Molecular elevation of insulin receptor signaling improves memory recall in aged Fischer 344 rats

Abstract As demonstrated by increased hippocampal insulin receptor density following learning in animal models and decreased insulin signaling, receptor density, and memory decline in aging and Alzheimer's diseases, numerous studies have emphasized the importance of insulin in learning and memory processes. This has been further supported by work showing that intranasal delivery of insulin can enhance insulin receptor signaling, alter cerebral blood flow, and improve memory recall. Additionally, inhibition of insulin receptor function or expression using molecular techniques has been associated with reduced learning. Here, we sought a different approach to increase insulin receptor activity without the need for administering the ligand. A constitutively active, modified human insulin receptor (IRβ) was delivered to the hippocampus of young (2 months) and aged (18 months) male Fischer 344 rats in vivo. The impact of increasing hippocampal insulin receptor expression was investigated using several outcome measures, including Morris water maze and ambulatory gait performance, immunofluorescence, immunohistochemistry, and Western immunoblotting. In aged animals, the IRβ construct was associated with enhanced performance on the Morris water maze task, suggesting that this receptor was able to improve memory recall. Additionally, in both age‐groups, a reduced stride length was noted in IRβ‐treated animals along with elevated hippocampal insulin receptor levels. These results provide new insights into the potential impact of increasing neuronal insulin signaling in the hippocampus of aged animals and support the efficacy of molecularly elevating insulin receptor activity in vivo in the absence of the ligand to directly study this process.


| INTRODUC TI ON
Insulin and insulin receptor (IR) signaling is known to be an integral component of healthy brain function, and numerous cell types, including endothelial cells, astrocytes, and neurons, express IRs throughout the brain (Zhang et al., 2014). This is particularly important with regard to learning and memory processes. Indeed, the hippocampus, an area of the brain integral for these functions, has consistently shown robust IR expression (Dore, Kar, Rowe, & Quirion, 1997;Unger, Livingston, & Moss, 1991). In hippocampal neurons, these receptors are often localized to postsynaptic densities (particularly in field CA1) where they have been reported to modulate AMPA and NMDA receptors, improve synaptic plasticity, increase hippocampal metabolism, and activate genes and pathways involved in both long-and short-term memory encoding (Adzovic & Domenici, 2014;Pearson-Leary, Jahagirdar, Sage, & McNay, 2018;Zhao, Chen, Quon, & Alkon, 2004). Additionally, learning has been shown to have a direct impact on hippocampal IR activity, with multiple studies reporting that rats had elevated receptor density and markers of IR signaling after training on the Morris water maze (MWM) behavioral task (Zhao et al., 1999;Zhao et al., 2004).
These results have been corroborated by numerous clinical studies which have also reported positive effects of INI administration on learning, verbal memory performance, and functional abilities (ADCS ADL scores) in older AD and MCI patients (Freiherr et al., 2013). However, while it is clear that insulin is closely tied to learning and memory and that enhancing IR activity can help ameliorate age-and AD-associated cognitive impairments, the precise pathways underlying these effects are still unclear. For this reason, more mechanistic investigations into CNS insulin actions in animal models using molecular methods are needed.
Studies focused on molecular modification of IR activity without involvement of the ligand have recently been employed to study the impact of receptor inhibition or loss of function on learning and memory. Some of the molecular techniques that are commonly used to accomplish this include genetic knockout of IR or associated signaling molecules (Bruning et al., 2000;Cai et al., 2018;Costello et al., 2012;Garcia-Caceres et al., 2016), the lentiviral delivery of an IR antisense sequence (LV-IRAS) (Grillo et al., 2015), and the introduction of IR inhibitory peptides (Luckett, Frielle, Wolfgang, & Stocker, 2013;Maimaiti et al., 2017;Paranjape et al., 2010). Results of these studies thoroughly support the importance of IR signaling in cognitive processes, with one showing that CNS-specific deletion of insulin receptor substrate 2 (IRS-2), an adaptor protein involved in downstream IR signaling, is associated with the loss of hippocampal metaplasticity in mice (Costello et al., 2012). Specifically, these IRS-2 knockout mice had impaired long-term potentiation, reduced hippocampal glutamatergic transmission at CA1 synapses, and decreased levels of IR signaling molecules such as GSK-3 and pAkt. Interestingly, the deleterious effects of molecular IR inhibition do not appear to be limited to only neurons, as one study has also shown that astrocyte-specific deletion of IR in mice is associated with reduced astrocytic exocytosis of ATP, decreased purinergic signaling of dopaminergic neurons, and elevated anxiety-and depressive-like behaviors (Cai et al., 2018). Similarly, work from a different group, also in mice, showed that postnatal ablation of astrocytic IRs (Cre/lox system), as well as knockout of astrocytic IRs in the hypothalamus, is associated with reduced astrocytic glucose availability, diminished glucose-induced activation of pro-opiomelanocortin (POMC) neurons, and impaired responses to peripheral glucose (Garcia-Caceres et al., 2016).
However, while these studies have provided significant insights into the processes impacted by impaired IR activity in the brain (a phenomenon strongly implicated in the progression of pathological aging and AD), they do not directly report on the potential mechanisms targeted by exogenous insulin. As the use of INI to elevate IR activity is currently being explored as a therapeutic, it is imperative that molecular approaches that induce a gain of function of the brain-specific IR are also investigated. Here, we attempted to address this need by employing a molecular approach to increase, rather than decrease, IR activity using a constitutively active human IR (IRβ). We expressed this modified receptor in the hippocampus of young and aged Fischer 344 (F344) rats to test the impact of elevated insulin signaling on ambulatory gait performance, receptor expression, and spatial learning and memory. We show that the molecular induction of increased IR signaling mildly altered gait performance and improved memory recall in aged animals performing the MWM task. IRβ expression was also associated with elevated downstream IR signaling markers and total IR density in the hippocampus of aged animals, indicating that prolonged receptor activation does not appear to trigger the downregulation of receptor levels or signaling in these animals. Surprisingly, animals that received the IRβ construct also showed increased nuclear staining (DAPI) reflecting a potential beneficial impact on cell health, although not likely through an enhancement in the number of neurons. The results presented here imply that this alternative approach to increase neuronal IR activity in the hippocampus in vivo without the need for the ligand is safe, effective, and able to alleviate memory impairment in aged animals.
This work also highlights the importance of using novel molecular techniques to study IR signaling in the context of age-related cognitive decline.

| Hippocampal expression of the constitutively active IRβ does not impact overall animal health or gait performance
Animals were assessed for changes in body weight following 5 weeks of constitutive insulin signaling driven by the IRβ construct. While a main effect of age was noted (2-way ANOVA; F (1,29) = 247.10, p < 0.0001), the AAV treatment did not cause any significant changes in weight in either age-group (Table 1; p > 0.05). Ambulatory performance was measured across age-and treatment groups to assess the impact of constitutive hippocampal IR signaling on gait ( Figure 1).

| Expression of the IRβ receptor is associated with improved MWM performance in aged animals
Animals were assessed for learning and memory performance using the MWM task. All groups showed significant learning across the three training days (Figure 2a; 3-way RM ANOVA; F (1.82,36.41) = 22.67, p < 0.001). As expected, aged animals performed more poorly than young, learning at a slower rate (3-way RM ANOVA; F (1,20) = 7.97, p = 0.012). Aged animals also had significantly longer path lengths to the goal proximity ring on the memory probe day (Figure 2b; 2-way ANOVA; F (1,20) = 6.92, p = 0.016). Interestingly, a significant interaction term was noted on this measure in animals that received the IRβ construct, with older animals displaying evidence of improved memory recall (2-way ANOVA; F (1,20) = 5.82, p = 0.026). This suggests that expression of the constitutively active IRβ construct and increases in IR signaling can beneficially impact memory processes in aging.

| Aged animals expressing hippocampal IRβ have elevations in downstream IR signaling markers
Hippocampal homogenates were probed for the HA-tag, the β-subunit of the IR, pAkt, and Akt expression using Western immunoblot techniques. The endogenous IR is predicted to have a molecular weight of ~230 kDa, with an ~95 kDa β-subunit. Based on the combination of the HA-tag structure (~100 amino acids) and the structure of the modified IRβ receptor used here (~470 amino acids), the total predicted molecular weight of our construct is ~60 kDa. However, analysis of Western immunoblots using an antibody specific to the IR β-subunit revealed the presence of three, rather than two, distinct bands: the endogenous IR band at ~95 kDa, the modified IRβ band at ~60 kDa, and an additional band at ~85 kDa (Figure 3a, right). While the ~85 kDa band was unexpected, it appears to align with an additional HA-tag band at a similar molecular weight (Figure 3a, left). This suggests that the proteins at ~85 kDa are either a by-product of our IRβ construct or a post-translational modification (i.e., ubiquitination or glycosylation) of the ~60 kDa IRβ protein to target it for degradation and thus are TA B L E 1 Measures of animal weight and ambulatory performance in young and aged animals treated with control or IRβ AAVs  suggest that, at least in the F344 animal, the hippocampus does not appear to experience a compensatory reduction in endogenous IR production, even in the presence of elevated signaling.

| IRβ expression is associated with increases in immunostained IR-positive area in the hippocampus
To investigate the impact of the IRβ construct on total IR expression, we measured the immunopositive area for the β-subunit of the IR (endogenous and exogenous) in hippocampal sections. We initially focused on modifying IR expression in primary projecting neurons of the hippocampus (stratum pyramidale of field CA1), but also included results measured in field CA3 to investigate potential changes across the dorsal-ventral axis of the hippocampus. In order to account for the impact of cell number when quantifying immunopositive signals, we normalized the thresholded % area covered of all FITC signals (green) measured in fields CA1 and CA3 to the thresholded % area covered of DAPI (a marker of adenine-thymine-rich regions of DNA that labels nuclei; blue) measured in stratum pyramidale of the corresponding field.
We show that IRβ expression was associated with a significant in- Analysis of immunopositive DAPI (mean gray value) in fields CA1 and CA3 revealed that IRβ expression was associated with elevated signal intensity across all fields and subfields tested (Table 2). Of note, DAPI signals measured in CA1 stratum pyramidale were nearly 2-fold higher in tissue from IRβ-treated animals compared to tissue from controls (2way ANOVA; F (1,12) = 11.92, p = 0.005). Surprisingly, these results suggest that elevated insulin signaling in the hippocampus may improve cellular health or growth (albeit not through increasing neuron numbers; see NeuN immunohistochemistry below), and also clearly show that our molecular approach was able to alter receptor dynamics.
As our AAV treatment was targeted to CA1, the presence of immunopositive signal in stratum pyramidale of field CA3 could be attributed to neuronal terminals incorporating the virus, leading to expression in the somatic region of this field. However, we did note significantly smaller levels of immunopositive signal in CA3 compared to CA1 (3way ANOVA; F (1,24) = 15.04, p = 0.001), which is perhaps not surprising given our intent to target field CA1. Interestingly, we also report significant main effects of age (3-way ANOVA; F (1,24) = 12.82, p = 0.002) and IRβ treatment (F (1,24) = 22.41, p < 0.0001) in both fields.

| NeuN immunohistochemistry of hippocampal sections
To investigate the impact of IRβ expression and increased IR signaling on neuronal density or the potential infiltration and expansion of other cell types (i.e., astrocytes or microglia), we quantified F I G U R E 1 Timeline for AAV injections and behavioral measures performed in young and aged F344 animals. Stereotaxic AAV injections began on Week 1, 3 days after animals arrived. Animals were then given 2-4 weeks of recovery prior to the initiation of gait (Week 5) and MWM (Week 6) measures. Following completion of the MWM, animals were perfused, and tissue was harvested for further analyses NeuN-positive neurons in subfield stratum pyramidale of hippocampal fields CA1 and CA3. Quantification of field CA3 showed a significant effect of age (Table 3; 2-way ANOVA; F (1,12) = 18.90, p = 0.001), as well as a trend for a reduction in neuron number in IRβ-treated animals compared to controls (F (1,12) = 4.17, p = 0.064). Surprisingly, this was not reflected in field CA1, as no effect of age or IRβ was detected (p > 0.05, respectively), suggesting that the IRβ-associated increase in immunopositive DAPI signals reported in Table 2 may be mediated through mechanisms independent of neurogenesis.

| D ISCUSS I ON
In addition to its role in spatial learning and memory, the hippocampus is also associated with the processing of information necessary for controlled ambulatory performance. In fact, age-associated changes in the control of movement are well recognized in the clinic (Beauchet, Allali, Launay, Herrmann, & Annweiler, 2013), where greater variability in these measures is correlated with structural or functional differences not only in the somatosensory cortex, but also in the hippocampus, anterior cingulate gyrus, and basal ganglia (Tian et al., 2017). For this reason, we sought to explore whether elevating hippocampal insulin signaling in young and aged animals through the use of a neuron-specific molecular approach that does not require administration of the ligand (Frazier et al., , 2020Lebwohl, Nunez, Chan, & Rosen, 1991) is a viable method to improve memory processes or ambulatory performance. We also attempted to corroborate our previous work using INI delivery in the same animal model (F344 rats) at comparable ages F I G U R E 2 Morris water maze performance in young and aged animals treated with control or IRβ AAVs. (a) Path length to goal across learning (training days) and memory (probe day). All animals (young control n = 8, young IRβ n = 5, aged control n = 5, aged IRβ n = 6) learned the task successfully, as indicated by reductions in path length across the training days (3-way RM ANOVA; F (1.82,36.41) = 22.67, p < 0.001).
As expected, aged animals learned at a slower rate than young animals ( (Table 1). However, this result does highlight the importance of the hippocampus in controlling at least some aspects of gait (Tian et al., 2017) and points to potential insulin-mediated reductions in the afterhyperpolarization (Maimaiti et al., 2016;Pancani et al., 2013) which could perhaps alter network activity in the microcircuitry guiding motor control. Additionally, reductions in the number of midline crossovers, stride length, and offset differentials in the aged group were noted (Table 1). This may reflect greater inquisitive behavior in the younger animals, or perhaps an underestimation of mobility (e.g., wider haunches) of larger animals in a relatively confined space. Interestingly, these results were corroborated by similar reductions in offset differentials from midline in the aged group, likely due to more focused and determined ambulatory behavior in these animals. Clearly, more investigations into the role of hippocampal IR activity in the control of gait are warranted.
The potential mechanisms by which IRβ expression improved spatial memory may have been driven by several factors. One consideration could be the enhancement of insulin signaling beyond  Further, another study using a different antibody for the β-subunit of the IR in this same animal model at later ages (17-18 months) also showed no aging differences in IR expression (Pancani et al., 2013).
Discrepancies between our findings and those of others could be attributed to multiple factors, including the study type performed Thus, it could be that F344 rats do not experience a reduction in IR expression with age but still benefit from a supraoptimal enhancement of insulin signaling through yet unknown mechanisms.
In addition to enhancing downstream IR signaling, the positive impact of elevating IRβ expression on spatial learning and memory could also be attributed to insulin's ability to reduce neuroinflam-

F I G U R E 4
Western immunoblots probing for downstream IR signaling markers in hippocampal tissue from control and IRβtreated animals. (a) Representative photomicrograph of Western immunoblots performed on hippocampal tissue taken from 30 animals (young control = 8, young IRβ n = 5, aged control n = 9, aged IRβ n = 8) probing for pAkt (left) or Akt (right). (b-c) Quantification of pAkt (b) and Akt (c) expression. No differences across age or treatment were detected for either marker (2-way ANOVA; p > 0.05). (d) Quantification of pAkt/Akt ratios indicate that our manipulation was able to impact downstream IR signaling in the hippocampus, as indicated by a significant interaction (2-way ANOVA; F (1,26) = 8.09, p = 0.009). Only aged animals responded with increased IR signaling following IRβ treatment. All data were normalized to total protein levels (Ponceau S staining). Asterisks (*) indicate significance at p < 0.05. Data represent means ± SEM Other mechanisms that could potentially underly the impact of  A robust effect of IRβ treatment on field CA1 immunopositive DAPI signal across the three subfields measured was detected in both young and aged animals: stratum pyramidale (F (1,12) = 11.92, p = 0.005), stratum oriens (F (1,12) = 10.66, p = 0.007), and stratum radiatum (F (1,12) = 13.6, p = 0.003).
Similarly, IRβ was also associated with increased DAPI signals in subfield stratum pyramidale of field CA3 (F (1,12) = 12.11, p = 0.005). No effect of age was noted in either of the fields measured (p > 0.05). Non-significance is indicated by "n.s." Data represent means ± SEM. Stella, Bryson, & Thoreson, 2001;Thibault et al., 2013), as these processes have a rich history of association with mechanisms tied to cognitive decline in aging or AD. Additionally, while the AAV delivery initially targeted field CA1, we did note alterations in IR density in field CA3 ( Figure 5). However, neuronal communication in field CA3 neurons is often associated with short-term spatial working memory (Duncan, Tompary, & Davachi, 2014;Lee & Kesner, 2003) and thus is not likely to have participated in the enhancement of long-term memory recall during the more delayed 24-h probe trial. It is also interesting to note that only two subfields (strata pyramidale and radiatum) of CA1 responded to the IRβ treatment used here. One potential explanation for a lack of effect in stratum oriens is the presence of numerous interneurons or basket cells in this region, as these cells are important for trace encoding and fear conditioning-type behaviors (Lovett-Barron et al., 2014) and are less likely to be targeted by our construct. Further, strata pyramidale and radiatum form the last synaptic structures within the hippocampal circuit and are therefore more likely to impact spatial navigation and performance (i.e., the MWM task). Finally, the separation of these subfields by mere microns suggests that the IRβ construct did not have a generalized effect throughout the hippocampus; nevertheless, future studies regarding the impact of insulin signaling between hippocampal fields and subfields are still warranted.
Although the findings reported here indicate that our molecular approach was safe, as highlighted by the lack of noticeable ill effects on the overall health of the animals (Table 1), the possibility of aberrant cellular growth caused by sustained signaling of a growth hormone must be considered. However, given that only a small IRβ-associated increase in hippocampal DAPI signals was detected (Table 2) along with no significant IRβ-associated changes in NeuN-positive neurons (Table 3), it is unlikely that our treatment led to uncontrolled growth.
Interestingly, our results also indicated that neither endogenous IR expression nor downstream IR signaling was reduced following several weeks of sustained insulin activity driven by the modified IRβ receptor (Figures 3 and 4). While the lack of endogenous IR downregulation is surprising, it could be attributed to functional differences between peripheral and CNS receptor isoforms. In fact, some work has shown that the brain-specific IR isoform (IR-A) does not downregulate as quickly as the peripheral IR-B isoform following prolonged incubation with the ligand (Boyd & Raizada, 1983). Additionally, while our results show that the constitutively active IRβ construct had very little impact on gait in the current study, this treatment did not critically impair ambulatory performance either, further supporting the overall safety of this technique.
In summary, the work presented here indicates that our molecular approach for increasing neuronal IR activity in the hippocampus in vivo without the need for the ligand is well-tolerated, effective, and able to alleviate age-dependent spatial memory impairments on the MWM task. Importantly, our findings also highlight the value of exploring novel molecular techniques to assess the impact of elevated insulin signaling in the CNS and directly study the specific mechanisms and pathways associated with IR activation in the hippocampus across different cell types.

| Animal models and stereotaxic AAV delivery
Male F344 rats, aged 2 (n = 14) or 18 months (n = 19), were obtained from the National Institute on Aging colony. Animals were housed in TA B L E 3 Number of neurons in fields CA1 and CA3 of hippocampal sections following NeuN immunohistochemistry. No effect of IRβ treatment or age was detected on neuron numbers in stratum pyramidale of field CA1 following NeuN immunohistochemistry of hippocampal sections (p > 0.05). In field CA3, a significant age-associated decrease in the number of NeuN-positive neurons was noted (2way ANOVA; F (1,12) = 18.90, p = 0.001), as well as a trend for reduced neuron numbers in IRβ-treated animals compared to controls (F (1,12) = 4.17, p = 0.064). Non-significance is indicated by "n.s." A superscript hash ( # ) indicates a trend at p < 0.10. Data represent means ± SEM.

Number of NeuN-positive neurons per 100 µm
pairs, tail marked for identification, maintained on a 12 h ON/12 h OFF light schedule, and fed Teklad global 18% protein rodent diet (2018; Harlan Laboratories, Madison, WI) ad libitum. For the next week, beginning on the fourth day after arrival, all animals received injections (2 µl per side, 0.2 µl/min) of either a control AAV containing a neuronspecific synapsin promoter and a fluorescent marker (mCherry) or an experimental AAV containing the synapsin promoter, mCherry, and the constitutively active IRβ receptor (Lebwohl et al., 1991)

| Gait behavioral tests
Animals were assessed for gait measures over 2 days on the fifth week post-surgery. Each animal was given one untracked test run to acclimate to the task, which consisted of a single, continuous, evenspeed walk down a corridor ending at their home cage. The corridor was then lined with paper and the animal's fore or hind paws were coated in nontoxic black tempera paint (Crayola LLC, Easton, PA).
Animal tracks were recorded once for fore paw positions and once for hind paw positions on each of the two testing days. Tracks were measured for stride, number of crossovers passing the midline, and offset differentials (defined as the sum of the absolute differences between the left and right fore paws' offsets from the midline).
Stride data were derived from an average of at least seven steps of both fore and hind paws across the two testing days, while measures of crossovers and offset differentials were derived from fore paws only. A number of crossovers passing the midline were then normalized to step number for each animal. For stride and offset differentials, data were normalized to the average width of haunches based on 6 young and 6 aged animals (in cm: young 7.37 ± 0.14, aged 8.74 ± 0.8). We present gait results from 30 animals (young control n = 8, young IRβ n = 5, aged control n = 9, aged IRβ n = 8).

| MWM behavioral tests
Spatial learning and memory were tested using the MWM. was removed based on both our visual acuity filter and the HA-tag filter. We present MWM performance measures from 24 animals (young control n = 8, young IRβ n = 5, aged control n = 5, aged IRβ n = 6).

| Western immunoblots
All Western immunoblots were conducted using 8% fresh-cast gels with overnight wet transfer at 15°C and overnight primary antibody incubation at 15°C. For HA-tag immunoblots, membranes were placed in 5% BSA in TBST for both blocking and antibody incubations. For pAkt, Akt, and total IR β-subunit immunoblots, membranes were blocked in 8% nonfat dry milk in TBST, while antibodies were diluted in 1% nonfat dry milk in TBST. The same protein concentration (50 µg) was loaded for each sample.  for imaging using a spectral analysis software package and camera (Nuance ® , CRi Inc., Hopkinton, MA) with the ability to scan fluorescence wavelengths in 10 nm increments using long-pass emission filters (DAPI: >400 nm LP; FITC: >500 nm LP). Excitation filters were centered on the DAPI (350 ± 50 nm) and the FITC (470 ± 40 nm) fluorophores. Settings for image acquisition were the same across all sections imaged. Pseudo-colored images were converted to grayscale prior to analysis. ImageJ was used to quantify the % area covered of immunopositive DAPI and FITC signals in fields CA1 and CA3 across hippocampal sections (3 sections/ subject). The same region of interest (ROI) was used for each tissue section. To account for changes in cell number, FITC signals were normalized to the immunopositive DAPI measured in stratum pyramidale of their respective fields (CA1 or CA3), as this subfield is comprised of neuronal soma and thus provides the most accurate representation of neuronal cell density. As an additional measure of cell number, the mean gray value of immunopositive DAPI signal only was also measured in stratum pyramidale of both fields. We present immunofluorescence results from a total of 16 animals (young control n = 4, young IRβ n = 3, aged control n = 5, aged IRβ n = 4).

| NeuN immunohistochemistry of hippocampal sections
Hippocampal tissue sections were probed using an anti-NeuN pri-

| Statistical analysis
All results were derived (GraphPad Prism v8.4, GSL Biotech LLC, Chicago, IL) using either a two-way analysis of variance (2-way ANOVA), a three-way analysis of variance (3-way ANOVA), or a three-way analysis of variance with repeated measures (3-way RM ANOVA). Tukey's post hoc tests were used for all ANOVAs. All data are reported as means ±SEM. Significance for all measures was set at p < 0.05.

CO N FLI C T O F I NTE R E S T
The authors report no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available as supplementary material.