*2.4. The E*ff*ect of AG on Fear Memory*

The effect of AG on fear memory was measured using the fear conditioning paradigm (Figure 4). The acquisition of fear memory in the TBI mouse model was not statistically higher than that of the control animals. AG treatment lowered the acquisition of fear memory in the TBI model mice (Figure 4A). There was no significant effect of TBI and AG on the consolidated contextual and cued fear memory (Figure 4B,C). The acquisition of fear memory in the CMS mice was lower than that of controls. There was no difference in the fear acquisition after AG treatment in CMS mice (Figure 4D). Compared to the normal mice, the CMS, AG, and CMS + AG mouse groups had lower levels of consolidated contextual fear memory. However, there were no differences among the CMS, AG, and CMS + AG groups (Figure 4E). Compared to normal mice, the AG and CMS + AG groups had lower levels of consolidated cued fear memory. In addition, AG reduced the consolidated cued fear memory in the CMS group of mice (Figure 4F).

**Figure 4.** Fear memory by AG in TBI and CMS mice, as measured by the freezing time in the fear conditioning test. (**A**) Fear acquisition in the TBI model. There were time (within) effects [F(5,140) = 45.7, *p* < 0.001], treatment group (between) effects [F(3,28) = 3.25, *p* = 0.03], but no within–between interaction differences [F(15,140) = 1.1, *p* = 0.36]. There were no differences between the treatment groups at base, cue1, cue2, cue3, or cue4. However, the differences between the TBI and TBI + AG groups were significant (*p* = 0.039). In addition, the differences between the TBI and control groups (*p* = 0.16), and TBI vs. AG (*p* = 0.13) groups were not significant. (**B**) Consolidated contextual fear memory in TBI model. There were no group differences in the freezing response to the context [F(3,28) = 0.36, *p* = 0.78). (**C**) Consolidated cued fear in the TBI model. There were procedure (within) effects [F(1, 28) = 61.6, *p* < 0.001], no treatment group (between) effects [F(3,28) = 0.29, *p* = 0.83], or within–between interactions [F(3,28) = 0.64, *p* = 0.59]. There was a significant increase in the freezing comparing cue vs. precue in the control (*p* < 0.001), AG (*p* < 0.001), TBI (*p* = 0.005), and TBI + AG groups (*p* < 0.001). (**D**) Fear acquisition in the CMS model. There were time (within) effects [F(5,95) = 27.8, *p* < 0.001], treatment group (between) effects [F(3,19) = 4.9, *p* = 0.011], but no within–between interaction effects [F(15,95) = 0.47, *p* = 0.95]. Fear acquisition in AG (*p* = 0.021), CMS (*p* = 0.035), and CMS + AG (*p* = 0.001) were lower than controls throughout the fear acquisition procedure. There were no statistical differences among the AG, CMS, and CMS + AG groups. (**E**) Consolidated contextual fear memory in the CMS model. There were group differences in the freezing response to the context [F(3,19) = 9.8, *p* < 0.001). The contextual fear in AG (*p* = 0.002), CMS (*p* = 0.008), and CMS + AG (*p* < 0.001) was lower than those of the controls. (**F**) Consolidated cued fear in the CMS model. There were procedure (within) effects [F(1,19) = 54.7, *p* < 0.001], treatment group (between) effects [F(3,19) = 9.3, *p* = 0.001], and within–between interaction differences [F(3,19) = 4.8, *p* = 0.012]. There was a significant increase in freezing between cue and precue in the control (*p* < 0.001), CMS (*p* < 0.001), and CMS + AG (*p* = 0.002) groups, but not in the AG group (*p* = 0.31). At pre-cue, the control vs. CMS + AG (*p* = 0.001) groups were statistically different. At cue, the control vs. AG, (*p* = 0.004); control vs. CMS + AG, (*p* < 0.001); CMS vs. CMS + AG, (*p* = 0.017) were all statistically different. All data were normally distributed and are represented as means ± S.E.M. Control: vehicle (DW) treated; AG: *Angelica gigas* 1 mg/kg; TBI: vehicle treated + traumatic brain injury; TBI + AG: *Angelica gigas* 1 mg/kg + traumatic brain injury; CMS: vehicle treated + chronic mild stress; CMS + AG: *Angelica gigas* 1 mg/kg + chronic mild stress. \*\* *p* < 0.01, \*\*\* *p* < 0.001 vs. control. # *p* < 0.05 vs. CMS. ANOVA and repeated measures ANOVA, Tukey's HSD post-hoc test.

#### **3. Discussion**

We investigated whether AG extract could prevent the progression of cognitive decline or improve memory in MCI models. We found that AG improved spatial learning and working memory but suppressed fear memory in the TBI model. In the CMS model, AG improved spatial learning and suppressed the cued fear memory.

MCI is a condition that increases a patient's risk of developing dementia. Therefore, it is important that MCI is diagnosed early in order to prevent or limit dementia development [4]. Methods to improve sleep disorders, depression, one's social network, and physical exercise are considered for the treatment of MCI [28–30]. Pharmacological interventions, such as cholinesterase inhibitors, nonsteroidal anti-inflammatory drugs, estrogen replacement therapy, Gingko biloba, and vitamin E, have not shown to prevent MCI progression to dementia [4].

The roots of AG have been used in traditional medicine to improve blood flow and anemia. They have also been used for their analgesic properties [31]. AG improves spatial memory, avoidance memory, and working memory in dementia models [27]. Among the components of AG, decursinol showed the highest inhibitory activity toward acetylcholinesterase [32]. Therefore, AG is expected to improve cognitive impairment in animal models.

In order to develop a valid MCI animal model, there must be subtle memory impairment [5]. The presence of depression-like symptoms may also be an important factor in the MCI model. Dementia and depression are known to have many associations. For instance, 60% of MCI cases that progress to AD are accompanied by depression [33].

The defect in a TBI model varies depending on the hitting area and velocity [14]. In contrast with severe TBI, mild TBI has minimal histological changes but apparent cognitive and emotional problems [8]. This association suggests that mild TBI can be used as an MCI model [34]. Such spatial memory deficit occurred in our TBI model mice, but other reflexes (such as the paw withdrawal reflex, righting reflex, and corneal reflex) were maintained (data not shown). Also, TBI is commonly known to cause disturbances in working memory. TBI in the parietotemporal regions has been shown to cause working memory deficits in mice [35]. However, working memory was normal in the mild TBI model made by repeated frontal impact; therefore, working memory is not always compromised in TBI [36]. Our measurements of working memory did not differ in the TBI model. Another common problem in TBI is the inability to recognize the source of information, such as facial recognition [37]. Face recognition memory corresponds to animal object recognition memory. There was a deficit in animal object recognition memory in the TBI models [38]. The TBI mice in this study also had decreased object recognition memory. These model mice also showed heightened fear memory; however, the results are controversial. One group found that there was no difference between normal animals and a mild repeated frontal TBI model with regard to conditioned fear after a severe CCI impact to the left parietal cortex [36]. However, a single impact above the skull increased anxiety and contextual fear in rats [39]. In mice, hippocampal-dependent fear memory decreased, but cued-dependent fear memory was not affected by TBI [40]. In our results, there was an enhanced acquisition of contextual fear in the TBI mice. However, after 24 h, the consolidated level of contextual and cued memory was not different with control animals.

Stress has a variety of effects on cognitive function [41]. CMS is a depression model also characterized by a decrease in cognitive function associated with neuroimmune, neuroendocrine, and neurogenesis functions [4]. In this experiment, the CMS animals displayed anhedonia-like behaviors in the sucrose preference test and deficits in spatial memory. In other studies using CMS models, working memory was reduced in rats but was maintained in mice [42,43]. Our study similarly found that there was no working memory deficit in CMS mice. Reduction in object recognition memory in mice and increased contextual fear in rats were reported previously [42,44]. In the present study, CMS did not affect object recognition memory, but fear acquisition and contextual consolidation were reduced. These findings were inconsistent; however, rats that were exposed to social instability stress (daily 1-h isolation, change of cage partners) showed deficits in contextual and cued memory [45,46]. Taken

together, in the TBI mice, spatial learning and object recognition memory were degraded. However, spatial memory, working memory, and fear memory were intact. In the CMS model, spatial memory was degraded, but working and object recognition memory were not affected, and cued fear memory was reduced.

AG has known antibacterial, immune-stimulating, antiplatelet aggregation, neuroprotective, anti-inflammatory, and antioxidant properties [31]. We found that AG did not improve spatial memory in normal mice. However, it did improve memory in TBI and CMS animals. In addition, AG treatment led to improved working memory in normal mice of the TBI cohort. AG did not increase the recognition memory of the normal mice. However, recognition memory was only abnormal in the TBI mice at baseline. In the TBI and AG co-administration group, recognition memory was maintained. This result suggests that AG can prevent TBI-induced deficit. In the CMS cohort, however, the object memory of the CMS group was not affected. Therefore, AG also did not improve object memory in either the normal or the CMS mice. With regard to fear memory, the contextual and cued consolidated fear memory of normal and CMS mice were reduced in the CMS cohort. However, consolidated fear levels were not affected in the TBI cohort. Among AG's known effects, its cholinesterase inhibition, improved blood flow, and anti-inflammatory properties may prevent cognitive decline and improve memory in these mouse models [31,32].

In conclusion, AG prevented the deterioration of spatial learning and object recognition memory in a mouse TBI model. AG also prevented the deterioration of spatial learning in the CMS model mice and improved working memory in normal mice. As TBI is a cognitive impairment gradually progresses, and chronic stress can cause AD-like pathologies, these findings suggest that AG may also prevent progressive cognitive decline in MCI animal models, which may be worth further research.
