From Determining Brain Insulin Resistance in a Sporadic Alzheimer’s Disease Model to Exploring the Region-Dependent Effect of Intranasal Insulin

Impaired response to insulin has been linked to many neurodegenerative disorders like Alzheimer’s disease (AD). Animal model of sporadic AD has been developed by intracerebroventricular (icv) administration of streptozotocin (STZ), which given peripherally causes insulin resistance. Difficulty in demonstrating insulin resistance in this model led to our aim: to determine brain regional and peripheral response after intranasal (IN) administration of insulin in control and STZ-icv rats, by exploring peripheral and central metabolic parameters. One month after STZ-icv or vehicle-icv administration to 3-month-old male Wistar rats, cognitive status was determined after which rats received 2 IU of fast-acting insulin aspart intranasally (CTR + INS; STZ + INS) or saline only (CTR and STZ). Rats were sacrificed 2 h after administration and metabolic and glutamatergic parameters were measured in plasma, CSF, and the brain. Insulin and STZ increased amyloid-β concentration in plasma (CTR + INS and STZ vs CTR), while there was no effect on glucose and insulin plasma and CSF levels. INS normalized the levels of c-fos in temporal cortex of STZ + INS vs STZ (co-localized with neurons), while hypothalamic c-fos was found co-localized with the microglial marker. STZ and insulin brain region specifically altered the levels and activity of proteins involved in cell metabolism and glutamate signaling. Central changes found after INS in STZ-icv rats suggest hippocampal and cortical insulin sensitivity. Altered hypothalamic metabolic parameters of STZ-icv rats were not normalized by INS, indicating possible hypothalamic insulin insensitivity. Brain insulin sensitivity depends on the affected brain region and presence of metabolic dysfunction induced by STZ-icv administration.


Introduction
Our knowledge regarding the effect of insulin in the brain is still modest and inadequate. Impaired response to insulin in the brain has been linked to many neurodegenerative disorders like Alzheimer's disease (AD) [1][2][3][4][5], suggesting a possible link between type 2 diabetes and AD [6][7][8][9][10]. A common postulation is that the brain is considered an insulin-insensitive organ, but emerging evidence suggests that insulin has a role in glucose metabolism and mitochondrial functioning in the brain [11][12][13]. The glucose transporter GLUT4 (insulin-dependent) has been found co-localized with neurons expressing the insulin-independent transporter GLUT3 in several brain regions, such as the basal forebrain, hippocampus, amygdala, cerebral cortex, cerebellum [14], and hypothalamus [15]. In addition, GLUT4 is found distributed in brain areas enriched with insulin receptors [16], indicating the role of insulin in GLUT4-mediated glucose transport in the brain. Intracerebroventricular (icv) administration of streptozotocin (STZ) induces cognitive deficit in rats [17][18][19], presumably through the induction of brain insulin resistance; hence, the STZ-icv rat model has been proposed as a non-transgenic AD model [20]. Given peripherally at high doses, STZ selectively destroys insulin-producing/insulinsecreting cells in the pancreas and causes type 1 diabetes mellitus [21], while given in low to moderate doses, it causes insulin resistance and type 2 diabetes mellitus [22]. Demonstration of brain insulin resistance in this model remains a problem. Brain insulin resistance is defined in literature as poor signaling of insulin receptors, reduced insulin levels in the brain, and/or reduced trafficking of insulin into the brain and other, so far, unknown processes [23]. Can we really consider the level of insulin in the brain as an indicator of insulin resistance and can we acknowledge insulin resistance based solely on changed protein levels and activity of the insulin receptor signaling cascade [24][25][26]? Insulin propagates signaling via two main branches: the PI3K-PDK-1-Akt and the Grb2-SOS-Ras-MAPK pathways. These signaling pathways contain several points of regulation, signal divergence, and cross talks with other signaling cascades. Many steps are negatively regulated by the action of phosphatases or inhibitory proteins [27]. Therefore, due to the complexity in the signaling system, we cannot be sure if changes in levels of phosphorylated and total protein forms involved in insulin signaling can give us a definite answer: "This is insulin resistance." It has been shown that insulin causes an increase of c-fos expression via activation of the ERK isoforms of MAP kinases [28,29], and as an immediate early gene, c-fos has long been known as a molecular marker of neural activity [30]. To explore the possible presence of brain insulin resistance in the STZ-icv model, we measured c-fos levels after insulin administration. Insulin was administrated intranasally (IN), because this route enables the delivery of insulin to the central nervous system with the relative absence of systemic uptake and related peripheral side effects [31].
Recently, intranasally administered insulin has been investigated in clinical trials in individuals with mild cognitive impairment and AD and research demonstrated enhancement of memory performance [32]. Most of the clinical studies did not explore the exact underlying molecular mechanisms involved in the beneficial effects of intranasal insulin on cognition and glucose metabolism [31]. In the brain, insulin contributes to diverse neuronal signaling mechanisms, including ion channel trafficking and the regulation of receptors for neurotransmitters (for example N-methyl-D-aspartate receptor [NMDAR] and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor [AMPAR]), as well as activity-dependent processes of synaptic plasticity (for example, long-term potentiation [LTP] and long-term depression [LTD]) [31]. Insulin has been shown to modulate LTD by inducing glutamatergic AMPAR internalization [33]. Downregulation of AMPAR activity in excitatory synapses of hippocampal CA1 neurons is fundamental for insulin-induced LTD, which is a key step for memory consolidation and flexibility [34]. Moreover, insulin potentiates NMDAR activity by delivery of NMDARs to the cell surface and NMDAR phosphorylation, processes that may trigger long-lasting meta-plastic changes [32]. On the other hand, in the hypothalamus (HPT), the activation of the PI3K-Akt pathway leads to reduced food intake and induces anorexigenic effects, all together resulting in a reduction of body weight [35]. Since insulin affects brain energy homeostasis and its malfunction can lead to numerous disease, including neurodegenerative disorders [12,36], it is of great interest to further advance the knowledge on the precise role of insulin in the brain. Therefore, in addition to investigating insulin resistance in the brain of the rat sAD model by measuring c-fos expression, we wanted to explore the acute effect of intranasal insulin on peripheral plasma levels of glucose, insulin, and amyloid β, as well as on glutamatergic and metabolic parameters in different brain regions in cognitively normal rats in comparison to rats with cognitive deficit (STZ-icv rat model of sAD).

Animals
Adult (3 months old) male Wistar rats weighing 280-350 g (Department of Pharmacology, University of Zagreb School of Medicine) were used in the experiment. Male rats were used because of the sex variation in response to STZ-icv injection; STZ-icv did not induce learning and memory impairment in female rats [37] and majority of our previous research has been done on male rats, so we can use the results for comparison. All animals were housed in cages (2-3 rats per cage) in the licensed animal facility at the department, kept on standardized food pellets and water ad libitum, and maintained under a 12/12 h light/dark cycle.

Ethics
The experiments were carried out in compliance with current institutional (University of Zagreb School of Medicine), national (Animal Protection Act, NN 102/17), and international (Directive 2010/63/EU) guidelines on the use of experimental animals. The experiments were approved by the national regulatory body, the Croatian Ministry of Agriculture (license number EP 186/2018).

Materials
Streptozotocin, monoclonal anti-glial fibrillary acidic protein (GFAP) antibody produced in mouse (clone G-A-5, #G3893, Rodriguez et al., 2008), monoclonal anti-β-actin antibody produced in mouse (clone AC-15, #A5441), paraformaldehyde, Fluoroshield with DAPI, normal goat serum, PhosSTOP phosphatase inhibitor tablets, and protease inhibitor cocktail were purchased from Sigma-Aldrich (St. Louis, MO, USA). The glucose measuring kit (Greiner Diagnostic Glucose GOD-PAP) was acquired from Dijagnostika (Sisak, Croatia). The chemiluminescent western blot detection kit (SuperSignal West Femto Chemiluminescent Substrate) was from Thermo Scientific (Rockford, IL, USA). The ELISA kit for rat/mouse insulin, monoclonal anti-NeuN antibody produced in mouse, polyclonal anti-AMPA receptor antibody produced in rabbit, polyclonal anti-AMPA receptor, phos-phoSer845, produced in rabbit, monoclonal anti-NMDAR1 antibody produced in rabbit, polyclonal anti-NMDAR1 (Ser897) produced in rabbit, monoclonal anti-insulin receptor antibody produced in mouse, and polyclonal antiglucose transporter GLUT-4 antibody produced in rabbit were acquired from Merck Millipore (Billerica, USA). The ELISA kit for glutamate was acquired from Abnova (Taipei City, Taiwan). Polyclonal anti-c-Fos antibody produced in rabbit was purchased from Abcam (Cambridge, UK). TGX FastCast Acrylamide Solution, monoclonal anti-CD11b antibody produced in mouse, was purchased from Bio-Rad (Hercules, CA, USA). Monoclonal anti-glucose transporter GLUT3 antibody produced in mouse was purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Monoclonal anti-phospho-AMPKα (Thr172) produced in rabbit, polyclonal anti-AMPKα produced in rabbit, anti-mouse IgG horseradish peroxidase-linked antibody, anti-rabbit IgG horseradish peroxidase-linked antibody, and anti-mouse and anti-rabbit IgG antibodies linked with Alexa Fluor 488 and 555 were acquired from Cell Signaling (Beverly, MA, USA). Colorimetric and fluorometric measurements were performed on an Infinite F200 PRO multimodal microplate reader (Tecan, Switzerland). EthoVision XT animal video tracking software was purchased from Noldus Information Technology (Wageningen, the Netherlands). Rapid acting insulin aspart NovoLog was purchased from Novo Nordisk (Bagsvaerd, Denmark).

Streptozotocin Treatment
Rats were subjected to general anesthesia (ketamine 70 mg/ kg; 7 mg/kg xylazine), followed by icv injection of STZ (3 mg/kg, dissolved in 0.05 M citrate buffer, pH 4.5, bilaterally 2 µL/ventricle, split in two doses administered on days 1 and 3), according to the procedure first described by Noble et al. [38] and used in our previous experiments [17,18,26,39,40]. Control (CTR) animals were given an equal volume of vehicle (0.05 M citrate buffer, pH 4.5) icv by the same procedure on days 1 and 3 (Suppl 1).

Insulin Treatment
Rats were equally divided in four groups based on their results in the MWM test. Two groups (CTR + INS and STZ + INS) received a single dose of 2 IU of fast-acting insulin aspart (half-life 81 min) intranasally (10 µL/1 IU per nostril), while their respective control groups (CTR and STZ) received saline only (10 µL in every nostril) 1 month after STZ-icv administration. The dose used in this research is a standard dose normally used in animal studies [24,41,42]. All animals were sacrificed 2 h after intranasal administration (time needed for c-fos protein expression induction [43,44]). There were 9 animals per group, and 4 groups in total (Suppl 1).

Morris Water Maze Swimming Test (MWM)
The procedure of MWM consisted of 5-day learning and memory training trials, and a probe trial on day 6, performed 1 month after STZ-icv administration, in a 180-cm-diameter round pool, 60 cm deep, with water temperature set at 25 ± 1 °C. On days 1-5, rats were trained to escape the water by finding a hidden platform (15 cm diameter) submerged about 2 cm below the water surface and placed in the northwest/NW quadrant. Remaining on the platform in order to memorize its location was allowed for 5 s. Four consecutive trials were performed per day, each from a different starting position (southwest/SW, south/S, east/E and northeast/NE), separated by a 30-min rest period. The time needed to find the platform, number of non-target entries (entries into the quadrants other than the NW quadrant in which the platform was located), and rat swimming velocity were recorded during the training trials, which tested the capacity of learning memory. A probe trial which tested memory retention was performed (from quadrant southeast/SE) with the platform removed from the pool, where the time spent in search of the platform within the NW quadrant, rat swimming velocity, and non-target entries were recorded. The cutoff time was 1 min. The data was recorded, tracked, and analyzed using the EthoVision XT video tracking software (Noldus Information Technology).

Tissue Preparation and Blood/CSF Sampling
Sacrification was done under the deep general anesthesia with thiopental and diazepam (60 mg/kg and 6 mg/kg). For CSF sampling, the heads were flexed downwards at an angle of approximately 45°. A small-gauge needle (29 G) was inserted into the cisterna magna by using the occipital bone as a landmark, and CSF was sampled by syringe aspiration until its color changed from transparent to reddish (approximately 200 µL per animal). Blood was sampled from the retro-orbital sinus in tubes with heparin and centrifuged for 10 min at 3000 rpm at 4 °C. The supernatant (plasma) was collected and stored at − 80 °C. Six out of 9 animals per group were decapitated after which brains were quickly removed, and the hippocampus (HPC), hypothalamus (HPT), and temporal cortex (TC) dissected out and frozen in liquid nitrogen, and stored at − 80 °C. Brain tissue samples for western blot analysis were thawed and homogenized with three volumes of lysis buffer containing 10 mM HEPES, 1 mM EDTA, 100 mM KCl, 1% Triton X-100, pH 7.5, protease inhibitor cocktail (1:100), and phosphatase inhibitor tablets and the homogenates were centrifuged at 13,000 rpm for 10 min at 4 °C. The supernatants were frozen and stored at − 80 °C. Protein concentration was measured by Lowry protein assay. The remaining 3 animals per group were perfused with 4% paraformaldehyde, pH 7.4. Brains were quickly removed and cryoprotected with sucrose (series of 15% and 30%) and stored at − 80 °C. Brains were sliced using a cryostat (16 µm), mounted on slides, and dried overnight on 37 °C before the immunofluorescence procedure.

Glucose, Insulin, Glutamate, and Amyloid β1-42 Measurements
Blood/CSF glucose, insulin and Aβ1-42 concentration, and brain tissue insulin and glutamate levels were measured spectrophotometrically in strict compliance with the manufacturer's protocol.

Western Blot Analysis
Equal amounts of total protein in the HPT (25 µg per sample), TC, and HPC (35 µg per sample) were separated by sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE) electrophoresis using 9% polyacrylamide gels or by TGX Stain-Free 9% gels, and transferred to nitrocellulose membranes. The nitrocellulose membranes were blocked for 1 h at RT in 5% non-fat milk, added to a low-salt washing buffer (LSWB) containing 10 mM Tris, 150 mM NaCl, pH 7.5, and 0.5% Tween 20. The blocked blots were incubated with primary anti-c-fos, anti-phospho-NMDAR, anti-total-NMDAR, anti-phospho-AMPAR, anti-total-AMPAR, anti-IR, anti-GLUT3, anti-GLUT4, anti-phospho-AMPK, and anti-total-AMPK antibodies (1:1000) overnight at 4 °C. After incubation, the membranes were washed 3 × with LSWB and incubated for 1 h at RT with secondary antibody solution (anti-rabbit or anti-mouse IgG, 1:2000). After 3 × washing in LSWB, signals were visualized using a chemiluminescence western blotting detection reagent. The signals were captured and visualized with a MicroChemi video camera system (DNR Bio-Imaging Systems). The membranes were washed 3 × with LSWB, blocked the same way, and incubated overnight with loading control β-actin (1:2000) at 4 °C. The membranes were then washed and incubated for 1 h at RT with a secondary antibody solution (anti-mouse IgG, 1:2000), washed in LSWB, and immunostained using a chemiluminescence reagent. The signals were captured and visualized with a MicroChemi video camera system. If TGX stain-free gels were used, gels were visualized after electrophoresis using the ChemiDoc Imaging System (Bio-rad) and transfer efficiency was checked after transfer using the same technology. All the proteins in the sample were used for normalization and analysis. All western blot analysis was done using the ImageJ software and images are provided in Supplement 4.

Immunofluorescence
Previously sliced and dried sections were washed 3 × with PBS-T buffer (0.025% Triton X-100 in PBS), incubated with 200 µL of 10 × normal goat serum for 1 h at RT, and incubated with primary anti-c-fos antibody (1:500, diluted in 1 × normal goat serum) overnight at 4 °C. After incubation, the sections were washed 3 × with PBS-T and incubated for 2 h at RT in the dark with anti-rabbit IgG secondary antibody, Alexa Fluor 555 or 488. After washing with PBS, slides were again incubated with primary anti-CD11, anti-GFAP, or anti-NeuN antibodies (1:500) overnight at 4 °C. After incubation, the sections were washed 3 × with PBS-T and incubated for 2 h at RT in the dark with an anti-mouse IgG secondary antibody, Alexa Fluor 555 or 488. Finally, the sections were washed 3 × with PBS-T and cover slipped with Fluoroshield mounting medium with DAPI, viewed, and analyzed using an Olympus BX51 microscope and CellSense Dimension software. Briefly, brain regions were identified in coronal sections using the rat stereotaxic atlas [45]. Pictures were taken through appropriate filters with DP-70 camera and channels were merged using CellSense Dimension software.

Statistics
Protein levels were presented as vertical scatter plots with mean ± SD and probe trial MWM results were expressed as box plots with between-group differences tested by two-tailed Kruskal-Wallis one-way analysis of variance, followed by the Mann-Whitney U test, with p < 0.05 considered statistically significant, using the GraphPad Prism 5 statistical software. MWM test results for learning trials were expressed as mean ± SD and analyzed by two-way ANOVA for repeated measures with Bonferroni post hoc correction, with p < 0.05 considered statistically significant, using the GraphPad Prism 5 statistical software. Principal component analysis was used for multivariate exploratory data analysis following z-score normalization to rescale each attribute to obtain a mean (µ) of 0 and SD of 1 (z = (x-µ)/ SD). using the Jamovi software (2.2.2). The results were presented as coordinates of individuals with respect to the biplot and vectors of variables indicating contributions to the first two principal components (PC). Monotonic associations were analyzed by calculating Spearman's rank correlation coefficients (ρ) for pairs of measured variables. Correlations were reported as heatmaps visualized in the JASP software (0.15).

STZ-icv Administration Induced Spatial Memory Deficit
One month after icv administration, STZ-induced cognitive deficits ( Fig. 1) were measured using the MWM test. During the learning period, STZ-icv rats needed a longer time to find the hidden platform (+ 35.7-+ 63%) and had a higher number of non-target entries (+ 35.7-+ 88.9%) in comparison to the control animals ( Fig. 1A). Furthermore, during the probe trial, STZ-icv-treated rats spent less time in search of the removed platform in the targeted quadrant (− 42.4%, p = 0.0353), compared to the control rats (Fig. 1B). There were no differences in animal velocity during the learning and probe trials and non-target entries during the probe trial ( Fig. 1).

STZ-icv Administration and Intranasal Insulin Caused an Increase in Plasma Amyloid β1-42 Levels
Streptozotocin and insulin treatment did not alter the concentrations of glucose ( Fig. 2A) and insulin (Fig. 2B) in plasma and CSF. One month after STZ-icv, plasma Aβ concentration was found increased compared to the controls (+ 83.8%, p = 0.007). Two hours after intranasal insulin administration, plasma amyloid-β concentration was found increased in CTR animals (+ 103.2%, p = 0.0401), while there was no change in the STZ group (Fig. 2C). CSF amyloid β1-42 levels were found unaltered by the treatments (Fig. 2C). During the learning trials, through 5 consecutive days, the time needed to find the platform, number of non-target entries (entries into the quadrants other than the NW quadrant in which the platform was located) and rat swimming velocity were recorded (A), with a cutoff time of 1 min. Each dot represents a group value expressed as mean ± SD. Group comparisons were analyzed by two-way ANOVA for repeated measures with Bonferroni post hoc test. Statistically significant changes between CTR and STZ in the same time point were marked. On day 6, the platform was removed from the pool, and time spent in search of the platform within the targeted quadrant, rat swimming velocity, and non-target entries were recorded (B). The cutoff time was 1 min. Data were analyzed by non-parametric Mann-Whitney U test and presented with a box plot with min and max values (*p < 0.05; **p < 0.01; ***p < 0.001 vs CTR) was no difference in HPT and HPC c-fos levels between CTR and STZ animals (Fig. 3A). Intranasal insulin increased c-fos levels in the TC (+ 115.7%, p = 0.0022) and decreased it in the HPC (− 66.8%, p = 0.0043) in STZ-icv rats (Fig. 3A).

C-fos Co-localizes with the Microglial Marker (CD11b) in the HPT and Intranasal Insulin Normalizes Decreased C-fos Levels Found in the TC
The marker for neuronal activity, c-fos, was found co-localized with NeuN (neuronal marker) in the TC (Fig. 3B, Suppl 2). In contrast to TC co-localization, in the HPT, there was some c-fos signal co-localized with the microglial marker CD11b (Fig. 3C, Suppl 2). Selected images are presented in Fig. 3B and C, while the rest of the images are available in Supplement 2.

The Temporal Cortex and Hypothalamus Exhibit Opposite Directions of Changes in Levels of IR and Glucose Transporters After STZ-icv or Intranasal Insulin Treatment
Opposite to the decreased levels of IR (− 52.1%, p = 0.0022) and GLUT3 (− 61.2%, p = 0.0087) found in the HPT (Fig. 4E) Fig. 4F) compared to controls. There was no change in protein levels after intranasal insulin treatment in the STZ group compared to STZ-icv alone, only a slight increment of insulin concentration in the TC (+ 13%, p = 0.0411) and decrement of GLUT4 levels (− 42.7%, p = 0.0043) in the HPC (Fig. 4C, D). AMPK activity was found altered only in the HPT, increased after STZ-icv injection compared to controls, and decreased by insulin treatment only in the STZ but not in the CTR group (Fig. 4E). Despite between-group insignificant variations in AMPK activity (TC) tested by two-tailed Kruskal-Wallis one-way analysis of variance, there was a clear tendency of AMPK activity decrement in CTR rats after insulin treatment (Fig. 4F), which was confirmed with Mann-Whitney U test (p = 0.0152). Because of insignificant between-group analysis, this significance is not presented on the graphs.

Streptozotocin and Insulin Treatment Alter the Activity and the Levels of Proteins Involved in Glutamate Signaling
STZ treatment significantly decreased glutamate concentration in the HPT (− 91.5%, p = 0.0043), compared to the control group (Fig. 5B), while there was no change in the HPC and TC (Fig. 5A and C). Insulin treatment increased AMPAR activation in the HPT (+ 78.9%, p = 0.0087; Fig. 5E) and TC (+ 137.3%, p = 0.0303) and NMDAR activation in the TC (+ 60.4%, p = 0.0260; Fig. 5F) compared to the control, while glutamate concentration was found decreased in the HPT of control animals (− 93.3%, p = 0.0043; Fig. 5B). Two hours after intranasal insulin administration in the STZ group, NMDAR activity and glutamate concentration were found increased in the HPC (+ 228.6%, p = 0.0087; + 44.6%, p = 0.0087 respectively; Fig. 5A and D) and an increased NMDAR activity was found in the TC (+ 143.6%, p = 0.0079; Fig. 5F) compared to the STZ group.
In the HPC, the first two components explain 25.34% and 19.30% of the variance respectively, with c-fos, GLUT3, and NMDAR being the largest contributors to the first PC and insulin and AMPK to the second one ( Fig. 6A and B,  Supplement 3). Clustering of the groups, predominantly STZ and STZ + INS, was most prominent with respect to the 1st PC in the HPC (Fig. 6A). In the HPT, two components explained 25.36% and 24.45% of the variance, with GLUT4, NMDAR, and insulin being the largest contributors to the first PC and IR and c-fos to the second one ( Fig. 7A and B, Supplement 3). Insulin treatment in both the CTR and STZ group showed the most pronounced effect with respect to the 1st PC in the HPT and STZ treatment showed the most pronounced effect with respect to the 2nd PC in the HPT (Fig. 7A). In the TC, the first two components explain 30.45% and 17.66% of the variance with GLUT3 and GLUT4 as the largest contributors to the first PC and IR and glutamate to the second one ( Fig. 8A and B, Supplement 3). Insulin treatment had the most pronounced effect in CTRs with respect to the 1st PC and in the STZ group with respect to the 2nd one compared to CTR and STZ alone (Fig. 8A).

Discussion
Up to date, preclinical data is scarce and cannot provide an answer on how IN insulin really works in the brain. In clinical studies, the only possibility to discern this mechanism lies in the post-mortem analysis of brains in the end stage of sAD. It is easier for us scientists to imagine the brain as one distinct organ, but it is in fact a mosaic of different regions involved in a variety of functions. Therefore, a question arises; does insulin cause changes in all brain regions and how does this change affect the metabolic state of the brain? Exploratory immunofluorescence analysis (Fig. 3) showed a decrease in c-fos expression in cognitively impaired rats ( Fig. 1) in the TC. Confirmation of these results was found with WB analysis of the brain tissue (Fig. 3A). Besides the HPC and HPT, the TC was also chosen for further analysis. Interestingly, the decreased c-fos expression found in our analysis is in concordance with our previous findings of increased AT8 immunoreactivity (tau protein phosphorylated at Ser202/Thr205 sites), appearing first in the TC region 1 month after STZ-icv treatment [17], with no differences in HPC and HPT regions. One month after STZ-icv treatment is still considered to be an acute phase of neurodegenerative changes, where cognitive deficit and other impaired neurodegenerative parameters can still be reversed [46]. Intranasal insulin increased c-fos expression in the TC but only in STZ-icv rats (Fig. 3A), while there was no difference in the control group. Surprisingly, insulin induced decrements of c-fos levels in the HPC of STZ-icv-treated rats, suggesting an impaired metabolic status. Can we say that the brain is not insulin resistant at this post-injection stage? STZ-icv rats are more responsive to insulin in comparison to CTR rats, implying possible metabolic imbalance, while CTR animals have a preserved compensational response to maintain normal neuronal activity even after insulin administration. Insulin in the TC of STZ-icv rats returned c-fos levels to the levels of controls, suggesting that decreased neuronal activity can be overcome by intranasal insulin treatment. Since both regions (HPC and TC) responded to insulin administration, but in a different direction, we can presume that in this stage after STZ-icv application, the cells are sensitive to insulin. It would be of interest to investigate the acute effect of insulin in later stages after STZ-icv, when diverse therapeutic strategies failed to elicit a positive response [39,46]. In contrast to that, most of the tested treatments were initiated in early stages of dysfunction progression in the STZ-icv model and had a positive therapeutic outcome in the form of improvement in cognitive functions [46]. It has been shown that c-fos expression following behavioral training correlates with learning, performance, and cellular A ensembles that are activated following memory retrieval, and correlates with fear memory persistence [47]. It was therefore questioned whether c-fos has a role in memory formation or maintenance of cellular homeostasis. It is more likely that c-fos has a role in cell metabolism maintenance. C-fos expression was found co-localized with the neuronal marker NeuN, but surprisingly, not every c-fos signal in the HPT was found in neurons (Fig. 3, Suppl 2). Therefore, we wanted to check which cells c-fos co-localizes with in the HPT. Interestingly, c-fos was co-expressed with the microglial marker CD11b, especially with activated microglial cells present in STZ-icv rats. As it was already seen in other studies [48,49], c-fos expression is not limited only to the neuronal population of cells. Other research indicated that c-fos expression by glial cells may be under the influence of proliferation, differentiation, growth, inflammation, and other conditions. Many cytokine genes are regulated by a transcription factor complex that includes AP-1, and it was suggested that c-fos could be an accurate sensor of inflammation in the brain and a reliable marker for the progression of some inflammatory diseases [50]. Microglial c-fos is present in immune/inflammatory response, as we have seen in HPT of STZ animals. As was anticipated, 2 h after intranasal insulin is too short time to reverse the present gliosis. To be more precise in interpreting the obtained results, changes in c-fos should always be accompanied by co-localization with specific cell populations.
Glucose and insulin concentrations in plasma and CSF were found unaltered after STZ and/or insulin treatment indicating that intranasal insulin was not absorbed systemically into the bloodstream and that it presumably traveled directly through the olfactory epithelium and/or along the trigeminal nerve branches [32]. If there were any increments of plasma insulin concentration, we should have been able to detect it, because the half-life of insulin aspart is approximately 1.5 h and it has a low binding affinity to plasma proteins [51]. Insulin concentration was found increased in the HPT and TC, but not in the HPC (Fig. 4) after IN insulin administration. While there was almost a fourfold insulin concentration increment in the HPT of control rats, 2 h after IN insulin administration, there was no change in concentration in STZ rats. There are several possible reasons for the changed insulin concentrations in the selected brain region. Insulin captured with the commercial insulin kit could also have been the exogenous insulin aspart applied intranasally, suggesting the entrance of insulin in hypothalamic cells and their intracellular or extracellular degradation by the insulin-degrading enzyme (IDE) in STZ-icv rats, while the insulin in the treated control group remained non-degraded. In addition, insulin treatment induced decrement of glutamate, IR, and GLUT3 levels but only in CTR rats, while there was no effect of insulin in HPT of STZ-icv rats on these parameters. There is also a possibility that exogenous insulin induced local endogenous insulin secretion, but that is highly unlikely. In the TC, insulin rose minimally and similarly in both the CTR + INS and STZ + INS group, while in the HPC, there was no change in insulin concentration.
The most pronounced effect of STZ or insulin treatment was on glutamate concentration in the HPT, the change of which follows a similar pattern of changes as the levels of IR and GLUT3. In control rats, intranasal insulin tremendously decreased levels of glutamate, IR, and GLUT3 in the HPT. In addition, the same pattern has been found in STZ-icv rats compared to controls, while intranasal insulin did not affect the levels of the aforementioned proteins and glutamate in STZ-icv rats, suggesting possible insulin insensitivity in the HPT. An interplay between insulin and glutamate signaling in the brain and its connection to neurodegenerative disease has not yet been clarified. There are numerous studies that show abnormally high concentrations of glutamate in neurodegenerative conditions [52] as well as after brain injury [53,54]. Moreover, glutamate stimulates insulin secretion in pancreatic β cells [55], but there are no data, to our knowledge, on the effect of insulin on glutamate levels in the brain. It was shown that insulin significantly reduces blood glutamate levels after its intraperitoneal administration [56]. It is also possible that the applied insulin was mostly concentrated in the HPT and TC, where the majority of changes have been found. However, the observed changes in the HPC of STZicv rats after IN insulin (decreased c-fos and GLUT4 and increased glutamate and NMDAR), compared to STZ alone, suggest that insulin also affected this brain region (though probably indirectly). The effect of insulin is highly dependent on the investigated brain region and studies that explored Fig. 3 Co-localization of c-fos signal with neuronal, astroglial, and microglial markers in the brain. One month after intracerebroventricular (icv) streptozotocin (STZ) or vehicle (CTR) administration, rats were injected intranasally with 2 IU of fast-acting insulin aspart (CTR + INS and STZ + INS) or saline only (CTR and STZ). Animals were sacrificed 120 min after intranasal administration. The hippocampus (HPC; N per group = 6), hypothalamus (HPT; N per group = 6 except CTR + INS: N = 5), and temporal cortex (TC; N per group = 6) were dissected out of six animals per group, homogenized/ sonicated, and protein concentration was measured. C-fos levels in the HPC, HPT, and TC were measured by western blot analysis. Values are presented as vertical scatter plots with mean + / − SD (A) and data was analyzed by a non-parametric Kruskal-Wallis oneway ANOVA test followed by a Mann-Whitney U test (*p < 0.05; **p < 0.01). Slides with brain Sects. its effect on glutamatergic function are contradictory. Some studies reported that insulin induces LTD and produces AMPAR endocytosis, while other studies stated that insulin elicits LTP and enhances membrane trafficking of glutamate receptor subunits in hippocampal synapses [57]. Our results indicate opposite insulin response: a decreased cell activity in the HPT (decreased glutamate, IR, and GLUT3 levels) and an increment in the TC (increased IR, GLUT3, and GLUT4 levels; AMPAR and NMDAR activity) after IN insulin in control rats; while in the HPC, there was no change in the CTR-treated group, suggesting the aforementioned compensational response to preserve normal neuronal activity. As HPT is a brain region involved in whole body metabolism [58], there is a possibility of negative compensational response to increased amounts of insulin that causes decrement of IR levels in this region, which is opposite to changes found in TC region, suggesting region-dependent role of insulin.
It is now well-established that one of the fundamental functions of astrocytes is the uptake most of synaptically released glutamate, which optimizes neuronal functions and prevents glutamate excitotoxicity [59]. Recently, insulin's action in astrocytes has received special emphasis, given its newly discovered regulatory role in brain glucose uptake, which until recently was assumed to be an insulin-independent process [60]. Besides metabolic homeostasis, in astrocytes, insulin also plays a role in the regulation of cognition and mood [60]. Mechanistically, most of these insulin effects on cognition and mood were attributed to its action on neurons [61,62]; however, the growing knowledge about the activity of insulin in astrocytes resulted in new studies showing that these glial cells mediate an insulin-dependent behavior [60]. Therefore, glial cells could maintain a role in the aforementioned region-dependent divergent response to insulin. AMPK acts as a cellular energy sensor and regulator in both the central nervous system and peripheral organs, and it is well known that hypothalamic AMPK restores the energy balance by promoting feeding behavior to increase sustenance intake, thereby increasing glucose production, and reducing thermogenesis to decrease energy output [63]. The observed changes in AMPK activity were only found in the HPT; STZ treatment increased the activity of AMPK, indicating an imbalance in energy homeostasis (decreased) in this brain region. We cannot conclude whether the HPT is affected directly by STZ or whether the energy decrement is merely a response to changes found throughout the brain. It was pointed out by Iliff et al. [64] that substances injected into the ventricular CSF only minimally enter the brain parenchyma and remain confined to the immediate periventricular, the lateral, but also the third ventricular region (proximity of the HPT). Therefore, STZ could act directly on the HPT. It seems that the concentration of STZ injected centrally is high enough to cause brain dysfunction, but the same amount is not enough to elicit a response in the periphery. Intranasal insulin decreased AMPK activity in the HPT of STZ-icv rats, as was seen after icv insulin treatment in diabetic rats [65]. Alterations of IR, GLUT3, GLUT4, and glutamate levels and activity of AMPAR and NMDAR in the HPT did not follow the decrement of AMPK activity in STZ-icv rats, indicating other possible routes of insulin action on AMPK.
Measurements of soluble amyloid β1-42 in plasma and CSF showed that STZ-icv animals have increased plasma levels of the circulating abnormal amyloid form in control animals. In our previous research, we have found accumulation of Aβ in the brain of STZ-icv animals starting from 3 months after the icv injection [17]. There is a possibility that Aβ clearance to the blood is still working 1 month after STZ-icv administration, and Aβ metabolism differs on disease stage progress. Furthermore, insulin has been shown to increase the concentration of circulating amyloid in plasma, both in CTR and STZ animals. A possible biochemical explanation for this result lies in the fact that proteins in the insulin receptor signaling pathway interact with the BACE enzyme (beta secretase 1) responsible for excretion of APP (amyloid precursor protein), and regulate trafficking and Aβ secretion outside the cell so that the concentration of Aβ decreases within the neuron. After that, soluble Aβ can be cleared by different mechanisms from the brain to plasma: transport to the CSF with subsequent reabsorption into the venous blood, and direct transport across the blood-brain barrier (BBB) into the venous blood [66]; in addition, increase in plasma Aβ was found in patients treated with intranasal insulin [67,68]. There is a possibility that in this time point after intranasal administration, insulin increased the clearance of Aβ from CSF to plasma. Also, increased insulin concentration (after intranasal insulin administration or in chronic hyperinsulinemia in insulin resistance) leads to competition with amyloid beta for the enzyme IDE (insulin Fig. 4 Insulin concentrations and the levels of insulin receptor and glucose transporters in the brain. One month after intracerebroventricular (icv) streptozotocin (STZ) or vehicle (CTR) administration, rats were injected intranasally with 2 IU of fast-acting insulin aspart (CTR + INS and STZ + INS) or saline only (CTR and STZ). Animals were sacrificed 120 min after intranasal administration. The hippocampus (HPC; N per group = 6), hypothalamus (HPT; N per group = 6 except CTR + INS: N = 5), and temporal cortex (TC; N per group = 6) were dissected out, homogenized/sonicated, and protein concentration was measured. Insulin concentration was measured using a commercial rat/mouse insulin ELISA kit (A, B, and C). Insulin receptor (IR) and glucose transporter 3 and 4 levels in the HPC (D), HPT (E), and TC (F) were measured by western blot analysis. AMP-activated protein kinase activity (AMPK) was measured indirectly as a ratio of phosphorylated/total levels of protein in the HPC (D), HPT (E), and TC (F) by western blot analysis. Values are presented as vertical scatter plots with mean + / − SD and data was analyzed by a non-parametric Kruskal-Wallis one-way ANOVA test followed by a Mann-Whitney U test (*p < 0.05; **p < 0.01) ◂ degrading enzyme), which breaks down both insulin and Aβ due to which Aβ levels can increase after intranasal insulin administration [69].
In summary, principal component analysis showed a region-dependent response to IN insulin and STZ-icv treatment (Figs,6,7,8). The most pronounced clustering of groups was seen in the TC, indicating increased IR, GLUT3, and GLUT4 levels and decreased glutamate levels and AMPK activity after IN insulin in control rats. Interestingly, STZ-icv animals did not show the same response to insulin treatment as control rats; insulin did not have any effect on GLUT3, GLUT4, and AMPK levels. It seems that the TC is more susceptible to IN insulin in comparison to the HPC of CTR rats. Interestingly, in both the HPT and TC, there is a strong correlation between GLUT3 and GLUT4 independently of the treatment, while in the HPC, the correlation was insignificant, suggesting the importance of GLUT3 and GLUT4 in these regions. GLUT3 has always been considered an insulin-insensitive transporter, but our study shows that insulin increases GLUT3 in TC. . Spearman's rank correlation of parameters in the temporal cortex is represented as Spearman's rho heatmap (C) with statistically significant correlations indicated with *p < 0.05, **p < 0.001, ***p < 0.001. Dim 1, 1st dimension; Dim 2, 2nd dimension; IR, insulin receptor; GLUT3 and GLUT4, glucose transporters 3 and 4; AMPAR, α-amino-3hydroxy-5-methyl-4-isoxazolepropionic acid receptor; NMDAR, N-methyl-D-aspartate receptor; AMPK, AMP-activated protein kinase Based on our results and other literature data, insulin can regulate neuronal glucose uptake by promoting translocation of GLUT3 to the plasma membrane [70] and by increasing GLUT3 expression [71]. Therefore, the increase of GLUT3 levels by insulin could have an effect on neuronal glucose uptake in response to increased neuronal metabolic demand. STZ can cause a metabolic demand increment of impaired neurons. STZ-impaired mitochondrial function induces oxidative stress, so the reactive increase of GLUT3 and GLUT4 could lead to reparation and restoration of energy supply [72][73][74]. Needless to say, glucose supply is important for challenging memory task performances [73,75]. It seems that the clearance of insulin depends on the brain region and cognitive status of the animal (STZ treated vs untreated), provided insulin spread in the same manner to all the observed regions. It is more likely that insulin spread directly through the olfactory bulb to the entorhinal and olfactory cortex [76] (increased insulin levels in TC) and probably through the trigeminal nerve to the hypothalamus. The TC samples taken included both the entorhinal and perirhinal cortex, components of the medial temporal lobe memory system, which have a role in memory and perception [77]. The entorhinal cortex appears uniquely sensitive to a number of disorders including AD. AD-related neuropathology appears to emerge first in the transentorhinal region [78], which was also seen in STZ-icv rats (increased tau phosphorylation [17]). In addition to its role in memory and cognition, the lateral entorhinal cortex is also a component of the olfactory cortex, receiving input from both the main olfactory bulb and piriform cortex [76]. It is widely accepted that trigeminal sensory information can reach the hypothalamus via multisynaptic pathways through the brainstem, thalamus, and cortex. Recently however, anatomical and electrophysiological studies showed that a substantial number of spinal trigeminal nucleus caudalis neurons send their axons directly to hypothalamic regions [79]. The absence of changes in insulin concentration and IR levels after insulin treatment in HPC suggests indirect insulin-mediated alteration in c-fos, glutamate, AMPAR, and NMDAR levels/ activities (only in STZ-icv rats) possibly through neuronal pathways from other directly insulin-affected brain regions. In conclusion, there seems to be a lot of different factors that contribute to and determine the response to insulin, depending on the brain region, type of cells, and dysfunction of the affected cells. Direction of future research should be on investigation of long-term intranasal insulin therapy as well as on examination of acute IN insulin effect in later stages of the STZ-icv rat model of sAD. Additionally, to better determine acute intranasal response to insulin in different brain regions, more time points should be included in research. For a possible therapeutic approach of IN insulin, it would be of interest to determine the exact role of insulin signaling depending on the brain cell type and region and target specific areas/cells of interest.
Funding This work was funded by the Croatian Science Foundation (IP-2018-01-8938) and co-financed by the Scientific Centre of Excellence for Basic, Clinical and Translational Neuroscience (project "Experimental and clinical research of hypoxic-ischemic damage in perinatal and adult brain"; GA KK01.1.1.01.0007 funded by the European Union through the European Regional Development Fund).

Ethics Approval
The experiments were carried out in compliance with current institutional (University of Zagreb School of Medicine), national (Animal Protection Act, NN 102/17), and international (Directive 2010/63/EU) guidelines on the use of experimental animals. The experiments were approved by the national regulatory body, the Croatian Ministry of Agriculture (license number EP 186/2018).

Competing Interests
The authors declare no competing interests.