**4. Discussion**

The present study demonstrates that memory decline in T1DM and T2DM in middle age occurs in an Aβ-independent manner, which is consistent with the findings of previous studies [34–36]. Similarly, age-related memory impairments arise with no alteration in Aβ levels.

In middle-aged DIO and aged mice, Aβ-unrelated memory deficits are accompanied by increased basal phosphorylation of hippocampal IRS1 at specific Ser residues, whereas Aβ-unrelated memory impairments develop in STZ mice without the modification of IRS1. Although p70S6K activation accompanied by unchanged AMPK activity in the hippocampus is mutually observed with Aβ-unrelated memory deficits in middle-aged DIO, STZ, and aged mice regardless of the presence or absence of the phosphorylation of IRS1 at Ser residues, different alterations in other downstream components were observed in these mouse models. Furthermore, we found that, in APPKINL-G-F mice, a concomitant increase in the basal phosphorylation of hippocampal IRS1 at multiple Ser sites with the activation of AMPK and elevation of Aβ42 level had already arisen at a young age, before the onset of memory decline, and that memory decline in middle age was accompanied by the persistence of these changes with age-related increase in the phosphorylation of IRS1 at specific Ser sites and in Aβ42 level.

HFD and genetically obese animals exhibit increased phosphorylation at mSer307 and p70S6K-induced phosphorylation at mSer1097 in peripheral tissues [37,38]. In the central nervous system, the involvement of non-activated AMPK with/without phosphorylation of IRS1 at mSer307 accompanied by p70S6K activation or different combinations of alterations in Akt and GSK3β activity (i.e., increased or decreased phosphorylation) in cognitive impairment in 40–45% HFD-fed and STZ mice has been reported [21,22,27,28,32,34,39,40], whereas the monotonous levels of phosphorylation of IRS1 at Ser sites and of downstream kinases including Akt, AMPK, GSK3β, and p70S6K are observed in 45% HFD-induced cognitive deficits [41]. Consistent with these results, memory decline in middle-aged DIO mice is accompanied by a concomitant increase in the basal phosphorylation of IRS1 at mSer307 and mSer1097 with p70S6K activation and monotonous activity of Akt, GSK3β, and AMPK in the hippocampus. On the other hand, in aged mice, the concomitant activation of Akt and GSK3β with an increase in the basal phosphorylation of hippocampal IRS1 at mSer307, mSer612, and mSer632/635 with p70S6K activation is observed, which is consistent with previous studies showing that the increased phosphorylation of mSer632/635 or mSer1097 on IRS1 is associated with p70S6K [5,37,38]. Given that the activation of mSer612 on IRS1 may negatively correlate with intracellular signaling and that the activation of Ser632/635 on IRS1 may occur independently of Akt [5], it is likely that, in aged mice, the increased phosphorylation of IRS1 at mSer307, mSer612, and mSer632/635 with the activation of multiple downstream factors other than AMPK may arise through reciprocal feedback regulation.

Additionally, HFD- and STZ-induced diabetes increases the phosphorylation of JNK accompanied by the activation of mSer307 on IRS1 in the brain [21,32]. The activity of JNK is elevated in aged mice and transgenic mouse models of AD, although the involvement of phosphorylation of IRS1 at Ser sites in the altered activation of JNK has not been reported [16,42]. Furthermore, the activation of aPKC ζ/λ is associated with the phosphorylation of IRS1 at Ser sites, including mSer1097 [5,20]. However, in all types of mouse models of memory impairment in the present study, the basal phosphorylation levels of JNK and aPKC ζ/λ consistently remained unchanged under our experimental conditions. These discrepancies may be due to the differences in observation time, animal species, and HFD or STZ protocols, such as age, duration of exposure, diet fat content, and drug dosages. Taken together, these results suggest that the reciprocal effects of the phosphorylation of IRS1 at T2DM- or age-related Ser sites and downstream components via feedback loops that may lead to the common alterations in the activity of p70S6K and AMPK are involved in Aβ-unrelated memory decline; however, the modification of IRS1 through Ser sites is not required for the onset of memory deficits in STZ-induced T1DM mice.

Although a link between the phosphorylation of neural IRS1 at mSer307, mSer612, and mSer632/635 and brain insulin resistance has been proposed [14,16,43], brain insulin resistance is not yet defined, and the phosphorylation of IRS1 at Ser sites emerges in insulin-dependent and insulin-independent manners [5,6]. Given that diabetes- and obesity-induced memory impairments were accompanied by the phosphorylation of IRS1 at mSer307 [5], we found that young APPKINL-G-F mice showing normal metabolism and memory function displayed increased phosphorylation of hippocampal IRS1 at three Ser residues including mSer307, mSer612, and mSer1097, in which the Aβ42 level had already increased before the onset of memory decline. Consistent with previous studies [13–16], the increased phosphorylation at mSer612 and mSer 632/635 and persistent activation of mSer1097 are observed in the hippocampus of middle-aged APPKINL-G-F mice exhibiting memory decline with age-associated increase in the Aβ42 level.

Interestingly, in both young and middle-aged APPKINL-G-F mice, the concomitant phosphorylation of hippocampal IRS1 at mSer612 and mSer1097 with the activation of AMPK is constantly observed regardless of the presence or absence of the activation of mSer307 or mSer 632/635, suggesting that persistent phosphorylation of mSer612 and mSer1097 accompanied by AMPK activation may contribute to memory dysfunction in AD.

Reportedly, the Arctic mutation in APPKINL-G-F mice reduces immunoreactivity in ELISA because the location of this mutation on the Aβ sequences is overlapped with the binding region of monoclonal antibody to Aβ used in human/rat/mouse Aβ sandwich ELISA [19]; however, T-PER-extractable Aβ40 and Aβ42 levels were successfully determined by human/rat/mouse Aβ sandwich ELISA using BNT77 (binds to the location of the Arctic mutation) and human Aβ sandwich ELISA using BAN50 (does not bind to the location of the Arctic mutation). We confirmed that the elevation of Aβ42 level is evident in young APPKINL-G-F mice and that the Aβ42 level increased in middle-aged APPKINL-G-F mice. Owing to the elevation of Aβ42 level in the hippocampi of APPKINL-G-F mice from a young age, it is likely that the increased phosphorylation of IRS1 at Ser sites is associated with Aβ42 level but not with memory decline. Consistent with this finding, a previous study showed that the phosphorylation of IRS1 at mSer307/hSer312 and mSer612/hSer616 increased in cultured hippocampal neurons exposed to Aβ oligomers prepared from synthetic Aβ1–42 peptide and in the AβO-injected hippocampi of non-human primates [16]. Interestingly, the increased basal phosphorylation of hippocampal IRS1 at multiple Ser sites in APPKINL-G-F mice is accompanied by AMPK activation regardless of the presence or absence of memory decline. Consistent with our findings, recent studies suggest that AMPK activation is implicated in brain aging and development of neurodegenerative diseases, including AD [44,45]. These results suggest that the elevation of Aβ42 accompanied by AMPK activation induced by energy depletion that occurs from a young age before the onset of memory decline is associated with increased phosphorylation of hippocampal IRS1 at multiple Ser sites and that sustained activation of these factors contributes to the onset of memory decline in middle-aged APPKINL-G-F mice.

In summary, whether the modification of IRS1 through its phosphorylation at Ser sites in the hippocampus has a pathogenic or an adaptive function remains unknown because the phosphorylation of hippocampal IRS1 at mSer307, mSer612, and mSer1097 increased when metformin improves memory deficits in middle-aged DIO mice [20,46]. Studies using mutant mice with IRS1 Ser residues in the brain will help us to understand the roles of Ser sites in memory function and to identify unrecognized downstream pathways.
