We used male 8–12 weeks old Long-Evans rats (Institute for Animal Reproduction, Kasumigaura, Ibaraki, Japan, http://www.iar.or.jp) weighing 300–500 g each. These rats were housed in groups in a temperature- and humidity-controlled room under a 12-h-light/12-h-dark cycle with food and water available ad libitum. Efforts were made to minimize the number of animals used, as well as their pain and discomfort. All animal treatments were approved by the Tohoku University Committee for Animal Research, Seiryo Campus (Sendai, Japan).

To block axonal transport, colchicine (Sigma-Aldrich, Tokyo, Japan) dissolved in saline (100 μg/10 μl) was stereotactically injected 5 μl of solution into the bilateral lateral ventricle of the brain. Control group was injected equal (5 μl) volume of saline into the bilateral lateral ventricle of the brain. The surgery was performed under inhalation anesthesia with isoflurane (1–3%, Pfizer, Tokyo, Japan). 2 days later, the brains were processed for immunohistochemical investigation, as described below.

Under deep nembutal anesthesia (Somnopentyl, 250 mg/kg body weight, intraperitoneally, Kyoritsu Seiyaku Co., Ltd., Tokyo Japan), the rats were transcardially perfused with saline containing 2 IU/mL heparin and then with 4% paraformaldehyde in 0.1M phosphate buffer (pH 7.4). The brains were removed and post-fixed in formaldehyde solution for 24 h at 4°C. Then, the brains were immersed in 30% sucrose solution in phosphate-buffered saline (PBS) for at least 24 h, and coronal sections (30 μm thick) were cut using a cryostat (Leica CM1950, Leica Microsystems, Nussloch, Germany).

The sections were washed in PBS for 10 min. After blocking with 3% bovine serum albumin in PBS for 1 h, the sections were incubated with primary antibodies overnight at 4°C and washed with PBS. The following antibodies were used in this study: mouse monoclonal anti-GAD65 antibody (1:300, BD Biosciences, San Jose, CA, United States), rabbit polyclonal anti-parvalbumin antibody (1:500, Abcam, Cambridge, United Kingdom), rabbit polyclonal anti-somatostatin antibody (1:500; Peninsula Lab, San Carlos, CA, United States), rabbit polyclonal anti-calretinin antibody (1:500, Proteintech group, Inc., Chicago, IL, United States), rabbit polyclonal anti-cholecystokinin antibody (1:100, Cloud-Clone Corp, Carlsbad, CA, United States), rabbit polyclonal anti-neuropeptide Y antibody (1:100, Proteintech group, Inc.), rabbit polyclonal anti-neuronal nitric oxide synthase antibody (1:800, Merck-Millipore, Temecula, CA, United States) and rabbit polyclonal anti-vesicular GABA transporter (VGAT) antibody (1:500, Gene Tex, Los Angeles, CA, United States). After washing three times with PBS, the brain sections were incubated with an appropriate species-specific secondary antibody for 1 h−Alexa Fluor 488-conjugated goat anti-mouse immunoglobin (Ig) G (1:200, Abcam) or Alexa Fluor 647-conjugated goat anti-rabbit IgG (1:200, Merck-Millipore)−and then with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) solution (1:500, Dojindo, Kamimashiki, Japan). After three washes with PBS, the sections were mounted using Fluoromount (Diagnostic BioSystems, Pleasanton, CA, United States).

We used IBM SPSS Statistics for Windows software (version 21.0 [released 2012], IBM Corp., Armonk, NY, United States) for statistical analysis and performed Welch’s t-test, Mann–Whitney U Test, two-way analysis of variance (ANOVA) followed by Bonferroni’s post hoc tests, and Pearson’s correlation test. Values are presented as the mean ± standard error of the mean (SEM). Groups were compared using thresholds for significance of 0.05 and 0.01.

We immunostained for GAD65 in the cerebral cortex; a typical image captured via confocal microscopy is shown in Figure 2A (GAD65: green; DAPI: blue). GAD65 expression was detected throughout the imaged region, except in the corpus callosum. All six GABAergic subtypes were detected across layers II–VI. On the other hand, only CR⁺ and nNOS⁺ neurons were detected in layer I. We present representative high-magnification images of the six GABAergic subtypes in the superficial (II–IV) layer in Figures 2B–G, which show triple signals of GAD65 immunofluorescence (green), the GABAergic subtype (magenta), and DAPI staining (blue).

We plotted GAD65 intensity among the six subtypes present in those three layers in Figure 3. Further details are provided in Table 1.

Next, we immunostained for GAD65 in the CA1 region of the hippocampus; a typical image from confocal microscopy can be seen in Figure 4A (GAD65: green; DAPI: blue). GAD65 expression was detected across the entire region excluding the corpus callosum. We detected all six GABAergic subtypes, and high-magnification examples of each subtype as observed in the superficial (SR) layer can be found in Figures 4B–G, which show triple signals of GAD65 immunofluorescence (green), the GABAergic subtype (magenta), and DAPI staining (blue).

We plotted the GAD65 staining intensity among the six subtypes and three layers in Figure 5. Further details are provided in Table 2.

We examined GAD65 expression in somata but not in axon terminals. Investigating GAD65 expression in axon terminals is difficult because the SOM, NPY, and nNOS proteins cannot be clearly detected in axon terminals under confocal microscopy. To detect GAD65 expression throughout the cytoplasm including in axon terminals, we injected colchicine, an inhibitor of microtubule polymerization, into the lateral ventricle to block axonal transport. At 2 days after colchicine injection, the GAD65 staining (green) intensity increased in somata in the cerebral cortex (Figures 6A-I, B-I, arrowhead) and hippocampus (Figures 6G-I, H-I, arrowhead). On the other hand, the intensity of GAD65 decreased in the neuropil region of the cerebral cortex (Figures 6A-II, B-II, arrows) and hippocampus (Figures 6G-II, H-II, arrows). Welch’s t-test showed the intensities of GAD65 clusters were significantly lower in colchicine-injected brains (Figure 6C, cerebral cortex: p = 0.001; Figure 6I, hippocampus: p = 0.000). In additions, the number of VGAT clusters (magenta) were not changed in colchicine-injected brains (Figure 6D, cerebral cortex; Figure 6J, hippocampus). VGAT is localized in inhibitory axon terminal (Chaudhry et al., 1998), suggesting that GAD65 was not transported to axon terminals and instead accumulated in somata. In these brains, we measured GAD65 expression in the six subtypes in the cerebral cortex (Figure 6E) and hippocampus (Figure 6K) and compared the results with control brains. In colchicine-injected brains, the intensities of GAD65 staining in somata were dramatically increased, but their distributions throughout the somata did not appear to show the major changes. Therefore, we determined that same quantitative method as control is applicable in them. GAD65 expression was compared using the mean value of the three layers in both colchicine-injected and control brains. Welch’s t-test showed the intensities of GAD65 in somata were significantly higher in colchicine-injected brains (Figure 6E, cerebral cortex: PV, p = 0.000; SOM, p = 0.004; CR, p = 0.026; CCK, p = 0.049; NPY, p = 0.000; and nNOS, p = 0.029; Figure 6K, hippocampus: PV, p = 0.001; CCK, p = 0.005; NPY, p = 0.003; and nNOS, p = 0.006). The data of CR did not show normal distribution, and was detected the significant difference using Mann–Whitney U Test (p = 0.008). There was no significant difference in SOM⁺ cells between colchicine-injected and control brains. Further details are provided in Table 3. Pearson’s test showed significant positive correlations in the expression profile between colchicine-injected and control brains (Figure 6F, cerebral cortex: r = 0.780, p = 0.003; Figure 6L, hippocampus: r = 0.889, p = 0.000).

Based on these results, GAD65 expression in the soma and cytoplasm exhibited similar tendencies across all GABAergic subtypes, and the level of GAD65 expression in the soma can be used as a proxy for the level of GAD65 expression in the cytoplasm for all GABAergic subtypes.

Our results revealed that nNOS⁺ and NPY⁺ neurons express GAD65 at particularly high levels, whereas some major subtypes (PV⁺/SOM⁺/CR⁺) express GAD65 at low levels in the cerebral cortex and hippocampus. This result is consistent with previous reports of PV⁺ neurons having lower GAD65 expression than non-PV neurons in the mouse hippocampus (Fukuda et al., 1997). Candidates of these non-PV neurons having high GAD65 may include the nNOS⁺ and NPY⁺ neurons based on our findings. There was another previous report comparing synaptic-GAD65 and GAD67 expression among PV⁺ basket neurons, PV⁺ chandelier neurons and Calbindin-positive (CB⁺) basket neurons in the monkey prefrontal cortex (Fish et al., 2011). They shows the higher synaptic expression of GAD65 in PV⁺ basket neurons than in PV⁺ chandelier neurons, and the same level of synaptic expression of GAD65 in PV⁺ basket neurons as in CB⁺ basket neurons. Because of different classification of GABAergic subtypes, it would be necessary to clarify distinct patterns of GAD65 expressions in synapse levels at multiple GABAergic subtype.

Neuropeptide Y is a neuromodulator with anti-seizure activity (Erickson et al., 1996; Baraban et al., 1997; Vezzani et al., 1999). Its anti-epileptic functions occur during the pre- and post-synaptic inhibition of excitatory neurons via Y1, Y2, and Y5 receptors (Baraban, 2004). NPY belongs to the G-protein-coupled inwardly rectifying potassium receptor family (Paredes et al., 2003). The functions of neuromodulator/GABA co-release remain unclear, but generally, GABA transmission contributes to fine-tuning of the synaptic effects of neuromodulators (Horvath et al., 2001; Yu et al., 2015; Tritsch et al., 2016). Many NPY⁺ neurons are excited by GABA in the hilus of the hippocampus (Fu and van den Pol, 2007), suggesting that NPY⁺ cells are activated by GABA, which results in the increased release of not only GABA but also the co-expressed inhibitory neuromodulator NPY. Based on these reports, we suggest that the neural inhibitory functions of NPY are regulated by GAD65-dependent GABA release in pathological situations.

The pharmacological functions of nitric oxide (NO) in epileptic seizures remain controversial, but NO has been considered to act as a mediator of neurotoxic effects in several reports (Dawson et al., 1991; Marangoz et al., 1994; Penix et al., 1994; Schuman and Madison, 1994). nNOS has the function of evoking the release of several neurotransmitters, including acetylcholine, catecholamines, neuroactive amino acids and GABA (Kahn et al., 1997a,b). NO-evoked neurotransmitter release is mediated by two distinct release systems, a Ca²⁺-dependent system and the reverse process of a Na⁺-dependent carrier-mediated transport system (Garthwaite, 1991; Kuriyama and Ohkuma, 1995). NO has effects of decreasing GABA transaminase activity and increasing GABA expression (Jayakumar et al., 1999; Vega Rasgado et al., 2018). The relationship between presynaptic NO release and GABA remains largely unknown, but as with NPY, NO, and GABA release may be related to the regulation of overexcited neurons in some pathological states. The high concentrations of GAD65 in nNOS⁺ and NPY⁺ neurons may be needed to support a rapid response to neuromodulators when necessary.

Additionally, SOM is an anti-epileptic neuropeptide in the hippocampus (Boehm and Betz, 1997; Schweitzer et al., 1998; Tallent and Siggins, 1999). GABA-mediated activation of NPY⁺ neurons may increase the release of SOM (Fu and van den Pol, 2007), as SOM/NPY co-expression has been observed in a subset of GABAergic neurons in the hippocampus (Köhler et al., 1987). However, GAD65 expression was not high in SOM⁺ neurons according to our results, indicating that the release of SOM is not associated with GAD65-dependent GABA release.

In the cerebral cortex, we found that GAD65 expression is higher in the superficial layer than in the other layers for all GABAergic subtypes. Pyramidal neurons project their dendrites into the superficial layer to receive amino-acid or monoamine signals from other parts of the cortex or extra-cortical regions. Inhibitory dendritic signals from local GABAergic neurons are essential for mediating these inputs and maintaining a normal neural state (Northoff and Mushiake, 2020). Therefore, high GAD65 expression in neurons in the superficial layer may play a crucial role in suppressing cortical over-excitation. This idea is consistent with a report stating that local GABAergic inhibition drives the elasticity of ictal progression and that cortical seizures occur in superficial layers prior to deep layers during lateral seizure spread (Wenzel et al., 2017). Additionally, GAD65 expression was found to be higher in the pyramidal layer of the visual cortex of 1-week-old cats (Mower and Guo, 2001), but higher expression was detected in the superficial layer in adults (Guo et al., 1997). This pattern change may reflect flexible regulation of GAD65 for maintaining a normal state during the growth of neural circuits.

In the hippocampus, we observed that GAD65 expression was higher in the superficial layer than in the deep layer only in nNOS⁺ and NPY⁺ neurons. Dendrites of hippocampal pyramidal neurons in the superficial layer receive glutamatergic Schaffer collateral inputs from the CA3 region. In the hippocampal CA1 region, GABAergic neurons expressing nNOS/NPY are known as Ivy cells (Price et al., 2005; Tricoire and Vitalis, 2012). Ivy cells provide widespread synaptic and extra-synaptic slow inhibition of the dendrites of CA1 pyramidal neurons (Fuentealba et al., 2008; Lapray et al., 2012). The number of Ivy cells was significantly reduced in the hippocampus of pilocarpine-induced epileptic rats. Interestingly, the number of PV⁺ neurons was not changed in this model (Orbán-Kis et al., 2015). These findings suggest that nNOS⁺ and NPY⁺ neurons in the superficial layer may play a critical role in suppressing ictal progression in the hippocampus. Same as nNOS⁺ and NPY⁺ neurons, in PV⁺, SOM⁺, CR⁺ and CCK⁺ neurons, they also showed the tendency of distinct GAD65 expression among layers. These results are consistent with a previous report stating that both PV⁺ and non-PV neurons exhibit distinct laminar distribution of GAD65 in the deep and pyramidal layer (Fukuda et al., 1997).

We noted that strong GAD65 fluorescence often accumulated in the perinuclear region. This localization may reflect GAD65 protein undergoing a hydrophobic posttranslational modification and becoming anchored to the cytosolic face of the Golgi membranes (Kanaani et al., 2008).

Based on this feature, we considered it inappropriate to compare GAD65 expression through measurement of its mean intensity inside a traced outline of the soma, as described in previous report (Jinno and Kosaka, 2009). GABAergic neurons have different soma size depending on subtypes (González-Albo et al., 2001). This affects the ratio of the Golgi membrane occupying the soma differs among GABAergic subtypes. To avoid this problem, we quantified GAD65 expression from the average intensity of four squares (details are provided in the section “Materials and Methods”); this method is considered less susceptible to the effects of soma size.

GAD65 localizes to both the soma and axon terminal, and thus, the amount of GAD65 in the soma does not necessarily reflect the total number of GAD65 molecules throughout the cytoplasm. As reported previously, there are two possible drivers of high GAD65 expression in the soma: a high synthesis rate or slow transport to the axon terminals (Esclapez et al., 1993). The inhibition of axonal transport by colchicine enhances GAD65 immunoreactivity in the soma (Wang et al., 2014). We found that GAD65 expression in the soma across neuronal subtypes was significantly and positively correlated between colchicine-treated and control brains, suggesting that GAD65 expression in the soma can be used as a proxy for the overall amount of GAD65 in the cytoplasm. It is important to consider the possibility that the colchicine injections might have caused abnormal effects on not only axonal transport but also protein synthesis or metabolism, as colchicine is neurotoxic (Sutula et al., 1983). Neural excitation caused by colchicine might affect the GAD65 expression, because temporal lobe epilepsy caused by pilocarpine is reported to increase GAD65 expression (Esclapez and Houser, 1999). In additions, GABAergic neurons may have differing colchicine sensitivities based on their morphologies and localization patterns. In this study, we injected colchicine into the lateral ventricle of the brain. Layers near the lateral ventricle might have thus been vulnerable to drug effects. In future research, synaptic protein levels can be detected directly using super-resolution fluorescence microscopy. Complementary data should be obtained through several methods.

High GAD65 expression in NOS⁺ and NPY⁺ neurons has important implications. As noted in the introduction, GAD65 is associated with activity-dependent GABA release, whereas GAD67 is associated with spontaneous GABA release (Tian et al., 1999; Jinno and Kosaka, 2009). Our findings suggest that nNOS⁺ and NPY⁺ neurons are closely associated with activity-dependent GABA release. Among GABAergic subtypes, PV⁺ neurons are considered a pivotal subtype based on previous studies using subtype-specific GAD67-knockout or -knockdown models (Kuki et al., 2015; Lazarus et al., 2015). However, no studies to date have examined the functions of GAD65 in nNOS⁺ and NPY⁺ neurons. In humans, electroconvulsive therapy increases the seizure threshold (Sackeim, 1999) and GABA expression in the brain (Sanacora et al., 2003). According to our findings, GAD65 expression may increase in nNOS⁺ and NPY⁺ neurons, resulting in increased GABA expression and a higher seizure threshold. This study demonstrates the baseline level of GAD65 expression at the resting state of various GABAergic subtypes. In the future, the relationships between GABAergic subtypes and pathological states should be studied, as GABAergic subtypes play distinct roles in excitatory-inhibitory balance under normal and pathological conditions, including epileptic seizures and mental disorders.