NRSF regulates age-dependently cognitive ability and its conditional knockout in APP/PS1 mice moderately alters AD-like pathology

Abstract

NRSF/REST (neuron-restrictive silencer element, also known as repressor element 1-silencing transcription factor), plays a key role in neuronal homeostasis as a transcriptional repressor of neuronal genes. NRSF/REST relates to cognitive preservation and longevity of humans, but its specific functions in age-dependent and Alzheimer’s disease (AD)-related memory deficits remain unclear. Here, we show that conditional NRSF/REST knockout either in the dorsal telencephalon or specially in neurons induced an age-dependently diminished retrieval performance in spatial or fear conditioning memory tasks and altered hippocampal synaptic transmission and activity-dependent synaptic plasticity. The NRSF/REST deficient mice were also characterized by an increase of activated glial cells, complement C3 protein and the transcription factor C/EBPβ in the cortex and hippocampus. Reduction of NRSF/REST by conditional depletion upregulated the activation of astrocytes in APP/PS1 mice, and increased the C3-positive glial cells, but did not alter the Aβ loads and memory retrieval performances of 6- and 12-month-old APP/PS1 mice. Simultaneously, overexpression of NRSF/REST improved cognitive abilities of aged wild type, but not in AD mice. These findings demonstrated that NRSF/REST is essential for the preservation of memory performance and activity-dependent synaptic plasticity during aging and takes potential roles in the onset of age-related memory impairments. However, while altering the glial activation, NRSF/REST deficiency does not interfere with the Aβ deposits and the electrophysiological and cognitive AD-like pathologies.

Introduction

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disorder and accounts for more than 60% of diagnosed dementia cases. Senile plaques (SP) assembled with the aggregates of misfolded insoluble amyloid-beta and neurofibrillary tangles (NFT) of the tau protein (1–3) are the pathological hallmarks in the brain of AD patients. Aging is the most definite risk factor in the pathogenesis of AD (4). Environmental factors that enhance the level of inflammation, oxidative stress or genetic mutations also contribute to the onset of AD (5). Moreover, the altered activity of distinct transcription factors promotes the pathogenesis of neurodegenerative diseases (6–8).

The transcription factor GLI-Krüppel class C2H2 zinc finger protein NRSF/REST (neuron restrictive silencing factor, also known as RE1-silencing transcription factor) binds a highly conserved 21–23 bp DNA sequence called NRSE (neuron-restrictive silencer element), also known as RE1 (repressor element-1). Genes targeted by NRSF/REST include synaptic ion channel receptors, synaptic vesicle proteins, neurotrophic factors and protein kinase mitogen-activated protein kinases (MAPKs) (9–11).

NRSF/REST is a transcription factor linking neuronal activity to intrinsic homeostasis and contributes to physiological levels of neuronal network activity (12). Dysfunction of NRSF/REST contributes to neurological disorders. Insults including stroke and epilepsy/seizures augment NRSF levels in specific brain regions, including the hippocampus (13–15). Moreover, NRSF/REST has been linked to neurodegenerative diseases like Huntington’s disease (16), Parkinson’s disease (7,17) and AD (18–20). NRSF/REST retains a protective role against Aβ toxicity and oxidative stress, likely by inhibiting the expression of genes involved in cell apoptosis and oxidative stress (19). NRSF/REST also contributes to extending the lifespan by mediating neural circuit activity (21). Moreover, NRSF/REST upregulates the neuroprotective neuroglobin in APP23 transgenic mice (22). Neuroglobin itself plays a role in the choline acetyltransferase (ChAT) expression (23). NRSF/REST has been reported to be inversely correlated to ChAT expression level in the human AD brain (18).

In our previous study, NRSF/REST expression dominated the survival of dopaminergic cells under MPP⁺ intoxication (17), while mice with neuronal NRSF/REST deficiency were more vulnerable to MPTP toxicity (7). These experiments pointed towards a protective role of NRSF/REST in Parkinson’s disease-related neurodegeneration. The expression level of NRSF/REST is not constant and increases with age but it decreases in AD patients (19). However, the complex interactions of NRSF/REST with neuronal development and the impact of NRSF/REST in aging and AD progression require further studies.

In this study, C57BL/6 mice, mice with NRSF/REST deficiency or APP/PS1 mutation and APP/PS1 transgenic mice with NRSF/REST deficiency, either in the dorsal telencephalon including neurons and astrocytes or specially in neurons were studied to elucidate the roles of NRSF/REST in aging and AD pathology. The cognitive performances and the levels of Aβ loads of different mouse genotypes, aged from 3 to 22 months, were analyzed. The loss of NRSF/REST resulted in significant memory performance deficits and synaptic impairments in 12-month-old mice, along with an increase of complement C3 and the upstream transcription factor CCAAT-enhancer-binding protein beta (C/EBPβ). The conditional depletion of NRSF/REST in either neurons and astrocytes of the dorsal telencephalon or specifically in neurons in 3- to 22-month-old APP/PS1 mice did not detectably attenuate or worsen the progress of AD-like pathologies, including the memory performance in a variety of behavioral tasks or the cerebral Aβ deposits. In line with this, NRSF/REST overexpression improved cognitive ability of aged wild type (WT) mice but not of APP/PS1 mice. However, reduction of NRSF/REST by conditional depletion induced the hyperactivation of astrocytes in APP/PS1 mice.

Results

NRSF/REST deficiency causes memory decline and gliosis in aging mice

To evaluate the role of NRSF/REST in aging and cognitive-related degenerative diseases, we exploited a mouse line with NRSF/REST deficiency in the dorsal telencephalon, including the cortex and hippocampus (Emx1-CKO). We examined the impact of NRSF/REST deficiency in 12-month-old mice on a variety of behavioral and biochemical parameters. The efficiency of NRSF/REST deletion was accessed by real-time PCR and immunofluorescence staining (Supplementary Material, Fig. S1). Behavioral tests including the open field test (OFT), fear conditioning (FC) test and Y-maze test were performed with Emx1-CKO mice and WT littermates. Compared with the WT counterparts, Emx1-CKO mice showed no difference neither in the total ambulatory distance nor the time spent in the center of the open field (Supplementary Material, Fig. S2A), indicating that NRSF/REST deficiency in mouse dorsal telencephalon did not induce anxiety-like behaviors.

The FC task was utilized to study processes related to fear memory (24). The aged Emx1-CKO mice showed a significantly lower percentage of freezing in response to context or cue (Fig. 1A, a). The Y-maze task was exploited to analyze the spatial memory performance of Emx1-CKO mice. Emx1-CKO mice showed lower alternation in the Y-maze, whereas the number of total entries did not differ between the two genotypes (Fig. 1A, b).

Figure 1

NRSF/REST deficiency in the dorsal telencephalon causes behavioral deficits of learning and memory in 12-month-old mice. (A) Memory performances of 12-month-old mice. (a) Fear conditioning test. Freezing percentage in the first 2 min in the contextual fear test, and in time blocks of the sound fear test. Control: n = 13 and Emx1-CKO: n = 24. (b) Y-maze test. Percentage of alternation and entries into arms (control: n = 9, Emx1-CKO: n = 13). (B) Glial cells in the brain of 12-month-old Control and Emx1-CKO mice. (a) Astrocytes: GFAP. (b) Microglia: Iba1. Nuclei: DAPI. Scale bar: 200 μm. (c) Glia: cell numbers. n = 3. (C) (a) WBs of NRSF and GFAP in the cortex of 6- and 12-month-old wild-type and APP/PS1 mice (β-actin as internal control). (b) Relative protein levels (n = 6 ~ 8). Unpaired t-test: ^*P < 0.05, ^**P < 0.01 and ^***P < 0.001.

NRSF/REST deficiency increases the number of astrocytes (7,19). Thus, we examined the degree of gliosis in the Emx1-CKO mice. We also observed an increase in the number of astrocytes and microglial cells in the cortex of 12-month-old Emx1-CKO mice. However, in the hippocampus of Emx1-CKO mice, the number of astrocytes, but not microglia, increased (Fig. 1B).

NRSF/REST deficiency has been previously linked to memory deficits (19); thus, we evaluated the NRSF/REST protein levels in aging WT, or cognitively impaired APP/PS1 mice. NRSF/REST expression did not change in the cortex of 6- and 12-month-old WT mice and did not differ in the cortex of 6-month-old WT and APP/PS1 mice. The expression of NRSF/REST decreased in the cortex of 12-month-old APP/PS1 mice compared with WT littermates (Fig. 1C), similar to human AD brains (19). An increase in GFAP expression was detected in the cortex of 12-month-old APP/PS1 mice compared with WT mice, indicating a hyperastrogliosis (Fig. 1C). Apoptosis is an important pathological mechanism in APP/PS1 and NRSF/REST-deficient mice (19,25). Therefore, the expression of Bax and Bcl2 proteins was examined. Bax protein levels increased in 12-month-old AD mice, and Bcl2 proteins did not differ. The ratio of Bcl2/Bax did not alter in the cortex of 6- or 12-month-old transgenic AD and WT mice (Supplementary Material, Fig. S2C).

We found that NRSF/REST deficiency in the telencephalon caused memory dysfunction and gliosis in aging mice. In addition, NSE-Cre: NRSF^flox/flox mice (NSE-CKO) were tested to learn whether specific knockout of NRSF/REST in neurons (7,26) also interferes with memory. In contrast to the Emx1-CKO mice, at the age of 12 months, NSE-CKO mice took significantly less ambulatory distance in the OFT, indicating a reduced autonomic activity (Supplementary Material, Fig. S2B, a). In comparison with WT littermates, neuronal-specific NRSF/REST-deficient mice showed reduced learning and memory performances in the Morris water maze (MWM) test, indicating a higher latency and the longer path to the platform on the seventh day (Fig. 2A, a and b). Transgenic mice also spent less time in the target quadrant but displayed similar target crossings in the probe test (Fig. 2A, c and d). However, in the Y-maze and fear conditioning tests, NSE-CKO mice showed no difference compared to the control littermates (Supplementary Material, Fig. S2B, b and c). On the synaptic level and at the age of 15 months, paired-pulse facilitation (PPF), long-term potentiation (LTP) and short-term potentiation (STP) were similar between NSE-CKO and control mice (Fig. 2B, a–c). To evaluate if the number of synapses was altered, the density of spines in the hippocampus was determined by Golgi staining. We observed that the number of spines was not affected by a neuronal NRSF/REST deficiency (Supplementary Material, Fig. S3). Therefore, NRSF/REST deficiency in the dorsal telecephalon or neurons both causes memory decline in aging mice, indicating poor performances in the learning and memory tests. Moreover, absence of NRSF/REST in the dorsal telecephalon leads to an increase of astrocytes and microglial cells in the cortex and hippocampus of aged mice.

Figure 2

The results of the Morris Water Maze test (A) and the synaptic plasticity of the hippocampal slices (B) in Control and NSE-CKO mice. (A) Performances of 12- month-old mice in the Morris Water Maze test. (a) Latency to the platform. (b) Path length to the platform. (c) Times of platform cross. (d) Percentage of time spent in the target quadrant (unpaired t-test. n = 22–23. ^*P < 0.05, ^**P < 0.01). (B) Synaptic plasticity of the hippocampal slices. (a) LTP in hippocampal slices. Data dots represent fEPSP slopes normalized to their respective averaged baseline values recorded before high-frequency stimulation. n = 3. (b) Paired-pulse ratio of amplitude recorded across different inter-stimulus intervals (0 ~ 200 ms). n = 6. (c) STP in hippocampal slices. n = 6. Brackets and ^*: P < 0.05 (Student t-test with Welch’s correction).

Enhanced NRSF/REST deficiency does not alter detectably impaired memory performance or Aβ loads in APP/PS1 mice

To determine whether NRSF/REST loss in the dorsal telencephalon interferes with Aβ-processing in APP/PS1 mice and to decipher the role of NRSF/REST in AD-like pathologies, a new mouse line (Emx1-CKO-AD) was generated by crossing APP/PS1 AD mice with Emx1-Cre:NRSF^flox/flox mice and their performance compared with mice of different genotypes (control, Emx1-CKO, APP/PS1).

We tested the memory performance of the new transgenic mouse strain utilizing the novel object recognition (NOR) and Y-maze tasks. At 3 months, the Emx1-CKO mice and the control mice did not differ in the NOR performance (Fig. 3A). However, APP/PS1 and Emx1-CKO-AD mice showed a significant memory decline as indicated by a decreased discrimination index and lower exploration time at the novel object in the NOR test. Notably, NRSF/REST loss did not enhance memory dysfunction of 3-month-old APP/PS1 mice (Fig. 3A). In addition, the data show that Emx1-CKO-AD and Emx1-CKO differed significantly in the NOR test, indicating that NRSF/REST conditional knockout does not prevent APP/PS1-induced memory decline in 3-month-old mice. In addition, the mice of all four groups scored similar between alternations and entries in the Y-maze test (Supplementary Material, Fig. S4A, a and b).

Figure 3

NRSF/REST deficiency has no impact on Aβ plaques or memory decline of APP/PS1 mice. (A-C) Behavioral performances of 3-, 6- and 12- months-old control and transgenic mice. (A) Exploration of 3-month-old transgenic mice in the Novel object recognition test. Behavioral performances of 6-month-old mice (B) and 12-month-old mice (C) in the Morris Water Maze test. (a) Latency to the platform. (b) Times of platform cross. (c) Percentage of time spent in the target quadrant. ^*P < 0.05 and ^**P < 0.01, (One-way or Two-way ANOVA with Fisher’s LSD multiple comparisons), ^#P < 0.05 and ^###P < 0.001, (One-Way ANOVA with Tukey’s multiple comparison test post hoc). The number of mice for each group ranges from 8 to 12 (A), 7 to 11 (B), 7 to 18 (C). (D) Aβ plaques in the cortex and hippocampus of 3 ~ 4 months old transgenic mice. β-amyloids were indicated by Thioflavin-S staining. Scale bar: 200 μm. (E) Aβ plaques in the cortex and hippocampus of 6 ~ 7 months old APP/PS1 transgenic mice. (a) β-amyloids were indicated by Thioflavin-S staining in mice. (b) Integrated optical density (IOD), area, and numbers of Aβ plaques in the indicated area marked with the red dotted line. n = 4. (F) Aβ plaques in 12 ~ 13-month-old transgenic mice. β-amyloid indicated by DAB staining with 6E10 antibody (a) and Thioflavin-S staining (b) in cortex and hippocampus of 12 ~ 13-month-old mice. Scale bar: 200 μm (a) and 500 μm (b). (c) Integrated optical density (IOD), area and numbers of Aβ plaques in indicated area marked with red dotted line in Thioflavin-S staining slices were analyzed. n = 6. (G) The expression levels of multiple proteins involved in the APP pathway were detected with Western Blot in the cortex (a) and hippocampus (c) of 12 ~ 13-month-old APP/PS1 and Emx1-CKO-ad mice. β-actin served as control. (b & d) The relative protein levels. n = 3. (Unpaired t-test).

At 6 months, the four genotypes of mice showed no difference in locomotor activity. The Emx1-CKO-AD mice even spent more time in the central area in the OFT arena (Supplementary Material, Fig. S4C). In the MWM test, Emx1-CKO mice behaved similarly to control mice, while APP/PS1 mice exhibited impaired learning and memory with higher escape latencies on the third, fifth and sixth day of the training phases. However, Emx1-CKO-AD mice had prolonged escape times only on the seventh day. Although APP/PS1 and Emx1-CKO-AD mice displayed reduced platform crossings in the single probe test, they spent a similar amount of time in the target quadrant (Fig. 3B). All experimental groups of 6-month-old mice showed similar performances in the Y-maze (Supplementary Material, Fig. S4A, c and d). However, in the NOR, APP/PS1 mice exhibited less exploration of the novel object, and had a smaller discrimination index, while performances of Emx1-CKO and Emx1-CKO-AD mice did not differ from control mice (Supplementary Material, Fig. S4B, a–c).

At 12 months, Emx1-CKO-AD displayed higher locomotor ability compared with control, Emx1-CKO and APP/PS1 mice in the OFT; however, the mice of the four genotypes did not differ in their anxiety levels as indicated by the similar sum time in the center (Supplementary Material, Fig. S4D). In the MWM test, longer escape latency was detected in Emx1-CKO mice on the sixth day of the training test compared to control mice (Fig. 3C). For the overall training time, APP/PS1 and Emx1-CKO-AD took a significantly longer escape time compared with the control mice. In the probe test, CKO, APP/PS1 and Emx1-CKO-AD mice spent similar time in the target quadrant compared with control mice. However, the platform crossings of the other three genotypes of mice decreased markedly compared with control mice. The MWM performance of Emx1-CKO, APP/PS1 and Emx1-CKO-AD indicated impaired learning and memory (Fig. 3C). Similarly, compared with control mice, AD and Emx1-CKO-AD mice showed fewer alternations in the Y-maze test. Notably, Emx1-CKO-AD mice showed more entries into the Y-maze arms than control and Emx1-CKO mice, indicating hyperactivity of the 12-month-old Emx1-CKO-AD mice (Supplementary Material, Fig. S4A, e and f). There were no significant changes among all four groups of 12-month-old mice in the NOR test (Supplementary Material, Fig. S4B, d–f).

Regarding the effects of NRSF/REST deficiency on Aβ, additional experiments were conducted. We examined the level of Aβ loads in the transgenic mice aged from 3 to 12 months by Thioflavin-S staining. Aβ plaques were very spare in the cortex and hippocampus of 3-month-old APP/PS1 mice. Aβ plaques in age-matched Emx1-CKO-AD mice were also detected (Fig. 3D). In the brain of 6-month-old APP/PS1 mice, considerably more Aβ plaques were identified (Fig. 3E, a). Interestingly, reduction of NRSF/REST in APP/PS1 mice did not influence Aβ plaque appearance in the cortex or the hippocampus (Fig. 3E, b). Aβ plaques in the brain of 12-month-old APP/PS1 mice were examined by immunohistochemistry (Fig. 3F, a) and Thioflavin-S staining (Fig. 3F, b). Both approaches revealed similar amounts of plaques, but the NRSF/REST loss did not aggravate nor mitigate the plaque burden in APP/PS1 mice (Fig. 3F, c).

Further, the expression level of the Aβ precursor APP, β-secretase (BACE-1) and C-terminal fragments (CTFs) proteins were examined in the cortex and the hippocampus. Compared with control mice, NRSF/REST conditional knockout did not affect APP or BACE-1 expression in 3- to 6-month-old Emx1-CKO mice. However, in the brain of APP/PS1 mice with or without NRSF/REST, the expression of APP was remarkably enhanced, indicating that NRSF/REST does not contribute detectably to the altered APP expression in APP/PS1 mice. In addition, the levels of BACE-1 did not vary in the cortex and hippocampus of the four mouse groups (Supplementary Material, Fig. S4E and F).

Similar to the expression pattern of APP, CTFs were barely detected in the brain of control and Emx1-CKO mice but were highly expressed in 6- to 7-month-old APP/PS1 mice with or without NRSF/REST (Supplementary Material, Fig. S4F). We did not detect differences in the protein levels of APP and BACE-1 between 6- to 12-month-old APP/PS1 and Emx1-CKO-AD mice (Fig. 3G, Supplementary Material, Fig. S4F).

Mice with a neuronal NRSF/REST deficiency were crossed with APP/PS1 transgenic mice to study the impact of NRSF/REST reduction on AD-like pathological behaviors. Nine-month-old mice in the groups of control, NSE-CKO, APP/PS1 and NSE-CKO-AD showed no detectable differences in the MWM, Y-maze and FC tests (Supplementary Material, Fig. S5A–C) from each other. At 12 months of age, the four genotypes of mice exhibited similar locomotor ability, and AD mice even spent more time in the center compared to control, NSE-CKO and NSE-CKO-AD mice in the OFT, implying that there was no prominent difference in the levels of stress/fear/anxiety (Supplementary Material, Fig. S5D). However, in the MWM, mice in groups of NSE- CKO, APP/PS1 and NSE-CKO-AD showed impaired learning and memory, indicating a higher latency to find the platform during training sessions and fewer crossings in the single probe trial (Fig. 4A). In the FC, NSE-CKO-AD mice had worse performance (Fig. 4B). The discrimination index values were similar for all four groups in the NOR test (Fig. 4C). Regarding the amyloid loads in 9- to 18-month-old AD and NSE-CKO-AD mice, the deposits of Aβ in the cortex and the hippocampus remained elevated and did not differ between groups (Fig. 4D). Finally, addressing modulations of synaptic transmission and plasticity by electrophysiology, the slope of hippocampal fEPSPs (field excitatory post-synaptic potentials, fEPSP) in control and NSE-CKO mice were comparable (Fig. 4E, a). A difference between the LTP outcomes of AD and NSE-CKO-AD mice was not detectable (Fig. 4E, b). Collectively, absence of NRSF/REST in dorsal telecephalon or neurons has no prominent impact on the canonical AD pathologies including Aβ loads or memory-related behavioral performances of APP/PS1 mice aged from 3- to 12-month-old mice.

Figure 4

The effects of neuronal NRSF/REST deficiency on Aβ deposition, behaviors and synaptic plasticity in 9–18-month-old mice. (A) Behavior performances in the MWM test. (a) Latency to the platform during training. (b) Percentage of time spent in the target quadrant. (c) Times of platform crossing. n = 8 ~ 12. (B) The results of the FC test, n = 10, 12, 3, and 5. (C) The results of the NOR test. n = 2 ~ 10. (D) Aβ deposition. Scale bar: 1 μm. (a) Thioflavin-S staining. (b, c) Percentages of Thioflavin-S positive area in the cortex (b) and the hippocampus (c) in the indicated area marked with the red dotted line. The representative Aβ plaques were indicated with white arrows. For the 9–14-month-old mice, n = 6; for the 18-month-old mice, n = 4. (E) Basal synaptic transmission and synaptic plasticity of hippocampal slices of 12-month-old mice. (a) Input/output (I/O) relationship obtained by plotting the fEPSP slope in the CA1 area of the hippocampus of Control and NSE-CKO mice as a function of the stimulus intensity (from 0 to 80 μA). (b) LTP in hippocampal slices. Data dots represent fEPSP slopes normalized to their respective averaged baseline values recorded before high-frequency stimulation. n = 10 & 8 respectively for Control and NSE-CKO mice. Statistical analyses: One-way or Two-way ANOVA with Fisher’s LSD multiple comparison test post hoc (A-C) and Student t test (D-E). ^*P < 0.05, ^**p < 0.01.

NRSF/REST deficiency attenuated synaptic plasticity in 12-month-old Emx1-CKO mice but not in APP/PS1 mice

We described before the effects of NRSF/REST deficiency in neurons on synaptic transmission and plasticity. Here, we highlight and compare data obtained by electrophysiological experiments with Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice. We were eager to learn if NRSF/REST deficiency in a different set of cell types (neurons and astrocytes) interferes with APP/PS1-like pathologies. To this end, we studied the level of synaptic transmission in the hippocampal CA1 area of 12-month-old mice. Compared with control mice, the slope and amplitude of fEPSPs were reduced significantly in the hippocampus of Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice. The slope and amplitude of fEPSPs of the three transgenic groups did not differ from each other (Fig. 5A). PPF of synaptic transmission was analyzed and revealed that the paired-pulse ratio of the slope and amplitude was reduced in Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice at stimulation intensities of 30 and 45% of the maximum slope (Fig. 5B) compared with WT mice. However, differences among the three genotypes were not detectable. LTP was impaired in Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice and did not differ among the three genotypes (Fig. 5C).

Figure 5

Synaptic plasticity of the hippocampal slices in 12-month-old Control (Ctrl), Emx1-CKO (KO), APP/PS1 (ad) and Emx1-CKO-ad (KA) mice. (A) Basal synaptic transmission of hippocampal slices. (a) Respective fEPSP traces per group (Ctrl: 9 slices from 3 mice; KO: 9 slices from 3 mice; ad: 7 slices from 3 mice; KA: 9 slices from 3 mice, Scale bar: 1 mV/5 ms). Input/output (I/O) relationship obtained by plotting the fEPSP slope (b) and peak amplitude (c) in the CA1 area of the hippocampus as a function of the stimulation intensity (from 0 to 5 V). Brackets indicate the interval of significant differences between two groups (pairs of the circle at the beginning of the respective bracket indicate compared groups). (B) Short-term plasticity in hippocampal slices. (a & b) Traces of representative fEPSPs per group. Scale bar: 1 mV/5 ms. The paired-pulse ratio of slope (c & d) and the paired-pulse ratio of amplitude (e & f) were recorded across different inter-stimulus intervals (0 ~ 300 ms). The paired-pulse facilitation was observed in the hippocampus as a function of the stimulation intensity 30% of maximum slope (c & e) and 45% of maximum slope (d & f). (C) Impairments of LTP in hippocampal slices. LTP was evoked by three high-frequency stimulation (100 Hz/5 min). (a) Data dots represent fEPSP slopes normalized to their respective averaged baseline values recorded before high-frequency stimulation. (b) fEPSPs recorded before and after high-frequency stimulation are shown (Scale bar: 1 mV/5 ms). Brackets and asterisks: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 (Student unpaired t-test). (D) Protein levels of GluR2, SAP102 and PSD95 in the cortex of 12-month-old transgenic mice. β-actin and GAPDH served as the control, n = 3. (E) Protein levels of GluR2, SAP102 and PSD95 in the hippocampus of 12-month-old transgenic mice. β-actin and GAPDH served as control unpaired t-test. n = 3 ~ 9. ^*P < 0.05).

To complement the electrophysiological analysis, the expression levels of postsynaptic protein SAP102 and GluR2 were tested. The expression levels of both proteins remained comparable in all tested groups and were not altered by NRSF/REST deficiency, whether in the cortex or the hippocampus. However, a reduction of PSD95 expression levels was observed in the hippocampus of 12-month-old Emx1-CKO mice (Fig. 5D and E). Collectively, absence of NRSF/REST in dorsal telecephalon incuces synaptic impairments in 12-month-old mice but not in APP/PS1 mice.

NRSF/REST deficiency upregulated the activation of astrocytes and enhanced C3 expression in 6- and 12-month-old Emx1-CKO-AD mice

Immunohistochemical analyses were conducted to study the impact of NRSF/REST deficiency in neurons and astrocytes on processes related to inflammation. To this end, astrocytes and microglial cells in the four genotypes of mice at 3 to 12 months of age were examined. At 3 months of age, the astrocytes in the cortex and hippocampus of APP/PS1 mice did not differ from control mice. In CKO and Emx1-CKO-AD mice, NRSF/REST deficiency enhanced the GFAP expression of astrocytes within cortical areas above the corpus callosum and the molecular layer of the hippocampus (Supplementary Material, Fig. S6A). The enhanced GFAP expression was confirmed by western blot analysis (Supplementary Material, Fig. S6F). Microglial cells showed no difference among the four genotypes of mice in both the cortex and hippocampus (Supplementary Material, Fig. S6B).

At 6 months of age, Aβ plaques were observed to be surrounded by the GFAP-positive astrocytes and Iba1-positive microglial cells in the cortex and hippocampus of APP/PS1 mice with or without NRSF/REST deficiency (Fig. 6A, Supplementary Material, Fig. S6C). In addition, the cortical area near the corpus callosum exhibited a significant increase in the number of astrocytes in the Emx1-CKO and Emx1-CKO-AD mice, but not in APP/PS1 mice (Fig. 6A, a and c). However, the number of hippocampal astrocytes in Emx1-CKO-AD mice was elevated compared to control and APP/PS1 mice (Fig. 6A, b and d). Despite the clustered cells around the plaques in the brain of APP/PS1, and Emx1-CKO-AD mice, the microglial cells in Emx1-CKO mice remained similar to the control mice (Supplementary Material, Fig. S6C). Cortical GFAP protein levels were markedly enhanced in Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice. Especially, Emx1-CKO-AD mice had a higher expression of GFAP proteins in the cortex compared to APP/PS1 mice, suggesting enhanced NRSF/REST deficiency promoting cortical GFAP protein expression in 6-month-old APP/PS1 mice. The hippocampal GFAP protein levels were significantly higher in Emx1-CKO-AD mice compared with the control mice (Fig. 6B). Astrocytes consist of A1 and A2 phenotypes, in which A1 astrocytes are recognizable by the expression of the complement 3 (C3). C3 has been linked with the induction of aggravated inflammation (27). By C3 staining and quantification, we observed that the number of C3-positive A1 astrocytes was increased significantly in the cortical area near the corpus callosum of 6-month-old Emx1-CKO and Emx1-CKO-AD mice compared with control and APP/PS1 transgenic mice. An increase in the number of C3-positive astrocytes was only detected in the hippocampus of Emx1-CKO-AD mice (Fig. 6A, c and d).

Figure 6

Persistent increase of astrocytes and C3 complement in the dorsal telencephalon NRSF/REST deficiency CKO mice. (A) Aβ plaques and the surrounding astrocytes in the cortex (a) and hippocampus (b) of 6 ~ 7-month-old control and transgenic mice. Astrocytes were double-stained with GFAP (green) and complement C3 (red) while β-amyloid were indicated in 6E10 (white) and the nucleus was indicated in DAPI (blue). The pictures depicted in the red square frame were enlarged and shown in Zoom. Scale bar: 200 μm. (c, d) The numbers of astrocytes in the cortex and hippocampus. Protein levels of GFAP in the cortex (a) and hippocampus (b) of 6 ~ 7-month-old (B) and 12-month-old (C) transgenic mice. β-actin served as the control. Statistical analyses of the protein expression have been summarized in the bar graphs. (D) Astrocytes in the cortex (a) and hippocampus (b) of 12-month-old transgenic mice. Astrocytes were double-stained with GFAP (Green) and complement C3 (Red). (c, d) The numbers of astrocytes in the cortex and hippocampus. Scale bar: 50 μm. (E) Astrocytes and Aβ plaques in the cortex of 12-month-old APP/PS1 mice with or without NRSF/REST deficiency. Astrocytes were double-stained with GFAP (green) and complement C3 (red) while β-amyloid was indicated in 6E10 (white) in the cortex. Scale bar: 50 μm. All statistical analyses were performed with One-way ANOVA with Fisher’s LSD multiple comparison test post hoc, n = 3. ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

At 12 months of age, the level of GFAP proteins in the cortex of Emx1-CKO mice was markedly upregulated compared with control mice (Fig. 6C). Moreover, compared with control and Emx1-CKO mice, both the cortical and hippocampal GFAP protein levels were significantly higher in APP/PS1 mice, and the expression of GFAP protein was further elevated in the cortex of Emx1-CKO-AD mice (Fig. 6C). In an overall view, the number of C3-positive A1 astrocytes increased in the cortex and hippocampus of Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice compared to control mice (Supplementary Material, Fig. S6D). C3-positive astrocytes and whole astrocytes in the cortical area near the corpus callosum and the molecular layer of the hippocampus were analyzed (Fig. 6D). The number of cortical astrocytes in Emx1-CKO and Emx1-CKO-AD was increased compared with control and APP/PS1 mice (Fig. 6D, c). The number of C3-positive astrocytes in the other three genotypes of mice compared with control mice increased significantly. In addition, the C3-positive astrocyte number in Emx1-CKO-AD mice was higher than in APP/PS1 mice (Fig. 6D, c). In the molecular layer of the hippocampus, the number of GFAP-positive cells remained similar in all the experimental groups. However, compared with control mice, there were more C3-positive astrocytes in the other three groups of mice. Consistent with the results in the cortex, the number of C3-positive astrocytes in the hippocampus of Emx1-CKO-AD mice was also higher than that of APP/PS1 mice (Fig. 6D, d). Besides the number, we also analyzed the morphological properties of C3-positive A1 astrocytes surrounding the Aβ plaques in 12-month-old mice. We observed that the morphological parameters of A1 astrocytes surrounding the plaques did not vary between APP/PS1 and Emx1-CKO-AD mice (Fig. 6E). Cerebral Iba1-positive microglial cells were detected by DAB staining. Clustered microglial cells in the cortex and hippocampus of APP/PS1 and Emx1-CKO-AD mice exhibited no difference (Supplementary Material, Fig. S6E). Collectively, NRSF/REST deficiency upregulates the activation of astrocytes in Emx1-CKO-AD mice.

Upregulation of C/EBPβ correlates with NRFS/REST deficiency-induced C3 activation in 12-month-old Emx1-CKO mice

Complement C3 is a central component in the complement system (28). Astrocytic release of complement C3 can be activated by NFκB p65 signaling (29). We observed that the NFκB p65 subunit was expressed in the cytoplasm of neurons, and only a few signals were detectable in nuclei (Supplementary Material, Fig. S7A). NRSF/REST loss did not alter the subcellular localization of p65 in the brains of 12-month-old transgenic mice. C3 expression is known to be directly regulated by the CCAAT/enhancer-binding protein β (C/EBPβ) (30). Thus, we detected the C/EBPβ expression in the cortex and hippocampus of 12-month-old mice by immunofluorescence staining (Fig. 7A, a and b). Compared with control mice, the number of C/EBPβ-positive cells in the cortex increased significantly in the other three mouse groups. Moreover, the numbers of C/EBPβ-positive cells in the hippocampus of APP/PS1 and Emx1-CKO-AD mice were markedly elevated compared to control mice. However, the numbers of C/EBPβ-positive cells were similar in the cortex and hippocampus of Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice (Fig. 7A, c and d).

Figure 7

Changes of upstream and downstream regulators of NRSF/REST in 12-month-old transgenic mice. (A) Staining of C/EBPβ in the cortex (a) and hippocampus (b) of 12-month-old transgenic mice. C/EBPβ was totally expressed in the nucleus indicated with DAPI. Scale bar: 200 μm. (c, d) Statistics for the numbers of C/EBPβ-positive cells in the cortex (c) and hippocampus (d). n = 3–4. Statistical analyses: One-way or ANOVA with Fisher’s LSD: ^*P < 0.05 and ^**P < 0.01. (B-E) Western blotting of MAPK signaling pathway including p38 (B), JNK (C) and ERK (D) and Tau (E) proteins in the 12-month-old mice. β-actin served as the control. Statistics of the protein levels were shown at the right panel. n = 3. Statistical analyses: unpaired t test. ^*P < 0.05 P < 0.05 and ^**P < 0.01.

Genes involved in the MAPK signaling containing the RE-1 element are targets of NRSF/REST (31) and activate C/EBPβ. We observed that MAPK/p38 activity indicated by the ratio of phosphorylated-p38 to total p38 proteins increased significantly in the cortex but not in the hippocampus of 12-month-old Emx1-CKO mice compared with control mice (Fig. 7B).

Protein kinases, including JNK, mediate Tau phosphorylation (32). The phosphorylation of JNK and phosphorylated-JNK and JNK protein ratio dramatically decreased in the hippocampus but not in the cortex of CKO mice (Fig. 7C). Phosphorylated ERK was barely detectable in the cortex and did not differ between the hippocampi of experimental mice (Fig. 7D).

Another hallmark of AD pathology is the enhanced expression level of the Tau protein and its phosphorylation. Surprisingly, we observed a significant increase in Tau phosphorylation in the cortex but not in the hippocampus of 12-month-old Emx1-CKO mice compared with the control mice (Fig. 7E).

Cerebral NRSF/REST deficiency was reported to accelerate apoptosis in the brain of 8-month-old mice (19). Here, the apoptotic cells in the cortex and hippocampus of 12-month-old mice were detected by TUNEL (Terminal-deoxynucleoitidyl Transferase Mediated Nick End Labeling) assay and cleaved caspase-3 staining. The results revealed no detectable difference in the cortex and hippocampus between control and Emx1-CKO mice (Supplementary Material, Fig. S7B). Apoptosis-related proteins were further determined by western blot. The protein levels of Bcl2 and Bax and their ratio did not differ between control and Emx1-CKO mice (Supplementary Material, Fig. S7C). Therefore, deficiency of NRSF/REST in dorsal telecephalon leads to an upstream increase of C/EBPβ and the overphosphorylation of MAPK/p38 and Tau in the cortex of 12-month-old mice, which may account for the underlying mechanisms of the brain injuries.

Inflammation and senescence in over 20-month-old Emx1-driven NRSF/REST-deficient APP/PS1 mice

We presented data concerning mice at 3 to 12 months of age that correlates with 20- to 40-year-old humans. However, aging is a defined pathological risk factor in the onset of AD (4). Thus, we studied 20-month-old mice that correlate with humans about 60 years of age (Jackson Laboratory, Bar Harbor, ME, USA).

We applied similar sets of methods like behavioral, cellular and protein biochemical analysis. We observed that the mice of the four genotypes at 20 months of age were characterized by similar behaviors in the OFT (Supplementary Material, Fig. S8A). Analysis of the spatial memory by the Y-maze test revealed that APP/PS1 and Emx1-CKO-AD mice exhibited a slightly decreased alternation compared with control mice (P = 0.063 and P = 0.056, versus control, respectively, Fig. 8A, a). In the NOR test, the short-term memory remained similarly among the four groups (Fig. 8A, b).

Figure 8

The effects of telencephalon NRSF/REST deficiency on Aβ deposition, behavior, protein levels, inflammatory molecules and senescence markers in 20–22 months old mice. (A) The results of behavior tests. (a) Y-maze test. Total entries and percentage of alteration were shown. (b) NOR test. Short-term (5 min) distinguish indexes were shown. n = 7 ~ 14. (B) Thioflavin-S staining. Scale bar: 1 μm. The representative Aβ plaques in the indicated area marked with the red dotted line were indicated by white arrows. (C) APP, phosphorylated Tau and synaptophysin protein expression in the cortex (a, c, e) and hippocampus (b, d, f) of mice. n = 3–9. (D) GFAP and Iba1 protein expression in the cortex and hippocampus of mice. n = 3–9. (E) The mRNA expression of BDNF, C3, IL1β, TNFα and Pai-1 in the cortex (a) and hippocampus (b) of mice. n = 5–6. (F) The mRNA expression of p19, p21 and BubR1 in the cortex (a) and hippocampus (b) of mice. n = 4–6. Statistical analyses: One-Way ANOVA with Fisher’s LSD: ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

Aβ plaques in the brain of 22-month-old mice were examined by Thioflavin-S staining (Fig. 8B). There was no detectable difference between APP/PS1 and Emx1-CKO-AD mice. The astrocytes increased in the cortex and the hippocampus of Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice compared with the control (Supplementary Material, Fig. S8B). Moreover, microglial cells displayed a similar tendency (Supplementary Material, Fig. S8C). APP, phosphorylated-Tau (p-Tau) and synaptophysin expressions were studied by western blot assays. We observed that the protein levels of APP in the cortex and hippocampus of APP/PS1 and Emx1-CKO-AD mice were markedly increased (Fig. 8C, a and b). Phosphorylated-Tau proteins among four groups of mice were expressed in a similar manner (Fig. 8C, c and d). The cortical synaptophysin protein level in Emx1-CKO-AD mice was significantly lower than in control and APP/PS1 mice. However, the synaptophysin protein level in the hippocampus did not differ among experimental groups (Fig. 8C, e and f).

Furthermore, glial activation was examined in 20-month-old mouse brains. Western blot analysis revealed an increase in GFAP protein expression in the cortex and hippocampus of CKO, APP/PS1 and Emx1-CKO-AD mice compared with the Control mice (Fig. 8D, a and b). Regarding the expression level of the microglia marker, we observed that the expression of Iba1 protein was markedly upregulated in the cortex of APP/PS1 and Emx1-CKO-AD mice compared with control mice. Differences in the Iba1 protein level in the hippocampus among the four transgenic mouse groups were not detectable (Fig. 8D, c and d).

Quantitative PCR analyses were conducted to learn if the transgenic mouse groups differ in the transcription of some genes involved in memory and inflammation. The transcripts of BDNF and inflammatory molecules like C3, IL-1β, TNF-α and PAI-1 were assessed. BDNF levels in the cortex and hippocampus of Emx1-CKO-AD mice were downregulated compared with control mice. In addition, BDNF transcripts were also reduced in the hippocampus of Emx1-CKO mice. C3 transcripts were increased in the cortex of APP/PS1 compared with the control. IL-1β expression levels were significantly elevated in the cortex of APP/PS1 and Emx1-CKO-AD mice. In addition, IL-1β transcripts increased in the hippocampus of Emx1-CKO and APP/PS1 mice. The expression of TNF-α was elevated only in the hippocampus of Emx1-CKO-AD mice compared with control mice. PAI-1 transcripts increased in the hippocampus of APP/PS1 mice (Fig. 8E).

Furthermore, the relations between NRSF/REST and cell senescence were evaluated. Utilizing β-gal staining and quantification to monitor cell senescence, we showed that the number of senescent cells was increased in the perirhinal cortex of APP/PS1 and Emx1-CKO-AD mice compared with control mice, whereas in the CA3 regions, senescent cells were slightly elevated in Emx1-CKO-AD mice (P = 0.088, versus control) (Supplementary Material, Fig. S8D and E). The transcripts of aging-related molecules p19, p21 and Bub1b were determined. The expression of p19 was increased dramatically only in the cortex of Emx1-CKO-AD mice. Expression levels of p19 in the hippocampus, p21 and Bub1b in the cortex and hippocampus did not differ between groups (Fig. 8F). Collectively, in the over 20-month-old mice, deficiency of NRSF/REST in dorsal telecephalon shows no impact on the Aβ loads and memory performances of APP/PS1 mice, but induces more remarkable increase of cell senescence and aging-related molecule p19 expression in the cortex.

Aged WT and AD mice overexpressing NRSF/REST show different changes in cognitive abilities

NRSF/REST deficiency in aged mice impaired cognitive abilities. However, an additional reduction of NRSF/REST in APP/PS1 transgenic mice did not restore detectable alterations of AD-like pathologies. To characterize if a gain of function of NRSF/REST alters the cognition abilities of WT and AD mice, NRSF was overexpressed in the hippocampus utilizing the AAV expression system. The AAV-GFP or AAV-NRSF viral particles were injected into the hippocampus. GFP or NRSF overexpression (NRSF OE) was confirmed by western blotting, immunohistochemistry and fluorescence imaging (Supplementary Material, Fig. S9A–C).

Mouse behaviors were tested 2 weeks after injection of AAV-NRSF. In the novel location recognition test, 7-month-old WT and AD mice overexpressing NRSF displayed impaired performance compared to their respective GFP-overexpression controls. However, at the age of 15 months, AAV-NRSF-injected WT performed significantly better compared with control GFP-overexpressing WT mice. Overexpression of NRSF in 15-month-old AD mice did not cause detectable improvements compared with AAV-GFP-injected AD mice (Supplementary Material, Fig. S10A,a and B,a). Notably, overexpression of NRSF increased total exploration in 7-month-old WT, but not in 15-month-old WT mice, neither in 7- nor 15-month-old APP/PS1 mice (Supplementary Material, Fig. S10A,b and B,b).

In the NOR test, 15-month-old AAV-GFP-injected WT mice performed worse than 7-month-old AAV-GFP-injected WT mice. AAV-GFP-injected APP/PS1 mice at the age of 7 months and 15 months performed similarly. Fifteen-month-old WT mice with NRSF OE showed better performance. A similar effect was not detected in 7-month-old WT mice, and 7-month- or 15-month-old AD mice (Supplementary Material, Fig. S10C, a and D, a). Similarly, overexpression of NRSF increased total exploration in 7-month-old WT, but not in 15-month-old WT mice and 7- or 15-month-old APP/PS1 mice (Supplementary Material, Fig. S10C, b and D,b).

Moreover, in the Y-maze test, AAV-GFP- or AAV-NRSF-injected WT and APP/PS1 mice at the age of 7 or 15 months had no difference in the alternation percentage. However, overexpression of NRSF increased total entries in 7-month-old WT mice (Supplementary Material, Fig. S10E and F).

At 3 months after AAV viral particle injection, Aβ deposits in the hippocampus and the cortex of APP/PS1 mice were measured by Thioflavin-S staining. Aβ plaques showed no remarkable difference in both ages of mice (Supplementary Material, Fig. S11). Therefore, overexpression of NRSF/REST in the hippocampus improvesthe memory performances of 15-month-old mice but not in APP/PS1 mice. The Aβ loads of APP/PS1 mice are not impacted by NRSF/REST overexpression.

Discussion

NRSF/REST is critically involved in the maintenance of neuronal functionality. Therefore, downregulation of NRSF/REST in mature neurons has severe consequences for cognitive abilities (33). Interestingly, NRSF/REST re-acquires enhanced expression levels in the frontal cortex during aging, seemingly compensating for an age-depend decline of cognitive abilities by altering the expression of genes involved in cell death, inflammation and stress (19).

In this study, 12-month-old mice with NRSF deficiency in the telencephalon displayed attenuated fear and spatial memory performances in the FC test and Y-maze tests (Fig. 1), respectively. Decreases in fear, spatial and novel object recognition memories were detectable in neuron-specific NRSF/REST knockout mice at 12 months of age in the FC test, MWM test and NOR tests (Figs 2 and 4), respectively. We observed that the expression of NRSF/REST in the cortex of 12-month-old APP/PS1 mice decreased significantly compared with age-matched control mice.

Enhanced NRSF/REST deficiency in APP/PS1 (dorsal telencephalon) mice aged 3 to 22 months had no detectable impact on the appearance or number of Aβ plaques as well as crucial proteins of Aβ processing pathways in comparison with APP/PS1 mice. These findings correlated with the absence of detectable cognitive deficits in Emx1-CKO-AD compared with APP/PS1 mice. NRSF/REST deficiency in neurons of 9- to 18-month-old APP/PS1 mice did not alter Aβ plaque appearance in native APP/PS1 mice. Although induced NRSF/REST deficiency in WT mice had a significant effect on synaptic transmission, memory performance and other cellular parameters, it did not contribute to most of the AD-like pathologies of APP/PS1 mice.

Notably, only 12-month-old CKO mice showed impaired learning and memory, which suggests that NRSF/REST is indispensable for normal brain functions at this specific stage. The model of LTP reflects the ability of synaptic plasticity and might be an underlying mechanism of learning and memory (34). NRSF/REST is essential for the experience-dependent fine-tuning of genes involved in synaptic plasticity (35). The synaptic transmission, PPF and LTP of 12-month-old Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice were significantly declined compared with the controls. However, there was no significant difference between the two genotypes of APP/PS1 and Emx1-CKO-AD mice.

Thus, we concluded that NRSF/REST deficiency reduced synaptic plasticity and learning and memory in 12-month-old mice. Lu et al. (19) reported that aging individuals with NRSF/REST highly expressed kept rather well cognition and longevity despite the AD pathological changes observed in their brain. Therefore, it could be that NRSF/REST protects the cognitive functions by regulating inflammation, cell deaths and other biological processes in aged mice, which do not work on mitigating Aβ production in AD pathology.

Glial cells play critical roles in the central nervous system and neurodegenerative diseases (36–38). Astrocytes, especially A1-type astrocytes marked by complement C3, are neurotoxic to cortical neurons. A1-type astrocytes have been also observed in the brains of patients with neurodegenerative diseases (27). Sustained activation of astrocytes has been commonly detected in several lines of NRSF/REST conditional deficient mouse strains including neuron-specific knockout mice (7), brain-specific knockout mice (39) and dorsal telencephalon knockout mice generated in this study. Astrocytes without NRSF/REST have a higher capacity response to inflammatory stimuli (40). Here, C3-positive A1 astrocytes significantly increased in the forebrain of 6- and 12-month-old Emx1-CKO mice. Therefore, the responses between the increased microglial cells and A1 astrocytes in the cortex of 12-month-old Emx1-CKO mice could orchestrate the dysfunction of C3 complement and thus contribute to memory deficits. Since NRSF/REST deficiency did not detectably affect the decline of memory or synaptic plasticity in APP/PS1 mice, we speculated that NRSF/REST-deficiency induced C3-mediated neurotoxicity bypassed the effects of Aβ.

The changes of transgenic mice in comparison with controls (Emx1-CKO versus Control and Emx1-CKO-AD versus APP/PS1) at the age of 3–22 months were summarized in Supplementary Material, Table S4. Generally, NRSF/REST deficiency barely interfered with the Aβ production in APP/PS1 mice and other AD-like pathologies. Nevertheless, NRSF deficiency in the dorsal telencephalon or neurons binds up with the age-dependent brain injuries indicated by impaired memory functions and abnormal glial cells (Supplementary Material, Table S4).

We observed that NRSF/REST deficiency induced memory dysfunction in the 12-month-old mice. NRSF/REST regulates the expression of a large set of synaptic proteins (41), suggesting that the absence of NRSF/REST might contribute to the synapse injury. Therefore, the reduced PSD95 protein expression in the hippocampus might have contributed to the impairments of hippocampal LTP of 12-month-old Emx1-CKO mice. However, the expression of PSD95 did not differ in the cortex of Control and Emx1-CKO mice. Thus, the absence of NRSF/REST exerts region-specific impacts on synaptic protein expression and functions. C/EBPβ and C3 expressions increased similarly in the cortex of 12-month-old Emx1-CKO mice. The enhancer-binding factor C/EBPβ directly regulates the C3 expression (30). In addition, C/EBPβ regulates gene expression involved in inflammation and cerebral injuries (42). Thus, upregulation of C/EBPβ coordinated with an increased C3 expression might be attributable to neuronal deficits in 12-month-old CKO, APP/PS1 and Emx1-CKO-AD mice.

Tau phosphorylation in the brain is a sensitive response to various exogenous and endogenous stresses (43–46). Neurofibrillary tangles containing over-phosphorylated Tau proteins have been linked to neuronal degeneration and cognitive impairments (47,48). A1 astrocytes release inflammatory cytokines like IL-1 which could induce abnormal phosphorylation of Tau (27). Depletion of C/EBPβ repressed tau expression while its overexpression accelerated Tau phosphorylation in 3xTg mice (49,50). Consistent with increased C/EBPβ and C3 expressions, Tau phosphorylation was significantly enhanced in the cortex of 12-month-old Emx1-CKO mice in our experiments. Protein kinases, including MAPK, especially JNK, mediate Tau phosphorylation (51). We detected that the phosphorylation of JNK (Thr183/Tyr185) significantly decreased in the hippocampus of 12-month-old Emx1-CKO mice, which might eliminate Tau phosphorylation. On the other hand, phosphorylation of p38 increased significantly in the cortex of 12-month-old Emx1-CKO mice. As p38 activation upregulates C/EBPβ and C3 (52), an increase in p38 activity might result in an enhanced C/EBPβ/C3/p-Tau signaling and contribute to brain injuries induced by NRSF/REST deficiency in 12-month-old Emx1-CKO mice.

At the very late stage of aging and AD, higher expression of inflammatory molecules and lower expression of BDNF were detected in the brain of transgenic mice. Although mice showed approximately similar behavioral performance independent of genotypes, which indicates that aging could be the most important determinant factor, we still observed the reduced synaptophysin protein, enhanced p19 expression and elevated senescent cells in the telencephalon of NRSF/REST deficiency AD mice. However, the underlying mechanism linking the appearance of senescent cells and cognitive decline warrants further studies.

Moreover, compared with the control mice, enhanced GFAP protein levels in the forebrain were observed in 22-month-old Emx1-CKO, APP/PS1 and Emx1-CKO-AD mice. However, the four genotypes of mice exhibited similar performances in learning and memory, indicating that inflammation might not be closely related to memory behaviors at the very late stage of aging.

As we have shown, 6- to 7-month-old WT and AD mice express similar levels of NRSF/REST. Overexpression of NRSF/REST at these ages attenuated the performance of mice in both groups, indicating that the balance of NRSF expression level needs to be maintained. Age-dependent decline of cognition ability in WT mice was observed in the NOR test. However, NRSF OE had a supportive effect on the cognitive ability of 15-month-old WT, but not of AD mice. Thus, the NRSF pathway seems to be widely bypassed in AD mice because additional NRSF reduction or NRSF OE in aged AD mice did not alter in a detectable manner of AD-like pathologies in the behavioral tasks. Other dominant factors like changes in the oxidative and inflammatory microenvironments might represent drivers for the onset of AD. Thus, our data are not entirely consistent with previous publications showing that insufficient NRSF/REST activity is well connected with AD pathologies (19).

In this study, we presented data showing that NRSF/REST deficiency caused alterations of A1 astrocytes, along with an increase of C/EBPβ and Tau over-phosphorylation. The p38/C/EBPβ/C3/p-Tau cascade in the cortex of 12-month-old Emx1-CKO mice offered a paradigm for NRSF/REST deficiency-induced brain injuries. The unbalanced network of the synaptic proteins, together with the increase of neurotoxic astrocytes and microglial cells, might account for NRSF/REST deficiency induced impairments of cognitive abilities and impaired synaptic plasticity in the 12-month-old Emx1-CKO and NSE-CKO mice. In addition, overexpression of NRSF/REST in aged WT mice showed a beneficial effect on cognitive ability. Moreover, our data demonstrated that except the hyperactivation of astrocytes, NRSF/REST absence in APP/PS1 mice had no prominent impact on the Aβ deposits, and the electrophysiological and cognitive AD-like pathologies. However, there exists some limitations in the study. The observed mechanisms of NRSF deficiency leading to age-dependent aggravated recognition disability and the role of NRSF in AD pathogenesis require further studies to combine into one molecular framework. Translational alterations of the senescence-related genes in the 22-month-old mice are absent. The sample size in some of the experiments allows detecting only large differences between groups.

Conclusively, NRSF/REST is essential for the preservation of memory performance and activity-dependent synaptic plasticity during aging and takes potential roles in the onset of age-related memory impairments. However, the memory impairments and Aβ loads of APP/PS1 mice are not altered by either NRSF/REST deficiency or overexpression, suggesting additional cellular mechanisms in the onset of AD-like pathologies.

Materials and Methods

Key resources

Drugs, antibodies and information of other key resources were listed in Supplementary Material, Table S1. Primers for transgene identification and qPCR were shown in Supplementary Material, Table S2.

Animals

Transgenic mouse strains

C57BL/6 mice were obtained from Shanghai Model Organisms Center, INC. A telencephalon-specific NRSF/REST conditional knockout mouse strain (Emx1-CKO) was generated by crossing NRSF^flox/flox (26) with Emx1-Cre mice (53). A neuron specific knockout (NSE-CKO) mouse strain was generated by crossing NRSF^flox/flox (26) with NSE-Cre transgenic mice (26). The Emx1-CKO and NSE-CKO mouse strains were crossed with APP/PS1 AD transgenic mice [B6C3-Tg (APPswe, PSEN1-dE9)85Dbo, The Jackson Laboratory]. The offspring of either sex was genotyped by primers for APP, PS1, Emx1, Cre and LoxP. Emx1-Cre:NRSF^flox/flox:APP/PS1 mouse strain was referred as Emx1-CKO-AD and NSE-Cre:NRSF^flox/flox:APP/PS1 mouse strain was named as NSE-CKO-AD. Mice were housed in a 12 h light/dark cycle with food and water ad libitum. The various experiments were summarized in Supplementary Material, Table S3.

Thioflavin-S staining

Amyloid deposits in tissue sections were visualized with fluorescent Thioflavin-S staining (54). Brain sections were rinsed with TBS (Tris Buffered Saline) buffer, incubated in 0.125% Thioflavin-S for 20 min at 37°C and then washed in 50% ethanol every 5 min for four times. Sections were rinsed in TBS and mounted in glycerin jelly. Images were acquired utilizing a fluorescence microscope (Olympus, Japan). The Aβ loads in each brain were determined from every fifth consecutive section. The number, area and the integrated fluorescence intensity of plaques (only dot-like structures were taken into account outside fiber bundles) were analyzed by Image J (NIH, USA).

Immunocytochemistry

Immunofluorescence staining

Mouse brains were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.2) overnight, soaked in 20% sucrose solution for 24 h and stored in 30% sucrose solution for 24–48 h at 4°C. Brain sections (30 μm) were prepared utilizing a cryomicrotome (Leica CM3050 S, Germany). The sections were permeabilized and blocked in 0.01 M PBS (phosphate buffer saline) containing 10% goat serum and 0.5% Triton X-100 at 37°C for 1 h and incubated at 37°C for 2 h and 4°C overnight with primary antibodies. Sections were then incubated with Alexa Fluor secondary antibodies (Alexa Fluor Plus, Invitrogen, USA) at 37°C for 1 h (55). Then the sections were mounted with AQUA-MOUNT™ medium (Thermo Scientific, USA) and fluorescence images were acquired with a Nikon Eclipse epi-microscope (Nikon, Japan).

Immunohistochemical staining

Brain sections were permeabilized with 0.6% H₂O₂ for 45 min to eliminate endogenous peroxidases. After washing in PBS, the sections were incubated in blocking solution of PBS containing 0.5% Triton X-100 and 10% goat serum for 1 h at 37°C and further incubated with a primary antibody for 2 h and overnight at 37 and 4°C, respectively. The brain sections were then kept in a PBS containing the secondary antibody (1:200, Vector Laboratories, CA, USA) for 1 h at 37°C followed by an AB peroxidase incubation (1: 200, Vector Laboratories) for 45 min at 37°C. DAB solution (Vector Laboratories) was applied to induce a color reaction to visualize the labeling.

Immunoblotting

Protein extraction and western blot analysis

Tissues were lysed in RIPA (Radio Immunoprecipitation Assay) buffer (50 mM Tris–HCl, pH 7.5; 150 mM NaCl; 1% NP-40; 0.5% sodium deoxycholate; 0.1% sodium dodecyl sulfate) containing a protease inhibitor cocktail (Thermo Scientific) and sonicated for five times of 5 s. After centrifugation for 15 min at 12000 g, the supernatants were taken and mixed with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) Sample Loading Buffer (Beyotime, China) and boiled for 5 min. Protein samples were separated on a 10% glycine SDS-PAGE and blotted onto a PVDF (Polyvinylidene Fluoride) membrane (Immobilon-P; Millipore, USA). Membranes were blocked by 5% non-fat dried milk in TBS-T (10 mM Tris–HCl, 150 mM NaCl and 0.1% Tween, pH 8.0) for 1 h and incubated with primary antibodies for 2 h at room temperature then at 4°C overnight (56). The membranes were washed with TBS-T and incubated with IRDye 800cw or 688cw immunoglobulin G (1:20000 dilutions in TBS-T), and fluorescence signals were detected by Odyssey infrared imaging system (LI-COR, USA). The protein levels were quantified by densitometry analysis using Quantity One 4.5.2 software (Bio-Rad, USA).

Cell counting

Fluorescence images were acquired by a confocal microscope with a 20x magnifying objective lens (Nikon, Japan). For each mouse brain, images at similar regions of every third consecutive sections were taken and analyzed. For each section, numbers of cells were counted and averaged with three representative squares of 200 μm × 200 μm at a similar position with Image J.

Hippocampal LTP and extracellular recordings

After isoflurane anesthesia, the brain was quickly removed and immersed in oxygenated ice-cold solution (95% O₂/5% CO₂ and composition in mM: 130 NaCl, 4.9 KCl, 1.5 CaCl₂.2H₂O, 0.3 MgSO₄ 7H₂O, 11 MgCl₂-6H₂O, 0.23 KH₂PO₄, 0.8 Na₂HPO₄, 5 glucose, 25 HEPES, 22 NaHCO₃, pH 7.32). Transverse hippocampal slices (350 μm) were prepared using a vibratome (Vibratome3000, St Louis, MO, USA), transferred to a custom-made interface recording chamber and incubated for at least 2 h at 32°C under constant perfusion with oxygenated artificial cerebrospinal fluid (95% O₂/5% CO₂ and a composition in mM: 110 NaCl, 5 KCl, 2.5 CaCl₂·2H₂O, 1.5 MgSO₄·7H₂O, 1.24 KH₂PO₄, 10 glucose, 27.4 NaHCO₃, pH 7.3). fEPSPs were evoked by stimulation of Schaffer-collateral fibers in the stratum radiatum (str. rad.) of the CA1 region by biphasic rectangular current pulses (100 μs/polarity) in the range of 15–30 μA through stainless steel electrodes (A-M Systems, Inc., Sequim, WA, USA). Evoked fEPSPs were recorded with stainless steel electrodes and a differential amplifier (EXT-20F; NPI electronic GmbH, Tamm, Germany) using 3 kHz low-pass and 0.1 Hz high-pass filters, and digitized at a sampling frequency of 10 kHz by a CED 1400 plus AD/DA converter (Cambridge Electronics Design, Cambridge, UK). Stimulation strength was adjusted to 30–45% of the maximum of fEPSP slope values. A second stimulation electrode (S2) was placed opposite to the first stimulation electrode (S1) in str. rad. as a control input. LTP was induced by three 1-s 100 Hz trains at an intertrain interval of 10 min while STP was induced by a one-time one-second 100 Hz train (57). The fEPSP slopes were determined and expressed as a percentage of baseline values (58–60).

Behavior tests

MWM test

The MWM task has been widely used to evaluate spatial memory performance in rodents. In our experiments, the animals were not allowed to swim longer than 60 s. In general, the MWM test was performed as previously described (61). The water maze consisted of a circular pool (120 cm in diameter, 62.5 cm in height) with a whitewall. The escape platform (10 cm in diameter) was localized at equidistant away from the center and the wall and submerged 1 cm below the surface. The tank was located in a room surrounded by various visual cues on the wall. The recorded swimming activities of the animal to the platform within 60 s were analyzed by behavioral-tracking software (Anymaze, Stoelting Co., USA). The animal was allowed to stay for 15 s on the hidden platform after arrival. During the learning trials of the platform position, the animals were gently guided to the platform and allowed to stay there for 15 s in case they could not find the platform within 60 s. Each mouse performed over 7 days four trials daily with an intertrial interval of 60 min at randomly chosen starting points. Probe sessions were performed 24 h after the last trial. The probe test included a single probe trial in which the platform was removed from the tank and each mouse was allowed to swim for 60 s in the maze from the furthest starting point to the platform.

Y-Maze test

Mice were placed in a chamber of Y-shape with three equal arms (A, B, C, 50 cm x 10 cm for each). Mice were placed in the center of the maze and allowed to explore the maze for 10 min. The number and sequence of entries into the different Y-maze arms were recorded. Three subsequent entries into three different arms were defined as one alternation. The ratio between alternations and all single entries was calculated and presented in percentage (62). The percentage has been taken as a score of spatial memory performance. For example, a sequence like ACBABACBAB consists of 10 entries with five alternations [ACB (1), CBA (2), BAC (3), ACB (4), CBA (5)] and a score of 50%.

FC test

Mice were placed in the conditioning chamber (MED Associates, USA) for a 2-min free exploration, and then presented with five sounds (CS, 2800 Hz, 85 dB, 1 s) that co-terminated with five electric foot shock (US, 0.40 mA, 1 s) every 30 s. The context and cue tests were conducted 24 h after conditioning. Mice were placed in the conditioning chamber for 5 min and the total freezing time determined to evaluate the degree of contextual fear memory. For cued fear memory, mice were placed in a novel chamber for 5 min and meanwhile sound was presented (CS, 2800 Hz, 85 dB, 1 s, Pre-cue). The freezing time was determined semi-automatically by software.

NOR test

To habituate mice for the object recognition memory test, the animals were allowed to explore the open-field box for 10 min per day on three consecutive days. On the fourth day, the animals were trained to memorize two objects (20 cm distance) within 10 min. After a 24 h retention interval, a novel object with a different shape and color replaced one of the objects. The exploration time for the novel and the familiar objects was measured over a 10 min interval and the values were used to calculate the discrimination index (DI). The DI was calculated as follows (60): (Time exploring the novel object—Time exploring the familiar object)/(Time exploring novel object + Time exploring familiar object) ^*100%.

Quantitative real-time PCR

RNA extraction, cDNA synthesis and quantitative real time PCR (qRT-PCR) from mouse brains were performed as previously described (56). Total RNA was isolated from brain tissues with TRIzol reagent (TIANGEN, China). cDNA was synthesized using a reverse transcription kit (TIANGEN, China). Real-time qRT-PCR was performed on the quantitative thermal cycler (Mastercycler ep realplex, Eppendorf, Germany) using SYBR Green agent (TIANGEN, China).

Statistical analysis

All values have been presented as mean ± SEM. Statistical analyses were performed using the GraphPad Prism 8.0 software (GraphPad Software, USA).

Data comparison of multiple groups were performed by one-way ANOVA, or two-way ANOVA, and followed by Fisher’s LSD (Least Significant Difference) multiple comparisons. Means between two groups were compared by the two-tailed unpaired Student t-test with Welch’s correction. P ≤ 0.05 was considered statistically significant.

Conflict of Interest statement. None declared.

Funding

National Natural Science Foundation of China (31970908 and 31671043 to F.H., and 8126120568 and 81572776 to J.F., and 31871076 to T.B., and 82104640 to Y.Y.F.); Shanghai Sailing Program (No. 19YF1448500); China Postdoctoral Science Foundation (No. 2019 M651562); Shanghai Municipal Science and Technology Major Project (No. 2018SHZDZX01); ZJLab, Science and Technology Commission of Shanghai Municipality (19DZ2280500, 18DZ2293500); Open Project of State Key Laboratory of Medical Neurobiology (SKLMN2003); innovative research team of high-level local university in Shanghai; Shanghai Center for Brain Science and Brain-Inspired Technology.

Author notes

Yufang Yang, Xiaoshuang Zhang, Dongxue Li and Rong Fang authors contributed equally.