Genetic deletion of the 67‐kDa isoform of glutamate decarboxylase alters conditioned fear behavior in rats

Inhibitory neurotransmitter GABA is synthesized by two enzymes, the glutamate decarboxylase 67‐kDa isoform (GAD67) and the 65‐kDa isoform (GAD65). Knocking out the Gad1 gene, which encodes GAD67, is lethal in mice, and therefore, the association between Gad1 and emotional function was unclear. In the present study, we assessed the impact of Gad1 knockout on fear expression and fear memory for the first time, using newly generated knockout rats.

and mood disorders [19], suggesting that decreased GABA neurotransmission may underlie the neural basis of anxiety in these disorders. This is also consistent with the findings that benzodiazepines, which are stimulants of GABA A receptors, reduce anxiety [20].
The mechanism by which the GABA-producing system contributes to anxiety and fear has been studied primarily in Gad2 knockout (KO) mice [15,[21][22][23][24][25][26][27]. This is partly because Gad1 KO mice exhibit a cleft palate and die on the first postnatal day [28]. In Gad2 KO mice, phasic inhibition in the basolateral amygdala, which is related to fear memory, is significantly reduced [15]. Consistent with these clinical findings, Gad2 KO mice exhibit increased anxiety-like behavior [22] and enhanced active defensive behavior during the fear-conditioning test [23]. Gad2 KO mice also display generalized fear [15,24] and impaired fear extinction [25]. In summary, enhanced fear response or fear memory has been consistently reported in Gad2 KO mice.
However, the role of Gad1 (and GAD67) in anxiety and fear is unclear. A study reported that the expression of GAD67 and GAD65 is differentially controlled in fear extinction; GAD67 is upregulated in the prefrontal cortex, while GAD65 is downregulated in the hippocampus during fear memory extinction, suggesting that they have a distinct role in the processing of fear memory [29]. Although fear conditioning is yet to be assessed, Gad1 'heterozygous' KO mice show anxiety levels comparable to those of wild-type (WT) mice [30,31]. It is unclear whether this suggests that GAD67 does not play a significant role in fear and anxiety, or whether a slight reduction in GAD67 levels in heterozygous KO mice is not sufficient to affect their behavior.
We recently obtained a novel tool to address this issue. Using genome editing technology [32], we successfully developed a Gad1 KO rat line in the Long-Evans background (Fujihara et al. [33]). The perinatal lethality of Gad1 KO rats was not as severe as that of Gad1 KO mice; approximately 33% of KO rats could grow up to adulthood. Therefore, in the present study, we characterized the behavioral phenotype of Gad1 KO rats using the fear-conditioning test to reveal the role of GAD67 in fear expression, fear memory, and fear extinction.

Animals
Gad1 KO rats were generated by a previously reported method [32]. The details of the generation procedure are reported elsewhere (Fujihara et al. [33]). The animals are available from NBRP-Rat (https://www.anim.med.kyotou.ac.jp/nbr/Default.aspx). Briefly, the exon 6 of the Gad1 gene was deleted in the Long-Evans background using the CRISPR/Cas9 system [34]. The rats used in the present study were obtained by crossing male and female Gad1 +/− rats. Only male Gad1 −/− (Gad1 KO) rats were used for behavioral analysis. Littermate Gad1 +/+ (Gad1 WT) rats were also used as a control group. Approximately 66% of these rats died at P0, while the remaining 33% survived and were able to grow into adulthood. None of the rats had epileptic seizures during the observation period. Genotyping PCR was performed using the following primers [33]: The rats were housed in a room maintained at 22 AE 3°C with a 12-h light-dark cycle (lights on at 6:00, lights off at 18:00). Food (CLEA Rodent Diet CE-2, Clea Japan, Meguro, Tokyo, Japan) and water were provided ad libitum. All the experiments in this study were approved by the Animal Care and Experimentation Committees of Gunma University and the Animal Research Committee of Osaka University. Every effort was made to minimize the number of animals used and their suffering.

Western blot analysis
Brain tissue was taken from adult rats, and the S1 fraction was prepared to confirm the Gad1 knockout. Western blot analysis using a standard protocol was performed as described previously [35]. An antibody that recognizes both GAD65 and GAD67 proteins was used as the primary antibody (1 μgÁmL −1 ) [36]. β-Actin was detected as an internal control using a mouse anti-β-actin antibody [1 : 10 000, #AC-15 (ab6276); Abcam, Cambridge, UK]. The results of the analysis are shown in Fig. S1, and details of the knockout validation are reported elsewhere [33].

Behavioral analysis
Fear-conditioning test for contextual and cued memory (C-C experiment) We refer to this experiment as the C-C (contextual and cued memory) experiment to discriminate it from another experiment. Five-month-old rats were used for the analysis (WT, n = 12; KO, n = 11). We used a modified version of our previous protocol for mice [35]. In the modification process, we also referred to a protocol by Ueno et al. [37]. The contextual and cued fear-conditioning test was conducted in a box surrounded by a sound-attenuated chamber (O'Hara & Co., Ltd., Tokyo, Japan). This test consisted of a conditioning trial (day 1), a contextual test trial (day 2), and 4-day cued test trials (day 3 to day 6) (Fig. 1). The rats were placed in a clear acrylic conditioning chamber (33 cm × 25 cm × 28 cm) equipped with a stainless-steel grid floor connected to an electric shock generator. The brightness of the chamber was 120 lux. On day 1, rats were habituated to the conditioning chamber for 180 s. Immediately after the habituation, they received five auditory tone presentations (CS; 10 kHz, 65 dB, duration: 20 s) with an 80-s interval, each of which was delivered concomitantly with a foot shock at the end of it (US; 0.7 mA, duration: 1 s). Because the acoustic startle response was comparable between Gad1 KO and WT rats [33], we can assume that the auditory function is not affected in Gad1 KO rats. One hundred seconds after the last foot shock, the rats were returned to their home cages (Fig. 1A). A day after the conditioning training, a contextual test (day 2) was performed in the same conditioning chamber for 600 s in the absence of the tone-shock pairings. Furthermore, a cued test was performed in a novel chamber with different contexts from those of the conditioning chamber to minimize freezing caused by contextual fear conditioning. The test chamber differed from the conditioning chamber in brightness (30 lux), color (black), and shape (triangular prism, 33 cm on one side, and 31.5 cm high). The cued test (day 3) was conducted 24 h after the contextual test. The schedule for the cue test was the same as that for the conditioning, except that no foot shock was administered. In the first 3 min of the exploration period (no CS), we evaluated nonspecific contextual fear. The acquired cued fear was tested through 5 CS periods (no foot shock). On day 4 to day 5, we repeated the same cued test to assess extinction learning (Fig. 1B). The percentage of time that mice exhibited a freezing response (immobility excluding respiration and heartbeat) was measured as an index of fear memory. The data were collected and analyzed using TimeFZ1 (O'Hara & Co., Ltd.). The freezing time was determined automatically by the software every 10 s.

Fear-conditioning test specialized in contextual memory (CTX experiment)
We refer to this experiment as the CTX (contextual memory) experiment to discriminate it from the C-C experiment. Different cohorts other than the C-C group were used (WT, n = 8; KO, n = 8). At least 1 week before the test, we performed an open field test, elevated plus-maze test, and flinch-jump test as described in the following. We used the same apparatus as the C-C experiment. This test consisted of a conditioning trial (day 1) and 4-day contextual test trials (day 2 to day 5) (Fig. 6). In the CTX experiment, we assessed only contextual memory. Therefore, we omitted auditory cues, and the electric shocks were given at random intervals (210, 370, 470, 520, and 615 s from the start of the experiment). The total duration of the experiment on day 1 was the same as that in the C-C experiment (700 s). The contextual tests (day 2 to day 5) were performed in the same conditioning chamber for 600 s in the absence of electric shocks to evaluate fear extinction.

Open field test
A 90 cm × 90 cm open field box (O'Hara) was used for the test. The brightness at the center was 120 lux. Each rat was allowed to walk freely for 10 min in the box. The behavior of the rats was recorded by a CCD camera attached to the ceiling. The distance travelled during a session and the stay time in the central area (25% of the field) were measured using the automated software TIMEOFCR1 (O'Hara).  From the third day to the sixth day, the same tones as the first day were delivered to the rats in a different chamber with a different shape and context (cued test and fear extinction).

Elevated plus-maze test
were allowed to walk freely on the entire maze for

Flinch-Jump test
Pain sensitivity can affect the results of the fear-conditioning test. To determine the sensitivity to the electric shock, we carried out a flinch-jump test based on Lehner et al. [38]. Shock naïve rats were given a series of ascending foot shocks from 0.1 to 0.7 mA with a 0.1 mA increment to determine the threshold current for flinch and jump. The interval between shocks was 10 s, and each animal was tested only once. The test was carried out in a different environment from the fear-conditioning test.

Statistical analysis
Graphs were generated using GRAPHPAD PRISM (GraphPad Software, Inc., La Jolla, CA, USA). We compared the two genotypes using Welch's t-test and two-way ANOVA with a simple main effect test. For the ANOVA, we first examined the interaction between the genotype and intraindividual factors. If the interaction was statistically significant, we performed simple main effect tests with Bonferroni correction. If the interaction was not significant, we assessed the main effect of each factor. We also calculated effect size Cohen's d for the t-test and partial η-squared (η 2 p ) for ANOVA. Welch's t-test and ANOVA were performed using R software (https://www.r-project.org/) and SPSS (SPSS Inc., Chicago, IL, USA), respectively. In addition, we calculated a cross-correlation function between the time course of the tones and freezing to evaluate their synchronization. We also performed analyses using general linear models (GLM) to assess the relationship between two or more parameters. The cross-correlation function and GLM analyses were calculated using R software. P-values less than 0.05 were considered significant in all statistical tests in the present study.

Results
Altered freezing behavior of Gad1 KO rats during the conditioning process in the C-C experiment We confirmed the loss of GAD67 protein in the whole brain of Gad1 KO rats using western blot analysis ( Fig. S1; Fujihara et al. [33]). Then, we performed the fear-conditioning test, as shown in Fig. 1. On day 1, after 180 s of habituation to the conditioning chamber, the rats were exposed to five tone-shock pairings (Fig. 1A). During the first 180 s (baseline), no significant difference in freezing time was observed between the two genotypes ( Fig. 2A,D; t = 1.3622, df = 18.85, P = 0.1892, Cohen's d = 0.5751). However, the responsiveness to tone-shock pairings differed between the genotypes ( Fig. 2A). To compare the two genotypes more easily, the time courses of freezing around tone-shock pairings were averaged, as shown in Fig. 2B. WT rats showed increased freezing with the onset of tone, and freezing increased again after a transient decrease caused by the shock (Fig. 2A,B: blue line). In contrast, KO rats showed a partial 'antiphase' pattern; higher freezing during the pretone period and decreased freezing with the onset of the tone. Immediately after the shock, the freezing time of KO rats rapidly increased to a higher level than that of WT rats ( Fig. 2A,B: red line).
We assessed the above-mentioned differences using cross-correlation analysis. Cross-correlation is a wellknown tool for detecting the synchronicity of neuronal activities [24]. In the present study, we utilized this method to assess the similarity between the tone signal ( Fig. 2A: green line) and the time course of the freezing behavior. The cross-correlation function of WT rats had positive peaks every 100 s around lag = 0 s. This reflected the synchronous increases in the freezing time of WT rats with the tones. However, in KO rats, these positive peaks disappeared and had negative values. This is consistent with a downward trend in freezing with the onset of the tone. The shapes of the cross-correlation functions of the two genotypes were significantly different (genotype × lag, F (28, 588) = 6.628, P < 0.001, η 2 p = 0.249), and a simple main effect test revealed several areas that were significantly different between the two groups, mainly at the positive peaks in WT rats (Fig. 2C).
The mean freezing values during the pretone (20 s before the tone) and tone (20 s) periods are summarized in Fig. 2D. There was a significant interaction between genotype and time period (pretone or tone), with a large effect size (genotype × period, F (1, 21) = 17.335, P < 0.001, η 2 p = 0.452). A simple main effect test showed that pretone freezing was significantly higher in KO rats than in WT rats (P = 0.005). In addition, WT rats showed significantly increased freezing from the pretone to tone (P = 0.003), while KO rats showed significantly decreased freezing from pretone to tone (P = 0.018). The total freezing time during the entire conditioning process was also significantly longer in KO rats than in WT rats, with a large effect size (WT, 25.63 AE 16.27%; KO, 36.74 AE 7.55%; t = 2.1278, df = 15.811, P = 0.04945, Cohen's d = 0.8625). In summary, although responsiveness to the tone was unusual, the fear response itself was increased in KO rats. Furthermore, at the very least, Gad1 KO rats were considered to have normal hearing as they responded to the tones.
Increased freezing of Gad1 KO rats in the contextual test in the C-C experiment A contextual test was performed 24 h after the training (Fig. 1B). The total freezing time was significantly prolonged in KO rats compared with that in WT rats, and a very large effect size was observed ( Fig. 3A; t = 4.498, df = 19.866, P < 0.001, Cohen's d = 1.8499). The freezing time during the pretone period on day 1 was positively associated with the freezing time during the contextual test ( Fig. 3B; standardized regression coefficient B = 0.7958, t = 6.023, df = 21, P < 0.001, adjusted R 2 = 0.6159). Considering that the baseline freezing level on day 1 was slightly higher (although not statistically significant) in KO rats with a moderate effect size, the baseline activity might have affected the result on day 2. Thus, we performed multiple regression analyses using the baseline freezing level as a covariate. The standardized partial regression coefficient (β) for the pretone freezing time was still significant (β pretone = 0.9540, t = 5.349, df = 20, P < 0.001, Adjusted R 2 = 0.6279), and that for the baseline freezing was not significant (β baseline = −0.2311, t = 1.296, df = 20, P = 0.21). These results indicate that pretone freezing during conditioning was a good predictor of freezing in the contextual test; in other words, the fear response on day 1 was recalled in the contextual test.
Altered fear response of Gad1 KO rats during the cued test We conducted a cued test on day 3 (Fig. 1B). The tone schedule for the cued test was the same as that for the While the freezing behaviors during baseline (0-180 s) and the tone period in KO rats were comparable to those of WT rats, KO rats exhibited significantly excessive freezing during the pretone period. The results are presented as the average AE SEM. Data were analyzed using two-way ANOVA (C, D) or t-test (D; baseline only). *P < 0.05, **P < 0.01, ***P < 0.001. In (D), red and blue asterisks represent intragroup significant differences. WT, wild-type; KO, knockout.
training, except that no foot shock was given. Unlike on day 1, the baseline freezing time was significantly higher in KO rats than in WT rats (Fig. 4A,D; t = 5.0168, df = 17.884, P < 0.001, Cohen's d = 2.1260). Although WT rats showed an increase in freezing time after the tone, the trend was not clear in KO rats (Fig. 4A,B).
Cross-correlation analysis revealed a peak every 100 s around lag = 0 s in WT rats. This implied an increase in freezing in response to the tone. However, this synchronization was not observed in KO rats. The shape of the cross-correlation function of the two genotypes was significantly different (genotype × lag, F (28, 588) = 2.531, P < 0.001, η 2 p = 0.108), and the simple main effect test revealed several areas that were significantly different between the two groups, mainly at the peaks in WT rats (Fig. 4C).
The mean freezing values during the pretone period (20 s before the tone) and during the tone period (20 s) are summarized in Fig. 4D. There was a significant interaction between the genotype and time period (pretone or tone), with a large effect size (genotype × period, F (1, 21) = 5.783, P = 0.025, η 2 p = 0.216). Freezing during the tone period was significantly lower in KO rats than in WT rats (P = 0.042). Although WT rats showed increased freezing when transitioning from pretone to tone (P < 0.001), KO rats showed no change (P = 0.545).
Although the increase in freezing evoked by auditory cues was lost in Gad1 KO rats, the average freezing during the tone period significantly increased from the baseline in both genotypes (KO, paired t-test, t = 7.0176, df = 10, P < 0.001, Cohen's d = 2.4253; WT, t = 10.885, df = 11, P < 0.001, Cohen's d = 2.7307). These results suggest that KO rats may also express a certain amount of fear during the tone period.

Impaired contextual fear extinction of Gad1 KO rats during the CTX experiment
Next, we assessed fear extinction in contextual memory. Because the freezing level in the contextual test was relatively low in the C-C experiment (Fig. 3A), we modified the protocol to be specialized for contextual memory in the CTX experiment (Fig. 6). The baseline freezing was similar in both genotypes on day 1 (t = 0.0081, df = 13.937, P = 0.9937, Cohen's d = 0.0040). Although the average freezing time during the shock period (210-700 s) in KO rats was significantly lower than that in WT rats (t = 3.3198, df = 13.947, P < 0.01, Cohen's d = 1.6599), the freezing time in both genotypes increased after the shocks compared with their baseline levels (Fig. 7A,B).
A contextual test was performed 24 h after conditioning, as in the C-C experiment (Fig. 6A). Unlike in the C-C experiment, the total freezing time in KO rats was comparable to that in WT rats ( Fig. 7C; t = 0.1174, df = 10.683, P = 0.9087, Cohen's d = 0.0587). Repetition of the contextual tests for a total of 4 days up to day 6 resulted in a significant decrease in freezing time in WT rats. On the other hand, KO rats showed no decrease in freezing time for 4 days (Fig. 7D). We found that the genotype and day had significant main effects, and there is a subthreshold interaction between genotype and day (genotype effect, F (1, 14) = 6.622, P = 0.022, η 2 p = 0.321; day, F (3, 42) = 3.341, P = 0.028, η 2 p = 0.193; genotype × day, F (3, 42) = 2.739, P = 0.055, η 2 p = 0.164). Simple main effect tests showed a subthreshold difference on day 3 (P = 0.071) and significant differences between the two groups on day 4 (P = 0.031) and day 5 (P = 0.004). These results indicated that fear extinction of contextual memory was impaired in Gad1 KO rats.

Open field test and elevated plus-maze test
Because locomotor activity and anxiety level can affect fear-related behaviors, we performed an open field test and elevated plus-maze test in the same subjects as in the CTX experiment. Although Gad1 KO rats showed significantly reduced locomotor activity in the open field test, their anxiety-like behaviors in the elevated plus-maze test were not altered (Fig. S2).
To adjust for possible confounding by the difference in locomotor activity between the two genotypes, we reanalyzed the data of contextual fear extinction using GLMs, which include the distance travelled in the open field test as a covariate. The results are shown in Table S1. Particularly on day 5 in the CTX experiment, the genotype still had a significant effect (P < 0.05) after adjusting for the distance travelled. On the other hand, the distance travelled itself had no significant effect on freezing (P = 0.8132). Although we should acknowledge the confounding effect of locomotor activity on freezing as a limitation of the present study, these GLM results suggest that the higher level of freezing might be independent of hypoactivity in KO rats.

Flinch-jump test
Shock sensitivity is also a possible confounding factor in the fear-conditioning test. We assessed shock sensitivity in the same animals as in the CTX test using the flinch-jump test. The thresholds for flinch and jump were not significantly different between the two genotypes (Table 1).

Discussion
In the present study, we assessed fear expression and fear memory for the first time in Gad1 KO animals and found the following results. In Gad1 KO rats, (a) the overall freezing time was prolonged, but (b) the freezing level decreased with the onset of tone during the training in the C-C experiment. (c) The freezing level was higher than that in WT rats during the contextual test of the C-C experiment. (d) The baseline freezing was increased in KO rats in the cued test. (e) KO rats displayed no increase in freezing level in response to the tones in the cued test, as if the cued memory was specifically disturbed. (f) KO rats displayed no change in freezing behavior during the fear extinction session in cued fear memory. (g) In the CTX experiment, KO rats displayed slightly lower freezing during the training, while (h) they showed comparable freezing in the contextual test. (i) KO rats showed an impaired contextual fear extinction.
In the C-C experiment, the responses of Gad1 KO rats during training for fear-conditioning and contextual test differed from those of Gad2 KO mice [15,[21][22][23][24][25][26][27]. Two studies have reported that Gad2 KO mice exhibited a decreased freezing and increased active defensive behavior and risk assessment behavior during both training and retrieval, as mentioned above [22,24]. Another study reported a comparable level of freezing between Gad2 KO mice and WT controls using a different protocol [25]. Unlike these findings, Gad1 KO rats exhibited a unique pattern of increased freezing during the pretone period and decreased freezing from the pretone level with the onset of the tone. On the other hand, it is of note that Gad1 KO rats showed lower levels of freezing than WT rats during the training of the CTX experiment as in Gad2 KO mice. In Gad2 KO mice, the inhibitory input is impaired in the basolateral amygdala [15]. The  basolateral amygdala consists mainly of excitatory neurons and a small population of GABAergic neurons. Hyperexcitability of the basolateral amygdala is associated with pathological anxiety [39]. However, almost all neurons in the central amygdala are GABAergic, and both its inhibition and its excitation cause freezing [39]. These circuit-level findings are consistent with the higher level of freezing observed in Gad1 KO rats in the C-C training, although the reason for these differences in freezing behaviors between Gad1 KO rats and Gad2 KO mice is unknown at present. Species differences between mice and rats should be considered, as this can affect the results. Gad2 knockout rats in the Long-Evans background are desirable for a direct comparison. GAD65 is known as an activity-dependent GAD that produces GABA in response to increased neuronal activity in the GABAergic terminal [11]. However, GAD67 is primarily distributed in the soma and constantly produces GABA. At the very least, the present results imply that this baseline production of GABA is also required for regulating an adequate level of fear expression.
Gad1 KO rats showed abnormally higher freezing in the contextual test of the C-C experiment, and normal freezing levels in that of the CTX condition. Generally, the freezing level of the contextual test in the CTX experiment is higher than that in the C-C experiment [40][41][42]. This phenomenon is associated with 'context salience', which is controlled by hilar perforant path-associated (HIPP) cells in the dentate gyrus. Under the suppression of HIPP cells, mice show abnormally high freezing in the C-C condition without affecting freezing in the CTX condition [40]. This phenotype is similar to that observed in Gad1 KO rats. Because HIPP cells are GABAergic, a deficit in the neurotransmission of the cells might contribute to the phenotype of Gad1 KO rats.
During the first 180 s of the cued test in the C-C experiment, Gad1 KO rats showed remarkably higher freezing than WT rats. Since the freezing behaviors during the baseline period on day 1 did not greatly differ between the two genotypes, we can speculate that the experimental procedure affected the increased freezing on day 3. One possible explanation is fear generalization as seen in Gad2 KO mice [25], although not necessarily in the same pattern. Gad2 KO mice display fear generalization to a tone that was not associated with an unconditioned stimulus during training, while Gad1 KO rats showed higher freezing in the context different from that of training. In any case, these findings suggest that impaired production of GABA is associated with enhanced fear.
Contextual fear memory is dependent on both the hippocampus and the amygdala, while cued fear memory is amygdala-dependent [43]. The amygdala contains an abundance of GABAergic neurons and forms a complex circuit [39]. This circuit may be the site responsible for the altered response in the cued test observed in Gad1 KO rats. However, a well-known lesion study reported that destroying the amygdala caused profound contextual memory deficits simultaneously [43]. Although we should be cautious in drawing conclusions, the peculiar result from the current study may suggest the existence of GABAergic neurons that are largely dependent on GAD67 and are specifically responsible for the association between the tones and electric stimuli. However, cued fear memory was not altered in amygdala-specific Gad1 knockdown mice [44], which is inconsistent with our current results. This discrepancy may be attributed to the fact that more than half of the GAD67 proteins remained in the amygdala of the knockdown mice. Due to technical problems, it is still impossible to determine the effects of selective and complete GAD67 elimination in the amygdala. Of course, we should note that there is a possibility that the alteration in freezing in the cued test was due to an alteration in fear expression rather than a deficit in the cued memory itself. Further studies are needed to overcome this limitation of the present experiment.
The contextual fear extinction was severely impaired in Gad1 KO rats in the present study. Fear extinction is also impaired in amygdala-specific Gad1 knockdown mice, although this is assessed in cued memory rather than in contextual memory [44]. Interestingly, parvalbumin neuron-specific Gad1 knockdown mice have the same phenotype [45]. Parvalbumin neurons are the largest population in cortical GABAergic neurons [46] and are implicated in the pathogenesis of schizophrenia [9,35]. In addition, Gad2 KO mice also show a delay in fear extinction of cued memory [25]. These findings suggest that both GAD67 and GAD65 are crucial for extinction learning.
Gad1 KO rats are also important in other aspects. Very recently, patients with intellectual disability with a null mutation of GAD1 were described for the first time [47,48]. The phenotype of patients differed from Gad1 KO mice; some patients survived up to 10 years of age, and 30% of patients did not have a cleft palate. These patients were epileptic, whereas no seizures were exhibited in Gad1 KO rats as far as we have observed. Although the patients also had joint contractures and pes equinovarus, Gad1 KO rats did not show these comorbidities and are capable of walking, running, and even swimming [33]. Taken together, the severity of the human GAD1 null mutation can be considered between that of Gad1 KO mice and Gad1 KO rats. It is also noteworthy that all known patients with Gad1 null mutations exhibit mental retardation.
It has been postulated that a reduction in GAD67 in the cerebral cortex and hippocampus in schizophrenia may cause cognitive impairment; this is consistent with patients with complete loss of GAD67 function showing mental retardation as a more severe form of cognitive impairment. Considering this, it is also possible that the impaired cued memory found in Gad1 KO rats reflects a more generalized memory impairment rather than a fear memory-selective one. Further characterization of the cognitive function of Gad1 KO rats will be needed.
In summary, the current study provides the first direct evidence that the loss of function of Gad1 altered conditioned fear behavior in adult rats. Furthermore, it is possible that it could have selectively impaired cued fear memory. Utilizing this novel model animal, further research should be conducted to reveal the differential role of GAD65 and GAD67 in the regulation of fear.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Western blot analysis of GAD67 and GAD65 (whole brain taken from 3-months-old rats). GAD67 protein was undetectable in Gad1 knockout (KO) rat, while GAD65 remains expressed. β-Actin was also evaluated as an internal control. +/+, wild-type; +/-, heterozygous Gad1 KO; -/homozygous Gad1 KO.  Table S1. General linear models describing the relationships between the freezing time on each day and genotype in the CTX experiment. The distance traveled in the open field test was included as a covariate. β, standardized partial regression coefficient; SE, standard error. †p < 0.1, * p < 0.05.