VPA-induced autism impairs memory ability through disturbing neural oscillatory patterns in offspring rats

Abstract

Autism spectrum disorder (ASD) is a general neurodevelopmental disease characterized by unusual social communication and rigid, repetitive behavior patterns. The purpose of this study was to investigate the effects of ASD on the alteration of neural oscillatory patterns and synaptic plasticity, which commonly supported a wide range of basic and higher memory activities. Accordingly, a prenatal valproic acid (VPA) exposure rat model was established for studying autism. The behavioral experiments showed that the social orientation declined and the memory ability was significantly impaired in VPA rats, which was closely associated with the synaptic plasticity deficits. Neural oscillation is the rhythmic neuron-activity, and the pathological characteristics and neurological changes in autism may be peeped at the neural oscillatory analysis. Interestingly, neural oscillatory analysis showed that prenatal VPA exposure reduced the low-frequency power but increased high-frequency gamma (HG) power in the hippocampus CA1 area. Meanwhile, the coherence and synchronization between CA3 and CA1 were abnormally increased in the VPA group, especially in theta and HG rhythms. Furthermore, the cross-frequency coupling strength of theta-LG in the CA1 and CA3 → CA1 pathway was significantly attenuated, but the theta-HG coupling strength was increased. Additionally, prenatal VPA exposure inhibited the expression of SYP and NR2B but enhanced the expression of PSD-95 along with decreased synaptic plasticity. The neural oscillatory patterns in VPA-induced offspring were disturbed with the intensity and direction of neural information flow disordered, which are consistent with the changes in synaptic plasticity, suggesting that the decline in synaptic plasticity is the underlying mechanism.

Introduction

Autism spectrum disorder (ASD) is a widespread neurodevelopmental disease. Several studies show that most autistic patients have emotional disorder, social disorder and different degrees of cognitive and memory impairment (Polleux and Lauder 2004; Lord et al. 2018). In order to better study the disease mechanism of autism, Rodier et al. first have established an animal model of autism with prenatal valproic acid (VPA) exposure (Rodier et al. 1996), where the VPA is a kind of short-chain fatty acid with strong teratogenicity. It was found that exposure to VPA during early pregnancy could induce severe brain damage (Ornoy 2006), memory impairment (Gao et al. 2016), and greatly increase the risk of autism in offspring (Christensen et al. 2013; Bromley et al. 2013). In addition, at the level of the nervous system, the study has shown that autism can impair synaptic plasticity, which is manifested by abnormal synaptic pruning (Hutsler and Zhang 2010). It is also reported that the number of dendritic spines is abnormal in the cortex and hippocampus of VPA model rats (Bringas et al. 2013), indicating that their structural plasticity is changed abnormally. However, little is known about the change of neural activity pattern, and the relationship between the alteration of neural oscillatory patterns and memory ability deficits in VPA-induced autism.

Neural oscillation is a rhythmic neural activity in the nervous system, which is closely related to cognitive memory ability (Tort et al. 2009). And it builds a bridge between memory ability and molecular biological mechanism (Abel et al. 2013). The analysis based on neural oscillation can be conducted from multiple perspectives, including power analysis, synchronization analysis, cross-frequency coupling analysis, etc. (Li et al. 2015). These indicators signify the characteristics of neural oscillatory patterns in different dimensions. There are a number of studies on cognitive and memory function based on neural oscillation, which show that the different frequency oscillations are involved in controlling cognitive and memory processes (Başar et al. 2001; Gulbinaite et al. 2014). Furthermore, the theta-gamma cross-frequency coupling strength in the hippocampal CA1 region plays a key role in spatial cognition (Nishida et al. 2014). In autism study, it reports that autistic patients have abnormal gamma rhythms (Rippon 2017; Gandal et al. 2010), and they have abnormal coherence in the resting state (Wang et al. 2013). Meanwhile, autism can impact neural oscillations (Simon and Wallace 2016; Takarae and Sweeney 2017). It strongly suggests that there are some disorders of neural activity and cognitive and memory impairment in the case of autism.

In normal animal investigations, a report has indicated that the neural pathways of MEC-CA1 and CA3-CA1 exhibit the strongest coherence in the theta rhythm, as well as significant coherence in the gamma rhythm (Colgin et al. 2009). In addition, theta rhythms periodically reproduced in working memory (Fuentemilla et al. 2010), and theta-gamma coupled oscillations are closely related to information exchange and cognitive memory across brain regions in the hippocampus (Lisman and Jensen 2013). In awake state, the low gamma rhythm of CA1 in freely exploring rats is related to the extraction of social experience (Zhu et al. 2022). However, the multidimensional measurement of autistic neural oscillation and its relationship with memory ability are rarely mentioned. Therefore, it is necessary to do a more comprehensive neural oscillatory analysis in the autism-model to classify how the pattern changes of neural oscillation are associated to memory ability through synaptic plasticity.

In the present study, a hypothesis was raised that prenatal exposure to VPA significantly disturbed the neural oscillatory patterns of VPA model rats, especially inhibited cross-frequency neural information flow both in the internal neural circuits of hippocampus and the pathway of entorhinal cortex (EC)-hippocampus, which was closely associated to the decline of synaptic plasticity and the impairment of the memory abilities in the hippocampus. This was done by establishing prenatal VPA-induced autism model rats and giving birth to offspring. And then, the memory ability was carefully evaluated by novel object recognition (NOR) test and in vivo electrophysiological experiments were performed to record LTP and collect LFP signals. In addition, the expression level of synapse-related proteins was measured by Western blot.

Material and methods

Animals

In this experiment, eight SD pregnant rats (the 8th day of pregnancy) were purchased from the Laboratory Animal Center of Academy of Military Medical Science of People's Liberation Army and raised in the Medical College of Nankai University. During the feeding process, the pregnant rats were fed separately, the environment was kept at 25 ± 2 °C, and the light/dark cycle lasted for 12 h (the light was turned on at 7 a.m.). At the same time, suitable water and food were given. All animal experiments were approved by the Animal Research Ethics Committee, Nankai University (06-01-15). During the experiment, a great effort has been made to reduce the pain of animals.

VPA-induced autism modeling procedure

The purchased SD pregnant rats (N = 8) were randomly divided into control group and model group, with four pregnant rats in each group. On 12.5 days of pregnancy, animals in the model group were given a single subcutaneous injection of sodium valproate salt solution (500 mg/kg, 250 mg/mL) prepared by normal saline, while others in the control group were given a single subcutaneous injection of normal saline solution with the same dosage. After the completion of drug modeling, the pregnant rats were raised separately and recorded. After the delivery of the pregnant rats, the mother rats give birth to their offspring until weaning. After that, the seven male offspring in the control group were selected as the control group (Sham, N = 7), and the seven male offspring in the model group were selected as the model group (VPA, N = 7).

Three-chamber social test

The three-chamber social test was used to evaluate the social ability of rats, that is, the instinct tendency of animals to contact social partners frequently (Grabrucker et al. 2016). The main equipment of the experiment is a rectangular box with a length of 60 cm, a width of 40 cm and a height of 25 cm (Fig. 1a). The box is composed of two side chambers and a central chamber, in which there is a cylindrical cage surrounded by a fence. In the study, the three-chamber test was divided into adaptation and two test stages. In the adaptation stage, an animal was gently put into the central area and freely explored in three chambers for 10 min. Afterwards, a test was performed immediately. In the first stage, a male strange rat that was the same age of the adaptation stage animal (one month old) was put into the cage in the left side chamber, and then the test rat was gently put into the device from the central area, and freely explored in the three chambers for 10 min. Similarly, in the second test phase, a strange rat was put into the cage on the right side and the test rat was allowed to continue to explore freely for 10 min. During the experiment, the time of sniffing the cages in the left and right rooms and the time of interaction with strange rats were recorded accordingly. Equipment was wiped with 70% ethanol every two stages to eliminate the interference of odor cues.

Fig. 1.

The behavioral differences in prenatal VPA exposed rats (N = 7). a Schematic diagram of three-chamber social test. b The statistics of time spent in three spaces in the training stage. c The statistics of time spent in three spaces in the testing-1 stage. d The statistics of time spent in three spaces in the testing-2 stage. e Schematic diagram of NOR test. f The statistics of preference in NOR test. Data were expressed as mean ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001

Novel object recognition test

The novel object recognition (NOR) test was used to evaluate the learning and memory ability, including the tests of short-term memory and long-term memory (Antunes et al. 2012; Feng et al. 2018). The experimental device is a square box with a length of 50 cm, a width of 50 cm and a height of 36 cm (Fig. 1e). This experiment was mainly divided into four stages, including habituation stage, training stage, short-term memory test stage and long-term memory test stage. During the training phase, the test rats were gently placed in the equipment with two similar objects (object A1, object A2) and freely explored for 10 min. After 2 h, object A2 was replaced by object B and then gently place the rat in the device to explore freely for 10 min. And 24 h later, the object B was replaced by object C. Similarly, rats were free to explore in this device for 10 min. At the end of each phase, the equipment was carefully cleaned with 70% ethanol. During the experiment, the exploration time of the test rat for new objects and old objects was recorded, and finally the object preference of the test rat was calculated as the formula follows:

$$\begin{array}{c}Preference\left(\%\right)=\frac{Tim{e}_{\mathit{novel}}}{Tim{e}_{\mathit{novel}}+Tim{e}_{\mathit{old}}}\ast 100\%\end{array}$$ 1

In vivo electrophysiological recording

After the NOR test, in vivo electrophysiological experiments were performed. Electrophysiological features were collected during the session in the form of the long-term potentiation (LTP), and local field potentials (LFPs) in both the left hippocampus CA1 region and Schaffer collateral branch (CA3). The specific steps can be divided into: 1. weighed the rats and intraperitoneal injection of 30% urethane (1.2 g/kg body weight) for anesthesia. 2. anesthetized rats were fixed on the brain stereotaxic apparatus and covered with a layer of cotton for warmth. After that, the hair on the brain was removed, the scalp was cut, and the periosteum was peeled to expose the skull completely. 3. found the anterior fontanel under the stereoscopic microscope, and located it according to the atlas of rat brain, with the anterior fontanel as the origin. Schaffer's lateral branches are 4.2 mm behind the anterior fontanel and 3.5 mm laterally open; CA1 is 3.5 mm behind the front fontanel and 2.5 mm laterally open. According to the positioning, circle an area, and polish the skull with a dental drill to fully expose the cerebral cortex in this area. 4. the electrodes were positioned according to the anterior fontanel again. A stimulating electrode (Schaffer lateral branch) and a recording electrode (CA1) were slowly implanted in the corresponding position. According to the depth range of Schaffer's lateral branches (2.3–2.6 mm below dura mater) and CA1 (2.0–2.2 mm below dura mater), slowly adjust the electrode depth manually to find the best position. A single pulse stimulation was delivered to observe the change of the slope of the evoked field excitatory postsynaptic potentials (fEPSPs) curve under the stimulation of 0.1–1.0 mA, that is, the I/O curve. The current intensity whose slope is about 50% of the maximum slope was selected as the stimulation intensity of this operation. 5. A stimulative optimal intensity that could evoke 70% response of its maximum amplitude (ranged from 0.3 to 0.5 mA) was delivered at single-pulse stimulation to record a 20-min baseline (Scope software, PowerLab). The LTP was recorded every 60 s for 1 h, after theta burst stimulation (TBS) including of 30 trains of 12 pulses (200 Hz) at 5 Hz. The original analysis of LTP data was conducted by using Clampfit 10.2 (Molecular Devices, Sunnyvale, CA, USA). The local field potentials (LFPs) signals were recorded at a sampling rate of 1 kHz for 20–25 min.

Western blot assay

The rats were decapitated after LFPs recording, and then the hippocampus was quickly removed and stored at − 80 °C. Hippocampal tissue was triturated and lysed in 600 μL of lysis buffer (Beyotime Biotechnology, Haimen, China) containing protease inhibitor cocktail (1:100 dilution). The homogenate was then centrifuged at 12,000 rpm at 4 °C for 15 min, and the supernatant was acquired. Afterwards, the protein concentration was measured using BCA assay kit (Beyotime Biotechnology, Haimen, China), separated an equal amount of protein loaded by SDS-PAGE on a 10–13% gels and transferred it to a PVDF membrane using electrophoresis (Millipore, USA). The PVDF membranes were blocked in 10% skim milk for 1 h at room temperature and incubated with primary antibodies overnight at 4 °C. Next day, they were washed 4 times with Tris-buffered saline containing 0.05% Tween 20 (TBST). The horseradish peroxidase coupled secondary antibody (1:4000 dilution; Promega, USA) was incubated at room temperature for 2 h, and then washed 4 times again. Then, they were detected and analyzed and beta-tubulin was used as an internal control. A computerized chemiluminescent imaging system (Tanon 5500, Tanon Science & Technology, China) was applied to define the protein band intensities. Band intensities were quantified with the NIH Image J program. All antibodies are as follows: anti-beta tubulin (1:5000, Abcam, UK), anti-SYP (1:1000, Abcam, UK), anti-PSD95 (1:1000, Abcam, UK), anti-NR2B (1:1000, Abcam, UK).

Neural oscillatory analysis

First of all, the original LFP signals were filtered at 0.5–100 Hz. Neural oscillatory measurements were performed in the MATLAB (R2021b, MathWorks) platform.

Power spectra analysis

The multi window spectral estimation method was used to calculate the power spectrum of the signal, which was measured through the program of Chronux toolbox (Bokil et al. 2010). The relevant parameters are as follows: sliding windows of 10 s with 50% overlap and the Slepian tapers field as (5–9). Furthermore, different frequency bands are divided: delta (1–3 Hz), theta (3–8 Hz), low gamma (30–50 Hz) and high gamma (50–100 Hz).

Coherence analysis

The coherence is a measure of the correlation between two signals, which is based on multi window power spectrum analysis (Bokil et al. 2010). We assume that the signals of the two channels are x and y, and then $S_x(f)$ and $S_y(f)$ are the power spectrum estimates of the signals x and y. $S_{\mathit{xy}}(f)$ is the cross term of the power spectrum estimates of the signals x and y. The following formulas show their power spectrum estimation:

$$\begin{array}{c}{S}_x\left(f\right)=\frac{1}{K}\sum _1^K{\left|{\tilde{x}}_K\left(f\right)\right|}^2\end{array}$$ 2

$$\begin{array}{c}{S}_y\left(f\right)=\frac{1}{K}\sum _1^K{\left|{\tilde{y}}_K\left(f\right)\right|}^2\end{array}$$ 3

$$\begin{array}{c}{S}_{\mathit{xy}}\left(f\right)=\frac{1}{K}\sum _1^K{\tilde{y}}_K{\tilde{x}}_K^{\ast }\left(f\right)\end{array}$$ 4

where K represents the number of windows, K = 2 T * W − 1, T represents the window length, and W represents the frequency bandwidth. Further, the formula for measuring coherence is as follows:

$$\begin{array}{c}Co{h}_{\mathit{xy}}\left(f\right)=\frac{S_{\mathit{xy}}\left(f\right)}{\sqrt{S_x\left(f\right)S_y\left(f\right)}}\end{array}$$ 5

The variation range of the coherence value is between 0 and 1, where 1 means the two signals are completely linear coherent at this frequency, and 0 means that they are completely linear incoherent. It is a non-directional signal analysis index. The coherence was obtained by the means of Multitaper coherency from Chronux toolbox, using 10-s windows with 50% overlap.

Phase synchronization analysis

Phase locking value (PLV) is an index commonly used to evaluate the synchronization of the same rhythm, focusing on the phase synchronization intensity between different neural oscillations (Rosenblum et al. 1996). The specific method is to extract the phase of the two signals by Hilbert transform and record them as ${\mathrm{\varnothing }}_1$ and ${\mathrm{\varnothing }}_2$, respectively. Then it is obtained by the following formula:

$$\begin{array}{c}\text{PLV}=\left|\frac{1}{N}\sum _{j=1}^N\text{exp}\left(i\left[{\mathrm{\varnothing }}_1\left(j\mathrm{\Delta }t\right)-{\mathrm{\varnothing }}_2\left(j\mathrm{\Delta }t\right)\right]\right)\right|\end{array}$$ 6

where $N$ is the length of signal and 1/Δt is the sampling frequency. The value range of PLV is [0, 1]. When the value of PLV is 1, the two signals achieve complete phase synchronization, and value of 0 indicates complete non synchronization. In this experiment, the length of sliding window is 10,000 (10 s), and the overlap rate of sliding window is 50%.

Directional coupling analysis

Directional coupling analysis can more accurately represent the intensity and direction of neural information flow. Conditional mutual information (CMI) is a classical directional coupling analysis method, which is defined as the expected value of mutual information of two random variables under given conditions (Wyner 1978). Paluš et al. (2003) proposed that CMI can be used to analyze the directivity of neural information flow. This is an analysis algorithm based on the theory of phase dynamics and Shannon entropy. First, the mutual information $I\left(X,;,Y\right)$ of two random variables X and Y is defined as:

$$\begin{array}{c}I\left(X;Y\right)=H\left(X\right)+H\left(Y\right)-H\left(X,Y\right)\end{array}$$ 7

where $H\left(X\right)$ and $H\left(Y\right)$ represent the Shannon entropy of X and Y respectively, and $H\left(X,,,Y\right)$ is the Shannon entropy of the joint distribution of $(X,Y)$. Given the variable Z, the conditional mutual information can be expressed as:

$$\begin{array}{c}I\left(X;Y|Z\right)=H\left(X|Z\right)+H\left(Y|Z\right)-H\left(X,Y|Z\right)\end{array}$$ 8

After extracting the instantaneous phases of both signals X and Y, $I\left({\mathrm{\varnothing }}_y,;,{\mathrm{\varnothing }}_{x\tau },|,{\mathrm{\varnothing }}_x\right)$ is used to estimate the information of X included in the Y process in the future t time as a measure of the coupling direction, and vice versa. Therefore, conditional mutual information can be defined as:

$$\begin{array}{c}{i}_{x,y}=I\left({\mathrm{\varnothing }}_{y,x}\left(t\right),{\mathrm{\Delta }}_{\tau }{\mathrm{\varnothing }}_{x,y}|{\mathrm{\varnothing }}_{x,y}\left(t\right)\right)\end{array}$$ 9

For each time point k, the phase increment is defined as ${\mathrm{\Delta }}_{x,y}\left(k\right)=\mathrm{\varnothing }\left(t_k,+,\tau \right)-\mathrm{\varnothing }(t_k)$, the coupling direction index of the CMI algorithm is defined as $d=(i_y-i_x)/(i_y+i_x)$. When 0 < d < 1, it indicates that the information transmission direction X → Y. When − 1 < d < 0, it indicates that the information flow direction is opposite. Moreover, i represents the amount of information transmitted in a single direction.

Cross-frequency coupling analysis

Cross-frequency coupling (CFC) is used to measure the interaction between different frequency bands, which can reflect the transmission of neural information flow and changes in synaptic plasticity. Phase amplitude coupling (PAC) is one of the most representative methods of CFC, which means that the amplitude (power) of high frequency signal is locked in the phase of low frequency signal, or is modulated by the phase of low frequency signal. PAC_PLV algorithm is often used to analyze cross-frequency coupling (Penny et al. 2008). Firstly, the low-frequency rhythm phase and high-frequency rhythm amplitude of the signal are extracted by Hilbert transform, and then the high-frequency rhythm amplitude is extracted by Hilbert transform to obtain its phase. Therefore, the mathematical definition of PAC_PLV is expressed as:

$$\begin{array}{c}{\text{PAC}}_{\text{PLV}}=\left|\frac{1}{N}\sum _{j=1}^N\text{exp}\left(i\left[{\mathrm{\varnothing }}_{\text{low}}\left(j\mathrm{\Delta }t\right)-{\mathrm{\varnothing }}_{\text{high}}\left(j\mathrm{\Delta }t\right)\right]\right)\right|\end{array}$$ 10

where $N$ is the length of signal and 1/Δt is the sampling frequency. ${\mathrm{\varnothing }}_{\mathit{low}}$ and ${\mathrm{\varnothing }}_{\mathit{high}}$ represent the phase of the low-frequency rhythm and the phase of the high-frequency rhythm, respectively. Generally, the PAC value was focused on the low-frequency theta rhythm with low gamma (LG) and high gamma (HG). The procedure was performed for a time window of 40 s with 50% overlap. The value range of PAC_PLV is [0, 1], where 1 indicates completely synchronization, and 0 indicates no-synchronization relationship.

Data and statistical analysis

The statistical analyses were performed by using SPSS 19.0 (SPSS Inc., Chicago, USA) and presented as mean ± S.E.M. The student’s t test evaluated the statistical significance of the difference between the two groups. When p < 0.05, the differences were considered to be statistically significant.

Results

VPA-induced autism caused autistic behavior and memory impairment

Previous studies showed that prenatal VPA exposure was an effective approach to autism modeling (Guo et al. 2018; Kim et al. 2014). First, the three-chamber social test was used to verify whether the rat-model of autism was successfully established (Fig. 1a). In the training phase, the time spent in the three chambers in the VPA group was not significantly different from that in the Sham group (Fig. 1b). However, in the testing-1 phase, the time spent in chamber-1 in the VPA group was significantly decreased (Fig. 1c-left, $t_{(12)}$ = 3.9857, p < 0.01), and considerably increased in the chamber-2 (Fig. 1c-right, $t_{(12)}$ = − 4.4485, p < 0.001) compared to that in the Sham group. Furthermore, in the testing-2 phase, another strange rat was put in the chamber-2. The time spent in the chamber-1 was distinctly increased (Fig. 1d-left, $t_{(12)}$ = − 6.4621, p < 0.001), while the time spent in the chamber-2 was significantly reduced in the VPA group (Fig. 1d-right, $t_{(12)}$ = 2.4138, p < 0.05). Therefore, according to the three-chamber social test, it was shown that the VPA rats produced social avoidance behavior, suggesting that the autism model was successfully established. After that, a NOR test was performed to assess the animal’s memory ability (Fig. 1e), including 4 stages which are habituation, training, 2 h testing and 24 h testing. The purpose was to examine the short-term memory and long-term memory. It was found that the preference in the VPA group was significantly reduced in the long-term testing compared to that in the Sham group (Fig. 1f-right, $t_{(12)}$ = 3.4882, p < 0.01), suggesting that the memory ability was impaired in VPA rats.

VPA-induced autism caused synaptic plasticity damage

It is well known that the synaptic plasticity is closely related to neural oscillations (Quan et al. 2011). Figure 2a showed that once delivering TBS, the slope of fEPSPs was increased in the both groups. However, there was obvious difference of the increase between these two groups. The statistic of LTP illustrated that the functional synaptic plasticity was lower in the VPA group than that in the Sham group (Fig. 2b, $t_{(8)}$ = 4.0416, p < 0.01). At the same time, the expression of synaptic associated proteins is closely related to changes in synaptic plasticity (Bruneau et al. 2009). Therefore, Western blotting was applied to detect synaptic related proteins in the hippocampus and attempt to show the molecular level of synaptic plasticity damage caused by autism. Thus, the expressions of NR2B, PSD-95 and SYP were quantified. NR2B is an NMDAR receptor subunit whose expression is correlated with synaptic plasticity. SYP is a protein closely to synaptic function that widely exists in nerve endings and presynaptic membranes. PSD-95 is a postsynaptic scaffold protein, which plays a key role in the process of synaptic transmission and dendritic spine formation. There were the representative immunoreactive bands (Fig. 2c). It could be seen that the expressions of SYP and NR2B were significantly decreased in the VPA group compared to that in the Sham group (Fig. 2d, $t_{(4)}$ = 4.9962, p < 0.01; Fig. 2e, $t_{(4)}$ = 4.2090, p < 0.05). Meanwhile, the level of PSD-95 was considerably increased in the VPA group compared to that in the Sham group (Fig. 2f, $t_{(4)}$ = − 4.7160, p < 0.01).

Fig. 2.

The alteration of synaptic plasticity in prenatal VPA exposed rats. a The changes of time coursing in the fEPSPs slope in LTP. b) The statistic of LTP (N = 5). c The representative immunoreactive bands of NR2B, PSD95 and SYP in Sham and VPA group. d The statistics of SYP protein expression level (N = 3). e The statistics of NR2B protein expression level (N = 3). f The statistics of PSD-95 protein expression level (N = 3). Data were expressed as mean ± SEM. *p < 0.05, **p < 0.01

VPA-induced autism disturbed power distributions

The power spectrum is the most widely used index to measure the power intensity of neural activity signal. Figure 3a showed the normalized power distribution in the CA1 at 1–100 Hz. From low frequency to high frequency rhythms, we mainly calculated the power value of delta, LG and HG rhythms. It was found that power in delta rhythm was significantly reduced, but it was increased in HG frequency band in the VPA group compared to that in the Sham group (Fig. 3b-left, $t_{(12)}$ = 3.4969, p < 0.01; Fig. 3b-right, $t_{(12)}$ = − 2.1926, p < 0.05). There was no statistical difference between these two groups in LG frequency band.

Fig. 3.

The power spectrum, coherence, phase synchronization, CMI in hippocampal CA3 and CA1 area (N = 7). a 1–100 Hz mean power spectrum in CA1 area. b The statistical results of power distribution in delta, LG, HG frequency bands in CA1 area. c 1–100 Hz mean coherence between CA3 and CA1 area. d The statistical results of coherence in theta, LG, HG frequency bands between CA3 and CA1 area. e The mean PLV in the range of 1–100 Hz. f The statistical results of PLV in theta, LG, HG frequency bands. g The mean CMI value from CA3 to CA1 in 1–100 Hz. h The statistical results of CMI in theta, LG, HG frequency bands. Data were expressed as mean ± SEM. *p < 0.05, **p < 0.01

VPA-induced autism undermined neural information communication

After performing power spectral analysis, the equivalent rhythm coupling analysis was executed, in which the computational algorithms included coherence, PLV and CMI. It could be seen that the mean coherence value was higher in the VPA group than that in the Sham group (Fig. 3c). The statistical data showed that the coherence value of VPA-rats was significantly increased in theta and HG frequency bands (Fig. 3d-left, $t_{(12)}$ = − 2.5370, p < 0.05; Fig. 3d-right, $t_{(12)}$ = − 3.2577, p < 0.01), but there was no significant difference in LG frequency band compared with that of Sham-rats. Moreover, the value of PLV based on phase synchronization was measured (Fig. 3e, f), which were consistent with the results of coherence. The data showed that the PLV value was higher in the VPA group at theta and HG frequency bands than that in the Sham group (Fig. 3f-left, $t_{(12)}$ = − 2.4954, p < 0.05; Fig. 3f-right, $t_{(12)}$ = − 2.3382, p < 0.05), but there was no significant difference of PLV between these two groups at LG frequency band. Figure 3g shows the CMI value in CA3 → CA1 pathway. And the results illustrated that the CMI value was lower in the VPA group than that in the Sham group at theta frequency band (Fig. 3h-left, $t_{(12)}$ = 2.8724, p < 0.05), however, the strength was higher in the VPA group at HG frequency band than that in the Sham group (Fig. 3h-right, $t_{(12)}$ = − 2.3855, p < 0.05). Interestingly, there was also no statistical difference between these two groups at the LG frequency band.

Furthermore, the PAC_PLV was adopted to evaluate the strength of cross-frequency coupling. There is the mean PAC_PLV between low frequency rhythm (1–8 Hz) and high frequency rhythm (30–100 Hz) in CA1 region and in CA3-CA1 pathway (Fig. 4a, c), where the red color indicates the strongest coupling strength and the blue represents the minimum coupling strength. The data showed that both in the hippocampal CA1 region and CA3 → CA1 pathway, the theta-LG PAC value was lower in the VPA group than that in the Sham group (Fig. 4b-middle, $t_{(12)}$ = 3.0983, p < 0.01; Fig. 4b-right, $t_{(12)}$ = 3.2252, p < 0.01). However, it was the opposite of the PAC value in the theta-HG (Fig. 4d-middle, $t_{(12)}$ = − 3.1109, p < 0.01; Fig. 4d-right, $t_{(12)}$ = − 2.2416, p < 0.05). There was no significant difference of the PAC in theta-LG and theta-HG in CA3 area. It suggests that the neural information flow in the VPA group was significantly disturbed, which was possibly related to the different destruction degree of multiple information transmission paths within the hippocampus and between EC and hippocampus.

Fig. 4.

The cross-frequency coupling analysis in hippocampal CA3 and CA1 area. a The mean PAC between low frequency rhythm (1–8 Hz) and high frequency rhythm (30–100 Hz) in CA1 area. b The statistical results of theta (3–8 Hz)-LG (30–50 Hz) PAC. c The mean PAC between low frequency rhythm (1–8 Hz) and high frequency rhythm (30–100 Hz) from CA3 to CA1 area. d The statistical results of theta (3–8 Hz)-HG (50–100 Hz) PAC. Data were expressed as mean ± SEM. *p < 0.05, **p < 0.01

Discussion

In this study, in order to evaluate whether autism declined memory ability through impairing the synaptic plasticity, which was closely associated with disturbing the pattern of neural oscillations, we established the VPA-induced autism rat model. Our data showed that LTP was significantly inhibited and synaptic related proteins were considerably altered in VPA-rats. Meanwhile, based on analyzing the pattern-alteration of neural oscillations in the hippocampal CA3 & CA1 regions and CA3-CA1 pathways, it represented that (1) The LFP power distribution was significantly distorted in VPA offspring rats. (2) The coherence and synchronization were abnormally enhanced, meanwhile the directional information flow from CA3 to CA1 was manifestly inhibited. (3) The theta-LG cross-frequency coupling strength either in the hippocampal CA1 region or on the CA3 → CA1 pathway were considerably reduced, but the strength of theta-HG were significantly increased, which was greatly related to spontaneous generation of low-frequency gamma rhythm in hippocampal CA3 and the input of high-frequency gamma rhythm via EC to CA1.

It is reported that social disorder is a typical symptom of autism, and its intuitive manifestation is the reduction of social interaction (Lord et al. 2018). The three-chamber social experiment can effectively evaluate the social behavior of rats, and our results are consistent with this point (Fig. 1a–d). In addition, more than 70% of autistic patients have some degree of learning and memory impairment (Wu et al. 2017). At the same time, it was reported in the literature that VPA model mice showed obvious impairment of spatial learning and memory in Morris water maze test (Wu et al. 2017; Hajisoltani et al. 2019). Similarly, our NOR results were also consistent with the data obtained from the above studies (Fig. 1e, f).

The pattern of neural oscillation is the characteristic of neural activity in the central nervous system, and different frequencies correspond to different physiological functions. Generally, theta oscillation is related to the memory ability (Sauseng et al. 2010), and gamma rhythm is related to sleep and advanced cognition (Buzsáki and Wang 2012). To be sure, these two rhythms will work together more to achieve excitation inhibition (E/I) balance (Buzsáki and Wang 2012). A previous study showed that the power of the high gamma rhythm in the hippocampus of VPA-model rats was increased while the low frequency power was decreased (Cheaha et al. 2015), which was consistent with our data (Fig. 3a, b). Furthermore, the result of CMI also represents that the information flow from CA3 to CA1 was significantly inhibited in theta band, but enhanced in HG. It is worth noting that the values of both coherence and PLV were increased in theta and HG frequency bands (Fig. 3d, f). Interestingly, a clinical study employed VPA as a drug to treat epilepsy, but the synchronization of neural oscillations has been abnormally increased (Zou et al. 2021), which is consistent with our coherence and PLV results. To sum up, all neural oscillatory measurements, including power spectra, coherence, PLV and CMI, suggest that the pattern of neural activity has been significantly disturbed.

Furthermore, cross-frequency coupling analysis has been performed, which could evaluate the interaction between low-frequency and high-frequency rhythms (Li et al. 2016; Zheng and Zhang 2013). In the present study, the PAC_PLV algorithm was applied to analyze theta-gamma coupling. It generates the locked value of the amplitude of high-frequency rhythm on the phase of low-frequency rhythm (Tort et al. 2009; Kumari et al. 2019). The data showed that the theta-LG PAC value was lower in the VPA group than that in the Sham group either in the hippocampus CA1 region or on the CA3 → CA1 pathway (Fig. 4b, d). In fact, the gamma oscillations in the hippocampal CA1 region originate in two oscillators, namely CA3 and medial entorhinal cortex (MEC) (Colgin and Moser 2010). In addition, LG and HG rhythms in CA1 are derived from CA3 and MEC, respectively, where the high gamma rhythm originated in MEC is directly transmitted to CA1 area through the perforant pathway (Colgin et al. 2009). Therefore, our PAC results revealed that the neural information flow from CA3 to CA1 was significantly inhibited. Meanwhile, the information flow from MEC to CA1 was enhanced, which was possibly related to the dynamic regulation of the neural circuit.

The oscillations in neural network temporally links many neurons into assemblies, cooperatively facilitates synaptic plasticity (Buzsaki and Draguhn 2004). Synaptic plasticity is the cellular mechanism of learning and cognition (Bliss and Collingridge 1993). In the study, LTP was recorded which is the most effective measure of synaptic functional plasticity (Nicoll 2017). The results showed that the functional synaptic plasticity from CA3 to CA1 was significantly inhibited (Fig. 2a&b), which is consistent with the idea that severe synaptic damage exists in autism (Blaylock 2008). In addition, the expression of some synaptic structure related proteins was also evaluated, and the data showed that the expression of SYP and NR2B was reduced, while the expression of PSD-95 increased (Fig. 2c–f). The significant decrease of NR2B results in the loss of excitatory receptors in neurons, thereby leads to the inhibition of synaptic plasticity. SYP is a protein that widely exists in nerve endings and presynaptic membranes, and the reduced level of SYP represents the decreased number of neurons and abnormal synaptic pruning particularly. Furthermore, PSD-95 is a postsynaptic scaffold protein, which plays a key role in the process of synaptic transmission and dendritic spine formation. However, the overexpression of PSD-95 also blocks LTP, causing synaptic plasticity damage (Stein et al. 2003). It has been reported that the synaptic pruning function of autistic-like rats could be impaired, resulting that there is an abnormal increase in the density of dendritic spines. As we know that PSD-95 is the postsynaptic membrane scaffold protein, which is positively correlated with dendritic spine density (Hutsler and Zhang 2010; Bringas et al. 2013). In all, the results of synaptic function and structural plasticity suggest that the synaptic plasticity from CA3 to CA1 has been seriously damaged in the VPA group.

The hippocampus is a key part of cognition and memory. A study has shown that the hippocampus of children with autism is abnormally enlarged (Groen et al. 2010). There is also a study showing an increase in microglia in the DG region of autistic model mice (Matta et al. 2020). Moreover, neuroinflammation occurs in VPA-induced autism rats, and inhibiting neuroinflammation could improve autism-like symptoms (Lucchina and Depino 2014). Therefore, it is necessary to explore the treatment and interference of autism in future research, as well as analyze the pattern of neural activity under different treatments.

Conclusion

VPA-induced autism significantly damages the memory ability through seriously inhibiting the hippocampal synaptic plasticity, which is thoroughly relevant to the disturbed patterns of neural oscillation, especially distorted the pattern of cross-frequency neural information flow both in the internal neural circuits of the hippocampus and the pathway of EC-hippocampus. The impairment of functional and structural of synaptic plasticity with the disturbance of neural oscillations confirms each other, and both coincide with the impairment of cognitive and memory ability in autistic-like rats. In addition, as a fast and convenient technology, diagnosing autism can be more effectively achieved by using neural oscillation analysis.

Author contributions

Conceived and designed the experiment: TZ, ZY; Performed the experiments: BC, XXX, CHL, YW; Analyzed the data: BC, XXX; Wrote the original manuscript and the revised version of manuscript: BC, TZ.

Data availability

The data that support the findings of this study are available from the corresponding author (TZ) upon reasonable request.

Declarations

Competing interests

The authors declare that they have no competing interests.