3.3.1. Catalase

Catalase activity was measured in the cerebral cortex, hippocampus, cerebellum, and midbrain and is presented in Figure S4 (due to low *n*, sexes were combined). Overall, the activity was ranked as follows: hippocampus > cerebellum, midbrain > cortex, supported by a main effect of region (*p* < 0.001). There was an overall effect of treatment (*p* = 0.003) which was not region-dependent (all *p* > 0.96), but no effect of genotype (all *p* > 0.35). The effect of treatment was due mostly to effects of Ex-Aox in all regions (not significant for cerebellum).

#### 3.3.2. Inflammatory Markers

Levels of IL6, TNFα, and IL10 were measured in the plasma and are reported in Figure 6. For IL6, the effects of treatments were seen in the GFAP-ApoE3 females, with a sex by treatment interaction approaching significance (*p* = 0.055). The female groups that exercised had lower levels of IL6 than the control groups. There was no effect of sex, strain, or treatment on TNFα levels (all *p* > 0.09), though the Ex-Aox GFAP-ApoE4 females had higher levels than the controls (*p* < 0.05).

**Figure 6.** Effect of exercise and antioxidants on markers of inflammation in male and female GFAP-ApoE3 and GFAP-ApoE4 mice. Each value represents mean±SEM, *n*=5–9. \* *p*<0.05 vs. sex- and strain-matched Sed-Con groups; † *p* < 0.05 comparing strain-matched Sed-Con males and females.

#### **4. Discussion**

The main findings of this study were: (1) there were strain differences for most motor functions, but no major differences for strength, balance, or cognition; (2) there were strain and test-dependent differences in response to the treatments: GFAP-ApoE3 mice were responsive on bridge and active avoidance, while GFAP-ApoE3 mice were on spatial learning and memory; (4) the most effective treatment was exercise, and no major additive or antagonistic effects were observed with antioxidant intake.

The current study provided a comprehensive phenotype of this mouse model, and its response to non-conventional therapeutic interventions. Anxiety, the most common non-cognitive symptom of AD [32] and associated with impaired daily activities [33], is often managed with benzodiazepines, which can lead to further cognitive and motor function declines [34,35]. Therefore, identifying non-pharmacotherapeutic agents to reduce anxiety could be an overall positive approach towards managing anxiety symptoms in AD without the added risk of furthering the functional declines. Previously, we determined that at 4 months of age, GFAP-ApoE4 mice were less anxious than GFAP-ApoE3 mice [30]; however, in the current study of 14-month-old mice, that difference subsided. This may indicate a pleiotropic effect of the ApoE genotype on anxiety levels, as has been described with cognitive function [30,36]. These data contrast with a previous study reporting higher anxiety among adult GFAP-ApoE4 mice [37]. Neither antioxidant nor exercise treatment affected the anxiety levels of the mice. In young mice, antioxidants increased the anxiety of the GFAP-ApoE4 mice, suggesting an age-dependent response to interventions [30]. Overall, the GFAP-ApoE3 mice were more active than the GFAP-ApoE4 mice, which is consistent with a previous report [38], and the treatments had no to minimal effects. Other complementary and alternative medicines have also been studied as non-traditional therapies alone or in combination with exercise. More specifically, in a pilot study, depression in AD patients was reduced when they combined Shiatsu with physical activity for 10 months [39].

During its early phase, AD is often associated with motor function impairments. Pathological changes in the motor cortex, striatum, cerebellum, or substantia nigra might be responsible for motor decline in AD [40,41]. The presence of ApoE4 doubles the rate of motor decline associated with aging [7]. Motor declines were observed in our mouse model with GFAP-ApoE4 mice exhibiting decreased activity, reflexes, and coordination. Similar effects on coordinated running were reported, in which the *APOE*4 mice performed poorly [42]. Repetitive transcranial magnetic stimulation suggested that ApoE4 is a critical determinant of the response to conditioning insults, especially for motor and cognitive brain networks [43]. Exercise and/or antioxidants improved the motor learning of the GFAP-ApoE3 males but not of the females or GFAP-ApoE4 mice. At younger ages, exercise training was associated with a reversal of the ApoE4-associated deficits in coordinated running [31]. The lack of response to treatment at an older age again suggests an age-dependent response to treatment. Interestingly, treatments improved the balance of the GFAP-ApoE3 mice, but not of the GFAP-ApoE4 mice. This strain-dependent effect was previously observed in younger mice [31], though the effects were not as large, most likely due to better performance of the young mice. Furthermore, exercise was more successful at improving motor function than antioxidants alone, indicative of a more promising therapeutic. Combining exercise and antioxidants did not lead to additive or antagonistic effects, as seen at younger ages on motor function [31].

Different mouse models expressing human ApoE4 exhibited poor spatial learning and memory [36,38,42]. Our data also suggest a significant decline in spatial learning and memory in GFAP-ApoE4 mice compared to GFAP-ApoE3 mice, a difference that was not observed in younger mice [30]. Interestingly, the majority of that effect comes from the males, as females did not show any differences. This may reflect sex differences in performance or that perhaps a water maze is not useful to detect cognitive differences in females. The treatments did not improve the performance of the GFAP-ApoE3 mice but did improve that of the GFAP-ApoE4 mice. Exercise improved more aspects of water maze performance than antioxidants alone, again supporting the fact that exercise is a better therapy. Improvements in cognition with exercise associated with the ApoE4 genotype have been shown previously [44]. Recent studies have implicated a potential cross-talk between the brain and muscles, with exercise-induced mediators being released, leading to the neuroprotection and brain function improvements associated with exercise. FNDC5/irisin is a myokine released upon exercising that has been associated with improvements in synaptic activity and memory in the APP/PS1 AD mouse model [45]. While our exercise paradigm differed from that study, it would be of interest to determine in future studies using our training paradigm if the FNDC5/irisin pathway is also activated in our regimen.

AD symptoms start with the loss of non-spatial short-term/working memory [46]. While we found strain differences in the spatial task, both GFAP-ApoE3 and ApoE4 mice performed similarly in the non-spatial task used. This is in contrast to previous studies that have identified genotype differences in working memory, which deteriorated in an ε4 allele dose-dependent manner in humans [47]. The GFAP-ApoE3 mice were more responsive to the treatments than the GFAP-ApoE4 mice, and additive effects were observed in the reversal phase for males. Other studies have also determined that exercise can improve short-term/working memory task performance [48,49]. These treatment effects were also observed previously with young GFAP-ApoE3 mice [30], however, while the treatments improved the performance of the young GFAP-ApoE4 mice [30], they did not have any effect on the older mice in the current study. These outcomes on cognitive function suggest that the type of memory, along with the age and genotype, can affect the outcome of interventions.

Contrary to expectation, catalase activity was reduced in brain regions in mice that were exercised, especially the ones combined with antioxidants. Reports on the effects of exercise on catalase activity in the brain are conflicting: from no effect [50] to up-regulation as a result of the oxidative bursts associated with acute exercise [51]. Catalase may be activated upon acute but not chronic exercise. Furthermore, supplementation with antioxidants may have led to a sparing of catalase or a feedback down-regulation of catalase. Plasma-based inflammatory biomarkers are often suggested to be future biomarkers for AD progression and treatment response. The effects of exercise on inflammation are ambiguous, and depend on the focus of inflammation (peripheral (plasma) vs. central (brain)). Studies focused on long-term exercise observed higher IL6 in the brain [52] compared to plasma, while no such change was noticed in TNFα. Some studies demonstrated lower IL6 in the brain after prolonged exercise without affecting plasma IL6 levels [53]. IL6 was positively related to IL10 in the exercise training context [54]. We also observed similar a relation of lowered IL6 in the presence of chronic exercise and associated lowered IL10 levels in GFAP-ApoE3 female mice. On the contrary, treadmill training previously reported higher IL10 and lower TNFα in plasma [55]. Differences between the sexes may be due to hormone influences, as females see a decrease in circulating hormones, which can have repercussion on the inflammatory response [56], and/or may be due to differences in oxidative stress levels and homeostasis [57].

The presence of ApoE4 is associated with an increased risk of developing late-onset AD, however, clinically, it does not seem to always be true with AD patients not carrying the ε4 and vice versa, suggesting a more complicated picture and gene involvement in AD etiology [58]. Furthermore, the presence of one ε4 allele may be insufficient to increase the risk in individuals 65 and younger [59]. Lastly, ApoE4 and other mutations of the known AD genes, such as APP, PSEN1, and PSEN2, only account for 5% of early-onset AD cases [58]. Therefore, the outcomes of this study would need to be further studied in other models of AD, especially relating to early-onset dementia in order to generalize them to all AD patients.

While it remains under debate that these non-pharmacological therapies are effective in reducing neurodegeneration, it is possible that they affect the vascular contribution to these diseases. By managing and controlling the vascular risk factor via a healthy lifestyle, such as exercising and improving one's diet, it is likely to reduce vascular cognitive impairments and reduce the advance to neurodegenerative diseases and dementias [60]. These non-pharmacological therapies may be working via the control and management of vascular risk, leading to improved cognition and diminished neurodegenerative disorders, such as AD.

Our study has several limitations: (1) the model used may not be generalizable to all AD patients, as it focused on the ApoE genotype and on a glial model of ApoE expression; (2) the limited oxidative stress measurements, as such glutathione levels or protein damage measurements would have strengthened the role of oxidative stress in the study [61]; (3) the involvement of other pathways, such as the irisin-dependent pathway involving muscle–brain cross-talks, were not studied and may be of importance to determine the mechanism of action of our exercise paradigm; (4) a dose–response component of our exercise paradigm would be of interest to determine the translational impact of our training regimen [62].
