a locus to store and recall extinction memory. However, the synaptic and cellular mechanisms underlying this process remain elusive.
Methods
We combined transgenic mice, electrophysiological recording, activity-dependent cell labeling, and chemogenetic manipulation to analyze the role of adaptor protein APPL1 in the vmPFC for fear extinction retrieval.
Results
We found that both constitutive and conditional APPL1 knockout decreases NMDA receptor (NMDAR) function in the vmPFC and impairs fear extinction retrieval. Moreover, APPL1 undergoes nuclear translocation during extinction retrieval. Blocking APPL1 nucleocytoplasmic translocation reduces NMDAR currents and disrupts extinction retrieval. We further identified a prefrontal neuronal ensemble that is both necessary and sufficient for the storage of extinction memory. Inducible APPL1 knockout in this ensemble abolishes NMDAR-dependent synaptic potentiation and disrupts extinction retrieval, while simultaneously chemogenetic activation of this ensemble rescues the impaired behaviors.
Conclusions
Therefore, our results indicate that a prefrontal neuronal ensemble stores extinction memory, and APPL1 signaling supports these neurons to retrieve extinction memory via controlling NMDAR-dependent potentiation.
To adapt to a continuously changing world, an organism must learn to mobilize the defensive response when confronted with cues that predict harm and learn to withdraw a defensive response when the cues no longer predict harm. The latter process depends on a new learning process called fear extinction (
An abundance of research suggests that the ventromedial prefrontal cortex (vmPFC) is a brain region critical for the storage and retrieval of fear extinction (
Long-term expression of human contextual fear and extinction memories involves amygdala, hippocampus and ventromedial prefrontal cortex: a reinstatement study in two independent samples.
Social cognitive and affective neuroscience.2014; 9: 1973-1983
Dissociable roles of prelimbic and infralimbic cortices, ventral hippocampus, and basolateral amygdala in the expression and extinction of conditioned fear.
Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology.2011; 36: 529-538
). Studies at the circuitry level reveal that vmPFC top-down controls fear responses through descending projections to the inhibitory circuits within the amygdala, and therefore, facilitates fear extinction (
). The causal links between these neuronal ensembles with their relevant events were then assessed, with the help of “engram” mouse lines or activity-dependent viral tools, by targeting those cells and manipulating their activity under specific behavioral circumstances (
). These findings provide a platform for investigating the neural mechanisms, at both cellular and synaptic levels, that enable long-term storage and recall of extinction memories within the prefrontal cortex.
Endosomal adaptor protein containing pleckstrin homology (PH) domain, phosphotyrosine-binding (PTB) domain, and leucine zipper motif 1 (APPL1), is acting not only in vesicle trafficking (
). Our previous work has identified APPL1 as a linker coupling neuronal activity with gene transcription via undergoing synapse-to-nucleus translocation in neurons (
). In this study, we identified a distinct prefrontal neuronal ensemble that stores extinction memories and undergoes NMDAR-dependent potentiation during extinction retrieval. Furthermore, APPL1 signaling in these cells supports extinction retrieval by preserving their capacity for NMDAR-dependent potentiation. These findings advance our understanding of the cellular and molecular basis of fear extinction and provide implications for the treatment of PTSD and other anxiety disorders.
Methods and Materials
Detailed methods are described in Supplement 1. Fear conditioning and extinction tests were used to assess fear memory and extinction memory. For robust activity marking (RAM) or Cre-dependent RAM (CRAM) labeling, mice were injected with adeno-associated virus (AAV) construct and were kept on Dox (60 mg/L) drinking after AAVs injection until behavior experiments.
Results
Constitutive APPL1 knockout mice exhibit impaired fear extinction
Here, we started to analyze the functional role of APPL1 in vivo by using germline Appl1-null (constitutive Appl1-/-) mice, which have been reported previously to grow normally with no obvious phenotypic deficiency under normal feeding conditions (
APPL1 counteracts obesity-induced vascular insulin resistance and endothelial dysfunction by modulating the endothelial production of nitric oxide and endothelin-1 in mice.
). We observed ubiquitous loss of APPL1 protein expression in different brain regions of Appl1-/- mice (Fig. S1A). APPL1 loss did not affect brain weight (Fig. S1B) or brain morphology (Fig. S1C). Moreover, the complexity and spine density of basal dendrites of prefrontal layer V pyramidal cells in Appl1-/- mice were unchanged compared to wild-type (WT) mice (Fig. S1D-F).
Next, we subjected constitutive Appl1-/- mice to a series of behavioral tests and observed that Appl1-/- mice and WT mice showed comparable performance in the rotarod test (Fig. S2A), temporal order memory test (Fig. S2B), object localization recognition test (Fig. S2C), novel object recognition test (Fig. S2D), and three-chamber social test (Fig. S2E). In contrast, Appl1-/- mice showed decreased number of times across the central area in the open-field test (Fig. S2F), less time in the brightly lit chamber in the light-dark box test (Fig. S2G), and less time in the open arms and reduced number of times to enter the open arms in elevated plus maze test (Fig. S2H) than WT mice, indicating that Appl1-/- mice displayed increased anxiety- and fear-like behaviors.
Fear conditioning and extinction have become influential animal models for the study of anxiety and fear (
). We then performed an auditory-cued fear conditioning and extinction protocol to test whether APPL1 is involved in fear learning and extinction. During fear conditioning (Cond.) on Day 1, mice were placed individually in a conditioning chamber (context A) and then exposed to 5 pure tones (CS) with foot shock (US). On Day 2, the conditioned mice were placed in a completely different test chamber (context B) and exposed to 5 CSs without the US. Appl1-/- mice and WT mice showed comparable CS-elicited freezing during fear learning (Day 1) and fear memory retrieval (Day 2) (Fig. S2I).
For fear extinction training, the conditioned mice were exposed to 20 CSs without the US in context B on Day 2 (Ext.). On Day 3 (the first extinction retrieval test, Ext. Retr.1) or Day 10 (the second extinction retrieval test, Ext. Retr.2), mice were placed in context B and exposed to 5 CSs to retrieve extinction memory (Fig. 1A). As shown in Fig. 1B, Appl1-/- mice exhibited normal freezing responses to CSs during fear extinction training, but higher freezing responses during extinction memory retrieval than WT mice, suggesting that APPL1 is specifically involved in fear extinction.
Fig. 1APPL1 in the vmPFC is necessary for learned extinction. (A) Experimental scheme of fear conditioning and extinction. CS, conditioned stimulus; US, unconditioned stimulus; Cond., fear conditioning; Ext., extinction training; Ext.Retr.1, the first extinction retrieval; Ext.Retr.2, the second extinction retrieval. (B) Constitutive Appl1-/- mice (Appl1-/-) showed impaired retrieval of fear extinction memory during Ext.Retr.1 and Ext.Retr.2. Left: time course of freezing in response to CS during different phases. BL: baseline of the freezing level. WT: wild-type mice. Each data point during extinction training on Day 2 represented the average of 2 CSs. Statistical analysis was performed by using two-way repeated-measures (RM) ANOVA followed by Sidak’s multiple comparisons tests; Cond., F1, 16 = 0.1120, P = 0.7423; Ext., F1, 16 = 0.08613, P = 0.7729; Ext.Retr.1, F1, 16 = 7.589; Ext.Retr.2, F1, 16 = 4.572. Right: average freezing level for all trials during extinction retrieval (WT, n = 10; Appl1-/-, n = 8 mice; unpaired t-test). (C to F) Fiber photometry in the vmPFC of Appl1-/- or WT mice during Conditioning (C), Extinction training (D), Ext.Retr.1 (E), and Ext.Retr.2 (F), respectively. Upper: heat maps show the dynamics of individual CS trials from 5 mice. CSs start at 0 s, and last 30 s. Lower: z-Score represents average time courses of all CS trials from 5 mice with shadows indicating SEM. The insets represent the average of peak values in response to CS (z-Score, n = 5 mice per group, unpaired t-test). (G) Diagram of bilateral injection of AAV9-hSyn-APPL1-GFP (APPL1) or AAV9-hSyn-GFP (GFP) virus into the vmPFC of Appl1-/- or WT mice; scale bar, 1 mm. (H) The abundance of APPL1 in the vmPFC (One-way ANOVA with Bonferroni’s Multiple Comparison Test, F2, 9 = 28.64, n = 4 mice per group). (I) Re-expression of APPL1 in the vmPFC rescued the impaired extinction memory recall. Left: Two-way RM ANOVA followed by Sidak’s multiple comparisons tests; Cond., F2, 21 = 0.4466, P = 0.6457; Ext., F2, 21 = 2.708, P = 0.0899; Ext.Retr., F2, 21 = 6.128. Right: One-way ANOVA with Bonferroni’s Multiple Comparison Test (WT + GFP, n = 9; Appl1-/- + GFP, n = 7; Appl1-/- + APPL1, n = 8 mice per group). Data are presented as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001.
). We started to investigate whether APPL1 in the vmPFC participates in fear extinction. We injected adeno-associated virus (AAV9)-hSyn-GCaMP7b into the vmPFC of constitutive Appl1-/- mice or WT mice (Fig. S3) and, 3 weeks later, monitored the neuronal activity in the vmPFC during different phases of fear conditioning and fear extinction by fiber photometry. On Day 1, both Appl1-/- mice and WT mice showed no obvious response to CS, but robust responses to US (Fig. 1C). On Day 2, Appl1-/- mice and WT mice showed no obvious increase in CS-related calcium signals (Fig. 1D). During Ext. Retr.1 (Fig. 1E) and Ext. Retr.2 (Fig. 1F), WT mice showed a significant increase in CS-elicited calcium signals, whereas Appl1-/- mice exhibited a lower signal, indicating that deficiency of APPL1 leads to reduced neuronal activity in the vmPFC during extinction recall.
Next, we injected AAV9-hSyn-APPL1-GFP into the vmPFC of constitutive Appl1-/- mice to re-express APPL1 (APPL1 group) or AAV9-hSyn-GFP into the vmPFC of Appl1-/- mice (GFP group) or WT mice (WT group) as control (Fig. 1G and 1H). 3 weeks later, we performed similar fear conditioning and extinction protocol and observed that mice in APPL1 group showed comparable freezing level with WT group during Ext. Retr. (Day 3), both of which were significantly lower than GFP group, indicating that expression of APPL1 in the vmPFC restores fear extinction retrieval in Appl1-/- mice (Fig. 1I). In contrast, Appl1-/- mice injected with AAV9-hSyn-APPL1-GFP into the ventral hippocampus (vHPC) showed similar freezing responses with those injected with AAV9-hSyn-GFP during extinction retrieval, both of which were higher than WT group (Fig. S4), indicating that re-expression of APPL1 in the vHPC could not rescue the impaired extinction retrieval in Appl1-/- mice.
We also co-injected AAV9-hSyn-GCaMP7b and AAV9-hSyn-APPL1 (APPL1 group) or AAV9-hSyn–mCherry (control group) into the vmPFC of Appl1-/- mice (Fig. S5A-B) and observed that the calcium signals during extinction retrieval on Day 3 or Day 10 significantly increased in APPL1 group compared with the control group (Fig. S5C-F), indicating that re-expression of APPL1 rescues the activity patterns of vmPFC in Appl1-/- mice.
Constitutive APPL1 knockout impairs NMDAR-dependent potentiation in the vmPFC
Next, we tested the electrical properties of layer V pyramidal neurons in the vmPFC of constitutive Appl1-/- mice by the patch-clamp recording of brain slice. Resting membrane potential (RMP), input resistance, Ithreshold, membrane time constant (Tau), and current injection elicited AP numbers showed no significant difference between WT and Appl1-/- mice (Fig. 2A-2B). We also measured mEPSC and mIPFC and no significant difference was observed between WT and Appl1-/- mice (Fig. 2C-2D). In the following, we utilized a protocol to induce NMDAR-dependent LTP in the vmPFC (
NMDA receptor-dependent long-term potentiation comprises a family of temporally overlapping forms of synaptic plasticity that are induced by different patterns of stimulation.
Philosophical transactions of the Royal Society of London Series B, Biological sciences.2014; 36920130131
) and observed that Appl1-/- mice exhibited a deficit in NMDAR-dependent LTP (Fig. 2E). Moreover, the abundance of GluN2B, but not other NMDAR subunits, was decreased in the vmPFC of Appl1-/- mice compared with that of WT mice (Fig. 2F). Together, these data indicate that APPL1 deficiency impairs NMDAR-dependent plasticity in prefrontal pyramidal neurons while leaving basic synaptic properties unaffected.
Fig. 2NMDAR-dependent LTP is abolished in vmPFC of constitutive Appl1-/- mice. (A) Resting membrane potential (RMP), input resistance, Ithreshold, and membrane time constant (Tau) of layer V pyramidal neurons in the vmPFC from WT or Appl1-/- mice were recorded (n = 10 neurons per group). (B) Left: representative traces of action potentials elicited by current injection in the layer V pyramidal neurons from WT or Appl1-/- mice. Right: statistical analysis of the number of action potentials elicited by current injection (n = 10 neurons per group). (C) Representative traces (upper), histogram, and cumulative graphs of mEPSC frequency and amplitude (lower) of pyramidal neurons from WT (left) or Appl1-/- mice (right) (WT, n = 9; Appl1-/-, n = 11 neurons). mEPSC was recorded at a holding membrane potential of -70 mV and APV was added to the bath solution. (D) Representative traces (upper), histogram, and cumulative graphs of mIPSC frequency and amplitude (lower) of pyramidal neurons from WT (left) or Appl1-/- mice (right) (WT, n = 14; Appl1-/-, n = 13 neurons). (E) High-frequency stimulation induced NMDAR-dependent LTP in the vmPFC of WT mice, but not Appl1-/- mice. Upper: representative LTP traces recorded. 1, baseline traces, and 2, post-stimulation traces. Lower: time course of normalized EPSC response (left) and summary graphs of LTP magnitude (right) (WT, n = 7; Appl1-/-, n = 8 mice). (F) Western blot analysis of NMDAR and AMPAR subunit expression in the mPFC from WT or Appl1-/- mice (n = 4 mice per group). Data are presented as mean ± SEM; *P < 0.05, ****P < 0.0001, unpaired t-test.
Conditional APPL1 knockout in the vmPFC impairs fear extinction retrieval and abolishes NMDAR-dependent LTP
To bypass developmental and compensatory mechanisms, we generated a conditional knockout mouse line Appl1fl/fl. AAV9-hSyn-Cre-mCherry was injected in the vmPFC of Appl1fl/fl mice to specifically knock out APPL1 in the vmPFC neurons (hSyn-Cre group) and AAV9-hSyn-mCherry was injected as control (Fig. 3A and Fig. S6A). The fluorescence intensity of APPL1 significantly decreased in the hSyn-Cre group compared with the control group (Fig. 3B). Next, we performed a similar fear extinction protocol and observed that hSyn-Cre group showed higher freezing level during Ext. Retr.1 in comparison with the control group (Fig. 3C). We also utilized a stronger extinction protocol with a 2-session extinction training and observed significantly impaired fear extinction retrieval in the hSyn-Cre group (Fig. S6B).
Fig. 3APPL1 is required for fear extinction retrieval and NMDAR-dependent LTP in the vmPFC. (A) Diagram of bilateral injection of AAV9-hSyn-Cre-mCherry or AAV9-hSyn-mCherry virus into the vmPFC of Appl1fl/fl mice. Scale bar, 1 mm. (B) Left: representative images showing mCherry labeling and APPL1 expression in the vmPFC of hSyn group (Upper) and hSyn-Cre group (Lower). White arrow indicates cells infected with virus; scale bar, 50 μm. Right: statistical analysis of APPL1 fluorescence intensity (hSyn, n = 5; hSyn-Cre, n = 8 mice, unpaired t-test). (C) Conditional APPL1 knockout in the vmPFC impaired extinction retrieval (n = 8 mice per group; Left: two-way RM ANOVA followed by Sidak’s multiple comparisons test; Cond., F1, 14 = 3.045, P = 0.1029; Ext., F1, 14 = 0.4651, P = 0.5064; Ext.Retr., F1, 14 = 6.277; Right: unpaired t-test). (D) Representative traces (upper), histogram, and cumulative graphs of mEPSC amplitude (left) and frequency (right) of hSyn-mCherry and hSyn-Cre-mCherry pyramidal neurons in the vmPFC of Appl1fl/fl mice (hSyn, n = 20; hSyn-Cre, n = 17 neurons, unpaired t-test). (E) Representative traces (upper), histogram, and cumulative graphs of mIPSC amplitude (left) and frequency (right) of hSyn-mCherry and hSyn-Cre-mCherry pyramidal neurons in the vmPFC of Appl1fl/fl mice (hSyn, n = 20; hSyn-Cre, n = 21 neurons, unpaired t-test). (F and G) Representative traces (left) and histogram graph (right) of the paired-pulse ratio (F) and AMPA/NMDA (A/N) ratio (G) of hSyn-mCherry and hSyn-Cre-mCherry pyramidal neurons in vmPFC of Appl1fl/fl mice (hSyn, n = 15 or 14; hSyn-Cre, n = 16 or 13 neurons, unpaired t-test). (H) Representative traces (left) and graph (right) of NMDAR input-output current of hSyn-mcherry and hSyn-Cre-mcherry pyramidal neurons in the vmPFC of Appl1fl/fl mice (hSyn, n = 13; hSyn-Cre, n = 12 neurons, unpaired t-test). (I) High-frequency stimulation induced NMDAR-dependent LTP of hSyn-mCherry and hSyn-Cre-mCherry pyramidal neurons in vmPFC of Appl1fl/fl mice. Upper: representative LTP traces recorded. 1, baseline traces, and 2, post-stimulation traces. Lower: time course of normalized EPSC response (left), summary graphs, and probability line plotting of LTP magnitude (right) (hSyn, n = 9; hSyn-Cre, n = 8 neurons, unpaired t-test). Data are presented as mean ± SEM; *P < 0.05, ***P < 0.001.
In the following, we detected the electrical properties of layer V pyramidal neurons in the vmPFC of conditional APPL1 knockout mice. RMP, input resistance, Ithreshold, decay time, mEPSC, mIPSC, paired-pulse ratio (PPR), and AMPA/NMDA (A/N) ratio showed no significant difference between the hSyn-Cre group and the control group (Fig. S7A-D, Fig. 3D-3G). In contrast, NMDAR input-output current (Fig. 3H) and the expression level of GluN2B (Fig. S7E) were reduced in the hSyn-Cre group compared to the control group. Moreover, a similar protocol induced LTP in the vmPFC of the control group but failed to maintain LTP in the hSyn-Cre group (Fig. 3I), confirming that APPL1 deficiency decreases NMDAR expression and impairs NMDAR-dependent plasticity.
APPL1 nuclear translocation is necessary for fear extinction retrieval
To investigate whether the nuclear translocation of APPL1 is involved in the regulation of NMDAR expression and fear extinction retrieval, we utilized a peptide TAT-APPL1NLS to specifically block APPL1 nuclear translocation via disrupting the interaction between APPL1 and Importin α1 (
). We injected peptide TAT-APPL1NLS (scramble peptide TAT-APPL1Scr as control) into the vmPFC of WT mice by cannula administration (Fig. 4A and 4B). As illustrated in Fig. 4C, mice treated with TAT-APPL1NLS showed statistically higher freezing levels during Ext. Retr.1 and Ext. Retr.2 compared with mice treated with TAT-APPL1Scr. Moreover, NMDAR input-output current was decreased in the pyramidal neurons of the vmPFC after constitutive administration of TAT-APPL1NLS, but not in TAT-APPL1Scr group (Fig. 4D).
Fig. 4Nuclear translocation of APPL1 is necessary for the maintenance of surface NMDAR abundance. (A-C) Bilaterally injection of Tat-APPLNLS into the vmPFC of WT mice impaired extinction memory recall. (A) Image of vmPFC with bilaterally implanted guide cannulas. Scale bar, 1 mm. (B) Behavioral and electrophysiological experiment procedure for local peptide injection into the vmPFC. (C) Tat-APPL1NLS impaired extinction retrieval on Day 10. Left: two-way RM ANOVA followed by Sidak’s multiple comparisons test; Cond., F1, 35 = 0.09818, P = 0.7559; Ext., F1, 35 = 1.309e-005, P = 0.9971; Ext.Retr.1, F1, 35 = 6.925; Ext.Retr.2, F1, 17 = 4.898). Right: average freezing level for all trials in extinction retrieval (Ext.Ret.1: Tat-APPL1Scr, n = 18; Tat-APPL1NLS, n = 19 mice. Ext.Ret.2: Tat-APPL1Scr, n = 9; Tat-APPL1NLS, n = 10 mice, unpaired t-test). (D) Representative traces (left) and graph (right) of NMDAR input-output current of pyramidal neurons in the vmPFC treated with Tat-APPL1Scr or Tat-APPL1NLS (n = 14 or 18 neurons per group, Mann Whitney test). (E) Extinction retrieval induces nuclear translocation of APPL1. Upper: experiment procedure; Lower (left): Western blot analysis of APPL1 in cytoplasmic and nuclear extract with Histone H3 and GAPDH as loading control for nuclear extract and cytoplasmic extract, respectively. Lower (right): Statistical analysis of nucleus/cytoplasm ratio of APPL1 with (Ext.Retr. group) or without (HC. group; Homecage control) fear extinction retrieval (n = 5 samples per group, the mPFC from two mice of same group were pooled to one sample, unpaired t-test). Data are presented as mean ± SEM; *P < 0.05, ****P < 0.0001.
). As shown in Fig. S8, a single injection of importazole after extinction learning on Day 2 significantly impaired fear extinction retrieval on Day 3. Furthermore, we examined the intracellular location of APPL1 after fear extinction retrieval and observed that the nuclear/cytoplasmic ratio of APPL1 in the vmPFC was significantly increased in Ext. Retr. group compared to the control group (Fig. 4E), indicating that APPL1 undergoes nuclear translocation during fear extinction.
A subpopulation of neurons in the vmPFC is repeatedly activated during extinction retrieval
In the following, we applied engram technology to examine whether the vmPFC contains neuronal assemblies that store fear extinction memory. We first employed TRAP2;Ai14 mice to label the neurons that were activated during the processing of fear extinction information. TRAP2;Ai14 mice were injected with 4-OHT immediately after extinction retrieval on Day 3 to isolate the putative neuronal ensemble associated with fear extinction and permanently tag them with tdTomato (Ext. Retr. group) (Fig. S9A). 7 days later, mice were subjected to extinction retrieval again (Day 10) and subsequently sacrificed for Fos staining. The control group went through the same protocol as Ext. Retr. group but was labeled in the homecage on Day 4 (HC group). As shown in (Fig. S9B), the number of tdTomato-positive (tdTomato+) cells in Ext. Retr. group, as well as the number of tdTomato and Fos double-positive cells (tdTomato+Fos+), was significantly larger than the HC group. Moreover, about 32% of the tdTomato+ cells activated during Ext. Retr.1 was reactivated during Ext. Retr.2 in Ext. Retr. group. These extinction-associated neurons were widely distributed from Layer II to Layer V of the vmPFC with diverse morphology (Fig. S9C).
Next, we employed another AAV9-based Fos-dependent Robust Activity Marking (RAM) reporter system (
) to identify and interrogate the prefrontal neuronal ensemble related to fear extinction retrieval (Fig. 5A). WT mice were injected with AAV9-RAM-GFP into the vmPFC and kept on Dox drinking 24 hr before and after viral injection, switched to Dox-free drinking 24 hr before and after Ext. Retr.1 on Day 3 (Ext. Retr. group) or Day 4 (homecage group), and placed back on Dox drinking until the following Ext. Retr. 2 test and Fos immunostaining on Day 10 (Fig. 5B). As shown in Fig. 5C and 5D, the number of cells labeled by RAM system (GFP+) in Ext. Retr. Group was significantly larger than the homecage group. Moreover, about 29% of the vmPFC neurons activated during Ext. Retr.1 on Day 3 was reactivated during Ext. Retr.2 on Day 10, similar to those labeled with the TRAP2 system. We also analyzed the identity of these trapped neurons by RNAscope in-situ hybridization and observed that around 83.7% of RAM+ neurons expressed vesicular glutamate transporter 1 (vGluT1) mRNA and 16.3% of them expressed vesicular GABA transporter (vGAT) mRNA (Fig. S10), indicating that most of the extinction-labeled neurons are excitatory.
Fig. 5Extinction-associated neurons in the vmPFC are necessary and sufficient for extinction memory recall. (A) Virus strategy for labeling. When the PRAM (an enhanced FOS promoter) promoter is activated, tTA is expressed in cells; tTA protein then binds to the TRE promoter, resulting in the expression of GFP (RAM) or Cre-GFP (CRAM). DOX prevents the binding of tTA to the TRE promoter and restricts the expression of downstream effector gene. (B) Experimental design for extinction-associated ensemble labeling. DOX was taken off 24 hours before Ext.Retr.1 or Homecage (HC), and placed back 24 hours after it. (C) Diagram to indicate the labeling of extinction-associated neurons with GFP. (D) Left: Images of neurons in the vmPFC that were activated during Ext.Retr.1 (GFP+) or Ext.Retr. 2 (Fos+). DAPI (blue) was used to label nuclei. White arrows indicate cells that were co-labeled with GFP and Fos (GFP+ Fos+). Right: Higher levels of neurons were reactivated (GFP+ Fos+) during Ext.Retr. 2 and activated (GFP+) during Ext.Retr. 1 in Ext.Retr. group (n = 12 mice) than HC. group (n = 6 mice), unpaired t-test. Scale bar, 200 μm. (E) Left: diagram to indicate the labeling of extinction-associated neurons with GFP and hM4Di. Right: representative images of neurons in the vmPFC co-expressing hM4Di and GFP. Scale bar, 100 μm. Inserted: validation of virus by electrophysiological recording. (F) Chemogenetic inhibition of extinction-associated neurons by i.p. injection of CNO impaired fear extinction retrieval on Day 10. Left: Two-way RM ANOVA followed by Sidak’s multiple comparisons test, Cond., F1, 14 = 2.778, P = 0.1178; Ext., F1, 14 = 0.8615, P = 0.3690; Ext.Retr.1, F1, 14 = 0.2386, P = 0.6328; Ext.Retr.2, F1, 14 = 5.362. Right: average freezing level for all trials in extinction retrieval (n = 8 mice per group, unpaired t-test). (G) Left: Diagram to indicate the labeling of extinction-associated neurons with GFP and hM3Dq. Right: representative images of vmPFC neurons co-expressing hM3Dq and GFP. Scale bar, 100 μm. Inserted: validation of virus by electrophysiological recording. (H) Chemogenetic activation of extinction-associated neurons by i.p. injection of CNO facilitated fear extinction retrieval on Day 10. Left: Two-way RM ANOVA followed by Sidak’s multiple comparisons test, Cond. F1, 15 = 0.7964, P = 0.3863; Ext., F1, 15 = 0.9534, P = 0.3443; Ext.Retr.1, F1, 15 = 0.5251, P = 0.4798; Ext.Retr.2, F1, 15 = 10.77. Right: average freezing level for all trials in extinction retrieval (mCherry, n = 8; hM3Dq, n = 9 mice, unpaired t test). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ****P < 0.0001.
Extinction-labeled neurons are necessary and sufficient for driving extinction memory recall
To investigate the role of these labeled neurons during fear extinction retrieval, we injected AAV9-hSyn-DIO-hM4Di-GFP (hM4Di group) or AAV9-hSyn-DIO-GFP (GFP group as control) into the vmPFC of TRAP2;Ai14 mice (Fig. S11A) and, 3 weeks later, subjected the mice to fear extinction protocol (Fig. S11B). Neurons activated during Ext. Retr.1 (tdTomato+) were tagged with hM4Di-GFP or GFP (Fig. S11C). During Ext. Retr.2 (Day 10), inhibition of these hM4Di-tagged neurons by CNO intraperitoneal injection elicited a higher freezing level than the GFP group, indicating impaired extinction retrieval (Fig. S11D).
Next, we employed a Cre-dependent RAM system and co-injected AAV9-CRAM-GFP and AAV9-hSyn-DIO-hM4Di-mCherry (hM4Di group) or AAV9-hSyn-DIO-mCherry (mCherry group as control) into the vmPFC of WT mice (Fig. S12A). 3 weeks later, two groups of mice were subjected to similar fear conditioning and extinction protocol and the neurons activated during Ext. Retr.1 (GFP+) were tagged with hM4Di-mCherry or mCherry (Fig. 5E). Intraperitoneal injection of CNO 30 min before Ext. Retr. 2 on Day 10 inhibited the hM4Di-tagged neurons and suppressed fear extinction retrieval in the hM4Di group when compared with the mCherry group (Fig. 5F).
Finally, we co-injected AAV9-CRAM-GFP and AAV9-hSyn-DIO-hM3Dq-mCherry (hM3Dq group) or AAV9-hSyn-DIO-mCherry (mCherry group as control) into the vmPFC of WT mice (Fig. S12B) and performed a sub-saturated extinction training protocol to avoid floor effect. Neurons activated during Ext. Retr.1 (GFP+) were tagged with hM3Dq-mCherry or mCherry (Fig. 5G). Activation of the neurons trapped during Ext. Retr. 1 significantly decreased the freezing level during Ext. Retr.2 in the hM3Dq group compared with the mCherry group (Fig. 5H), indicating that activation of the tagged neurons facilitates retrieval of fear extinction memory.
Inducible APPL1 knockout in the prefrontal neuronal ensemble inhibits fear extinction memory
To test whether APPL1 is required for the prefrontal neuronal ensemble to consolidate and retrieve extinction memory, we combined the CRAM system with Appl1fl/fl mice to specifically decrease the expression level of APPL1 in the extinction-tagged neurons (Fig. 6A). Appl1fl/fl mice were injected with AAV9-CRAM-GFP (CRAM group, Fig. S13A) or AAV9-RAM-GFP (RAM group as control, Fig. S13B) into the vmPFC and, 3 weeks later, underwent fear conditioning and extinction programs. The mice were kept on Dox drinking unless 24 hr before and after the Ext. Retr. 1 on Day 3, tested for Retr.2 on Day 10, and sacrificed for immunostaining 1.5 h after the behavioral experiment (Fig. 6B). In another control group, mice injected with AAV9-CRAM-GFP underwent similar treatment and were switched to Dox-free drinking in the homecage on Day 4 (CRAM-control group, Fig. S13C). The abundance of APPL1 in the labeled cell in the CRAM group significantly decreased when compared with the nearby unlabeled cells (Fig. 6C and 6D).
Fig. 6Inducible APPL1 knockout in extinction-associated neurons in the vmPFC impairs fear extinction memory recall. (A) Virus strategy for inducible APPL1 knockout in extinction-associated neurons in the vmPFC. (B) Behavioral protocol to label extinction-associated neurons. (C) Diagram to indicate that the extinction-associated neuron is labeled with GFP and, simultaneously, the level of APPL1 in this neuron decreases. (D) Left: Images of cells trapped during retrieval of extinction memory (GFP+) and immunostained with anti-APPL1 antibody (Red). Scale bar, 50 μm. Dotted circle indicates extinction-tagged neurons and solid circle indicates unlabeled neurons. Right: the abundance of APPL1 in the trapped cells decreased in CRAM group (n = 9 mice) compared with RAM group (n = 8 mice), unpaired t-test. (E) Inducible APPL1 knockout in extinction-tagged cells impaired fear extinction memory recall. Left: Two-way RM ANOVA followed by Tukey’s multiple comparisons test. F2, 26 = 2.141, P = 0.1377; Ext., F2, 26 = 0.2227, P = 0.8018; Ext.Retr.1, F2, 26 = 0.1312, P = 0.8776; Ext.Retr.2, F2, 26 = 3.604. Right: One-way ANOVA with Bonferroni’s Multiple Comparison Test (RAM, n = 10; CRAM, n = 9; CRAM Ctrl, n = 10 mice). (F) Inducible APPL1 knockout did not affect the activation of neurons in the vmPFC during extinction retrieval. Left: Images of neurons activated during Ext Retr. 1 (GFP+) and Ext Retr. 2 (Fos+) in RAM group (upper) and CRAM group (lower). White arrows indicate cells co-labeled with GFP and Fos (GFP+ Fos+). Scale bar, 200 μm. (G) Statistical analysis of the number of neurons activated during Ext.Retr.1 (Left, GFP labeling; RAM, n = 6; CRAM, n = 9 mice) or Ext.Retr. 2 (middle, Fos staining; RAM, n = 18; CRAM, n = 13 mice), or reactivated during Ext.Retr. 2 (Right, Colocalization ratio; RAM n = 13; CRAM, n = 10 mice), unpaired t-test. Data are presented as mean ± SEM. *P < 0.05, ***P < 0.001.
We next examined the behavioral effect of inducible APPL1 knockout in extinction-tagged neurons and observed that the freezing level of mice during Ext.Retr. 2 was significantly higher in the CRAM group than the RAM group or the CRAM-control group (Fig. 6E), suggesting that inducible APPL1 knockout in the prefrontal extinction-associated neurons suppresses retrieval of fear extinction memory. We further analyzed the number of neurons activated during Ext. Retr. 1 (GFP+) and during Ext. Retr. 2 (Fos+) and observed that the CRAM group and RAM group showed comparable levels of neuronal activation (Fig. 6F and 6G). Moreover, the ratio of neurons reactivated during Ext. Retr.2 to neurons activated during Ext. Retr.1 in the CRAM group was similar to the RAM group (Fig. 6G), suggesting that inducible APPL1 knockout has little effect on the reactivation of prefrontal extinction-associated neurons during extinction retrieval.
Inducible APPL1 knockout abolishes NMDAR-dependent potentiation in the extinction-associated neurons
To test the neural activity differences between extinction-trapped neurons and untrapped neurons, we combined RAM system with the electrophysiological recording (Fig. 7A) and observed that extinction-trapped neurons had higher mEPSC frequency and amplitude (Fig. 7B-C) than the neighboring untrapped neurons, while the mIPSC frequency and amplitude showed no difference (Fig. 7D-E). We also recorded PPR (Fig. 7F) and A/N ratio (Fig. 7G) and the data showed that extinction-trapped neurons had a higher A/N ratio than neighboring untrapped neurons. Taken together, it indicates that extinction-trapped neurons had enhanced synaptic strength. Moreover, we successfully induced LTP in untrapped pyramidal neurons, and this induction was blocked by pretreatment with ifenprodil (Fig. 7H-I). However, we failed to induce LTP in extinction-trapped pyramidal neurons (LTP occlusion). LTP occlusion has been interpreted as evidence that learning induced LTP early on, which can ultimately limit the ability to induce further LTP (
Fig. 7APPL1 is required for enhanced synaptic strength in prefrontal extinction-associated neurons. (A) Experimental scheme for recording labeled neurons and nearby unlabeled neurons after Ext.Retr. 2 on Day 10. AAV-RAM-GFP was injected into the vmPFC of WT mice and neurons activated during Ext.Retr. 1 was labeled with GFP. (B-C) Representative traces (left), histogram, and cumulative graphs (right) of mEPSC obtained from vmPFC extinction-labeled neurons and nearby unlabeled cells (RAM-, n = 44; RAM+, n = 35 neurons per group from WT mice, unpaired t-test). (D-E) Representative traces (left), histogram, and cumulative graphs (right) of mIPSC obtained from vmPFC extinction-labeled neurons and nearby unlabeled cells (RAM-, n = 35; RAM+, n = 30 neurons per group from WT mice, unpaired t-test). (F-G) Representative traces (left) and histogram (right) of paired-pulse ratio (F) and A/N ratio (G) from vmPFC extinction-labeled neurons and nearby unlabeled cells (RAM-, n = 10; RAM+, n = 9 neurons per group from WT mice, unpaired t-test). (H-I) High-frequency stimulation induced NMDAR-dependent LTP in RAM- neurons, which was completely blocked by pretreatment with ifenprodil (5 μM), but not in RAM+ neurons, in the vmPFC of WT mice. Upper: representative LTP traces recorded. 1, baseline traces, and 2, post-stimulation traces. Lower: time course of normalized EPSC response (left), summary graphs, and probability line plotting of LTP magnitude (right) (One-way ANOVA with Bonferroni’s multiple comparisons test, F2, 27 = 690.2. RAM-, n = 13; RAM+, n = 12; RAM- + ifenprodil, n = 9 neurons from WT mice). (J) Experimental scheme for recording labeled neurons with inducible APPL1 knockout and nearby unlabeled neurons with normal level of APPL1 during Ext.Retr. 2 on Day 10. AAV-CRAM-GFP was injected into the vmPFC of Appl1fl/fl mice and neurons activated during Ext.Retr. 1 was labeled with GFP. (K-L) Representative traces (left), histogram, and cumulative graphs (right) of mEPSC obtained from vmPFC extinction-labeled neurons and nearby unlabeled cells (CRAM-, n = 35; CRAM+, n = 34 neurons per group from Appl1fl/fl mice, unpaired t-test). (M-N) Representative traces (left), histogram, and cumulative graphs (right) of mIPSC obtained from vmPFC extinction-labeled neurons and nearby unlabeled cells (CRAM-, n = 24; CRAM+, n = 32 neurons per group from Appl1fl/fl mice, unpaired t-test). (O and P) Representative traces (left) and histogram (right) of paired-pulse ratio (O) and A/N ratio (P) from vmPFC extinction-labeled neurons and nearby unlabeled cells (CRAM-, n = 13; CRAM+, n = 11 neurons per group from Appl1fl/fl mice, unpaired t-test). (Q) High-frequency stimulation fails to induce NMDAR-dependent LTP in the CRAM+ neurons in the vmPFC of Appl1fl/fl mice. Upper: representative LTP traces recorded. 1, baseline traces, and 2, post-stimulation traces. Lower: time course of normalized EPSC response (left) and probability line plotting of LTP magnitude (right) (n = 9 neurons from Appl1fl/fl mice). (R) Representative traces (left) and graph (right) of NMDAR input-output of CRAM- and CRAM+ pyramidal neurons in vmPFC of Appl1fl/fl mice (n = 15 neurons per group from Appl1fl/fl mice, unpaired t-test). Data are presented as mean ± SEM; *P < 0.05, ** P < 0.01, ***P < 0.001, ****P < 0.0001.
Next, we started to examine whether APPL1 is required for synaptic strengthening of the prefrontal extinction-labeled neurons. AAV9-CRAM-GFP was injected into the vmPFC of Appl1fl/fl mice (Fig. 7J). As shown in Fig. 7K-L and Fig. 7M-N, extinction-trapped neurons and neighboring untrapped neurons in the vmPFC elicited similar frequency and amplitude of mEPSC and mIPSC. Moreover, no difference in PPR or A/N ratio was observed in the two groups (Fig. 7O and 7P). In contrast, the high-frequency stimulation (HFS) protocol failed to induce LTP in the extinction-trapped pyramidal neurons (Fig. 7Q) and NMDAR current was reduced in the extinction-trapped neurons compared to untrapped neurons (Fig. 7R). Combined, it indicates that APPL1 is required for synaptic potentiation in the trapped neurons.
Chemogenetic activation of prefrontal extinction-associated neurons with inducible APPL1 knockout restores impaired extinction recall.
To further examine whether artificial activation of extinction-associated neurons with insufficient APPL1 rescues impaired extinction memory, we co-injected AAV9-CRAM-GFP and AAV9-DIO-hM3Dq-mCherry into the vmPFC of Appl1fl/fl mice to decrease the expression level of APPL1 in the extinction-associated neurons and, concomitantly, tag these neurons with hM3Dq (hM3Dq group) (Fig. 8A, Fig. S13D). AAV9-CRAM-GFP and AAV9-DIO-mCherry were co-injected into the vmPFC of Appl1fl/fl mice as a control (Fig. 8A, Fig. S13D). Both groups of mice were kept on Dox drinking until 24 hr prior to the Ext. Retr. 1 to capture extinction-associated neurons and were intraperitoneally injected with CNO 30 min before Ext. Retr.2 (Fig. 8B). As shown in Figure 8C and 8D, mice in hM3Dq group showed less freezing behavior than control group during Ext. Retr.2, indicating that artificial activation of the extinction-associated neurons with inducible APPL1 knockout restores impaired extinction behaviors.
Fig. 8Chemogenetic activation of prefrontal extinction-associated neurons with inducible APPL1 knockout restores extinction memory recall. (A) Diagram to indicate the effect of virus-mediated inducible knockout of APPL1 and simultaneously labeling with hM3Dq in the extinction-associated neurons in Appl1f1/f1 mice. (B) Experimental protocol to capture extinction-associated neurons in the vmPFC of Appl1f1/f1 and label them with hM3Dq during Ext.Retr.1. (C) Representative images of neurons co-expressing hM3Dq and GFP in the vmPFC neurons of Appl1fl/fl mice. Inserted: validation of virus by electrophysiological recording. (D) Chemogenetic activation of extinction-associated neurons with inducible APPL1 knockout by i.p. injection of CNO facilitated fear extinction retrieval on Day 10. Left: Two-way RM ANOVA followed by Sidak’s multiple comparisons test, Cond., F1, 14 = 0.6823, P = 0.4226; Ext., F1, 14 = 0.06634, P = 0.8005; Ext.Retr.1, F1, 14 = 0.900, P = 0.3589; Ext.Retr.2, F1, 14 = 22.69. Right: average freezing level for all trials in extinction retrieval (n = 8 mice per group, unpaired t-test). Data are presented as mean ± SEM. ***P < 0.001. Scale bar, 100 μm. (E to J) Working model of fear extinction memory retrieval. In WT mice, natural retrieval cue triggers plastic changes of the extinction-associated neuronal ensemble, which contribute to successful recall of extinction memory (E). Artificial inhibition (F) or activation (G) of the neuronal ensemble impairs and facilitates fear extinction retrieval, respectively. Inducible knockout of APPL1 in the neuronal ensemble abolished synaptic potentiation and impaired extinction memory retrieval (H). Simultaneously artificial activation of the neuronal ensemble with inducible APPL1 knockout restores extinction memory recall (I).
A convergent body of human and non-human studies suggests that vmPFC mediates the extinction of conditioned fear and abnormal vmPFC function accounts for the symptoms of PTSD. In this work, we provide evidence indicating that a subset of vmPFC neurons holds engrams for fear extinction memory and APPL1 signaling supports these engram cells to build up NMDAR-dependent synaptic potentiation, which is required for the retrieval of extinction memory. Thus, our finding provides novel insight into the neurobiological mechanisms that underlie fear extinction, which is of utmost importance to understanding the working principles of memory and how it is affected in PTSD.
Here, by taking advantage of activity-dependent cell labeling techniques and functional manipulation, we captured a subset of neurons in the vmPFC that were repeatedly activated during retrieval of fear extinction memory (Fig. 8E). Functional manipulation revealed the necessity (Fig. 8F) and sufficiency (Fig. 8G) of these captured neurons for the retrieval of extinction memory. We also observed functional plastic changes and LTP occlusion in the tagged neurons after re-exposure to extinction retrieval cues. These findings indicate that vmPFC holds engrams for fear extinction memory. Notably, waiting for the labeling of extinction-associated neurons introduces time delays between extinction training and extinction test, which provides the opportunity for spontaneous recovery, and thus, we cannot exclude the possibility that manipulation of these engram cells may affect their susceptibility to spontaneous recovery.
Although engram cell-specific synaptic plasticity has been observed in several different preparations, including mPFC, hippocampus, and amygdala, following fear conditioning (
), the intracellular signaling pathways that mediate the plastic changes of engram cells have not yet been identified. Here, we found that APPL1 signaling is required for the maintenance of NMDAR expression and is necessary for the synaptic strengthening of extinction-trapped engram cells (Fig. 8H). Combined with our previous finding that nuclear APPL1 is a critical modulator of gene transcription (
), our results indicate that APPL1 signaling may link experience with gene transcription to maintain NMDAR expression, which modifies synaptic weights and facilitates memory retrieval. Consistently, fear memory and fear extinction have been found to be associated with changes at the level of transcription (
). Here, we found that chemogenetic activation of the engram cells with insufficient APPL1 signaling rescued the impaired recall of fear extinction (Fig. 8I), suggesting that augmented synaptic strength is crucial for the accessibility of the extinction memory, but not the storage of it. This finding is consistent with several recent studies showing that artificial synaptic potentiation of engram cells could rescue impaired memory recall (
In this study, we found that the freezing level of mice with APPL1 knockout in the vmPFC during fear extinction retrieval is still lower than the initial fear memory after fear conditioning, indicating that APPL1 signaling alone is insufficient for a full extinction memory retrieval. It is probably due to the existence of other signaling pathways or extinction-associated engrams located in other brain regions, such as the amygdala (
). Moreover, APPL1 is widely distributed in both neurons and glial cells and is involved in different functions, such as modulating neuronal synaptic plasticity (
). Although we found that APPL1 signaling-mediated synaptic potentiation in the pyramidal neurons of vmPFC is necessary for extinction retrieval, we cannot exclude the possibility that APPL1 in the inhibitory neurons or even in glial cells may also contribute to the handling of fear extinction information.
Taken together, our findings demonstrate the critical role of vmPFC engram cells in storing extinction memory and APPL1-mediated augmentation of synaptic strength in recalling extinction memory, which provide insights into the development of potential therapeutic approaches to treating PTSD.
Acknowledgements
We thank Dr. Aimin Xu (University of Hongkong) for providing constitutive Appl1-/- mice. We thank Dr. Hailan Hu (Zhejiang University) for providing Trap2 mice. We thank Yudong Zhou and Xiaodong Wang (Zhejiang University) for their valuable suggestions on the manuscript. We also thank the technical support of the Core Facilities, Zhejiang University School of Medicine.
Long-term expression of human contextual fear and extinction memories involves amygdala, hippocampus and ventromedial prefrontal cortex: a reinstatement study in two independent samples.
Social cognitive and affective neuroscience.2014; 9: 1973-1983
Dissociable roles of prelimbic and infralimbic cortices, ventral hippocampus, and basolateral amygdala in the expression and extinction of conditioned fear.
Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology.2011; 36: 529-538
APPL1 counteracts obesity-induced vascular insulin resistance and endothelial dysfunction by modulating the endothelial production of nitric oxide and endothelin-1 in mice.
NMDA receptor-dependent long-term potentiation comprises a family of temporally overlapping forms of synaptic plasticity that are induced by different patterns of stimulation.
Philosophical transactions of the Royal Society of London Series B, Biological sciences.2014; 36920130131
00087-2&id=gr1.jpg){kind=link}
00087-2&id=gr2.jpg){kind=link}
00087-2&id=gr3.jpg){kind=link}
00087-2&id=gr4.jpg){kind=link}
00087-2&id=gr5.jpg){kind=link}
00087-2&id=gr6.jpg){kind=link}
00087-2&id=gr7.jpg){kind=link}
00087-2&id=gr8.jpg){kind=link}