Chronic insulinopenia/hyperglycemia decrease cannabinoid CB₁ receptor density and impair glucose uptake in the mouse forebrain

Abstract

Both endocannabinoids and insulin regulate peripheral and cerebral glucose homeostasis via convergent signaling pathways that are impacted by diabetes. Here we asked how glucose metabolism and important facets of insulin signaling are affected in the forebrain of cannabinoid CB₁ receptor knockout mice (CB₁R-KO) and their wild-type (WT) littermates, seven weeks after the induction of insulinopenia/hyperglycemia (diabetes) with intraperitoneal streptozotocin injection. Sham-injected animals served as control. Diabetes caused milder weight loss in the WT mice compared to the phenotypically −11% leaner CB₁R-KO, while hyperglycemia was similar. Resting [³H]deoxyglucose uptake was significantly reduced by −20% in acute ex vivo frontocortical and hippocampal slices obtained from both the sham-injected CB₁R-KO and the diabetic WT mice. Surprisingly, the third cohort, the diabetic CB₁R-KO showed no further impairment in glucose uptake, as compared to the sham-injected CB₁R-KO. Depolarization-induced [³H]deoxyglucose uptake was proportional to the respective resting values only in the cortex in all four cohorts. The dissipative metabolism of [¹⁴C]-U-glucose remained largely unaffected in all cohorts of animals. However, diabetes reduced cortical CB₁R density by −20%, as assessed by Western blotting. Albeit the changes in insulin signaling did not reflect the glucose uptake profile in each cohort, there were significant interactions between diabetes and genotype. In conclusion, a chronic decrease or lack of CB₁R expression reduces glucose uptake in the mouse brain. Additionally, diabetes failed to cause further impairment in cerebral glucose uptake in the CB₁R-KO. These suggest that diabetic encephalopathy may be in part associated with lower CB₁R expression.

1. Introduction

Almost a century ago, it was discovered that diabetes was not only detrimental to peripheral nerves but could also cause cerebral neuropathy (Miles and Root 1922), i.e. diabetic encephalopathy (Reske-Nielsen et al. 1965). Diabetic encephalopathy is accompanied by attention and cognition deficits, resulting from complex neurophysiological and microanatomical alterations (Miles and Root 1922; Brands, 2004; Duarte 2015). Impaired brain glucoregulation could be one of them (de Ceballos and Köfalvi 2017).

Glucose is the primary energy source for the brain (Clarke and Sokoloff 1999; McCall 2004), but chronic hyperglycemia overactivates the sorbitol pathway, generates toxic glycation, oxidative stress and microvasculature damage with consequent neuroinflammation (Brands et al. 2004; McCall 2004; Duarte 2015). However, the proinsulin C-peptide also possesses important neuroprotecive roles (Sima et al. 2008).

Moreover, hyperglycemia augments phospholipase C (PLC) activity, resulting in increased cellular diacylglycerol content (Ramana et al. 2005). Importantly, PLCβ is a key enzyme in endocannabinoid signaling while diacylglycerol is the direct precursor of 2-arachidonoyl-glycerol (2-AG), a major endocannabinoid molecule (Harkany et al. 2008; Solymosi and Köfalvi 2017). Therefore, uncontrolled hyperglycemia may impair endocannabinoid signaling in nervous system tissues.

The latter may have further pathological consequences, since endocannabinoids regulate communication, proliferation, growth, migration and metabolism of brain cells, among other roles (Di Marzo et al. 2011; Solymosi and Köfalvi 2017). Since endocannabinoid production is associated with neural metabolic activity, it may also be linked with the modulation of glucose metabolism in these cells. Indeed, the activation of the cannabinoid CB₂ receptor increases astrocytic and neuronal glucose uptake with relevance to Alzheimer’s disease (Köfalvi et al. 2016) while the activation of cannabinoid CB₁ receptors (CB₁Rs) reduces mitochondrial oxidative glucose metabolism (Bénard et al. 2012; Duarte et al. 2012).

CB₁R can form functional heterodimers with the insulin and the insulin-like growth factor (IGF-1) receptors in neuronal cell culture (Dalton et al. 2012) and pancreatic β-cells (Kim et al. 2012). In addition, CB₁R overactivation in adipocytes and liver causes peripheral insulin resistance (Jourdan et al. 2017; Sidibeh et al. 2017), and similarly in the brain, CB₁Rs regulate insulin receptor-induced [³H]deoxyglucose uptake in the nucleus accumbens of the rat (Pinheiro et al. 2016). As a possible feed-back mechanism, insulin has been shown to induce 2-AG release and indirectly, activate CB₁R in the ventral tegmental area (Labouèbe et al. 2013). These findings therefore suggest that chronic changes in cannabinoid or insulin receptor activity may perturb cerebral glucose homeostasis, especially, if the two conditions co-occur. Hence, they may contribute to diabetic encephalopathy and the detrimental hypoglycemia unawareness in type-1 diabetic patients (Cranston et al. 2001; Duarte 2015).

We now asked how chronic insulinopenic diabetes affects forebrain glucoregulation in CB₁R knockout (KO) mice, as compared to their wild-type (WT) littermates. Glycemia and body weight changes were also compared. With the help of previously validated techniques we argue for an association between CB₁R expression and diabetes-induced metabolic distress.

Materials and methods

Animals

All studies were conducted in accordance with the principles and procedures outlined as “3Rs” in the guidelines of EU (86/609/EEC), FELASA, and the National Centre for the 3Rs (the ARRIVE; Kilkenny et al. 2010), and were approved by the Animal Care Committee of the Center for Neuroscience and Cell Biology of the University of Coimbra, Portugal. We also applied the ARRIVE guideline for the design and execution of in vitro experiments (see below), as well as for data management and interpretation (McGrath et al. 2010).

Sixteen CB₁ receptor null-mutant (knockout, CB₁R-KO) male mice with CD-I background (Ledent et al. 1999) and 16 wild-type littermates (10-week old) were randomly stratified in 2 × 2 cohorts of 8 mice. Mice were raised in the laboratory of Dr. Catherine Ledent (IRIBHM, Université Libre de Bruxelles, Brussels B-1070), where they were genotyped before shipping. The mice arrived with 6 weeks of age and were kept at the conventional animal facility until use. The mice were housed with 12 h light on/off cycles under controlled temperature (23 ± 2 °C), and ad libitum access to food and water. All efforts were made to minimize the number of animals used and to minimize their stress and discomfort.

Materials

The antibodies against phospho-(p)-(Ser21) GSK3α, total and p(Ser9)-GSK3β, total and p(Ser473)-Akt, were purchased from Cell Signaling Technologies (Danvers, MA, USA). The antibodies against IRβ, IGF-1R, GLUT1 and GLUT4 were purchased from Santa Cruz Biotechnology (Santa Cruz, California, USA). The antibody against total GSK3α was obtained from UpState (Lake Placid, NY) and the antibody against β-actin was bought from BioLegend (San Diego, CA). The guinea-pig anti-CB₁R antibody was obtained from Frontier Institute co., ltd. (Hokkaido, Japan). The chemifluorescence (ECF) reagent was acquired from GE Healthcare (Chalfont St. Giles, UK).

All other unspecified (in)organic reagents were from Merck (formerly, Calbiochem, Sigma-Aldrich, Merck-Millipore Corporation; Darmstadt, Germany).

Diabetes induction

Chronic insulinopenia (aka. unmedicated type-1 diabetes)/hyperglycemia was induced as described previously (Moura et al. 2014), following the protocol of Rossini et al. (1977). In brief, one cohort of 8 CB₁R-KO mice and one cohort of 8 wild-type littermates (WT) were intraperitoneally injected with streptozotocin (50 mg/kg body weight) dissolved in citrate buffer (pH 4.5), during 5 consecutive days. The remaining cohorts of 8 WTs and 8 CB₁R-KOs received only citrate buffer injection during 5 days. Four days after the last injection, blood glucose levels were measured by the Accu-Chek Aviva glucometer (Roche Diagnostics, Germany). Mice with blood glucose levels higher than 300 mg/dL were considered diabetic. All streptozotocin-injected mice became diabetic and none perished during the 7-week post-injection period. Blood glucose and body weight levels were regularly recorded in the diabetic cohorts (see Fig. 1). Seven weeks after the first injection, mice were deeply anesthetized (no reaction to tail pinch and handling while still breathing) with 2-bromo-2-chloro-1,1,1-trifluoroethane (halothane; 5%, 1 L/min flow rate) before decapitation with a scissor.

Fig. 1.

The effect of diabetes on body weight and plasma glucose levels in both genotypes. (A) In spite of being ~11% lighter than their wild-type (WT) littermates from the beginning, cannabinoid CB₁ receptor global knockout (CB₁R KO) mice lost significantly more body weight in both absolute and relative terms in the first month of hyperglycemia/insulinopenia, aka. STZ-induced type-1 diabetes. (B) Albeit the gradual development of hyperglycemia is virtually different in the two genotypes, neither the plateau nor the half maximum time to reach the plateau revealed significant difference, as assessed by the Student’s t-test. Symbols represent mean ± S.E.M. of 8 mice; *P < 0.05 vs. the preinjection value, as determined with one way ANOVA followed with Bonferroni’s multiple comparison post-hoc test; n.s., not significant.

In vitro tandem [³H]deoxyglucose/[¹⁴C]6-glucose uptake in brain slices

This assay was carried out as before (Lemos et al. 2012; Pinheiro et al. 2016), with slight modifications. Before decapitation with a scissor, the mice were deeply anesthetized with halothane vapor in air, using 1 mL liquid halothane in a 9000 cm³-volume box. The mice were killed around 2:00 PM each experimental day to reduce potential circadian hormonal effects, and their brain was immediately placed in ice-cold Krebs-HEPES assay medium of the following composition: (in mM: NaCl 133, KCl 3, KH₂PO₄ 1.2, MgSO₄ 1.2, CaCl₂ 2.5, NHCO₃ 25, glucose 5.5, HEPES 10 [pH 7.4]). The pairs of hippocampi and the frontal cortices were rapidly dissected and 400 μm-thick transverse slices were cut parallel to the coronal plane with the help of a McIlwain tissue chopper. The slices were gently separated in ice-cold assay solution (carboxygenated with 95% O₂ and 5% CO₂), then transferred to a slice incubator using 4 chambers, each containing four 10-mm diameter baskets with 80-μm-pore nylon mesh bottom and maintained at 37°C in 50 mL of carboxygenated assay solution until the end of the experiment. Each basket contained either 6 hippocampal or 3 frontocortical slices of the same animal, and each chamber had slices from all the four cohorts to allow statistical comparison using repeated measures ANOVA. A sham experiment was also run on ice to measure the nonspecific external labeling.

After 1 h incubation under gentle gassing at 37°C, 263 μL of 3 M NaCl solution or KCl solution was added to one of the two chambers, to create the osmotic control and the KCl-depolarized conditions. Subsequently, 2-[³H](N)-deoxy-D-glucose ([³H]DG; 1 nM; specific activity; 60 Ci/mM; American Radiolabeled Chemicals, USA) and [¹⁴C]₆-(uniformly labeled)-D-glucose (50 nM; 360 mCi/mM; Perkin Elmer, USA) were pipetted into the chambers. The idea behind this is that [³H]DG is a glucose analogue generally thought to represent only the net uptake process, taken that once it enters the cell it stays trapped phosphorylated. On the contrary, [¹⁴C]₆-glucose is subject to metabolism, thus a difference between the final [³H] and [¹⁴C] contents in the same slice gives a rough readout of the dissipative glucose metabolism (Lemos et al. 2012).

Thirty min later, the baskets with the slices were transferred to large volumes of ice-cold assay medium for extensive washing, and then the slices were removed to 1 mL of NaOH (0.5 M). Once the slices were dissolved, 800 μL of these samples were counted for [³H] and [¹⁴C] by a dual-label counting protocol with the help of a 2900TR Tricarb β-counter (PerkinElmer). The remaining 200 μL from the samples were used for protein quantification with the bicinchoninic acid assay (BCA, Smith-assay) method (Smith et al. 1985). Original bath samples were also counted to measure the precise starting concentrations of the two tracers.

Western blotting

Remaining frontocortical slices from each mouse, not used for the glucose uptake experiments were homogenized with RIP A buffer (50 mM Tris/HCl [pH 8.0], 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 2 mM EDTA, proteases inhibitor cocktail, phosphatase inhibitor cocktail and 1 mM dithiothreitol) and cortex protein lysates were denatured at 95°C, for 5 min, in sample buffer (0.125 mM Tris [pH 6.8], 2% w/v SDS, 100 mM dithiothreitol, 10% glycerol and bromophenol blue). Thirty μg of total protein was resolved on 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes. The membranes were blocked with 5% (w/v) fat-free dry milk in Tris-buffered saline containing 0.1% (v/v) Tween 20, for 1 h at room temperature.

After blocking and washing, membranes were incubated overnight at 4°C with the primary antibodies against the different proteins studied: IRβ (1:1000), IGF-1R (1:1000), GLUT1 (1:2000), GLUT4 (1:1000), p-GSK3α (1:1000), GSK3α (1:1000), p-GSK3β (1:1000), GSK3β (1:1000), p-Akt (1:500), Akt (1:1000), CB₁R (1:1000), and β-actin (1:1000). After incubation, membranes were washed and incubated for 1 h at room temperature, with alkaline phosphatase-conjugated anti-rabbit, anti-guinea pig or anti-mouse antibody (Santa Cruz Biotechnology; all at 1:5000). The membranes were exposed to ECF reagent followed by scanning for blue excited fluorescence on the VersaDoc (Bio-Rad Laboratories, Amadora, Portugal). The generated signals were analyzed using the Image-Quant TL software.

Data handling

For data presentation and manipulation, we followed the rules of Curtis et al. (2015). All data represent the mean and S.E.M. of n ≥ 5 individual observations (animals). Means were tested for normality with the Kolmogorov-Smirnov test. Statistical significance was determined by one-way ANOVA of repeated measures followed by Bonferroni’s post-hoc test or two-tailed Student’s t-test, and P < 0.05 was accepted for significant difference, according to Curtis et al. (2015). Tests were performed and figures were prepared using the GraphPad Prism 5.0 software package

Results

Body weight and blood glucose changes

Long-term cerebral effects on glucose metabolism due to an insulinopenic/hyperglycemic (diabetic) state have not been compared between wild-type (WT) and CB₁R knockout (CB₁R-KO) mice. Therefore, we now assessed these effects seven weeks after the last streptozotocin injection. During this period, we monitored both body weight and blood glucose levels in all cohorts studied. The starting body weight of CB₁R-KO mice was 10.7 ± 1.3% lower (n = 16; P < 0.001) than their (WT) littermates (Fig. 1A). Surprisingly, weight loss was more severe in the diabetic CB₁R-KO mice, compared to the diabetic WT. To note, between the first and the fourth weeks after diabetes induction, the average weight loss was −6.5 ± 0.4% in the WT (−3.0 ± 0.2 g) vs. −8.5 ± 0.3% (−3.4 ± 0.1 g) in the CB₁R-KO (P < 0.001 for both the absolute and the relative differences). Nevertheless, the CB₁R-KO also recovered faster from weight loss since their body weight was not statistically different from their starting body weight, unlike the WT (Fig. 1A).

Normoglycemia of the WT mice (7.55 ± 0.63 mM) was not statistically different from that of the CB₁R-KO (7.34 ± 0.51 mM) (n = 16; P > 0.05). After the last streptozotocin injection, the two genotypes had slightly different kinetics in the rise of their blood glucose levels (Fig. 1B). Fitting a hyperbolic function onto the data point of each rat, the WT cohort exhibited a plateau of 31.7 ± 2.3 mM vs. the CB₁R-KO (27.0 ± 1.1) (n = 8; P = 0.09), while the time to reach half-maximum hyperglycemia was 2.3 ± 0.5 days for the WT and 1.5 ± 0.4 days for the CB₁R-KO (P > 0.1) (Fig. 1B)

Glucose uptake in frontocortical and hippocampal slices

Resting glucose uptake, as calculated from the [³H]deoxyglucose uptake amounted to 66.3 ± 3.8 nmol/mg protein in the frontocortical slices and 65.5 ±3.1 nmol/mg protein in the hippocampal slices (WT sham; NaCl controls; Figs. 2A and 3A). On the other hand, diabetes significantly reduced the resting glucose uptake by −18.1 ± 3.8% in the frontal cortex (n = 8; P < 0.05; Fig. 2A WT STZ) and −12.1 ± 4.2% in the hippocampus (n = 8; P < 0.05; Fig. 3A WT STZ). Interestingly, there were similarly lower glucose uptake values observed in the frontal cortex (−21.8 ± 6.4%; P < 0.05) and the hippocampus (−17.0 ± 4.4%; P < 0.05) of the CB₁R-KO sham animals (Figs. 2A and 3A; KO sham). Surprisingly, diabetes failed to further reduce resting glucose uptake in the CB₁R-KO (Figs. 2A, 3A; KO STZ), as the frontocortical uptake remained 19.1 ± 4.0% (P < 0.05) and the hippocampal uptake 19.0 ± 3.95% smaller (P < 0.05) than in the WT sham. No statistical difference was revealed when comparing the two cohorts of CB₁R-KOs.

Fig. 2.

Glucose uptake and metabolism in the frontal cortex. (A) Seven weeks of STZ-induced diabetes and the genetic deletion of the CB₁R cause mutually occlusive impairment on both the resting and the 15 mM K⁺-depolarization induced uptake of glucose in the frontal cortex of male mice of the CD-I strain. For the resting condition, KCl was substituted with 15 mM extra NaCl. (B) Glucose metabolism was similar among the four cohorts both under resting (+15 mM NaCl) and depolarized conditions (+15 mM KCl). Bars represent mean + S.E.M. of 8 mice; *P < 0.05 when basal glucose uptake was compared to the WT sham, ^$P < 0.05 when the depolarized slices were compared to their respective resting control, and ^#P < 0.05 when the depolarized glucose uptake was compared to the depolarized WT sham slices, as determined with one way ANOVA followed with Bonferroni’s multiple comparison post-hoc test.

Fig. 3.

Glucose uptake and metabolism in the hippocampus. (A) Seven weeks of STZ-induced diabetes and the genetic deletion of the CB₁R cause mutually occlusive impairment of the resting uptake of glucose in acute hippocampal slices of male mice. (B) While resting glucose metabolism was similar among the four cohorts, and 15 mM KCl-depolarization also triggered carbon atom dissipation of similar magnitudes in the WT sham, WT STZ and KO sham mice, it failed to stimulate glucose metabolism in the hippocampus of the KO STZ mice, pointing to a failure to cope with increased energy needs. For the resting condition, KCl was substituted with 15 mM extra NaCl. Bars represent mean + S.E.M. of 8 mice; *P < 0.05 when basal glucose uptake was compared to the WT sham, and ^$P < 0.05 when the depolarized slices were compared to their respective resting control, as determined with one way ANOVA followed with Bonferroni’s multiple comparison post-hoc test.

Depolarization with 15 mM K⁺ stimulated glucose uptake by 60.7 ± 12.9% in the frontal cortex (WT sham; Fig. 2A). While depolarization also stimulated glucose uptake in the remaining three cohorts, some difference between the cortex and the hippocampus were observed. In the cortex, the stimulated glucose uptake was strongly proportional in all four cohorts to their respective resting glucose uptake value, i.e. +66.4 ± 12.2% (P < 0.05; WT STZ), +68.3 ± 16.0% (P < 0.05; KO sham) and +65.1 ± 14.3% (P < 0.05; KO STZ) above resting uptake, and consequently, these stimulated nmol/mg (i.e. absolute) uptake values were also significantly different (P < 0.05) from those of the WT sham (Fig. 2A).

In regard to the hippocampal slices, glucose uptake was stimulated by 53.4 ± 8.9% in the WT sham (P < 0.05 vs. resting value; Fig. 3A). Depolarization stimulated glucose uptake was slightly greater in the diabetic WT cohort (+63.2 ± 18.4%; P < 0.05; WT STZ) and in the CB₁R-KO sham (+65.0 ± 17.3%; P < 0.05; KO sham), but proportionally much larger in the diabetic CB₁R-KO mouse slices (+99.0 ± 28.9%; P < 0.05; KO STZ), as compared to their respective resting values. Therefore, the absolute values of stimulated glucose uptake did not differ statistically among the four cohorts (P > 0.05; Fig. 3A)

Glucose metabolism in frontocortical and hippocampal slices

In the above uptake experiments, we simultaneously measured the accumulation of [³H] label associated with [³H]deoxyglucose and [¹⁴C] label derived from [¹⁴C]-U(uniformly labelled)-glucose. While the former is not metabolized, the latter is, and the discrepancy between the final [³H] and [¹⁴C] contents indicates a gross metabolism of glucose that dissipates carbon atoms e.g. in the form of [¹⁴C]O₂ and [¹⁴C]lactate into the bath (for details see Lemos et al. 2012, and Methods).

The resting glucose-derived carbon loss amounted to 115.2 ± 9.5 nmol/mg in the frontocortical slices, and 117.1 ± 9.6 nmol/mg in the hippocampus of the WT sham (Figs. 3A, 3B). High-K⁺-depolarization stimulated carbon dissipation by 80.1 ± 17.8% (frontal cortex; P < 0.05) and by 56.4 ± 13.9% (hippocampus; P < 0.05) above the respective resting values (Figs. 3A, 3B). Neither the genetic deletion of the CB₁R nor diabetes affected significantly the dissipation of glucose carbon atoms under either resting or depolarization in the frontocortical and hippocampal slices (Figs. 3A, 3B). The dissipative metabolism was also unaltered in the cortex of the diabetic CB₁R-KO mice (Fig. 3A), however, their hippocampal slices were unable to correctly respond to high-K⁺ depolarization, as the [³H]/[¹⁴C] ratio remained similar to the resting value (Fig. 3B).

Diabetes decreases cortical CB₁R density, and affects the insulin pathway

The above neurochemical data indicate that the lack of CB₁Rs triggers a diabetic phenotype in cerebral glucoregulation. We hypothesized that diabetes may affect CB₁R density to trigger impaired glucoregulation. Indeed, CB₁R density in cortical membranes was 21.6 ± 9.1% smaller in diabetic mice (n = 8; P < 0.05; Fig. 4A,B). Unfortunately, there was not enough hippocampal tissue for both glucose uptake and Western blotting.

Fig. 4.

CB₁R density and insulin signaling in WT and CB₁R-KO animals with and without diabetes: (A) Representative Western blot images and (B) the resulting bar graph documenting that seven weeks of STZ-induced diabetes significantly reduced CB₁R density in protein extracts from frontocortical slices of CD-I mice (note that 2 WT sham and 2 WT STZ mice are shown in the uppermost left panel). The CB₁R-KO sham displayed significant reduction in insulin receptor β-chain density, while the diabetic CB₁R-KO mice show marked increase in IGF-1R β-chain density accompanied with a reduction in Akt phosphorylation at Ser473. Neither the phosphorylation of glycogen synthase kinases nor the density of insulin independent (GLUT1) and insulin-dependent (GLUT4) glucose transporters were different throughout the cohorts. Normalization controls were as indicated below each protein of interest. Bars represent mean + S.E.M. of 5-8 mice; *P < 0.05 as determined with one way ANOVA followed with Bonferroni’s multiple comparison post-hoc test.

While diabetes did not significantly affect insulin receptor density in either WT or KO mice as compared to sham (n = 8; P > 0.05; Fig. 4A,B), the CB₁R-KO mice had significantly less insulin receptor β-chain immunoreactivity in their frontal cortex (P < 0.05; InRβ in Fig. 4A,B). The insulin-like growth factor 1 receptor β density remained unaffected in both the diabetic and the CB₁R-KO (n = 8; P > 0.05; IGF-1Rβ in Fig. 4A,B), however, the two conditions combined caused a 43.0 ± 11.0% increase in IGF-1Rβ density (n = 8; P < 0.05; Fig. 4A,B).

Akt phosphorylation at Ser473 followed a similar trend in that it became significantly different only in the diabetic CB₁R-KO, with a marked 30.0 ± 4.1% drop in phosphorylation, as compared to WT sham (n = 5; P < 0.05; Fig. 4A,B). Nevertheless, the phosphorylation of the downstream kinases GSK3α at Ser21 and 3β at Ser9 remained unaffected in diabetic WT, CB₁R-KO sham and diabetic CB₁R-KO (n = 8; P > 0.05 for both; Fig. 4A,B). Similarly, there was no statistical difference throughout the cohorts in the density of glucose transporters 1 and 4.

Discussion

This study shed light on two major findings. First, chronic type-1 diabetes lowers glucose uptake in forebrain areas responsible for cognition, memory and mental health, even in the absence of circulating ketone bodies. Second, the lack of CB₁Rs either in specific subcellular compartments or in the entirety of tissue seems to be the downstream link between hyperglycemia/insulinopenia and impaired cerebral glucoregulation.

We compared two different metabolic phenotypes, chronically type-1 diabetic mice and CB₁R global KO mice with their sham-injected wild-type (WT) littermates. In the fourth cohort, we rendered the CB₁R-KO mice also diabetic. There is robust literature on the roles of the CB₁R in the control of central and peripheral energy metabolism (Matias et al. 2008; Piazza et al. 2017), including the regulation of glucose homeostasis (Bermúdez-Silva et al. 2016), brain energetics (Bénard et al. 2012; Duarte et al. 2012; Harkany and Horvath 2017) and insulin/IGF-1 signaling (Bouaboula et al. 1997; Dalton et al. 2012; Kim et al. 2012; Labouèbe et al. 2013). Furthermore, the blockade or the genetic deletion of CB₁Rs reduces food intake and the mass of white adipose tissue (WAT), with a consequent increase in energy expanditure (Cota et al. 2003; Ravinet Trillou et al. 2004).

Even though CB₁R-KO mice have a lean phenotype and smaller body weight due to reduced WAT mass (Cota et al. 2003; Ravinet Trillou et al. 2004), they lost significantly more weight both in relative and absolute terms in the first month of diabetes, as compared to their WT littermates. Streptozotocin-induced diabetes may atrophy skeletal muscle and WAT and decrease liver weight, both in rats (Pillion et al. 1988; Price et al. 1996) and in mice (Chen et al. 2009). Likely, diabetic CB₁R-KO mice wasted more skeletal muscle protein than their WT littermates taken that their WAT reserve is already smaller, however this was not measured. We speculate that this could be due to loss of anabolic function of the CB₁R in the skeletal muscle of the KO mice. Namely, CB₁Rs can transactivate and phosphorylate the insulin receptor in the absence of its ligand (Dalton and Howlett 2012; Kim et al. 2012), which theoretically could attenuate wasting in diabetic WT mice.

The major aim of this study was to assess glucose uptake and metabolism in the frontal cortex and the hippocampus under chronic hyperglycemia/insulinopenia in the absence and the presence of the CB₁R. Impaired cerebral glucoregulation contributes to or even triggers brain diseases (de Ceballos and Köfalvi 2017; Zilberter and Zilberter 2017). Encephalopathy is a possible consequence of both insulinopenia and impaired peripheral glucoregulation (Miles and Root 1922; Brands 2004; McCall 2004). Type-1 diabetes in humans is normally medicated with intermittent insulin administration that can make blood glucose levels oscillate with much greater amplitudes than in healthy subject. Hence, the chronic hyperglycemia/insulinopenia animal models have limited predictability for the human disease. Albeit more studies are necessary in diabetic patients, it appears that hypoglycemia-unaware patients also have compromised cerebral glucose uptake at euglycemia and hypoglycemia (Cranston et al. 2001) while they appear to have preserved glucose transport through the blood-brain barrier (Duarte 2015).

In streptozotocin-induced diabetic models (STZ-DM) in rats and mice, impairments in hippocampal neurogenesis, synaptic plasticity and learning, neuroinflammation, tau hyperphosphorylation and oxidative stress have been described (Stranahan et al. 2008; Duarte 2015: Elahi et al. 2018). No significant increase in cell death has been reported in the hippocampi of STZ-DM rats after one month with diabetes (Guven et al. 2009), even though such animals (in that case, rats) exhibited altered memory and learning deficits accompanied with strongly impaired expression of neural cell adhesion molecules in the hippocampus and cortex (Baydas et al. 2003). After two months of diabetes induction, we also found no significant change in vesicular GABA and glutamate transporter densities in the hippocampi of STZ-DM rats (Baptista et al. 2011), which together with the above data suggest that lower glucose uptake is unlikely a consequence of neuronal loss.

In accordance with our present ex vivo glucose uptake data, a previous in vivo study found a 24% decrease in frontocortical and a 48% decrease in hippocampal glucose uptake of systemically injected [³H]deoxyglucose, as assessed by post-mortem autoradiography in STZ-DM rats after 4 months with diabetes (Jakobsen et al. 1990). Another in vivo study obtained a similar 32% reduction in global cerebral [¹⁴C]-3-O-methylglucose uptake in STZ-DM rats 8 days after diabetes induction (Mooradian and Morin 1991). None of the two studies observed significant change in blood flow in the brain regions of interest. The question then is how does the long-term insulinopenia cause neuroglycopenia. Our ex vivo method has the advantage of eliminating the confounding factors in diabetes, including the possible differences in the permeability of the blood-brain barrier, in blood glucose levels and in blood flow. Consequently, the impairments detected in our assays are exclusive to the brain parenchyma.

At the first glance, it is logical to assume that the brain takes up less glucose when circulating insulin levels are low. Indeed, brain cell express insulin-sensitive glucose transporters including the GLUT3 and GLUT4, and insulin can stimulate glucose uptake in cultured astrocytes and neurons (Clarke et al. 1984; Uemura and Greenlee 2006; Bakirtzi et al. 2009; Pearson-Leary et al. 2018). However, we found that in experimental conditions similar to the present study, insulin (1 nM-1 μM) had negligible effect on glucose uptake in hippocampal and frontocortical slices (not shown), prepared from fed mice and fed, fasted and STZ-DM rats, which is in accordance with a previous study (Abdul-Ghani et al. 2007). Moreover, the STZ-DM model not only failed to reduce cerebral insulin content but also increased it 2-3 times within a month spent with diabetes (Havrankova et al. 1979). It is known now that there is not only a concentrating insulin transport into the brain but also brain cells themselves are capable of synthesizing insulin (Gashemi et al. 2013; Molnár et al. 2014). Therefore, it is less likely that cerebral insulinopenia causes reduced glucose uptake in the slices.

An alternative explanation for lower glucose uptake in the STZ-DM brain slices could be a lower density of the insulin receptor or hypophosphorylation of its downstream targets. Nevertheless, neither our results nor a previous milestone study (Havrankova et al. 1979) found significantly altered insulin receptor density in the cerebral cortex of STZ-DM mice and rats. Furthermore, we did not detect significant impairment in Akt or GSK3α/β phosphorylation in the STZ-DM mice. Taken that diabetes failed to reduce glucose uptake and metabolism in the CB₁R-KO mice which already had an STZ-DM-like metabolic phenotype, its is safe to hypothesize that STZ-DM causes an impairment in the putative glucoregulator role of the CB₁R as an intermediate step leading to cerebral hypometabolism. In line with this, Zhang et al. (2007) found that chronic hyperglycemia itself is sufficient to downregulate neuronal CB₁R expression in a model of diabetic neuropathy.

The CB₁R-insulin receptor interaction is further supported by our protein density data: while diabetes per se did not affect insulin receptor density, the genetic deletion of CB₁R had already significantly reduced it. Furthermore, under insulinopenia in the CB₁R-KO cortex, the phosphorylation of Akt at Ser473 was strongly diminished. Both insulin and IGF-1 enhance the expression of the monocarboxylate transporter 2 in neurons via the PI₃K-Akt pathway, which is crucial for the astrocyte-neuron lactate shuttle (Chenal et al. 2008). Thus the markedly elevated IGF-1R density in the diabetic CB₁R-KO mice suggests some kind of incomplete compensatory mechanism.

Glucose is the most abundant but not the only energy source for the brain. We found recently that ketone bodies in excess are preferred substrates over glucose for neuronal acetyl-coenzyme A generation in hippocampal neurons (Valente-Silva et al. 2015). Importantly, in STZ-DM rats, circulating ketone levels can go up to 160% after 4 weeks in diabetes (Hawkins et al. 1986). This leads to a greater cerebral uptake of ketone bodies (Hawkins et al. 1986), resulting in up to 60% increase in citrate levels that can inhibit phosphofructokinase, glycolysis, and consequently, the uptake and oxidation of glucose in the brain (Ruderman et al. 1974). Taken that the activation of CB₁Rs in the outer leaflet of the mitochondrion (Bénard et al. 2012; Harkany and Horvath 2017) reduces intermediary metabolism in both neurons and astrocytes (Duarte et al. 2012), the lack of CB₁Rs in the KO animal is expected to increase the velocity of the Krebs-Szentgyörgyi (aka. citrate) cycle. In the presence of enough ATP derived from ketone bodies, citrate would inhibit the most important enzyme in the glycolysis pathway, i.e. phosphofructokinase, leading to decreased glucose uptake. It is important to note that acute CB₁R activation slowed the citrate cycle (Duarte et al. 2012) rather than reducing glucose uptake (Lemos et al. 2011; Köfalvi et al. 2016) in hippocampal slices.

Against this plausible explanation, in the absence of ketoacidosis, the lack of CB₁Rs is expected to increase cerebral glucose oxidation, albeit acute CB₁R blockade with the CB₁R antagonist, AM251 did not affect the citrate cycle (Duarte et al. 2012). Also, CB₁R activation triggers ketone body synthesis in astrocytes to support energy metabolism (Blázquez et al. 1999), hence the lack of CB₁R should theoretically increase glucose utilization rather than reduce it.

Another possibility is that the lack of CB₁Rs may hamper the trafficking or the glucoregulator function or both of the CB₂R. Namely, the CB₁R and the CB₂R form a functional heterodimer in the brain (Callén et al. 2012) and there is evidence for concomitant reduction in both CB₁R and CB₂R density in disease conditions (Solymosi and Köfalvi 2017). Now taken that CB₂R activation stimulates neuronal and astrocytic glucose uptake in the frontal cortex and the hippocampus by ~30% (Köfalvi et al. 2016), this process may become impaired in the CB₁R-KO mice, and may lead to cerebral hypometabolism.

Finally, the last testable hypothesis is the change in the function of the cation pumps in diabetic encephalopathy. It has been shown before that the activity of Na⁺/K⁺-ATPases is decreased in both the peripheral nerves (Greene and Lattimer 1983; Catanzaro et al. 2013) and in the blood-brain barrier (Jakobsen et al. 1987). Importantly, Na⁺/K⁺-ATPases are the biggest consumers of ATP and consequently, glucose in the brain, even under resting conditions (Astrup et al. 1981; Attwell and Laughlin 2001). Remarkably, CB₁R activation via G_i/o proteins stimulates the activity of Na⁺/K^+>-ATPases in synaptoneurosomes (Araya et al. 2007). Furthermore, a solid body of evidence claims that insulin directly enhances Na⁺/K⁺-ATPase activity in skeletal muscle, liver, kidney, adipocytes, corneal endothelial cells, lymphocytes, among others (Hatou et al. 2010). Thus, it is quite possible that the action of both insulin and CB₁R is necessary to exert a converging stimulation on neuronal Na⁺/K⁺-ATPases, and as a result, increased basal and K⁺-evoked energy expenditure.

In summary, reduced/absent CB₁R expression is sufficient to mimic and to occlude insulinopenia/hyperglycemia-induced impairment in forebrain glucose metabolism. These findings are mirroring our skin ageing data in these four cohorts, namely that 7-week STZ-DM reduced CB₁R density in the skin of WT mice. Also, collagen levels and antioxidant capacity were decreased, and a plethora of inflammatory markers were increased in the skin of the WT STZ, CB₁R-KO, and KO STZ mice with similar amplitude, that is, reduced/absent CB₁R expression was sufficient to mimic and to occlude insulinopenia/hyperglycemia-induced skin ageing (Leal et al. 2018). These results prompt the idea that the hypoactivity of the CB₁R could be a druggable pharmacologic target to treat peripheral and central pathologies in type-1 diabetes. As an example, insulin exerts an important antidepressant role via IGF-1R activation in the brain (Mueller et al. 2018). Accordingly, de Morais and colleagues (2016) showed that a systemic injection of the endocannabinoid anandamide counteracts depressive-like behaviour in STZ-DM rats 4 weeks after diabetes induction, in a CB₁R-dependent fashion. Hence, further studies are invited to compare the pathophysiological importance of the CB₁R in the STZ-DM model, but especially when combined with insulin administration.

Highlights.

• Hippocampal and frontocortical resting glucose use is 20% smaller in the CB₁R-KO mice

• KCl-stimulated glucose uptake was also lesser in frontocortical slices of the CB₁R-KO

• Seven weeks of type-1 diabetes mimicked the impaired glucose use found in the CB₁R-KO

• Diabetes also reduced frontocortical CB₁R density, which could reduce glucose use

• Indeed, diabetes failed to affect glucose uptake in the forebrain of the CB₁R-KO mice

Abbreviations

[³H]DG

[³H]-2-deoxy-D-glucose

2-AG

2-arachidonoyl-glycerol

3Rs

Replacement, Refinement and Reduction of Animals in Research

ARRIVE

Animals in Research: Reporting In Vivo Experiments

CB₁R(s)

cannabinoid CB₁ receptor(s)

FELASA

Federation for Laboratory Animal Science Associations

HEPES

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

IGF-1R(s)

insulin-like growth factor-1 receptor(s)

RIPA

radioimmunoprecipitation assay

SDS-PAGE

sodium dodecyl sulphate-polyacrylamide gel electrophoresis