Mitochondrial antioxidant elamipretide improves learning and memory impairment induced by chronic sleep deprivation in mice

Male 2‐month‐old C57BL/6J mice were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. (specific pathogen‐free grade). The animals are kept in an environment where food and water are available ad libitum, with a 12‐h dark–light cycle, 21–23°C temperature and 50%–60% relative humidity.

The male C57BL/6J mice were randomly allocated to one of the following four groups (n = 8 per group): Control + saline, Control + SS‐31, SD + saline, and SD + SS‐31. SS‐31 or saline (5 mg/kg) was administered intraperitoneally to the mice 30 min prior to sleep deprivation using an adapted version of the BW‐NSD404 sleep deprivation machine (Shanghai Bio‐will Co., Ltd.). The principle of the machine is to place the mouse on a platform that is constantly moving to prevent the mouse from falling asleep (Zhang, Cheng, et al., 2022). After administration of SS‐31 or saline, the mice in the SD + SS‐31 and SD + saline groups were subjected to the sleep deprivation machine for 21 consecutive days from 2:00 p.m. to the following day at 8:00 a.m. (Konakanchi et al., 2023), whereas mice in the Control + saline and Control + SS‐31 groups were maintained on the static (not moving) machine during this same period (see Figure 1).

The new object experiment was used to assess short‐term cognitive function in mice. As previously performed by others (Leger et al., 2013), the experiment was divided into three parts: habituation, familiarization, and test phases. In the habituation phase, mice were placed in an empty open field and explored freely for 5 min. A period of 24 h later, mice were placed in an open field containing two identical objects for familiarization, and the experiment was discontinued when both objects had been explored for 20 s or when the 10‐min had expired. During the test phase, one of the objects in the open field was replaced with a new object and the mice were allowed to explore the old (D1) and new objects (D2) freely with 10 min. A 75% alcohol was used at the end of each experiment to eliminate the effect of odor. The memorization ability was quantified by novel object recognition index (NOI): NOI = D2/(D1 + D2).

To avoid the influence of the novel object recognition test on the subsequent Morris water maze test, the Morris water maze test was conducted on a new cohort of the mice. The protocol used in this study was similar to the one previously described and was used to assess the spatial learning and memory abilities of mice (Zhang, Cheng, et al., 2022). The test was divided into a learning phase and a memory phase. During the learning phase, mice were trained in a circular white pool with a diameter of 120 cm and a height of 30 cm. Each mouse performed four trials per day to find the hidden target platform and was allowed to rest on the platform for 30 s whether or not they found the hidden platform within 60 s. The learning phase lasted for a total of 7 days. The target underwater platform was removed 2 h after the last experiment on the last day of the learning phase. During the memory phase, after the hidden platform was removed, the mice entered the water from the opposite quadrant of the quadrant where the target platform was located and explored freely for 60 s. ANY‐maze tracking system was used for recording and analysis of all trials.

Total RNA was extracted from the hippocampal tissue by adding Trizol lysate. Spectrophotometer was used to assess the purity of the extracted RNA. The RNA was reverse‐transcribed to cDNA using the PrimeScript reverse transcription (RT) reagent Kit with gDNA Eraser (TaKaRa, RR047A). The transcripts were then amplified by quantitative polymerase chain reaction (qPCR). The cDNA was used as a template and the reaction system was as follows: 5 µL of 2× SYBR Green Mixture, 1 µL of forward primer, 1 µL of reverse primer, 2 µL of RNase‐free water, and 1 µL of cDNA. The reaction conditions were a single cycle of pre‐denaturation at 95°C for 1 min, followed by a total of 40 cycles at 95°C for 20 s and 60°C for 1 min. The primers are designed by Primer 3 software and synthesized by Sangon Biotech company. The amplification profile of PCR showed S curve and the level of mRNA was quantified using the 2^−ΔΔCt method. Table 1 presents the primer sequences.

The protocol of western blotting (WB) was performed as previously described (Wei et al., 2022). In brief, the hippocampal tissue was lysed in RIPA lysis buffer (Beyotime, P0013B) and centrifuged at 12,000 × g for 15 min to collect the supernatant. The protein samples were electrophoretically separated and then blotted onto polyvinylidene fluoride membranes (Millipore, IPVH00010). The protein levels were determined via incubation with antibodies against PGC‐1α (1:1000; abcam), SIRT1 (1:500; Santa Cruz), NF‐κB (1:5000; abcam), brain‐derived neurotrophic factor (BDNF; 1:1000; abcam), PSD‐95 (1:2000; abcam), SYN (1:1000; bioss), IL‐1β (1:500; bioss), IL‐6 (1:1000; Wanleibio), and TNF‐α (1:1000; bioss). The membranes were then incubated with horseradish peroxidase‐labeled secondary antibodies (1:20,000; Zsbio) according to the properties of the primary antibody. Protein detection was performed using an ECL luminescence kit (Thermo, 340958) and quantified with ImageJ software.

The concentrations of inflammatory factors (IL‐1β, IL‐6, and TNF‐α) were detected in the hippocampal samples using corresponding enzyme‐linked immunosorbent assay (ELISA) kits from Wuhan ColorfulGene Biological Technology Co., Ltd. (JYM0531Mo, JYM0012Mo, and JYM0218Mo, respectively). All experiments were performed in accordance with the manufacturer's instructions.

All data are normal and expressed as mean ± standard errors of the mean. Repeated‐measures analysis of variance was used to analyze the data from the learning period of the Morris water maze test, whereas two‐way analysis of variance was used to analyze the data from the memory period of the Morris water maze test, novel object recognition, WB, RT‐qPCR, and ELISA with the Tukey post hoc test to compare the differences among the four groups. The correlations between variables were determined using Pearson's correlation coefficient. All data analyses were performed using GraphPad 8.0 software. p < .05 was considered statistically different.

First, the novel object recognition test was used to evaluate whether SS‐31 could improve recognition memory deficits induced by CSD. During the familiarization phase, there was no significant difference in the exploration time for two same objects among the four groups (p > .05, Figure 2a,b). In the test phase, the NOI of SD + saline group was significantly lower than that in Control + saline group (treatment: F_{(1, 28)} = 8.37, p < .01; drug: F_{(1, 28)} = 4.71, p < .05; treatment × drug: F_{(1, 28)} = 6.81, p < .05, Figure 2C). Furthermore, SS‐31 treatment reversed this decline in the SD + SS‐31 when compared to SD + saline group (p < .05). These results suggested that SS‐31 treatment could ameliorate the recognition memory impairment induced by CSD.

The Morris water maze test was used to assess the influence of CSD and SS‐31 on hippocampal‐dependent learning and memory. On the first day of the learning phase, there was no difference in the time of searching hidden platform in the four quadrants, indicating that the mice from four groups did not show a preference for any quadrant at the beginning of the learning phase (supplementary file Table 1). The escape latency and distance gradually decreased with an increase in the training days (escape latency: F_(6,168) = 207.47, p < .01; distance: F_(6,168) = 158.24, p < .01; Figure 3a,b). With respect to treatment effects, there were significant differences in escape latency and distance among the four groups (escape latency: F_(3,28) = 10.10, p < .01; distance: F_(3,28) = 8.93, p < .01; Figure 3A,B). The post hoc analysis documented that the mice from the SD + saline group spent a longer time to find the hidden platform than the mice from the Control + saline and Control + SS‐31 groups (Ps < .1). There was also a significant difference between the SD + saline and SD + SS‐31 groups (p < .01). No differences were observed in swimming velocity among the different treatment groups (Figure 3C).

In the memory phase, the number of platform‐crossing, percent time, and distance showed significant differences among the four groups (number of platform‐crossing: treatment: F_{(1, 28)} = 9.93, p < .01; drug: F_{(1, 28)} = 8.73, p < .01; treatment × drug: F_{(1, 28)} = 6.55, p < .05; time: treatment: F_{(1, 28)} = 19.12, p < .01; drug: F_{(1, 28)} = 9.25, p < .01; treatment × drug: F_{(1, 28)} = 4.97, p < .05; distance: treatment: F_{(1, 28)} = 19.12, p < .01; drug: F_{(1, 28)} = 12.28, p < .01; treatment × drug: F_{(1, 28)} = 7.25, p < .05; Figure 3d–f, supplementary file Table 2). The post hoc analyses showed that the number of platform‐crossing, percent time, and distance were shorter in the SD + saline group than in the Control + saline group (p < .05). There was also a significant difference between the SD + saline and SD + SS‐31 groups (p < .05).

The mRNA levels of Sirt1, Pgc1a, and Nfkb were different among the four groups (Sirt1: treatment: F_{(1, 28)} = 46.93, p < .01; drug: F_{(1, 28)} = 9.65, p < .01; treatment × drug: F_{(1, 28)} = 10.48, p < .01; Pgc1a: treatment: F_{(1, 28)} = 75.10, p < .01; drug: F_{(1, 28)} = 12.83, p < .01; treatment × drug: F_{(1, 28)} = 3.28, p > .05; Nfkb: treatment: F_{(1, 28)} = 101.40, p < .01; drug: F_{(1, 28)} = 30.18, p < .01; treatment × drug: F_{(1, 28)} = 35.59, p < .01; Figure 4A–C, Figure S1). Post hoc analysis revealed that the mRNA expression of Sirt1 and Pgc1a was downregulated, whereas the expression of Nfkb mRNA was upregulated after CSD. Furthermore, SS‐31 treatment counteracted the effects induced by CSD.

Similarly, the protein levels of SIRT1, PGC‐1α, and NF‐κB were different among the four groups (SIRT1: treatment: F_{(1, 20)} = 74.25, p < .01; drug: F_{(1, 20)} = 10.04, p < .01; treatment × drug: F_{(1, 20)} = 11.13, p < .01; PGC‐1α: treatment: F_{(1, 20)} = 144.20, p < .01; drug: F_{(1, 20)} = 10.94, p < .01; treatment × drug: F_{(1, 20)} = 7.95, p < .05; NF‐κB: treatment: F_{(1, 20)} = 34.91, p < .01; drug: F_{(1, 20)} = 9.52, p < .01; treatment × drug: F_{(1, 20)} = 19.86, p < .01; Figure 4d–g, Figure S1). Post hoc analysis showed that the protein levels of SIRT1 and PGC‐1α were decreased after CSD, which were restored with SS‐31 treatment. The protein level of NF‐κB was increased in the SD + saline group when compared to that of the Control + saline group, and this effect was also reversed by SS‐31 treatment.

The expression levels of IL‐1β, IL‐6, and TNF‐α were significantly different among the four groups (ELISA: IL‐1β: treatment: F_{(1, 28)} = 26.30, p < .01; drug: F_{(1, 28)} = 16.17, p < .01; treatment × drug: F_{(1, 28)} = 8.25, p < .01; IL‐6: treatment: F_{(1, 28)} = 19.71, p < .01; drug: F_{(1, 28)} = 9.09, p < .01; treatment × drug: F_{(1, 28)} = 6.49, p < .05; TNF‐α: F_{(1, 28)} = 21.26, p < .01; drug: F_{(1, 28)} = 13.15, p < .01; treatment × drug: F_{(1, 28)} = 1.82, p > .05; Figure 5a–c; WB: IL‐1β: treatment: F_{(1, 20)} = 21.88, p < .01; drug: F_{(1, 20)} = 14.80, p < .01; treatment × drug: F_{(1, 20)} = 11.78, p < .01; IL‐6: treatment: F_{(1, 20)} = 20.54, p < .01; drug: F_{(1, 20)} = 8.40, p < .01; treatment × drug: F_{(1, 20)} = 8.40, p < .01; TNF‐α: F_{(1, 20)} = 29.58, p < .01; drug: F_{(1, 20)} = 21.33, p < .01; treatment × drug: F_{(1, 28)} = 13.78, p < .01; Figure 5d–g, Figure S2). Post hoc analyses showed upregulation expression of IL‐1β, IL‐6, and TNF‐α in the SD + saline group when compared to the Control + saline group; however, SS‐31 immediately improved the increase in the levels of IL‐1β, IL‐6, and TNF‐α.

The mRNA levels of Bdnf, Psd95, and Syn were significantly different among the four groups (Bdnf: treatment: F_{(1, 28)} = 68.71, p < .01; drug: F_{(1, 28)} = 9.75, p < .01; treatment × drug: F_{(1, 28)} = 2.17, p > .05; Psd95: treatment: F_{(1, 28)} = 62.22, p < .01; drug: F_{(1, 28)} = 2.66, p < .01; treatment × drug: F_{(1, 28)} = 9.91, p < .01; Syn: treatment: F_{(1, 28)} = 45.42, p < .01; drug: F_{(1, 28)} = 1.28, p > .05; treatment × drug: F_{(1, 28)} = 9.48, p < .01; Figure 6a–c). The post hoc analyses revealed that the mRNA levels of all three of these synaptic plasticity‐associated factors were decreased in the SD + saline group compared with those of the Control + saline and Control + SS‐31 groups (Ps < .01). SS‐31 treatment attenuated the decrease of Bdnf, Psd‐95, and Syn mRNA levels in the SD + SS‐31 group (Ps < .05).

The protein expression level of BDNF was also significantly different among the four groups (treatment: F_{(1, 20)} = 104.40, p < .01; drug: F_{(1, 20)} = 3.07, p > .05; treatment × drug: F_{(1, 20)} = 5.85, p < .05; Figure 6d,e; Figure S3). Post hoc analysis showed that the expression level of BDNF was decreased in the SD + saline group compared with that in the Control + saline group (p < .01), whereas SS‐31 treatment restored the BDNF protein content (p < .01). The protein expression levels of PSD‐95 and SYN were different among the four groups (PSD‐95: treatment: F_{(1, 20)} = 59.09, p < .01; drug: F_{(1, 20)} = 3.61, p > .05; treatment × drug: F_{(1, 20)} = 7.39, p < .05; SYN: treatment: F_{(1, 20)} = 106.90, p < .01; drug: F_{(1, 20)} = 3.50, p > .05; treatment × drug: F_{(1, 20)} = 6.73, p < .05; Figure 6f,g; Figure S3). The post hoc analysis showed that the protein expression of PSD‐95 and SYN was downregulated in the SD + saline group compared to that of the Control + saline group (Ps < .01), whereas SS‐31 significantly attenuated the decrease of PSD‐95 and SYN protein levels in the Control + SS‐31 group (Ps < .05).

The escape latency and distance were negatively correlated with the levels of SIRT1, PGC‐1α, BDNF, PSD‐95, SYN, and were positively correlated with the levels of NF‐κB and pro‐inflammatory cytokines including IL‐1β, IL‐6, and TNF‐α in the learning phase. In the memory phase, the percent time and distance were positively correlated with the levels of SIRT1, PGC‐1α, BDNF, PSD‐95, and SYN and were negatively correlated with the levels of NF‐κB and pro‐inflammatory cytokines, as shown in Tables 2, 3, 4.

We assessed cognitive impairment using the novel object recognition and Morris water maze tests in this study, as this is proven to be reliable tests that are strongly related with hippocampal‐dependent learning and memory (Zhang, Chen, et al., 2022). Previous studies showed that sleep dysfunction could affect hormone levels, neural circuits, dendrite density, and synaptic transmission, leading to cognitive decline (Brüning et al., 2019; Voldsbekk et al., 2022). Male Wistar rats subjected to sleep deprivation for 3 days showed spatial memory decline during the Morris water maze test (Duan et al., 2016). Short‐term sleep deprivation of 6 h significantly impaired the memory consolidation in male C57BL/6J mice during the object‐location memory paradigm (Raven et al., 2020). Consistently, we found that CSD for 21 days decreased the NOI during the NOI, increased escape latency and distance, and increased time and distance percent in the SD + SS‐31 group compared to the SD + saline group during the Morris water maze, indicating impaired learning and memory in CSD mice.

The cognitive protective effects of SS‐31 have been extensively studied in various pathological models. One study showed that SS‐31 pretreatment improved isoflurane‐induced cognitive decline through improving mitochondrial function (Wu, Li, et al., 2015). Another study showed that SS‐31 increased the downregulation in the percentage freezing time caused by sepsis‐associated encephalopathy during contextual fear conditioning (Wu, Zhang, et al., 2015). We found that treatment with SS‐31 improved CSD‐induced learning and memory impairment as observed in the novel object recognition and Morris water maze tests. Additionally, the mice from the Control + SS‐31 group showed no significant difference from those in the Control + saline group in the novel object recognition and Morris water maze tests. This suggests that SS‐31 might only have an effect on damaged mitochondria and not on normal mitochondria.

Accumulating evidence suggests that integrated mitochondrial function is closely related to cognitive function (Zhuang et al., 2021). The structure and number of mitochondria are not static, as mitochondria grow and divide continuously in a process known as mitochondrial biogenesis to respond to changing energy demands and stressful events. PGC‐1α is an important transcription factor involved in mitochondrial biogenesis, which helps to maintain mitochondrial function by regulating mitochondrial dynamics and energy (Chen et al., 2011; Zhu et al., 2010). PGC‐1α is transferred to the nucleus where it co‐activates nuclear respiratory factor (NRF) and mitochondrial transcription factor A (TFAM). The combination of NRF and TFAM then promotes the replication and expression of mitochondrial genes (Finck & Kelly, 2006; Scarpulla, 2008). Furthermore, SIRT1 is a pivotal factor in the regulation of mitochondrial metabolism, which activates PGC‐1α through histone deacetylation. Isoflurane‐induced anesthesia decreased the expression of SIRT1 and PGC‐1α and impaired cognitive function in aging rats (Yang et al., 2020). Triptolide activated the SIRT1/PGC‐1α signaling pathway to improve cognitive impairment in rats with vascular dementia (Yao et al., 2019). In the current study, we found that the mRNA and protein levels of SIRT1 and PGC‐1α were decreased in the SD + saline group after CSD, implying that mitochondrial biogenesis dysfunction was involved in the cognitive impairment induced by CSD. Moreover, SS‐31 increased the levels of SIRT1 and PGC‐1α under the sleep deprivation condition, suggesting that SS‐31 improves CSD‐induced cognitive and mitochondrial dysfunction by modulating the SIRT1/PGC‐1α signaling pathway.

Progressive neuroinflammation plays a primary role in the cognitive impairment related with neurodegenerative diseases such as Alzheimer's disease and Huntington's disease (Bains et al., 2022; Decandia et al., 2023). Neuroinflammation influences cell proliferation and differentiation, synaptic plasticity, and cognitive function (Bai et al., 2023; Cangalaya et al., 2023). The levels of pro‐inflammatory cytokines were found to be increased in the peripheral blood of patients with sleep disorders (Irwin et al., 2016; Xia et al., 2021). In addition, sleep deprivation significantly elevated the levels of proinflammatory cytokines including IL‐1β, IL‐6, and TNF‐α in the hippocampus and resulted in cognitive impairment during the novel object recognition test (Lu et al., 2022). In line with these previous studies, the results showed that the levels of IL‐1β, IL‐6, and TNF‐α were increased in the hippocampus of the mice exposed to CSD, which contributed to the observed cognitive decline.

Growing evidence suggests that mitochondria are directly linked to the immune system and inflammatory response. NF‐κB and NOD‐like receptor thermal protein domain associated protein 3 (NLPR3) are important transcription factors associated with the inflammatory response. The ROS produced by mitochondria activate NF‐κB and NLPR3, which in turn promotes the excessive release of the pro‐inflammatory cytokines (Nakahira et al., 2011; Sato et al., 2004; Zhou et al., 2011). Importantly, a previous study showed that treatment with SS‐31 significantly ameliorated isoflurane‐induced cognitive deficits and inflammatory responses by decreasing the levels of NF‐κB, NLPR3, and ROS (Wu et al., 2015). Our results showed that SS‐31 reduced the elevated levels of NF‐κB and proinflammatory cytokines in the SD + SS‐31 group, suggesting that SS‐31 could suppress the CSD‐induced inflammatory response by improving mitochondrial dysfunction.

BDNF is mainly expressed in the hippocampus, where it controls neuronal maturation and survival (Mosiołek et al., 2022). A large body of evidence suggests that BDNF plays an important role in synaptic plasticity and synaptic transmission that is considered to underlie learning and memory (Liu et al., 2022; Wang et al., 2023). SYN and PSD‐95 are synaptic proteins that are closely related to synaptic maturation, transmission, and remodeling (Aguiar et al., 2022; Reinés et al., 2008). Decreased expression levels of BDNF, SYN, and PSD‐95 have been associated with the pathogenesis of cognitive dysfunction (Wang & Zhao, 2016; Hong et al., 2020). The present study showed that CSD decreased the mRNA and protein levels of BDNF, SYN, and PSD‐95 in mice, which contributed to cognitive deficits induced by CSD. Furthermore, pretreatment with SS‐31 attenuated the reduction of BDNF, SYN, and PSD‐95 induced by CSD, which is consistent with a previous study showing that SS‐31 prevented the lipopolysaccharide‐induced downregulated of SYN and PSD‐95 through modulating the BDNF signaling pathway (Zhao et al., 2019).

Emerging evidence also suggests that mitochondrial biogenesis function is closely related to cognitive function. Our previous study showed that the decreased expression level of PGC‐1α was correlated with poor cognitive performance induced by prenatal inflammatory exposure during the Morris water maze test (Zhuang et al., 2021). In the current study, the downregulated expression of PGC‐1α was also considered to play a role in the learning and memory impairment induced by CSD, as reflected by the correlation between the expression level of PGC‐1α and indicators of the Morris water maze test in the SD + SD group. Clinical studies have found that the levels of anti‐inflammatory cytokines are positively correlated with cognitive performance associated with sleep disorders (He et al., 2021). Thus, our preclinical study confirms that the levels of pro‐inflammatory cytokines are correlated with the cognitive impairment induced by CSD. Additionally, the levels of synaptic plasticity‐associated proteins reflect the function of synaptic connection, synaptic transmission, and synaptic plasticity and are closely associated with cognitive function (Sadigh‐Eteghad et al., 2020). We also found that the decreased expression levels of BDNF, SYN, and PSD‐95 were correlated with CSD‐induced cognitive impairment.

Our study has some limitations. First, we have not further evaluated mitochondrial function and oxidative stress in the hippocampus, including the mitochondrial electron transport system, ROS, and antioxidant defense enzymes. Second, we determined the mitochondrial biogenesis in the whole hippocampus; thus, it was not possible to distinguish whether mitochondrial dysfunction in the neurons or glial cells was responsible for the inflammation and synaptic dysfunction induced by CSD. Third, we only selected male mice for behavioral experiment and did not evaluate the sex effects of CSD on cognitive function even though previous studies have reported that women are more susceptible to develop sleep disorders than men. Finally, the markers of mitochondrial function, inflammation, and synaptic function were only examined in the hippocampus and were not examined in other regions associated with cognitive function.

The peer review history for this article is available at https://publons.com/publon/10.1002/brb3.3508↗.