Glycogen metabolism is impaired in the brain of male type 2 diabetic Goto‐Kakizaki rats

Diabetes impacts the central nervous system predisposing to cognitive decline. While glucose is the main source of energy fueling the adult brain, brain glycogen is necessary for adequate neuronal function, synaptic plasticity and memory. In this study, we tested the hypothesis that brain glycogen metabolism is impaired in type 2 diabetes (T2D). 13C magnetic resonance spectroscopy (MRS) during [1‐13C]glucose i.v. infusion was employed to detect 13C incorporation into whole‐brain glycogen in male Goto‐Kakizaki (GK) rats, a lean model of T2D, and control Wistar rats. Labeling from [1‐13C]glucose into brain glycogen occurred at a rate of 0.25 ± 0.12 and 0.48 ± 0.22 µmol/g/h in GK and Wistar rats, respectively (p = 0.028), despite similar brain glycogen concentrations. In addition, the appearance of [1‐13C]glucose in the brain was used to evaluate glucose transport and consumption. T2D caused a 31% reduction (p = 0.031) of the apparent maximum transport rate (Tmax) and a tendency for reduced cerebral metabolic rate of glucose (CMRglc; −29%, p = 0.062), indicating impaired glucose utilization in T2D. After MRS in vivo, gas chromatography‐mass spectrometry was employed to measure regional 13C fractional enrichment of glucose and glycogen in the cortex, hippocampus, striatum, and hypothalamus. The diabetes‐induced reduction in glycogen labeling was most prominent in the hippocampus and hypothalamus, which are crucial for memory and energy homeostasis, respectively. These findings were further supported by changes in the phosphorylation rate of glycogen synthase, as analyzed by Western blotting. Altogether, the present results indicate that T2D is associated with impaired brain glycogen metabolism.

time known to be present in astrocytic processes that are adjacent to synapses (Koizumi & Shiraishi, 1970;Phelps, 1972). Recent work verified that glycogen occupies the finest perisynaptic processes of astrocytes (Calì et al., 2016;Oe, Baba, Ashida, Nakamura, & Hirase, 2016). This glycogen location points toward a special role in the metabolic support to synapses. Indeed, glucose cycling through the astrocytic glycogen pool was proposed to rapidly provide the energy for neurotransmitter clearance, which would be necessary upon rapid increases of neuronal activity (Shulman, Hyder, & Rothman, 2001). Indeed, glycogenolysis was found to be necessary for adequate glutamatergic neurotransmission in vivo (Gibbs, Lloyd, Santa, & Hertz, 2007) and in cultured cells (Sickmann, Walls, Schousboe, Bouman, & Waagepetersen, 2009). Accordingly, the brain isoform of glycogen phosphorylase was reported to be present in a protein complex with the most abundant high-affinity glutamate carrier, GLT-1, which is particularly enriched in small astrocytic processes near synapses (Genda et al., 2011). Given that glutamatergic neurotransmission is impaired in T2D (Girault, Sonnay, Gruetter, & Duarte, 2019, and references there in), it is likely that brain glycogen metabolism is also perturbed. Moreover, astrocytic glycogen has been demonstrated to be necessary for synaptic plasticity, learning and memory formation (e.g., Duran, Saez, Gruart, Guinovart, & Delgado-García, 2013;Gibbs, Anderson, & Hertz, 2006;Gibbs et al., 2007;Newman, Korol, & Gold, 2011;Suzuki et al., 2011). Goto-Kakizaki (GK) rats, a non-obese T2D model that was produced by selective breeding of nondiabetic Wistar rats showing poor glucose tolerance (Goto, Kakizaki, & Masaki, 1976), are known to perform poorly in memory and learning tasks Li et al., 2013;Xiang et al., 2015). Recently, we reported that GK rats display synaptic dysfunction in the hippocampus, impaired glutamate-glutamine cycle rate, and a rearrangement of oxidative metabolism pathways that fuel brain function Girault et al., 2019). In the present study, we tested the hypothesis that brain glycogen metabolism is impaired in insulinresistant GK rats. For that, we measured labeling incorporation from [1-13 C]glucose into pools of glucose and glycogen in the brain of GK and control Wistar rats, as well as brain levels of glycogen metabolizing enzymes.

| Animals
All procedures were performed in accordance with the Swiss federal law on animal experimentation, approved by the local committee (EXPANIM-SCAV), and reported following ARRIVE guidelines.
Male GK rats (6-months old, 346 ± 27 g, n = 11) and Wistar rats (4-6 months old, 453 ± 71 g, n = 9) from Charles River (L'Arbresle, France) were housed in pairs on a 12-hr light-dark cycle with lights on at 07:00, room temperature at 21-23°C, humidity at 55%-60%, and with tap water and food (Kliba Nafag 3436, Provimi-Kliba, Kaiseraugst, Switzerland) provided ad libitum. The present study focused on male rats due to major gender differences in the phenotype of the GK model (e.g., Desrois et al., 2004, and references therein), and because of females' differential steroid hormone profiles have an impact on the brain over the course of diabetes progression (Candeias et al., 2016). Based on previously reported variances for glycogen turnover in the rat brain (van Heeswijk, Pilloud, Morgenthaler, & Gruetter, 2012), we estimated the sample size required for a difference between independent means at α = 0.05 and power 1-β = 0.9, using G*Power version 3.1.9.2 (Faul, Erdfelder, Buchner, & Lang, 2009). Magnetic resonance spectroscopy (MRS) experiments were conducted randomly across several weeks (2-3 experiments per week). Researchers were not blind to the rat strain during experiments in vivo. There was blinding to the origin of tissue samples during processing.

Significance
The findings of the present study suggest that glycogen metabolism is impaired in the brain of insulin-resistant rats, which display synaptic dysfunction and memory impairment. It is known that brain glycogen is necessary to maintain proper glutamatergic neurotransmission, and it is important for certain forms of memory. Since insulin resistance has been proposed in other forms of dementia, such as Alzheimer's disease, it is plausible that glycogen metabolism is key in mediating actions of insulin in the brain.

| MRS experiments in vivo
Eight GK rats and six control rats were fasted overnight and anaesthetized with about 1%-2% isoflurane vaporized in 30% O 2 in air.
After surgery, rats were placed in a MR compatible holder, the head was stereotaxically fixed, and the radiofrequency coil was placed on top of the head. Body temperature was continuously monitored with a rectal temperature probe (SA Instruments, Stony Brook, NY, USA) and maintained at 37°C with a warm water circulation system. Heart rate, blood pressure, and breathing rate were also continuously recorded with the SA Instruments monitoring system. Throughout the MRS experiment, arterial pH, pressures of CO 2 (P a CO 2 ) and O 2 (P a O 2 ), and haematocrit were measured with a Cobas b121 system (Roche, Rotkreuz, Switzerland), plasma glucose was measured with a Reflotron Plus system (Roche), and plasma lactate with a GM7 Micro-Stat analyser (Analox Instruments, London, UK). Additional 50-µL plasma samples were collected hourly, and stored at −80°C for measurement of insulin concentration (ELISA kit #10-1250-01 from Mercodia, Uppsala, Sweden), and for MRS analyses in vitro.
Scans in vivo were performed on a 14.1 T/26 cm horizontal bore magnet (Magnex Scientific, Abingdon, UK), equipped with 12-cm gradients (400 mT/m in 120 μs) and interfaced to a Direct Drive console (Agilent Technologies, Palo Alto, CA, USA) using a surface coil with a 13 C linearly polarized loop and 1 H quadrature loops. Fast-spin-echo images were acquired to visualize anatomical landmarks for placement of a volume of interest (VOI) with 420-480 µl in the brain. The static magnetic field was homogenized by FAST(EST)MAP shimming, and localized 1 H MRS using STEAM with echo time of 2.8 ms, repetition time of 4 s, and mixing time of 20 ms was performed to determine baseline concentration of glucose in the brain (Duarte, Lanz, & Gruetter, 2011). 1 H spectra were analyzed using LCModel (Stephen Provencher Inc., Oakville, ON, Canada). Localized 13 C MRS was performed with a modified SIRENE pulse sequence including 3D outer volume suppression, 1D inversion nulling, 1D Image-Selected In vivo Spectroscopy (ISIS) module, and bilevel Waltz-16 nuclear Overhauser effect (NOE) and decoupling (van Heeswijk et al., 2012). Spectra were sequentially recorded with 128 scans and repetition time of 1 s for at least 9 hr. During 13 C MRS, [1-13 C]glucose (1.1 mmol/L) was administered as a bolus with 99.9% enrichment, followed by continuous infusion with a 70%-enriched solution (Duarte et al., 2011). Peak areas of 13 C spectra were quantified by line-fitting in NUTS (Acorn NMR, Livermore, CA USA). Glucose C1 was quantified from each recorded spectrum (128 scans), while glycogen C1 was quantified from the sum of 32 spectra (4,096 scans), resulting in temporal resolutions of 2.1 and 68.3 min, respectively.

| Brain tissue processing
After 13 C MRS, rats were immediately sacrificed with a focused microwave fixation device irradiating the brain for 2.3 s at 4 kW and 2,450 MHz (Gerling Applied Engineering, Modesto, CA, USA). The cerebrum was rapidly taken out of the skull and dissected. The hypothalamus, which was not comprised in the VOI for 13 C MRS, was dissected and stored. The cortex, hippocampus, and striatum were collected from one hemisphere to investigate regional variations of glycogen metabolism. The other hemisphere was stored to allow scaling the whole-brain 13 C enrichment curves measured in vivo to absolute concentration of glucose and glycogen. All samples were immediately stored at −80°C until further processing.
Frozen samples were powdered with a mortar and pestle in liquid N 2 , homogenized in 30 mmol/L HCl with an ultrasonic homogenizer Sonopuls HD 2070 with the microtip MS 73 (Bandelin, Berlin, Germany), and then boiled for 45 min at 90°C. After adjusting the pH to 5 with sodium acetate, samples and standards of glycogen (glycogen from oyster Type II, Sigma-Aldrich) were incubated in the presence or absence of 30 ng/ml amylo-α-1,4-α-1,6-glucosidase (Roche) for 2 hr at 37°C . After enzymatic digestion, pH was adjusted to 7. Samples were then centrifuged at 14,000 g for 15 min, and supernatants were collected for lyophilization.

| MRS in vitro
After processing, extracts from the whole hemisphere were lyophilized and then re-dissolved in 2 H 2 O (99.9% 2 H, Sigma-Aldrich).
Plasma samples were diluted 1:20 in 2 H 2 O. After adding sodium fumarate (1.2 µmol) as internal standard, water-suppressed 1 H MRS was performed as previously detailed (Duarte, Cunha, & Carvalho, 2007) on a DRX-600 spectrometer equipped with a 5-mm cryoprobe (Bruker BioSpin SA, Fallanden, Switzerland). In each sample, [ 12 C]glucose was determined from the 1 H resonance of α-glucose that consists of a doublet at 5.22 ppm, and [1-13 C]glucose was determined from the resonances arising from heteronuclear 1 H-13 C coupling (J HC = 184.5 Hz), which appear as satellites of the central α-glucose H1 doublet at 5.22 ppm.
Concentrations of unlabeled glucose and [1-13 C]glucose were determined in 1 H spectra of non-digested extracts of the whole hemisphere. These concentrations were subtracted from those in spectra of digested extracts to determine the concentrations of unlabeled glycogen and [1-13 C]glycogen in the brain. The concentrations of [1-13 C]glucose and [1-13 C]glycogen served to scale 13 C labeling curves detected in vivo.
Absolute concentrations of [1-13 C]glucose and [1-13 C]glycogen in the whole-brain hemisphere were used to scale 13 C signal areas measured in vivo during the 13 C MRS experiment. Plasma glucose (Glc plasma ) was assumed to be transported into the brain with a reversible Michaelis-Menten kinetics as previously described for the rat brain (Duarte et al., 2011). Therefore, brain glucose (Glc brain ) is defined by: where T max is the apparent maximal transport rate, K t is the apparent Michaelis constant for glucose transport, CMR glc is the cerebral metabolic rate of glucose consumption, and V d is the physical volume for glucose distribution in the brain (0.77 ml/g). Similarly, for [1− 13 C] glucose, transport is defined by As detailed previously (Duarte et al., 2011), Equations 1 and 2 were fitted to the concentration of [1-13 C]glucose in the brain as a function of total glucose and [1-13 C]glucose in plasma, and allowed determining T max , CMR glc , and K t .
For the analysis of glycogen metabolism, we assumed that variations of glycogen concentration during the experiment are negligible Morgenthaler, Koski, Kraftsik, Henry, & Gruetter, 2006), implying that synthesis and degradation fluxes are equal (V glyc ); and that glycolysis does not influence glycogen labeling rate because V glyc «CMR glc (Morgenthaler et al., 2009), and thus labeling of glucose resembles that of brain glucose-6-phosphate, from which glycogen is synthetized. Therefore, the rate of glycogen labeling is given by This equation was fitted to the curve of [1− 13 C]glycogen with [1-13 C]Glc and Glc obtained using Equations 1 and 2. The turnover time of glycogen (τ), i.e., the time required to replace all the glucosyl units, is obtained by dividing brain glycogen content by V glyc .

| Enzymatic determination of glucose and gas chromatography-mass spectrometry (GC-MS) analysis
Glucose concentration in processed extracts from the hypothalamus, cortex, hippocampus, and striatum was determined by the glucose oxidase method coupled to a peroxidase reaction to colorimetrically detect H 2 O 2 formation . Briefly, samples and glucose standards were diluted in 50 mmol/L sodium phosphate buffer (pH 7.4) and incubated in the presence of 2 U/ml glucose oxidase (Sigma-Aldrich), 0.2 U/ml horseradish peroxidase (Sigma-Aldrich) and 100 μmol/L AmplexRed (Invitrogen, Basel, Switzerland) for 30 min in the dark. Resorufin, the product of AmplexRed oxidation, was detected at 560 nm with a PowerWave XS microplate spectrophotometer (BioTek, Luzern, Switzerland). Glycogen concentration was reported in terms of glucosyl equivalents released after amylo-α-1,4-α-1,6-glucosidase digestion, and normalized to the tissue's wet weight.
Processed extracts from the hypothalamus, cortex, hippocampus, and striatum were lyophilized and subsequently derivatized and analyzed by GC-MS to establish 13 C-labeling in glycogen-derived glucose using a method adapted and modified from Jackson, Waterhouse, and Bluck (2007). Glucose butylboronate derivatives were prepared by adding 100 µl of a 10 mg/ml solution of 1-butylboronic (Sigmal-Aldrich) acid in pyridine (Sigma-Aldrich) to a maximum of 50 µg of 12 C-glucose (Sigma-Aldrich) (standards) or the fully dried extracts containing a minimum of 1 µg of glucose. Subsequently, the resulting mixtures were whirl-mixed, spun down, and heated at 80°C for 1 hr. The derivatization was completed by the addition of 10 µl of acetic anhydride (Sigma-Aldrich) to the cooled pyridine-mixture, whirl-mixing, and heating at 80°C for 1 hr. The samples were subsequently dried under nitrogen at 60°C, and the residue taken up in 100 µl of trimethylpentane (Sigma-Aldrich) containing 1% of acetic anhydride for stabilization. The samples were transferred to inserts and run on GC-MS using an Agilent Technologies 7820A chromatograph (Agilent J&W GC column HP-5MS, parts no. 19091S-433) coupled to Agilent Technologies 5977E mass spectrometer. Isotopic enrichment was corrected for natural abundance of 13 C in the standards and natural abundance of 10 B in the derivatization agent, and calculated according to Biemann (1962) and Walls, Bak, Sonnewald, Schousboe, and Waagepetersen (2014).
Total protein content was determined with the bicinchoninic acid assay from Pierce (Thermo Fisher Scientific, Göteborg, Sweden). (1) dGlc brain dt = T max Glc plasma − Glc brain V d Glc plasma + Glc brain V d to confirm that similar protein content was loaded in each gel lane.

| Statistics
Statistical analyses were performed using Prism 8.0.1 (GraphPad, San Diego, California, USA). Results are shown as mean ± SD. Data were assumed to follow normality, linearity, and equality of variance.
Significant differences between GK and Wistar rat groups were considered for p < 0.05 on Student t tests. Two-way ANOVA was applied to compare glycogen and glucose concentrations and labeling originating from multiple brain areas, and the Holm-Sidak's test was used for post hoc multiple comparisons.
Although physiology parameters were maintained within the normal range across the whole experimental protocol under isoflurane anesthesia, there were some differences between the two groups (Table 1). Namely, relative to controls, GK rats displayed significantly lower P a CO 2 (−11%, p = 0.031). Given their insulin resistance, GK rats required slower glucose infusion rate than controls to achieve similar glycaemia (−43%, p = 0.004). Since less [1-13 C]glucose solution is administered, GK rats displayed slightly larger hematocrit than controls (+9%, p = 0.024) without changes in total hemoglobin content.
Moreover, consistent with the slower glucose disposal rate, GK rats had lower circulating lactate levels than controls (−61%, p < 0.001).
Insulin concentration in circulation increased upon glucose administration, and was similar in GK and Wistar rats.
In Wistar rats, the fractional enrichment (FE) of plasma glucose was stable throughout the infusion protocol, reaching 0.61 ± 0.05 at the end of the 13 C experiment (Figure 1b) At the end of the glucose infusion protocol, whole-brain glucose concentration (determined in tissue extracts), was 4.4 ± 0.8 µmol/g in GK rats and 4.2 ± 1.6 µmol/g in Wistar rats (p = 0.794). Wholebrain glycogen concentration was also similar in the two groups: 5.0 ± 1.1 µmol/g in GK rats and 5.5 ± 2.3 µmol/g in Wistar rats (p = 0.681). Like for labeling of plasma and brain glucose, brain glycogen FE was lower in GK than Wistar rats (0.14 ± 0.02 vs. 0.24 ± 0.07, The labeling of brain [1-13 C]glucose and [1-13 C]glycogen was detected in real time by 13 C MRS (Figure 2). Glucose 13 C signals from α and β anomeres were quantified in each spectrum (128 scans), that is at a temporal resolution of about 2 min. In contrast, given the slow turnover of glycogen and its naturally wider linewidth, 32 spectra were summed for accurate quantification of its 13 C signal area at 100.5 ppm. Figure 3 shows typical curves of glucose concentration and 13 C labeling in plasma, as well as brain [1-13 C]glucose, in one GK and one Wistar rat during the 13 C MRS experiment. Glucose transport and consumption rates were estimated from these curves. T max was 1.14 ± 0.22 and 0.78 ± 0.30 µmol/g/min (p = 0.031), K t was 0.9 ± 1.0 and 1.4 ± 2.1 mM (p = 0.625), and CMR glc was 0.65 ± 0.13 and 0.46 ± 0.21 µmol/g/min (p = 0.062) in Wistar and GK rats, respectively. Interestingly, this resulted in similar ratio of T max to CMR glc : 1.8 ± 0.4 in controls and 1.8 ± 0.8 in GK rats (p = 0.763), suggesting that the relation of brain to plasma glucose levels is not changed in diabetic GK rats.
Labeling of brain glycogen from [1-13 C]glucose was slower in GK than Wistar rats (Figure 4). The rate of glycogen synthesis (V glyc ), which was two orders of magnitude slower than CMR glc , was 0.004 ± 0.002 and 0.008 ± 0.004 µmol/g/min in GK and Wistar rats, respectively (p = 0.028). In µmol/g/h, V glyc was 0.25 ± 0.12 in GK rats and 0.48 ± 0.22 in controls. Taking in account brain glycogen concentration, glycogen turnover time τ was 26.6 ± 13.6 hr in GK rats and 14.4 ± 4.2 hr in Wistar rats (p = 0.041).

| Regional glycogen labeling
We further investigated the regional distribution of glycogen in the brain (Figure 5a). Diabetes affected brain glycogen concentration (F 1,48 = 13.4, p = 0.001) and there was also a significant effect of brain region (F 3,48 = 7.1, p < 0.001), but no interaction (F 3,48 = 2.5, p = 0.062). Posttests revealed reduced glycogen in the striatum of GK rats compared to controls (−44%, p = 0.001). Diabetes also affected brain glucose concentration (F 1,48 = 11.8, p = 0.001), and there was a brain region effect (F 3,48 = 3.5, p = 0.020; interaction F 3,48 = 1.7, p = 0.178). Upon post hoc testing, the reduction of F I G U R E 2 Typical 13 C spectra acquired in vivo in the rat brain upon infusion of [1-13 C]glucose. Each spectrum is the accumulation of 4,096 scans (68 min). Lorentzian line broadening of 10 Hz was applied prior to Fourier transformation. The inset rat brain image shows the typical location of the VOI used for 13 C MRS glucose concentration was significant in the cortex (−37% vs. Wistar, p = 0.047) and striatum (−37% vs. Wistar, p = 0.010).
Whole-brain glycogen turnover was reduced in the brain of GK rats, compared to controls, which resulted in reduced brain glycogen enrichment at the end of the [1-13 C]glucose infusion.
In turn, when investigating 13 C labeling in extracts from specific brain areas by GC-MS, we observed significant effects of diabetes (F 1,48 = 50.4, p < 0.001) and brain region (F 3,48 = 17.0, p < 0.001), but no interaction (F 3,48 < 0.1, p = 0.979). In particular, there was a reduction of glycogen 13 C enrichment of 33% in the hypothalamus (p = 0.002), 29% in the hippocampus (p = 0.002), 25% in the striatum (p = 0.002), and 24% in the cortex (p = 0.002) of GK rats, relative to Wistar rats (Figure 5b). Since 13 C labeling of glucose in these brain areas was about 21%-22% lower in GK than controls (p < 0.001 for all posttests), we infer that the diabetes-induced reduction of glycogen labeling was most prominent in the hypothalamus and hippocampus.

| GLUT1 and glycogen metabolizing enzymes
Since glucose uptake from the blood stream is primarily mediated by GLUT1, the density of the glucose carrier GLUT 1 was analyzed by Western blot (Figure 6a), which detected the two glycosylation forms that are specifically present in endothelial cells (55 kDa) and brain parenchyma (45 kDa) (Simpson, Carruthers, & Vannucci, 2007).
The immunoreactivity of GLUT1 in the striatum and cortex was similar between GK and Wistar rats. In the hypothalamus, there was a Glycogen concentration is directly dependent on the activities of glycogen synthase and glycogen phosphorylase. The density of both muscle and brain isoforms of glycogen phosphorylase was similar in GK rats and controls (Figure 6b,c). Glycogen synthase levels were affected by diabetes. Namely, we observed a 21% reduction in the density of glycogen synthase in the hippocampus of GK rats, relative to controls (p = 0.015; Figure 6d). Moreover, we measured the density of phospho-glycogen synthase at the phosphorylation sites Ser460 and Ser641. The level of phospho-glycogen synthase in the cortex was significantly higher in GK than Wistar rats (+57%, p = 0.027, Figure 6e). The rate of glycogen synthase phosphorylation was higher in the cortex (+47%, p = 0.042) and hippocampus (+69%, p = 0.007), but lower in the striatum (−38%, p = 0.042) of GK rats, when compared to controls.
F I G U R E 3 Typical fit of brain [1-13 C]glucose time courses ([1-13 C]Glc brain ) as function of plasma glucose (Glc plasma ) concentration (a) and 13 C labeled Glc plasma (b). In the particular cases of this figure, best fit in panel c resulted in T max was 1.37 and 0.92 µmol/g/min and CMR glc was 0.77 and 0.65 µmol/g/min for Wistar and GK rats, respectively

| D ISCUSS I ON
The present results confirm previous findings that brain glucose utilization is reduced in the brain of diabetic GK rats, relative to controls (Girault et al., 2019), and provide evidence for impaired brain glycogen metabolism that might contribute to the learning and memory deficits in this diabetes model Li et al., 2013).

| Brain glucose utilization
The brain requires continuous glucose supply to sustain metabolic demands of neuronal activity. By analyzing the labeling of brain [1-13 C]glucose over time, we determined a 31% reduction in T max , the apparent maximum glucose transport rate, which is in accordance with reduced density of glucose carriers at the blood-brain-barrier (BBB). We further observed a tendency for reduced CMR glc (−29%) in GK rats relative to controls, which suggests a global glucose hypo-metabolism. A reduction in both CMR glc and T max resulted in identical ratio of T max to CMR glc , indicating that brain glucose levels are similar in GK and Wistar rats at any given glycemia Girault et al., 2019). Nevertheless, since GK rats display sustained hyperglycemia, their brains are exposed to excessive glucose levels when compared to normoglycemic rats, which may trigger a plethora of deleterious effects leading to neurodegeneration, namely through oxidative stress and non-enzymatic glycation of macromolecules (Duarte, 2015).
In GK rats, GLUT1 density was found reduced in total protein extracts of the hippocampus and hypothalamus, regions crucial for learning and memory and for regulation of glucose homeostasis, respectively. Interestingly, brain glucose content was particularly reduced in the cortex and striatum of GK rats compared to controls, while GLUT1 density was not modified by T2D in these brain areas. This clearly indicates that although total levels of GLUT1 may represent the availability of carriers to mediate glucose transport, they are not indicative of BBB transport capacity.
Accordingly, cerebral glucose utilization was found reduced in the brain of db/db mice, a model of T2D, relative to controls (Garris, Williams, Coleman, & Morgan, 1984;Vannucci, Gibbs, & Simpson, 1997). However, Vannucci et al. reported no changes in the density of the glucose carriers GLUT1, widely expressed in the brain and the main glucose carrier at the BBB, and GLUT3, mainly located in neuronal membranes, in brain preparations obtained from db/db mice, when compared to wild-type mice (Vannucci et al., 1997).
In contrast, a recent study reported lower levels of both GLUT1 and GLUT3 in whole-brain homogenates prepared from mice exposed to a diet rich in fat and sugar for 3 months, which develop an insulin-resistant phenotype, relative to mice fed a regular chow (Kothari et al., 2017). Altogether, these observations lead us to speculate that reduced glucose transport and utilization is caused by changes in metabolic regulation according to cellular needs and adaptations to a diabetic and neurodegenerative state rather than reduced glucose carrier concentrations. Nevertheless, deficiency of GLUT1 results in reduced brain glucose uptake, and was suggested to contribute to cerebral microvascular degeneration, impaired blood flow, reduced amyloid β (Aβ) peptide clearance, diminished neuronal activity, neurodegeneration, and behavioral deficits in an AD mouse model that overexpresses the Aβ precursor protein (Winkler et al., 2015). In line with this, depression of brain glucose metabolism is an early feature of AD (Duarte, 2015;Zilberter & Zilberter, 2017), demonstrated for example by reduced density of GLUT1 and GLUT3 in the hippocampus, caudate nucleus, and neocortex (Simpson, Chundu, Davies-Hill, Honer, & Davies, 1994), reduced binding of the non-competitive inhibitor of glucose carriers [ 3 H]cytochalasin B to neocortical and hippocampal tissue and microvessels (Kalaria & Harik, 1989), or reduced cortical glucose transport and CMR glc (Jagust et al., 1991).
F I G U R E 5 Regional distribution of glucose and glycogen concentrations (a) and 13 C fractional enrichment (b) in the brain of GK (n = 8) and Wistar (n = 6) rats. Results are shown in Tukey plots with mean value indicated by + and outliers by symbols (open circles for Wistar and filled triangles for GK rats). *p < 0.05, **p < 0.01, ***p < 0.001 for GK versus Wistar in Holm-Sidak's posttests after significant 2-way ANOVA

(a) (b)
Insulin was suggested to modulate astrocytic glucose metabolism by acting in concert with the insulin-like growth factor type 1 (IGF-1), namely via a mitogen-activated protein kinase/protein kinase D pathway that controls maintenance of GLUT1 at the cell membrane (Fernandez et al., 2017). While insulin is the main regulator of peripheral glucose homeostasis, it is traditionally considered to not directly modulate global brain glucose uptake and utilization.
Therefore, the extent of this insulin-IGF-1 interaction within glucose transport regulation remains unclear, although it might be of importance in neurodegenerative disorders that involve insulin resistance.
For example, while hyperinsulinemia within the normal physiologic range does not impact glucose transport or CMR glc (Hasselbalch et al., 1999), insulin was shown to stimulate brain glucose metabolism in subjects with impaired glucose tolerance (Hirvonen et al., 2011).
Moreover, insulin has a major role in multiple brain processes that require energy and thus insulin can indirectly modulate brain glucose homeostasis (reviewed in Duarte, 2015).

| Brain glycogen metabolism
Assuming that the whole glycogen pool is being metabolized at a rate of 0.48 µmol/g/h, we estimated that about 14 hr would be necessary for the brain glycogen pool to turnover in Wistar rats. Previous 13 C-MRS studies in rats under light α-chloralose anesthesia, have estimated similar (Choi, Tkác, Ugurbil, & Gruetter, 1999) or slightly faster (0.7 µmol/g/h, van Heeswijk, Morgenthaler, Xin, & Gruetter, 2010) glycogen turnover rates. The discrepancy on the measured turnover rates in our study and these previous publications could be caused by either the deepness of anesthesia that modulates brain activity and metabolism (Sonnay, Gruetter, & Duarte, 2017), or the brain F I G U R E 6 Western blot analysis of GLUT1 detected the highly glycosylated (top band, 55 kDa) and less glycosylated (lower band, 45 kDa) isoforms (a), which are present in endothelial cells and brain parenchyma, respectively. The relation between the normalized immunoreactivity of both GLUT1 forms is plotted together with the line of equality. Immunoreactivity of brain glycogen phosphorylase (b), muscle glycogen phosphorylase (c), and glycogen synthase (d) was detected at 97, 94, and 84 kDa, respectively. Panel e shows phosphoglycogen synthase immunoreactivity (84 kDa) and the ratio of phospho-to total glycogen synthase (GS). SDS-PAGE gels were loaded with 40 µg of protein for analysis of GLUT1, and with 100 µg for the remaining proteins. Data points are shown in symbols (open circles for Wistar, and filled triangles for GK rats), and bars indicate the mean value of n = 2-3, as depicted in the respective immunoreactivity pictures. The cortex from one GK rat was lost during sample preparation, thus only n = 2 was available for SDS-PAGE. *p < 0.05 from Student t test comparisons between GK and Wistar (a) levels of insulin that might stimulate glycogen synthesis (Fernandez et al., 2017;Muhič et al., 2015). Accordingly, administration of insulin to anaesthetized rats was previously reported to increase brain glycogen levels (Morgenthaler et al., 2006). In fact, circulating levels of insulin in the present study under isoflurane were much lower than those in similar experimental conditions but under α-chloralose anesthesia (Girault et al., 2019).
The 13 C incorporation from glucose into glycogen was conspicuously hampered in GK rats, relative to non-diabetic Wistar rats under identical experimental conditions. In the whole brain, glycogen synthesis rate was reduced by 48% in GK rats, relative to controls. This was further supported by the fact that the phosphorylation rate of glycogen synthase was increased in the cortex and hippocampus of GK rats, relative to controls, in line with reduced activity. However, whole-brain glycogen content was found similar. Therefore, we estimated brain glycogen turnover time to be increased by 85% in diabetic rats.
While total brain glycogen content in GK rats at the end of the hyperglycemic clamp was similar to that of Wistar rats, GK rats had lower glycogen concentration in the striatum than controls.
Another rat model of T2D, the Zucker diabetic fatty rat, displayed lower brain glycogen content than lean littermates, but higher than control Sprague-Dawley rats (Sickmann et al., 2010). Then, we found that glycogen labeling from glucose was reduced in the hippocampus and the hypothalamus. Furthermore, higher phosphorylation rate of glycogen synthase in GK rats than controls is in line with reduced brain glycogen synthase activity, namely in the cortex and hippocampus. The antibody anti-phospho-glycogen synthase used in this study targets two serine phosphorylation sites, the Ser640 and Ser641, in liver and brain isoforms (according to supplier). This site is phosphorylated by the serine/threonine protein kinase glycogen synthase kinase 3 (GSK-3), as well as Ser644 and Ser648, resulting in a potent inactivating effect (Roach, 1990;Skurat & Roach, 1995).
The hippocampus of GK rats showed prominently reduced levels of glycogen synthase. Notably, mice lacking glycogen synthase in the brain display impaired acquisition of an associative learning task, as well as impaired hippocampal long-term potentiation (Duran et al., 2013). In contrast to glycogen synthase, we found no changes in levels of glycogen phosphorylase muscle and brain isoforms. The liver isoform was not investigated due to its vestigial presence in the adult brain (Pfeiffer-Guglielmi, Bröer, Bröer, & Hamprecht, 2000).
Although it would be of interest to determine the phosphorylation rate of glycogen phosphorylase in the present study, to our knowledge there are no good antibodies against its phosphorylation sites.
Nevertheless, phosphorylation is not the main regulatory mechanism of the enzyme activity. Glycogen phosphorylase is regulated allosterically by AMP and activated by phosphorylation by the phosphorylase kinase, and the brain isoform in astrocytes responds more strongly to AMP than does the muscle isoform (Bak et al., 2018;Müller, Pedersen, Walls, Waagepetersen, & Bak, 2015). In sum, total protein level alterations (or their absence) may not directly translate into equivalent changes of enzyme activity.
Altogether, our findings suggest hampered dynamics of the brain glycogen pool, at least in the diabetic GK rat model. This has likely an impact on brain activity that depends on glycogen as an energy source, such as memory formation (Gibbs et al., 2007;Suzuki et al., 2011). Impaired learning and memory capabilities in GK rats have been associated with the degree of insulin resistance (Li et al., 2013), and it was also proposed that hippocampal insulin resistance impairs spatial learning and synaptic plasticity (Grillo et al., 2015) and is associated to astrogliosis and neuroinflammation (Calvo-Ochoa, Hernández-Ortega, Ferrera, Morimoto, & Arias, 2014).
Thus, we propose that hampered glycogen utilization in the diabetic brain may provide a biochemical link between cognitive impairment and insulin resistance. In line with this, noradrenergic dysfunction has been proposed to be an important component of AD (Hertz, 1989;Marien, Colpaert, & Rosenquist, 2004;Weinshenker, 2008 were suggested to modify brain noradrenaline metabolism and levels (Kwok & Juorio, 1988;Lacković, Salković, Kuci, & Relja, 1990;Trulson & Himmel, 1985), but the role of the noradrenergic system on the control of glycogen metabolism in the diabetic brain remains to be investigated.

| Limitations
In accordance with the insulin resistance state, when GK rats were infused a 13 C-enriched glucose solution, a strong dilution of its FE was observed in the plasma, when comparing with non-diabetic Wistar rats. This finding is in agreement with increased endogenous glucose production from gluconeogenesis in GK relative to Wistar rats (Sena, Barosa, Nunesa, Seiça, & Jones, 2007), which accounts for the unlabeled glucose in the blood stream. Another difference between the experimental groups is the younger age of the Wistar rats (4-6 months old) relative to GK rats (all 6 months old). This difference was because age-matched controls are often too large to fit in the fixation system used for MRS in this study. Nevertheless, energy metabolism differences are not expected to occur in the rat brain within this age range (Dienel, 2012).
Lower isoflurane levels were needed for anesthesia of GK rats, compared to controls, and the doses of isoflurane within this range show sigmoidal responses in animal immobility tests (Saab et al., 2010). In line with this, metabolic assessments under different anesthesia doses (29% less in GK than Wistar rats) may also bias our metabolic assessments. Any anesthesia reduces glucose uptake and consumption relative to the awake state, and we have shown that T max and CMRglc are similar under isoflurane and light α-chloralose anesthesia (Duarte & Gruetter, 2012). Therefore, the lighter anesthesia in GK than Wistar rats might result in higher metabolic activity in GK than Wistar rats, and thus lead to an underestimation of T2D-induced effects on brain glycogen metabolism. On the contrary, it is known that isoflurane inhibits the mitochondrial electron transport chain in a dose-dependent manner (Hanley, Ray, Brandt, & Daut, 2002;Zimin et al., 2018). This results in increased lactate in the brain, and also in circulation. Indeed, circulating lactate was similar at baseline, but lower in GK than Wistar rats during glucose infusion (Table 1). Accordingly, isoflurane anesthesia could eventually exacerbate glucose utilization at glycolytic level to fulfil eventual needs of ATP. Being possible that these isoflurane doses have effects on some assessed parameters but not others, future studies on GK rats should consider a dose-response curve for isoflurane and the determination of minimum alveolar concentration value in these strains.
With 13 C MRS in vivo, we measured brain glycogen turnover using [1-13 C]glucose as a precursor, which was infused for at least 8 hr. Our measurements assume that glycogen content doesn't change during the experiments and that, therefore, the appearance of [1-13 C]glycogen reflects turnover, i.e., replacement of glucosyl units in glycogen without net synthesis. Although glycogen in the brain of awake rats does not change with hyperglycemia , glycogen accumulation occurs upon glucose infusion under α-chloralose anesthesia, likely mediated by insulin (Morgenthaler et al., 2006). Glucose-stimulated insulin secretion in rats under isoflurane anesthesia resulted in circulating insulin concentration ( Table 1) that is one order of magnitude smaller than that under α-chloralose anesthesia (Girault et al., 2019). One might thus speculate that brain glycogen accumulation is much smaller than that reported previously (Morgenthaler et al., 2006). Moreover, circulating insulin was similar between GK rats and controls, as well as brain glycogen levels after the hyperglycemic clamp of our experimental protocol. Therefore, the conclusion that glycogen metabolism is slower in GK than Wistar rats remains valid, despite the possibility of net glycogen synthesis to occur.
It should also be noted the small sample sizes in this study, namely n = 2-3 for protein determination by Western blot, and n = 6-8 for 13 C tracing experiments. Despite the limited number of rats used, altogether the experiments of our study point in the direction of T2D-associated alterations of glycogen metabolism.
Metabolism of glucose and glycogen is linked to glutamatergic neurotransmission (Gibbs et al., 2007;Sickmann et al., 2009;Sonnay et al., 2017). However, the present results are not sufficient to determine whether brain hypo-metabolism and reduced glycogen turnover are cause or consequence of the previously reported reduced glutamate-glutamine cycle rate in the brain of GK rats (Girault et al., 2019).

D ECL A R ATI O N O F TR A N S PA R EN C Y
The authors, reviewers and editors affirm that in accordance to the policies set by the Journal of Neuroscience Research, this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present.

ACK N OWLED G M ENTS
The authors are grateful to Jordi Duran (Institute for Research in Biomedicine, Barcelona) for advice on Western blotting experiments for glycogen-metabolizing enzymes.

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

DATA AVA I L A B I L I T Y
The data that support the findings of this study are available from the corresponding author upon reasonable request.