Glucose transport and utilization are altered in the brain of rats deficient in n-3 polyunsaturated fatty acids

Abstract

Long-chain polyunsaturated (n-3) fatty acids have been reported to influence the efficiency of membrane receptors, transporters and enzymes. Because the brain is particularly rich in docosahexaenoic acid (DHA, 22:6 n-3), the present study addresses the question of whether the 22:6 n-3 fatty acid deficiency induces disorder in regulation of energy metabolism in the CNS. Three brain regions that share a high rate of energy metabolism were studied: fronto-parietal cortex, hippocampus and suprachiasmatic nucleus. The effect of the diet deficient in n-3 fatty acids resulted in a 30–50% decrease in DHA in membrane phospholipids. Moreover, a 30% decrease in glucose uptake and a 20–40% decrease in cytochrome oxidase activity were observed in the three brain regions. The n-3 deficient diet also altered the immunoreactivity of glucose transporters, namely GLUT1 in endothelial cells and GLUT3 in neurones. In n-3 fatty acid deficient rats, GLUT1-immunoreactivity readily detectable in microvessels became sparse, whereas the number of GLUT3 immunoreactive neurones was increased. However, western blot analysis showed no significant difference in GLUT1 and GLUT3 protein levels between rats deficient in n-3 fatty acids and control rats. The present results suggest that changes in energy metabolism induced by n-3 deficiency could result from functional alteration in glucose transporters.

Abbreviations

used

BBB, blood–brain barrier

CBF

cerebral blood flow

CO

cytochrome oxidase

2-DG

2-deoxy-d-[1–¹⁴C]glucose

LA

linoleic acid

LnA

α-linolenic acid

PBS

phosphate-buffered saline

PC

phosphatidylcholine

PE

phosphatidylethanolamine

PUFA

polyunsaturated fatty acid

ROD

relative optical density

SCN

suprachiasmatic nucleus.

A large proportion of membrane phospholipids in the CNS are composed of long-chain polyunsaturated fatty acids, namely arachidonic (20:4 n-6) and docosahexaenoic acid (22:6 n-3). These fatty acids must either be provided by diet or may be synthesized from linoleic (LA 18:2 n-6) and α-linolenic acid (LnA 18:3 n-3), respectively. In neural and retinal tissues, membrane phospholipids are particularly rich in 22:6 n-3 (Sinclair 1975; Bourre et al. 1990; McGee et al. 1994). Several studies have shown that an α-linolenic acid deficient diet lowers membrane 22:6 n-3 levels (Connor et al. 1990; Anderson and Connor 1994; Guesnet et al. 1997; Carriéet al. 2000), leading to behavioural alterations such as poor performance in learning tasks and brightness discrimination (Yamamoto et al. 1988; Bourre et al. 1989), as well as changes in the physiological properties of receptors, transporters and enzymes in neuronal membranes (Viani et al. 1991; Delion et al. 1994; Youdim et al. 2000). Moreover, Tsukada et al. (2000) reported that ‘dairy 22:6 n-3 supply restores the age-related impairment of the coupling between neuronal activation and regional cerebral blood flow (CBF) response to vibrotactile stimulation in the somatory cortex in monkey brain’.

As high levels of 22:6 n-3 have been reported in ;endothelial cells (Selivonchick and Roots 1976), mitochondrial and neuronal membranes (Youdim et al. 2000), we speculated that a modification in membrane polyunsaturated fatty acid (PUFA) composition could affect the function of transporters and of enzymes implicated in the coupling mechanism between CBF, metabolism and neuronal activity.

Glucose, the primary source of energy for mammalian brain, is transported across the endothelial cells of the blood–brain barrier (BBB) to neurones and astrocytes. In the brain, the most important glucose transporters are two isoforms of GLUT1, the 55-kDa and 45-kDa isoforms primarily detected in endothelial cells of the BBB and in glial cells, respectively (Harik et al. 1990; Bondy et al. 1992), and GLUT3 40 kDa, a neuronal glucose transporter (Nagamatsu et al. 1993). Brain regions with a high density of glucose transporters also possess high rates of utilization of cerebral glucose (e.g. cerebral cortex and hippocampus) (Maher et al. 1994). Sokoloff (1981) showed a linear relationship between local cerebral glucose utilization and CBF in the rat. The coupling between CBF and cerebral metabolism is used as an index of metabolic activation underlying the increase in neuronal activity (Gsell et al. 2000). According to physiological conditions, mitochondrial oxidative phosphorylation can vary widely in order to match ATP synthesis to energy demand. Cytochrome oxidase (CO), the last enzyme of the respiratory chain components inserted at the inner mitochondrial membrane, is regarded as a metabolic marker for neuronal functional activity (Wong-Riley 1989).

The purpose of this study was to investigate the effect of n-3 PUFA deficiency on glucose uptake and oxidative metabolism in the brain of rats fed on a diet deficient in LnA compared with rats fed on a diet with appropriate LnA supply. Three different brain regions were chosen: the fronto-parietal cortex, the hippocampus and the suprachiasmatic nucleus (SCN), which show a high rate of energy metabolism. The fatty acid composition of the two main phospholipid classes, phosphatidylcholine (PC) and phosphatidylethanolamine (PE), was studied in the three cerebral areas. The local expression of GLUT1 55 kDa and GLUT3 40 kDa in the three brain regions was investigated on control and n-3 fatty acid deficient rats. To find out whether the changes observed in immunoreactivity of the glucose transporters could be related to a modified utilization of glucose, experiments using autoradiographic 2-deoxyglucose method (Sokoloff et al. 1977) and an histochemical study (Wong-Riley 1979) were performed to evaluate the local rates of glucose uptake and of CO activity.

Materials and methods

Animals and diets

Wistar rats were housed in an air-conditioned room illuminated from 8 am to 8 pm with food and water ad libitum. In order to suppress some n-3 fatty acid maternal supply, two weeks before mating female rats were divided into two groups or fed different diets (UPAE, INRA Jouy en Josas, France): control and an LnA-deficient diet. All diets contained 6% lipids (6 g/100 g of diet) and differed only in oil vegetable content (Table 1). The control group lipid diet was a mixture of peanut oil and rapeseed oil containing ∼ 1200 mg of LA and ∼ 200 mg of LnA/100 g of diet. The LnA-deficient diet contained peanut oil with ∼ 1200 mg of LA per 100 g of diet. Pups were fed the same diet until adult age (3 months) when the experiments were performed. In this study we only used the first generation of rats fed an LnA-deficient diet over their life span. All experimental procedures were performed in the middle of the light phase when brain metabolic activity was the highest (Ximenes da Silva et al. 2000). All animal experiments were carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC).

Table 1.  Diet composition

	Control g/kg	n-3 deficient g/kg
^a Composition (g/kg of mineral mixture): CaHPO4−2H2O, 380; K2HPO4, 240; CaCO3, 180; NaCl, 69; MgO, 20; MgSO4−7H2O, 90; FeSO4−7H2O, 8.6; ZnSO4–H2O, 5; MnSO4–H2O, 5; CuSO4−5H2O, 1; NaF, 0.8; CrK(SO4)2−2H2O, 0.5; (NH4)6Mo7O24−4H2O, 0.02; KI, 0.04; CoCO3, 0.02; Na2SeO3−5H2O, 0.02. ^bComposition of vitamin supplement, triturated in dextrose (mg/kg of vitamin mixture): retinyl acetate (IU), 500 000; cholecalciferol (IU), 250 000; d,l-α-tocopherol acetate (IU), 5000; menadione (IU), 100; thiamine HCl (IU), 1000; riboflavin, 1000; nicotinic acid, 4500; d-calcium pantothenate, 3000; pyridoxine HCL, 1000; inositol, 5000; d-biotin, 20; folic acid, 200; ;cyanocobalamin, 1.35; l-ascorbic acid, 10 000; p-aminobenzoic acid, 5000; choline chlorhydrate, 75 000.
Casein	220	220
dL-Methionine	1.6	1.6
Corn starch	432.4	432.4
Saccharose	216	216
Cellulose	20	20
Mineral mixture^a	40	40
Vitamin mixture^b	10	10
Peanut oil	23.6	60
Rapeseed oil	36.4	–

Fatty acid analysis of phospholipid classes

Rats (n = 6 for each experimental group) were decapitated, and their brains were quickly removed and frozen in − 40°C isopentane. Thick coronal sections (300 µm) were cut on a freezing microtome. Under microscope, three sections (for each brain region) were selected to ensure that they covered cortex, SCN and hippocampus at the same plane in each animal. SCN and cortex were analysed in the same coronal sections. Hippocampus was studied over 900 µm from the first cells of the CA1 area.

Cerebral rat tissues were homogenized on ice with 2 mL of NaCl solution (9 g/L) containing butylhydroxytoluene (0.02 g/L), and their total lipids were then extracted according to Folch et al. (1957). Phosphatidylcholine (PC) and phosphatidylethanolamine (PE) were separated from total lipids by solid-phase extraction on a 500-mg aminopropyl-bonded silica column (BAKERBOND spe™ Amino; Mallinckrodt Baker Inc., Phillipsburg, NJ, USA) as described by Goustard-Langelier et al. (2001). Briefly, neutral lipids were eluted with isopropanol/chloroform (1 : 2, v/v) mixture from the total lipid extract deposited beforehand on the silica cartridge. A mixture of diethylether/acetic acid (98 : 2, v/v) then eluted free fatty acids and acetone-eluted PC. After elution of sphingomyelin with acetonitrile/n-propanol mixtures (1 : 1, v/v), PE was finally eluted by methanol.

After evaporation under nitrogen, PC and PE were transmethylated with 10% boron trifluoride (Fluka, Socolab, Paris, France) at 90°C for 20 min according to the method described by Morisson and Smith (1964). Fatty acid methyl esters were injected into a Carlo Erba gas chromatograph (HRGC 5300; Fisons Instruments, Arcueil-Cachan, France) equipped with an on-column injector, a flame ionization detector and a CP WAX 52 CB bonded fused-silica capillary column (50 m × 0.2 mm id with 0.20 µm film thickness) purchased from Chrompack (Les Ulis, France). Hydrogen was used as the carrier gas at 1–2 mL/min flow rate. Fatty acid methyl esters with a C8–C24 chain length and dimethyl acetals were identified by comparing equivalent chain lengths with those of commercial standards (Nu-Check-Prep Inc., Coger, Paris, France), and quantified with a computing integrator using the Nelson Analytical Program System (SRA, Gagny, France). Fatty acid composition was expressed as weight percentage.

Blood glucose levels

Animals were fasted for 12 h and then lightly anaesthetized with ketamine (100 mg/kg) before intracardiac blood collection (1 mL) into fluoride oxalate tubes. Blood glucose levels of control (n = 6) and n-3 fatty acid deficient (n = 8) animals were measured using a glucose enzymatic colour method (Biotrol Diagnostic, Lyon, France). This method is based on the glucose oxidase reaction, which catalyses the conversion of glucose (in the presence of oxygen) to gluconic acid and hydrogen peroxide so that the developed colour intensity is proportional to the starting concentration of glucose (Trinder 1969).

Immunocytochemistry

Rats were deeply anaesthetized with sodium pentobarbital(120 mg/kg) and perfused intracardially with a mixture of 0.9% (wt/vol) NaCl and 1% (wt/vol) nitrite at 37°C followed by 4%(wt/vol) cold paraformaldehyde in 0.1 m phosphate-buffered saline (PBS). Brains were removed and post-fixed in the same fixative for 4 h before being immersed in 30% sucrose in PBS until they sank. Serial 30 µm coronal sections were cut on a freezing microtome. For each animal, 24 sections containing cortex, SCN and hippocampus were selected under the microscope with dark-field illumination. SCN and cortex were analysed in the same coronal sections. Hippocampus was studied over 720 µm from the first cells of the CA1 area. Free-floating sections were processed for immunocytochemistry as follows: endogenous peroxidase was blocked by incubation in a solution of 0.3% (v/v) H₂O₂ for 30 min. After treatment in PBS containing 0.3% (v/v) Triton X−100 (PBST) and blocking serum (5% v/v normal horse serum) for 1 h, sections were incubated for 72 h at 4°C with the appropriate goat polyclonal primary antibody (either anti-GLUT1 at 1 : 1000 dilution or anti-GLUT3 at 1 : 250 dilution). Polyclonal antibodies against a synthetic 20 amino acid C-terminal sequence peptide of either GLUT1 or GLUT3 were used (Santa Cruz Biotechnology, Santa Cruz, CA, USA). After three rinses in PBST, sections were incubated in excess secondary antibody (biotinylated horse anti-goat, 1 : 400 dilution; Vector Laboratories, Burlingame, CA, USA) overnight at 4°C. Sections were then rinsed three times and incubated with the ABC kit (1 : 100 dilution; Vector) for 1 h. Finally, sections were rinsed and incubated with 0.02% (wt/vol) 3–3′diaminobenzadine tetrahydrochloride with 0.1% (v/v) hydrogen peroxide for 5 min. Sections were mounted on gelatin-coated slides and covered with montage medium. Negative controls were performed using standard procedural omission of the primary antibodies. GLUT immunolabelling was analysed every two sections using a video-based analysis system (Biocom, Les Ulis, France) equipped with an Olympus BH-2 microscope and a Cohu solid state camera. Immunoreactive GLUT3 cells were pointed in 12 serial sections from cortical and hippocampal regions for each animal in the two experimental groups (n = 5 for controls; n = 4 for n-3 fatty acid deficient rats). In the SCN, neurone staining with anti-GLUT3 was not adequately distinguishable. GLUT1 immunoreactivity (Ir) analysis was performed as previously described (Vilaplana and Lavialle 1999). Briefly, the image is automatically binarized, in order to subtract background noise and thereby retain only the staining, and the percentage of staining is determined for an area of 350 µm × 350 µm in parietal cortex, of 350 µm × 250 µm in hippocampus, and over the area of the whole SCN. The level of GLUT1 staining was evaluated using this semiquantitative method developed in order to provide similar results regardless of the observer. Measurements were performed on 12 sections for each brain region in each animal (n = 4 for each group).

Western blotting

Collection of tissues was performed as described above in ‘fatty analysis’ section. Cerebral cortex, hippocampus and SCN from decapitated animals (n = 4 for each group) were homogenized using a buffer consisting of 40 mm Tris-HCl, pH 7.4, 250 mm sucrose, 5 mm dithiothreitol, 2 mm EGTA and a cocktail of protease inhibitors (leupeptine, aprotinin and pepstatin at 2 µg/mL). Post-nuclear supernatants were obtained by centrifugation at 1000 g for 15 min at 4°C. Protein samples (30 µg) were resolved by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE) according the method of Laemmli (1970). Proteins were electrotransferred onto nitrocellulose membranes for 3 h at 300 mA (transfer buffer: 25 mm Tris, 192 mm glycine, 20% (v/v) methanol, pH 8.3). Membranes were blocked in PBS containing 0.3% (v/v) Tween 20 plus 3% (v/v) bovine serum albumin for 60 min, and then incubated with either GLUT1 or GLUT3 antibodies (1 : 1000 dilution; Santa Cruz Biotechnology) at 4°C overnight. After three washes in PBS Tween, membranes were incubated with peroxidase-labelled anti-goat secondary antibody (diluted 1 : 10000; Sigma) for 60 min at 20 ± 2°C. After extensive washes in PBS blots were developed using an enhanced chemiluminescence kit (Amersham Life Science Little Chalfont, Buckinghamshire, UK). Signals were quantified by densitometric scanning (IM1D; Pharmacia, Peapack, NJ, USA).

Measurement of cerebral glucose uptake

Cerebral glucose utilization was evaluated using the semiquantitative autoradiographic 2-deoxyglucose method (Sokoloff et al. 1977). Eight adult Wistar rats were used in each experimental group. They were injected intramuscularly with 125-µCi/kg 2-deoxy-d-[1–¹⁴C] glucose (2-DG; ICN Pharmaceuticals, Orsay, France) 45 min before decapitation. Brains were then rapidly dissected out and frozen in − 40°C isopentane and stored at − 80°C until sectioning. Coronal sections (20 µm) through the SCN, cortex and hippocampus were prepared and mounted on slides and exposed to BioMax MR film (Kodak, Rochester, NY, USA) for 6 days. Sections were selected as described for the immunohistochemistry experiment. Brain slices were counterstained with cresyl violet. Brain regions were delineated using the computer image analysis system (Biocom) and matched with autoradiogram images. 2-DG uptake was evaluated by measuring relative optical density (ROD), which was defined as the optical density measured over region of interest/optical density of reference. The corpus callosum was taken as the reference region because its metabolic activity does not vary significantly (Ximenes da Silva et al. 2000).

Cytochrome oxidase histochemistry

Cytochrome oxidase histochemistry was performed according to the methods of Wong-Riley (1979) and Gonzalez-Lima and Garrosa (1991) with modifications. Briefly, rats (n = 4 for each group) were deeply anaesthetized with sodium pentobarbital (120 mg/kg) and perfused intracardially with a mixture of 0.9% (wt/vol) NaCl and 1% (wt/vol) nitrite at 37°C, followed by 2.5% (wt/vol) paraformaldehyde and 1.5% (wt/vol) glutaraldehyde in 0.1 m PBS buffer at 4°C. Brains were dissected out and post-fixed for 1 h in the same fixation solution before immersion in 30% (wt/vol) sucrose in PBS until they sank. Coronal sections (30 µm) containing SCN, cortex and hippocampus were selected under the microscope with dark-field illumination as described for immunohistochemistry experiment. Twelve serial slices along the rostro-caudal axis of the SCN, cortex and hippocampus were processed for cytochrome oxidase histochemistry. Free-floating sections were initially incubated with 27.5 mg cobalt chloretum, 10 g sucrose, 50 µL DMSO, in 100 mL 0.1 m PBS. After two 5-minute PBS rinses, sections were incubated for 2 h at 37°C with 50 mg DAB, 5 g sucrose, 39 mg cytochrome c, in 100 mL 0.1 m PBS. The reaction was stopped with two 5-min rinses in 0.1 m PBS and sections were mounted. CO activity was evaluated using the image analysis system and ROD was measured as described for 2-DG uptake.

Statistical analysis

For phospholipid composition, two-way analysis of variance (anova) with diet and brain region, as main factors, was used to evaluate the differences between brain regions, and the effect of diet for three fatty acids (20:4 n-6, 22:5 n-6 and 22:6 n-3) in PC and PE. When significant effects (p < 0.05) were found, the Tukey test was performed for post-hoc comparison.

For immunocytochemistry, histochemistry and 2-DG results, 12 sections per region were examined in all the animals. The mean level for each animal in each brain region was estimated by averaging measures of the 12 sections. Data are expressed as mean ± SD. Control and deficient values were compared separately for each brain structure using Student's t-test. The level of significance was set at p < 0.05.

Results

Membrane phospholipid composition

Polyunsaturated fatty acid composition of PC and PE in the three brain regions studied was reported in Fig. 1. Only PUFAs that exhibited major changes, under LnA-deficient diet, were reported, i.e. 22:6 n-3, 20:4 n-6 and docosapentaenoic acid (22:5 n-6).

Figure 1. Effect of α-linolenic dietary deficiency on PUFA composition in rat brain. Mean levels of 20:4 n-6, 22:5 n-6 and 22:6 n-3 in phosphatidylcholine (left-hand column) and phosphatidylethanolamine (right-hand column) are reported for cortex, hippocampus and suprachiasmatic nucleus (SCN) as a weight percentage of total fatty acids (n = 6). **p < 0.01, ***p < 0.001 (anova), significant difference in the level of each fatty acid between control and n-3 fatty acid deficient groups.

Phosphatidylcholine

In both groups, the concentration of 20:4 n-6 and 22:6 n-3 varied significantly among the three brain regions (p < 0.001). In the control group, 20:4 n-6 was the major PUFA. The highest levels were observed in cortex in 20:4 n-6 (15.33 ± 0.3 weight percentage total fatty acid) and 22:6 n-3 (7.62 ± 0.5). SCN exhibited the lowest levels in both 20:4 n-6 (4.38 ± 1.2) and 22:6 n-3 (2.11 ± 0.5). A significant effect of diet (p < 0.001) was detected on the rate of 20:4 n-6, 22:5 n-6 and 22:6 n-3. In the n-3 fatty acid deficient group a significant increase of 22:5 n-6 was detected in all regions, although a significant decrease in 22:6 n-3 in cortex (73%) and hippocampus (65%) was revealed. No change in SCN was observed. On the other hand, levels of 20:4 n-6 significantly dropped in cortex (51%), increased in SCN (37%), and did not vary in hippocampus.

Phosphatidylethanolamine

In both groups, the concentration of 20:4 n-6 varied significantly among the three brain regions (p < 0.001). In the control group, 22:6 n-3 was the most important PUFA in the three regions. In the n-3 fatty acid deficient group, analysis of the PUFA profile showed a significant increase of 20:4 n-6 and particularly of 22:5 n-6 in membranes in the three brain regions but a remarkable decrease of 22:6 n-3. The percentage of 22:6 n-3 was reduced by 49% for cortex, 46% for hippocampus and 30% for SCN. The increase in 20:4 n-6 reached 31%, 25% and 11% in cortex, hippocampus and SCN, respectively.

The global effect of LnA-deficient diet on the PUFA composition of membrane phospholipids thus appears to induce a significant decrease in 22:6 n-3 and an increase in 22:5 n-6 in the two classes of phospholipids. In contrast, changes in 20:4 n-6 are not so clear, with higher or lower levels depending on the brain region and phospholipid class.

Physiological variables

Rats were weighed before each experiment. For all of the rats used, no significant difference in body weight was observed between the control (427.1 ± 45.6) and the n-3 fatty acid deficient (405.8 ± 58.3) group.

Plasma glucose levels were measured at the middle of light phase. Both groups fed the control or LnA-deficient diet showed normal glycemia (normal 0.9–1.92 g/L). However, a slight decrease was observed in the n-3 fatty acid deficient group compared with control group. In this experiment the body weights of the control and the n-3 fatty acid deficient groups showed a similar propensity (Table 2).

Effect of α-linolenic acid-deficient diet on brain glucose uptake

As shown in Fig. 2 (left column), [¹⁴C]2-DG uptake was decreased in the brain of n-3 fatty acid deficient rats. Autoradiograms showed a defined labelling in the three brain regions where SCN could be distinguished in control animals but not in n-3 fatty acid deficient animals. Glucose uptake was about 30% lower in cortex, hippocampus and 35% in SCN compared with controls.

Figure 2. Effect of α-linolenic dietary deficiency on glucose uptake (left-hand column) and cytochrome oxidase activity (right-hand column) within cortex, hippocampus and suprachiasmatic nucleus (SCN). Glucose uptake and CO were measured using the autoradiographic 2-deoxyglucose (2-DG) method and a histochemical study, respectively. Values are the mean ± SD of relative optical density measured in each region (12 sections per region) of each animal (n = 8 per group for 2-DG uptake; n = 4 per group for CO). *p < 0.05, **p < 0.01, significantly different from control group.

Effect of α-linolenic acid-deficient diet on oxidative phosphorylation

Optical density data revealed a significantly lower enzyme activity in the cerebral cortex, hippocampus and SCN of n-3 fatty acid deficient animals compared with controls (Fig. 2, right-hand column). CO histochemistry showed a differential reduction of staining in cortex, hippocampus and SCN with a lower staining of 20%, 30% and 40%, respectively.

Effect of α-linolenic acid-deficient diet on BBB and neuronal glucose transporters immunoreactivity

Western blot analysis (Fig. 3a) showed bands of GLUT1 and GLUT3 proteins that migrated, respectively, at 55 and 40 kDa, thus characterizing the glucose transporter of endothelial cells of the BBB and the neuronal glucose transporter, respectively. No significant change in expression of either glucose transporter protein was detected in the three brain areas studied in n-3 fatty acid deficient animals using this method (Fig. 3b).

Figure 3. Effect of α-linolenic dietary deficiency on GLUT1 and GLUT3 glucose transporters in rat brain. (a) Western blot analysis (30 µg of total protein), showing the protein expression of GLUT1 and GLUT3 transporters in cortex, hippocampus and suprachiasmatic nucleus (SCN) of control and n-3 fatty acid deficient rats. The polyclonal antibodies used (Santa Cruz Biotechnology) detected isoforms present in blood brain barrier for GLUT1 55 kDa and in neurones for GLUT3 40 kDa. Two independent experiments were performed. Two animals per group were analysed in each experiment. (b) Densitometric analysis of blots was performed using a Scan image analysis software (IM1D; Pharmacia). No difference was detectable between control (n = 4) and n-3 fatty acid deficient (n = 4) group in the three brain regions.

Figure 4 shows the GLUT1 and GLUT3 immunoreactivity (Ir) pattern within cortex. GLUT1 is distributed in a diffuse pattern throughout the brain, whereas GLUT3 has more discrete regional localization. Immunoreactivity for GLUT1 was detected in cerebral microvessels, whereas immunoreactivity for GLUT3 was detected in the cell body, axons and dendrites of neurones. However, GLUT1-Ir positive endothelial cells gave a lower percentage of staining area: about 50% lower in cortex, 36% lower in hippocampus, and 46% lower in SCN of n-3 fatty acid deficient animals compared with controls (Fig. 5, left column). Quantification of cells stained for GLUT3 revealed an opposite effect to that of GLUT1 with a dramatic increase in cortex (92%) and hippocampus (30%) of n-3 fatty acid deficient rats (p < 0.01; Fig. 5, right column). Within the SCN, because this area consists of parvocellular neurones, no labelling of the neuronal glucose transporter was detected under optical microscope.

Figure 4. Immunocytochemical staining of GLUT1 (a and b) and GLUT3 (c and d) in the cortex of control (a and c) and n-3 fatty acid deficient (b and d) rats. Endothelial cells and neurones glucose transporters were detected using GLUT1-C20 and GLUT3-C20 polyclonal antibodies (Santa Cruz Biotechnology) developed against a synthetic 20 amino acid C-terminal sequence peptide. Quantitative analysis is reported in Fig. 5. Scale bar = 50 µm.

Figure 5. Effect of α-linolenic dietary deficiency on immunoreactivity of GLUT1 and GLUT3 glucose transporters in rat brain. Percentage of GLUT1 immunoreactivity-(Ir) (left-hand column column), was measured on each section (n = 12) containing cortex, hippocampus and suprachiasmatic nucleus (SCN), in n-3 fatty acid deficient rats (n = 4) and controls (n = 4). Bars represent the mean ± SD of GLUT1-Ir percentage. GLUT3-Ir cell number (right-hand column) was measured in cortex and hippocampus of n-3 fatty acid deficient rats (n = 4) and controls (n = 4). Immunoreactive neurones were not distinguishable in the suprachiasmatic nucleus. Cell counts were performed in each section (n = 12) containing cortex and hippocampus. Data are shown as mean ± SD. *p < 0.05, **p < 0.001 significantly different from control group.

Discussion

The effect of LnA dietary deficiency on the phospholipid fatty acid composition in cerebral membranes is well established (Bourre et al. 1984). Several studies have reported that this alteration is associated with learning and behavioural deficits (Bourre et al. 1989; Wainwright 1992; Moriguchi et al. 2000), changes in local neurotransmission (Delion et al. 1994; Zimmer et al. 2000) or hormonal release (Gazzah et al. 1993). These effects were observed in animals maintained for two or three generations on an LnA-deficient diet. We show here that feeding rats an LnA-deficient diet for ;one generation is sufficient to induce disturbances in membrane PUFA composition, and consequently significant and important alterations in brain metabolic activity. In our ;study, rats showed a significant reduction of 22:6 n-3 fatty acid (30–50%) within brain membrane phospholipids, ;and a remarkable decrease in glucose uptake (30–35%) and ;CO activity (20–30%) in the three brain regions. We assume that these effects can be attributed primarily to the ;alteration of 22:6 n-3 in membrane phospholipid composition.

Though we did not measure PUFA phospholipid levels in the mitochondrial membrane, we suppose that our experimental procedure induced a notable reduction of 22:6 n-3 mitochondrial phospholipids because several studies have shown dietary fat induced changes in mitochondrial phospholipid fatty acid composition (Tahin et al. 1981; Barzanti et al. 1994; Marteinsdottir et al. 1998). It has been proposed that phospholipids containing 22:6 n-3 could be necessary for the assembly of mitochondrial enzyme complexes at inner membrane, and that a reduction of 22:6 n-3 level would reduce oxidative phosphorylation (Infante 1987) and the activity of oxidative enzymes such as succinate dehydrogenase and cytochrome oxidase (Thomas et al. 1993). Here, in n-3 fatty acid deficient animals, we showed that a significant decrease in CO activity and also in glucose uptake could be explained as a consequence of the depressed oxidative phosphorylation.

Although glucose uptake decreased, GLUT3 immunoreactive cells increased, and immunoreactivity of GLUT1 declined. Similar results were reported in rats after one week of moderate hypoglycemia where glucose utilization was decreased, the mean density of glucose transporters GLUT1 remained unchanged whereas the mean density of GLUT3 increased slightly (Duelli et al. 1999). In our study, n-3 fatty acid deficient animals had normal blood glucose levels, which means that glycemia did not contribute to the decrease in BBB GLUT1-Ir during n-3 deficiency.

In n-3 fatty acid deficient rats, the reduced oxidative phosphorylation could function as a signal to neuronal cells to mobilize GLUT3 at the neuronal membrane surface. Thus, GLUT3 translocation could represent a mechanism for sustaining glucose uptake into neurones to ensure cellular brain metabolism and functionality. Nevertheless, the higher immunoreactivity demonstrated for GLUT3 in neuronal cells of n-3 fatty acid deficient rats did not result in normal glucose uptake into the cells.

In contrast to GLUT3, GLUT1-Ir levels were lower in endothelial cells of the cortex, hippocampus and SCN of n-3 fatty acid deficient rats. It has been demonstrated that endothelial cell GLUT1 has an asymmetric distribution across the luminal (blood side) and abluminal (brain side) membranes (Gerhart et al. 1989). An electron microscopic immunogold study in rat brain has reported that 12% is present at the luminal and 48% at the abluminal membrane, with 40% in the cytoplasmic compartment, suggesting that recruitment of the cytoplasmic pool could be involved in regulation of GLUT1 at the BBB (Farrell and Pardridge 1991). More recently, Simpson et al. (2001) suggested that the lower distribution of GLUT1 described at the luminal membrane could be explained by a conformational change in GLUT1 protein that makes immunodetection difficult. Nugent et al. (2001) reported a significant increase in basal glucose uptake in 3T3-L1 adipocyte-cultured cells supplemented with 20:4 n-6 or 22:6 n-3, and no effect on total cellular levels of glucose transporters, but significantly increased levels of GLUT1 at the plasma membrane. They hypothesized, either a translocation of glucose transporter to ;the plasma membrane, or a reduction of the rate of internalization. In our study, because we showed modifications in GLUT1 and GLUT3-Ir but no change in protein levels, we can address an equivalent hypothesis. A conformational change in GLUT1 protein or an alteration in the translocation mechanism for the regulation of transporter concentration in the luminal or abluminal membranes caused by membrane phospholipid modifications could make glucose transport through the BBB difficult.

Altogether, the decrease in immunoreactivity of GLUT1 at the BBB associated with enhanced GLUT3-Ir in neurones shows that there is a differential regulation of these different forms of glucose transporters for brain glucose uptake and utilization by neuronal cells in n-3 fatty acid deficient animals. GLUT3 glucose transporter is reported to possess a higher affinity to glucose than GLUT1. Therefore, it is thought that the higher GLUT3 expression in the neurones might improve glucose uptake from the extracellular space at low glucose concentrations (Gould and Seatter 1997).

In conclusion, we suggest that neurones tend to balance energy deficiency, and that structural changes in membranes do not allow glucose transporters to furnish the supply of glucose to the brain cells, and consequently changes observed in CO would not be a direct result of the changes in membrane composition, but would be secondary to change in glucose transport.

Although we did not evaluate the effects of the n-3-deficient diet on the glucose transporter or cellular metabolism in glial cells, astrocytes certainly play an important role in the metabolic changes observed as they participate in metabolic coupling with neurones (Magistretti et al. 1999) and in the delivery of essential fatty acids to the BBB and brain (Bernoud et al. 1998).

Acknowledgements

We are grateful to Marie-Thérèse Gozzelino, Régine Monnerie, Catherine Papillon for technical assistance, Manuel Costa da Silva for animal care and Alan Strickland for revising the English language.