Isoform-specific effects of neuronal repression of the AMPK catalytic subunit on cognitive function in aged mice

Mice

Mice were housed in a barrier facility at Wake Forest School of Medicine. Operation of the facility complies with standards of the US Department of Agriculture’s Animal Welfare Information Center (AWIC), and the NIH Guide for Care and Use of Laboratory Animals. The facility is on a 12-hour light/dark cycle, with a regular cage-cleaning and feeding schedule. Female and male mice were both used. Polymerase chain reaction (PCR) was performed to verify the genotypes.

Mouse behavioral studies

Mice for behavioral studies were handled for 5 days before behavioral testing and habituated for at least 1 hour prior to experiments. Open field (OF) test, Novel object recognition (NOR) test, Morris water maze (MWM) test, visible platform test, and passive avoidance (PA) test were carried out as described [16, 17].

Preparation of acute hippocampal slices and synaptic electrophysiology

Transverse hippocampal slices of 400 μm thick were prepared as described previously [18]. Slices were maintained at room temperature for at least 2 hours before experimentation in artificial cerebrospinal fluid (ACSF). For synaptic electrophysiology, slices were transferred to preheated (32° C) recording chambers where they were superfused with ACSF. High-frequency stimulation (HFS, consisting of two 1-sec 100 Hz trains separated by 60 sec) was delivered to induce late-LTP (L-LTP) [12].

Western blots

Brain tissues were dissected, and flash frozen on dry ice. Protocols for Western blot were as described [12]. Resources of antibodies were as described [12].

Surface sensing of translation (SUnSET) assay

Slices were maintained in ACS at room temperature for at least 2 hours before experiments. Slices were further incubated for 1 hour at 32° C in ACSF containing Puromycin (1 μg/mL). Micro-dissected area CA1 slices were used for Western blot analysis. Puromycin labeled proteins and de novo proteins synthesis were detected by using the anti-puromycin antibody, and quantified from the total lane (15 to 250 kDa).

Transmission electron microscopy (TEM)

Protocols for TEM sample preparation and imaging were described in our recent studies [12].

Mass spectrometry (MS)/ proteomic analysis

MS experiments were performed at Rutgers University center for advanced biotechnology and medicine using a Dionex rapid-separation liquid chromatography system interfaced with a QE HF (Thermo Fisher Scientific). Details for MS analysis were described previously [19].

Statistical analysis

Data are presented as mean ± SEM. A two-tailed unpaired Student’s t-test was used for two groups comparisons. One-way ANOVA and post hoc tests (when applicable) were used to compare multiple groups. p < 0.05 were considered statistically significant. Sample size was based on previous publication [12]. GraphPad Prism software was applied for data analysis. Outliers were assessed by Grubbs test.

Genetic suppression of the neuronal AMPKα1 isoform improves aging-related impairments of recognition memory

We developed the AMPKα1/α2 cKO mice as described. In these transgenic mice, expression of either the α1 or the α2 isoform is conditionally reduced in excitatory neurons [12]. The mice were aged to 19-22 months old, when multiple aspects of cognition decline in wild-type (WT) mice [2]. We performed a series of behavioral tasks to assess various cognitive functions in the aged AMPKα1 cKO and α2 cKO mice, with the Cre+/- mice as the control. In the open field (OF) test, we did not observe difference among all the three groups of mice in the measurement of periphery duration and travel distance or velocity, indicating unaltered locomotor activities and anxiety levels with suppression of either AMPKα isoforms (Figure 1A–1C). We next performed the novel object recognition (NOR) task to evaluate their long-term recognition memory (24 hours interval between the exploration day and testing day) [20]. In agreement with our previous studies [2], aged Cre+/- mice exhibited impaired recognition memory, as they did not show a preference for the novel object measured by object interactions (Figure 2A left). Notably, aged AMPKα1 cKO mice interacted more with novel than with the familiar objects (p<0.01), indicating normal recognition memory (Figure 2A middle). In comparison, aged AMPKα2 cKO mice displayed a deficiency in recognition memory (Figure 2A right). Interestingly, there is a trend that the aged AMPKα2 cKO mice spent more time with the familiar object (p=0.07). Results from the analysis of the interactions with the novel objects showed that the long-term recognition memory was improved in the aged AMPKα1 cKO mice compared to the AMPKα2 cKO mice (p<0.05) (Figure 2B). There was also a trending improvement of recognition memory in the aged AMPKα1 cKO mice compared to the control mice (p=0.08) (Figure 2B). We further analyzed the discrimination index (calculated as time spent with novel object – familiar object) /total interaction time) based on the NOR performance, and confirmed that the aged AMPKα1 cKO mice exhibited improved recognition memory (positive high index), compared to aging-related impaired cognition in the other two groups (Figure 2C).

Next, we examined the spatial learning/ memory of the mice with the classic Morris water maze (MWM) assay. Consistent with our recent report [2], aged Cre+/- mice exhibited defects in spatial learning and memory indicated by insignificant day-to-day decrease of escapes latency during the learning/training stage (Figure 3A). During the probe trial test, the aged Cre+/- mice also showed reduced target quadrant occupancy and “platform” crossing (Figure 3C and Supplementary Figure 1D). Unlike the results from the NOR test, both the AMPKα1 cKO and α2 cKO mice displayed similar spatial learning/memory impairments during the MWM test compared to the performance of the control group (Figure 3A, 3C and Supplementary Figure 1A, 1D). Further analysis revealed that the performance of the AMPKα1 cKO mice on day 3 of the MWM training phase was better than that of the AMPKα2 cKO mice (Figure 3B). In addition, we did not observe differences in swimming distance or velocity during the probe trial test among the three groups of mice (Supplementary Figure 1B, 1C). To evaluate the effects of AMPKα isoform suppression on other non-cognitive factors (e.g. vision, motivation, and swimming ability), we performed a visible platform task in which the escape platform is labeled with a flag and moved between trials [12]. We found that the escape latency during the visible platform test was unaltered in day 2 among all groups (Figure 3D). Interestingly, the Cre+/- mice exhibited shorter escape latency compared to the AMPKα2 cKO mice in day 1 (p<0.05) (Figure 3D).

Additionally, we investigated the effects of AMPKα isoform suppression on the performance of fear-associated learning and memory using the passive avoidance (PA) behavioral task. In the PA task, the mice learn to avoid an environment (measured by the entry latency on the testing day), where they had been shocked a day before (training day). The performance of all three groups of mice was similar during the training day. For the testing day, however, the aged AMPKα1 cKO mice showed a trend of improved fear learning/memory compared to the aged Cre+/- group (p=0.06). Moreover, there is no difference between the AMPKα2 cKO mice and the control group (Supplementary Figure 2). Combined with the results from the NOR and MWM tests, inhibition of the neuronal AMPKα1 isoform (but not the AMPKα2 isoform) can alleviate aging-related impairments of object recognition meanwhile does not improve spatial learning and memory deficits in aged mice. It is known that both the NOR and MWM tests are hippocampal-dependent. However, other brain regions such as prefrontal cortex might play distinct roles in these behavioral tasks [21, 22]. It would be intriguing for future studies to elucidate potential brain-region- or cell type-roles of AMPKα isoforms in aging-related regulation of cognitive function.

Genetic suppression of the neuronal AMPKα1 isoform does not alter hippocampal long-term synaptic plasticity in aged mice

Furthermore, we examined whether suppression of the AMPKα isoforms affects hippocampal long-term potentiation (LTP), a form of synaptic plasticity that is considered as a cellular model for learning and memory [23]. In this study, we induced protein synthesis-dependent late LTP by applying a strong electric stimulation protocol [2]. The LTP in the aged Cre+/- mice appeared to be normal (Figure 4), which is consistent with our previous findings in aged wild mice [2]. In addition, the AMPKα1 cKO and α2 cKO mice showed similar hippocampal LTP compared to the Cre+/- mice (Figure 4).

Phosphorylation levels of hippocampal eIF2α in aged mice are reduced by suppression of the neuronal AMPKα1 isoform in aged mice

We seek to understand molecular mechanisms underlying the behavioral phenotypes associated with the suppression of AMPKα isoforms in aged mice. It has been reported that inhibition of overall AMPK activity results in phosphorylation of eEF2 and/or activation of the mTORC1 signaling, which would increase general protein synthesis [11]. Interestingly, we did not observe alterations of eEF2 phosphorylation in the hippocampus of either the AMPKα1 cKO or α2 cKO mice compared to the control group (Figure 5A). In addition, suppression of either AMPKα isoform did not affect the mTORC1 signaling in the hippocampus of the aged mice, as assessed by phosphorylation levels of mTOR (at both Ser2448 and Ser2481 sites), and of the two downstream substrates: p70S6K and 4EBP (Figure 5A and Supplementary Figure 3). Consistently, we also did not observe any significant effects of AMPKα isoform suppression on the phosphorylation levels of AKT, a well-established upstream regulator of mTORC1 (Figure 5A).

We recently reported, in young mice, that reduction of AMPKα2 results in increased phosphorylation of the eIF2α, which is associated with inhibition of translational capacity [12]. In contrast, here we did not observe alterations of eIF2α phosphorylation in the hippocampus of the aged AMPKα2 cKO mice compared to the Cre+/- mice group (Figure 5B). Notably, phosphorylation levels of eIF2α in the hippocampus of the aged AMPKα1 cKO mice were significantly decreased compared to those in either the control or the AMPKα2 cKO mice (Figure 5B). We further investigated ATF4, whose expression might be associated with eIF2α phosphorylation [24, 25]. We did not observe any changes in the protein levels of hippocampal ATF4 in AMPKα1 or α2 cKO mice (Figure 5C). Additionally, protein levels of eIF2α kinase PERK were also unaltered among all the groups (Figure 5D).

Phosphorylation of eIF2α is considered to result in repression of general protein synthesis [26]. We thus examined de novo protein synthesis in living hippocampal slices by the SUnSET assay [16]. Surprisingly, the results indicate no change of general protein synthesis levels (assessed by puromycin incorporation) among all three groups (Figure 5E). We further carried out transmission electron microscopy experiments to investigate regulation of hippocampal dendritic polyribosomes, an indicator of new protein synthesis [27]. We found that the presence of polyribosomes was unaffected by the suppression of either AMPKα1 or α2, compared to the control group (Figure 5F). Potential mechanisms for the unaltered overall de novo protein synthesis (in consideration of the eIF2α phosphorylation regulation) may involve the regulation of the protein degradation process and a balance between both upregulation and downregulation of individual protein synthesis (see the proteomics results below). Taken together, suppression of the AMPKα isoform in neurons does not alter the general mRNA translation rate in the hippocampus of aged mice.

Proteomic analysis reveals distinct alterations of protein expression levels associated with suppression of the neuronal AMPKα isoform in aged mice

To gain insights into the potential regulation of the expression of individual proteins associated with the AMPKα isoform reduction, we performed mass spectrometry (MS)-based proteomic experiments. (Figure 6C, 6D). Most significantly changed hippocampal proteins (up-regulated or down-regulated) in the AMPKα1 cKO or α2 cKO mice compared to the Cre+/- mice were shown in Tables 1–4 (illustrated in Supplementary Figure 4). To understand the differences in proteomics between AMPKα1 cKO and AMPKα2 cKO mice, we further generated a “volcano” plot to show the fold changes of protein expression levels in the AMPKα2 cKO mice compared to those in the AMPKα1 cKO mice (Figure 6B). In brief, levels of 29 proteins were increased and 16 proteins were decreased in the AMPKα2 cKO mice compared to the AMPKα1 cKO mice. Detailed lists of these proteins were shown in Tables 5, 6. To further understand how these regulated proteins are involved in different biological processes, we carried out the Gene Ontology (GO) analysis and the Gene Set Enrichment Analysis (GSEA) in AMPKα2 cKO and AMPKα1 cKO mice (Figure 6C, 6D). GO analysis of the differentially regulated proteins between AMPKα1 cKO and AMPKα2 cKO mice suggested that expression of proteins involved in the biological processes related to ribosome subunit biology was suppressed in the AMPKα2 cKO mice compared to AMPKα1 cKO mice. In contrast, expression of proteins involved in the metabolism-related processes was activated in the AMPKα2 cKO mice compared to AMPKα1 cKO mice (Figure 6C). Consistent with GO analysis, GSEA analysis shows that ribosome-related pathways were suppressed, while metabolism-related pathways were activated in AMPKα2 cKO mice compared to AMPKα1 cKO mice (Figure 6D). Interestingly, we found COVID-19-related pathways were suppressed in the AMPKα2 cKO mice, suggesting AMPKα isoforms may play distinct role in viral infection (Figure 6D).