Activation of perineuronal net-expressing excitatory neurons during associative memory encoding and retrieval

Perineuronal nets (PNNs), proteoglycan-rich extracellular matrix structures, are thought to be expressed around inhibitory neurons and contribute to critical periods of brain function and synaptic plasticity. However, in some specific brain regions such as the amygdala, PNNs were predominantly expressed around excitatory neurons. These neurons were recruited during auditory fear conditioning and memory retrieval. Indeed, the activation of PNN-expressing excitatory neurons predicted cognitive performance.

regions. The brain regions are associated with emotional memory formation 14 . Therefore, we reasoned that PNNexpressing excitatory neurons contribute to fear memory.   Table S1), and all PNN-expressing PV-negative (PNN-PV − ) neurons expressed CaMKII ( Fig. 2A,B and Table S1) and neurogranin 15 (Fig. S1B), but showed no immunoreactivity for GABA (Fig. S1C). These data demonstrate that the population of PNN neurons in the LA and BA consists of two subpopulations: PNN-PV-positive (PV + ) inhibitory neurons and PNN-PV − excitatory neurons. Therefore, an anti-PV antibody was used in the following experiments to distinguish between PNN-expressing inhibitory neurons and PNN-expressing excitatory neurons in the LA and BA. There were more PNN-PV − neurons than PNN-PV + neurons in these regions, except for the rostral portions of the LA and BA (Fig. 2C,D). Moreover, a large percentage of PV + neurons in the medial and caudal portions of the BA were not sheathed by PNNs (Fig. S2), in contrast to previous descriptions of PV + neurons in the neocortex and the hippocampus 16,17 . Because the proteoglycan aggrecan is a major PNN component 18 , we next examined expression of aggrecan in WFA-labeled neurons in the LA and BA. Aggrecan immunoreactivity largely overlapped with WFA-labeling (Fig. 2E, upper). High-magnification images showed that PNN-CaMKII + neurons exhibited immunoreactivity for aggrecan (Fig. 2E, lower). The pattern resembled that seen for WFA (Fig. 2E, lower). Figure 2F summarizes the distribution of PNN-PV − and PNN-PV + neurons in 12 brain slices obtained from nine animals. PNNs in the medial and caudal regions of the amygdala exhibited an unusual distribution pattern. Whereas a large proportion of PNN-PV + neurons were localized in the dorsal LA, PNN-PV − neurons were spatially concentrated in the ventral LA and in the lateral portion of the BA (Fig. 2F). The mean distance r ( ) from a given PNN-PV − neuron to its nearest neighbor in the LA was 89.9 ± 7.5 μ m, and that in the BA was 84.9 ± 4.4 μ m (Fig. 2G). Assuming that PNN-PV − neurons are randomly distributed, the expected mean distance r ( ) E in the LA was 164.8 ± 11.4 μ m and that in the BA was 169.1 ± 9.3 μ m (LA: n = 8; t-test, t 7 = 3.97, P = 0.0081; BA: n = 9; t-test, t 8 = 6.04, P = 0.003; see Methods), suggesting that PNN-PV − neurons are spatially clustered. PNN-expressing excitatory neurons express c-Fos during fear conditioning. We next measured c-Fos expression in PNN-PV − neurons in the ventrolateral subdivision of the LA after fear conditioning and after a fear memory retrieval test (Fig. 3A). Mice were exposed to five pairings of a white-noise cue (conditioned stimulus, CS) either alone or combined with a mild foot shock (unconditioned stimulus, US), which resulted in an increase in freezing behavior with successive CS-US pairings, but not with CS-alone (Fig. 3B). Fear conditioning resulted in a significant increase in the total number of c-Fos + neurons (51.3 ± 3.7 neurons/ area) compared with that in the control groups (home cage: 12.2 ± 2.1 neurons/area, n = 6; one-way ANOVA,  Fig. S3). More PNN-CaMKII + neurons than CaMKII + neurons without PNNs expressed c-Fos after training (WFA + CaMKII + : 61.4% ± 6.3%, n = 8; WFA -CaMKII + : 9.65% ± 1.2%, n = 8; Wilcoxon paired signed-rank test, P = 0.000931; Fig. S4). On the other hand, c-Fos expression did not differ between PV + neurons with and without PNNs (WFA + PV + : 33.5% ± 5.4%, n = 6; WFA -PV + : 26.6% ± 2.0%, n = 6; Wilcoxon paired signed-rank test, P = 0.31; Fig. S4). Therefore, PNN-CaMKII + neurons are more likely to be recruited during fear conditioning.
Activation of PNN-expressing neurons correlates positively with fear memory. Twenty-four hours after fear conditioning, mice were re-exposed to the training cue.  Compared with animals in their HC and those exposed to CS-alone and IS, c-Fos is localized to PNN-PV − neurons in mice subjected to fear conditioning. Colors as in (D). Error bars indicate the SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; one-way ANOVA with Tukey's post hoc test.
PNN-PV − neurons during memory retrieval was negatively correlated with the activation of neighboring neurons in the CS-US group (Fig. S6A), but showed a weak but positive correlation in the CS-alone group (Fig. S6B). These findings indicate that PNN-PV − neurons are activated after fear memory retrieval, regardless of whether surrounding neurons are activated. Importantly, the normalized probability of c-Fos expression in PNN-PV − neurons correlated positively with freezing within the CS-US group (Fig. 4C), and even within both the CS-US and CS-alone groups (Fig. S6C), suggesting that greater activation of PNN-PV − neurons is linked to increased fear memory.

Discussion
To date, researchers have mainly examined the role of PNNs in inhibitory neurons 19 since PNNs were barely detectable in association with excitatory neurons [19][20][21][22] . However, we found that PNNs, as determined by aggrecan expression and WFA-labeling, were present mainly in association with excitatory neurons in brain areas related to emotional memory function, suggesting that PNNs modulate the function of excitatory neurons.
PNNs limit plasticity in the adult central nervous system 2,23 . However, this function seems to differ depending on the brain region and animal species. For example, PNN degradation allows the induction of long-term potentiation in the hippocampal CA2 area where plasticity is normally limited 5 , and, likewise, the enhancement of long-term depression in the perirhinal cortex 8 . By contrast, the digestion of PNNs leads to the impairment of long-term potentiation at thalamo-LA synapses 7 and the absence of enhanced plasticity in the feline visual cortex 24 . Since the types of PNN-expressing neurons differ according to the brain region, cell type-specific degradation of PNNs could provide insight into how PNNs play a role in plasticity, including fear memory.
We showed here that excitatory neurons expressing PNNs are functionally different from those lacking PNNs. PNN-PV − excitatory neurons are activated during fear conditioning, although the precise mechanism remains unclear. WFA recognizes 4-O-sulfation on chondroitin sulfate chains 25 . The deposition of chondroitin 4-sulfate on neurons causes membrane depolarization, which may help neurons reach the threshold for spike firing 26 , thereby lowering the threshold of c-Fos expression. Alternatively, since the meshwork of PNNs surrounds synaptic contacts at thalamocortical boutons 25 , synaptic inputs from both the thalamus and cortex may converge on PNN-PV − neurons. Our data do not rule out PNN-PV − neuron activation due to sensory stimuli because exposure to tone or foot shock alone resulted in the upregulation of c-Fos in these neurons. However, PNN-PV − neuron activation was affected by the activity of neighboring neurons in the CS-alone group, but not in the CS-US group. Furthermore, the population of PNN-PV − neurons activated during retrieval was higher in the paired CS-US group than in the unpaired CS-US group. Hence, these neurons may encode CS-US associations. While almost nothing is known about the participation of PNN-PV − neurons in memory traces, our data suggest that a memory engram is distributed in a subpopulation of neurons, PNN + neurons, in the LA.

Methods
Animals. Male C57BL/6J mice (aged 5-10 weeks; SLC, Hamamatsu, Japan) and GAD67-GFP knock-in mice 12 were used in this study. The mice were housed in plastic cages under a 12 h light/dark cycle at 24 °C and had free access to water and food. All procedures were approved by the animal experiment ethics committee of Hoshi University and performed in accordance with the Hoshi University guidelines for the care and use of laboratory animals. All experimental procedures minimized the number and suffering of the animals.
Immunohistochemistry. Mice were anesthetized using intraperitoneal urethane (1.25 g/kg) and transcardially perfused with phosphate-buffered saline (PBS; pH 7.4) followed by 4% paraformaldehyde in PBS. The brains were post-fixed overnight at 4 °C in the same fixative solution, transferred to 30% sucrose in PBS for 48 h, and then cut into 30 μ m coronal sections using a cryostat (CM1510-11, Leica Co., Wetzlar, Germany). The free-floating sections were blocked with 5% bovine serum albumin (BSA) and 0.3% Triton X-100 in PBS for 1 h at room temperature and incubated overnight at 4 °C with the following antibodies: mouse monoclonal anti-parvalbumin Behavioral procedures. The day before fear conditioning, male C57BL/6J mice were allowed 10 min to freely explore the conditioning chamber (box A), which consisted of a transparent plastic box with a stainless steel grid floor (16 × 14 × 12 cm). For training, mice were placed in box A, and, after 2 min, they were presented with five pairings of a 30 s white-noise CS (65 dB, 1 Hz) that co-terminated with a 1 s foot shock US (0.3 mA) at a variable interval of 20-120 s. Mice in the unpaired group received explicitly unpaired training with the same number and specifications of CS and US stimuli. Thirty seconds after the final CS-US for the paired group or the final CS for the unpaired group, mice were returned to their home cage. The CS-alone group was exposed to the same protocol without the US. Mice in the IS group were given three consecutive 1 s foot shocks (0.3 mA) at an interval of 1 s immediately after placement in the conditioning chamber and quickly returned to their home cage. The next day, mice were transported to a novel chamber (box B), which consisted of a white plastic box (17 × 10 × 10 cm), and tested for freezing behavior when re-exposed to the CS. All sessions were video-recorded for automatic scoring of freezing (TimeFZ software, O'HARA & Co., LTD., Tokyo, Japan), which was defined as a period of immobility, except for respiratory-related movements, lasting for at least 2 s. Analysis of c-Fos. Mice were euthanized for immunohistochemical analysis 90 min after completing the behavioral task. The c-Fos-immunostained sections from − 1.6 mm to − 2.0 mm posterior to bregma were imaged at magnification × 20 and × 40 using an FV1200 confocal laser scanning microscope (Olympus, Tokyo, Japan). The c-Fos + neurons in the ventrolateral subdivision of the LA were counted bilaterally using three slices from each mouse. The total number of c-Fos neurons in the LA was determined in an area of 59,862 ± 1,814 μ m 3 using ImageJ software (version 1.48; National Institutes of Health, Bethesda, MD, USA) and averaged within each mouse. The background fluorescence cut-off for all images was equal. The percentage of c-Fos + neurons in PNN-PV − neurons was calculated as the percentage of c-Fos + neurons per total PNN-PV − neurons in the LA, and was also normalized by the total number of c-Fos + neurons in the same region.
Statistical analysis. The results are expressed as the mean ± SEM. Statistical comparisons were analyzed with Student's t-test (paired data, normal distribution), Wilcoxon paired signed-rank test (non-normal distribution), or one-way ANOVA followed by Tukey's post hoc test (comparisons of multiple groups). Correlations were calculated using a linear correlation. No statistical methods were used to predetermine sample sizes, but the sample sizes used were similar to those generally reported in the field for similar experiments 9,17 . The criterion for statistical significance was set at P < 0.05.